Flexible high performance lithium ion battery electrode based on a free-standing TiO₂ nanocrystals/carbon cloth composite

Abstract

Wearable and flexible electronics is an emerging and popular area of research. However, the availability of flexible energy storage to accompany these devices is lacking. Herein, a flexible electrode based on a composite of TiO₂ nanocrystals and carbon cloth was synthesized by a simple drop-casting method followed by a heat treatment. The obtained TiO₂/carbon cloth electrode shows a reversible capacity of 280–310 mA h g⁻¹ at a current density of 100 mA g⁻¹. The composite electrode also exhibits excellent mechanical stability even after 100 cycles of flexion. A proof-of-concept flexible pouch half-cell was assembled and it was able to power an LED bulb under flexion.

---

1. Introduction

In recent years, significant attention has been focused on the search for superior electrochemical energy storage devices due to the popularity of hybrid electric vehicles, electric vehicles and wearable electronics.^1,2 Currently, the gold standard for electrochemical energy storage is the lithium-ion battery (LIBs).^2,3 LIBs are more popular compared to other forms of electrochemical storage due to their high energy density, relatively long cycle lives and low self-discharge. These advantages make LIBs perfect for many broad applications such as hybrid electric vehicles and portable electronics. However, one of the limitations of LIBs is their rigidity which limits their application in up and coming technologies such as flexible and wearable electronics.^4–6 Cui et al. has shown that the crucial factor for achieving flexible LIBs is the development of flexible electrode materials.⁷ There are several important parameters to consider when designing a flexible electrode for LIBs that has high performance. There has been numerous approaches to the development of flexible electrodes such as printable flexible batteries,⁴ paper based batteries^7,8 and carbon based batteries.^9–11 The performance of the electrode relies on several properties of the material such as high electron and ion diffusivities, high surface area and good electrical contact between the active material and current collector. As such, there has been a focus on the search for materials or combination of materials with the aforementioned properties. Amongst the numerous strategies to obtain such materials, synthesis of nanostructured materials have been promising. Nanostructuring allows for increased available surface area of the electrode in addition to shortening lithium ion diffusion path lengths, allowing for both greater energy and power density.¹² Carbon cloth has recently been a popular choice as a current collector and substrate for the growth of various nanostructured active material. Its flexibility, relatively high electron conductivity and robustness makes it a great alternative to metal foils,^10,11,13 which is traditionally used as a current collector in conventional LIBs. Titanium dioxide (TiO₂) is a promising candidate for electrode materials and is also used in a wide range of applications from photocatalysis^14,15 to solar cells^16,17 due to its low-cost, natural abundance, low toxicity and high stability in various solvents.¹⁸ Most recently, TiO₂ and its associated polymorphs have been investigated for use in LIBs and supercapacitors.^19–24 However, bulk TiO₂ has its own drawbacks such as poor electronic conductivity and slow lithium-ion diffusion kinetics.²⁰ Therefore, many strategies have been developed to overcome the limitations of TiO₂ including the synthesis of novel TiO₂ structures,^18,25 fabrication of TiO₂ composites,^10,14,26,27 and decreasing the particle size of TiO₂.^28,29 For example, Balogun et al. synthesized a TiO₂ and titanium nitride composite nanowire that was used as a flexible anode for LIBs with a reversible capacity of 240 mA h g⁻¹.¹⁰ Yang et al. doped TiO₂ with Nb in order to form nanoplate composites that have a reversible capacity of 220 mA h g⁻¹.²⁷ Chen et al. fabricated a free standing carbon nanotube and TiO₂ composite that delivers a reversible capacity of 270 mA h g⁻¹.³⁰ Wang et al. fabricated a hierarchical 3D structure of TiO₂ on carbon nanowire arrays, achieving a flexible battery anode with a high capacity and cycling stability.³¹

Herein, a novel flexible binder-free TiO₂ anode material has been successfully prepared by direct growth TiO₂ nanocrystals on carbon cloth using a solvothermal process with the assistance of oleic acid. The oleic acid allows the TiO₂ nanocrystals to be well dissolved in organic solvents which results in the ability to be dispersed more evenly on the surface of the substrate. This is an advantage as traditional electrode fabrication methods involves preparation of slurry which may have different active material dispersion, resulting in difference in performance from batch to batch. Carbon cloth was chosen as the substrate due to its flexibility, something that traditional current collectors such as copper foil collector lack. This resulting composite electrode has excellent reversible capacity of 270 mA h g⁻¹ at current density of 100 mA g⁻¹, which is close to the theoretical capacity of TiO₂. The excellent performances are attributed to the unique structure and the small size of TiO₂ nanocrystal. The method presented herein can provide a reference to fabricate other flexible energy storage devices.

2. Experimental

2.1 Materials preparation

All chemicals purchased from Sigma-Aldrich and were used as-received unless otherwise noted. Commercial carbon cloth was treated as follows. First, carbon cloth was sonicated in ethanol, methanol and isopropanol for 15 min, respectively, and then dried in an oven for half an hour. Next, the resulting clean carbon cloth was then heat-treated at 900 °C in a tube furnace under an inert argon atmosphere.

TiO₂ was synthesized using a two-phase solvothermal method reported elsewhere.³² The water phase was made by mixing 20 mL of distilled deionized water and 0.2 mL of tert-butylamine, while the oil phase was composed of 20 mL of toluene, 330 μL of titanium isopropoxide and 2 mL of oleic acid. The two separate phases were then transferred into a 100 mL steel autoclave with a Teflon liner. The autoclave was heated to 180 °C for 12 h and cooled down to room temperature. The resulting oleic acid-capped TiO₂ was washed with methanol and re-dispersed in toluene. This stock solution was then diluted to the desired concentration. The solution of nanocrystals in toluene was stable and did not observe any precipitates even after several weeks.

The fabrication process of flexible binder-free TiO₂ electrode was as follows. First, carbon cloth was cut into 12 mm circular electrodes using a commercial electrode punch. Then, TiO₂/toluene solution was added dropwise using a pipette onto the carbon cloth. The resulting composite material was then annealed in a tube furnace under air atmosphere at 450 °C for 3 hours at a heating rate of 1 °C min⁻¹ to remove the oleic acid capping agent. The approximate final TiO₂ loading of each electrode is 0.6 mg cm⁻². The composite electrode was then assembled into coin cells under an inert argon atmosphere with a cellulose separator and lithium metal counter-electrode.

2.2 Physical characterization

The morphologies and nanostructures of carbon cloth, TiO₂ nanocrystals and the composite electrodes were characterized using a field emission scanning electron microscope (FE-SEM, Zeiss LEO 1530, 10 kV). The crystal structure and purity of the synthesized TiO₂ crystals was determined using powder X-ray diffraction (XRD, Bruker AXS D8, 0.154 nm Cu-α source) and Raman spectroscopy (Brukker SENTERRA). Thermogravimetric analysis (TGA, TA Q500, Ar atmosphere, 1 °C min⁻¹ ramp rate) was done to confirm the actual TiO₂ and oleic acid mass ratios. The mechanical stability of the electrode was tested by subjecting the as-prepared composite electrodes to 100 bend cycles, where 1 cycle is composed of going from un-flexed (shown in Fig. S5(a)†), to 180° flexion (shown in Fig. S5(b)†) and back to the un-flexed position.

2.3 Electrochemical characterization

All of the electrochemical characterization in this work was conducted using a two-electrode system in the form of a CR3023 coin cell. The electrolyte consisted of 1 M of LiPF₆ salt dissolved in a 1 : 1 mixture of ethyl carbonate and dimethyl carbonate. Each cell was wetted with 50 μL of electrolyte solution. Galvanostatic charge and discharge (GCD) was performed using a Neware battery testing station from 1 V to 3 V at current densities ranging from 0.1C to 20C. To demonstrate another advantage of the binder-free flexible electrode, a flexible pouch half-cell was fabricated. As a comparison, electrodes were also made using the traditional slurry-based method using commercially available TiO₂ nanoparticles (Degussa P25) with an average diameter of 25 nm at a similar loading to the composite electrodes. Here, the TiO₂ nanoparticles, polyvinyldifluoride (PVDF) and conductive carbon with a ratio of 8 : 1 : 1 was homogeneously suspended in N-methyl-2-pyrrolidone.

3. Results and discussion

Schematic illustrations of the composite electrode synthesis is shown by Fig. 1(a). First, oleic acid capped-TiO₂ nanocrystals were drop casted onto bare carbon cloth. Then, the resulting composite was annealed in order to remove the oleic acid capping, which is non-conductive. Fig. 1(b)–(d) shows the surface of the composite electrode before annealing and Fig. 1(e)–(g) shows the surface of the TiO₂/CC composite electrode after annealing. It can be seen that the pre-annealed TiO₂ forms a uniform, smooth and polymer-like coating on the surface of the carbon fibers, which have a diameter of approximately 6 μm. This is due to the oleic acid capping that helps connect the TiO₂ particles seamlessly. When the electrode is annealed in air, the TiO₂ spheres are uniformly coated on the carbon fiber after the oleic acid capping removed, as shown by the difference of surface smoothness shown on Fig. 1(d) and (g) with the surface of the post-annealed electrode, as shown in Fig. 1(g), becoming more rough and bumpy, indicating the presence of particles. Fig. S1† shows the TGA curve of pure TiO₂ nanocrystals which supports the removal of the oleic acid capping agent at 450 °C, leaving behind TiO₂ particles. It can also be seen that the oleic acid comprises 20% of the mass of the nanocrystals. The mass of oleic acid was taken into account when calculating loading masses and all loading masses cited in this work are based on TiO₂. From Fig. 1(g), it appears that the TiO₂ nanocrystal spheres have a size of roughly 10 nm. The oleic acid helps distribute the particles uniformly on the carbon fibers which maximize the exposed surface area of the TiO₂ nanoparticles to electrolyte, and then allows greater lithium-ion access to the inner structure of TiO₂ which is beneficial for the improvement of specific capacity and rate capability. Additionally, the oleic acid also ensures intimate contact between the substrate and TiO₂, which can reduce the internal resistance and improve the cycling stability of the electrode. It is worth noting that there was no discernable change in mechanical properties of the carbon cloth before and after annealing at 450 °C in air. Fig. S2 shows the carbon cloth before it is annealed (Fig. S2(a)) and after it is annealed in air (Fig. S2(b)).† Both pieces of carbon cloth were identical in flexibility and apparent mechanical strength and the post-annealed piece of carbon cloth did not show any signs of being more brittle than the pre-annealed piece.

Fig. 1 (a) schematic illustration of the fabrication of the composite TiO₂/carbon cloth electrodes, (b)–(d) SEM images of pre-annealed TiO₂/carbon cloth, (e)–(g) SEM images of post-annealed TiO₂/carbon cloth.

Fig. 2 shows the Raman spectrograph and X-ray diffractograms of bare carbon cloth and the various stages of the composite electrode. The Raman spectrograph shows both the G and D bands at ∼1400 cm⁻¹ and ∼1600 cm⁻¹ which comes from the graphitic structure found in carbonaceous materials. There is a sharp peak at ∼150 cm⁻¹ which is characteristic of anatase TiO₂. The Raman spectrograph shows that there is no change in the TiO₂ crystal structure after the electrode is annealed as there are no shifts or additional peaks that appear in the curve for the annealed material. It is clear from the differences in peak between the bare carbon cloth and the composite electrode that TiO₂ is present on the carbon cloth. The XRD diffractogram also confirms the presence of anatase TiO₂ on the surface of the carbon cloth. The XRD pattern of the plain TiO₂ nanocrystals matches well with that of bulk anatase TiO₂ (JCPDS Card No. 21-1272).¹⁹ The diffractogram also shows that the crystal structure of the anatase TiO₂ remains unchanged after drop-casting and annealing to remove the oleic capping agent. The crystal domain size estimated from (101) peak using the Scherrer equation is about 10.8 nm. This is consistent with the SEM images previously shown previously.

Fig. 2 (a) Raman spectrograph of neat carbon cloth (black), pre-annealed composite electrode (red) and post-annealed composite electrode (blue) (b) X-ray diffractogram of neat carbon cloth (black), post-annealed TiO₂ particles (red), post-annealed composite electrode (blue).

Fig. 3 shows the electrochemical performance data of the composite electrode. Fig. 3(a) shows the discharge capacities of the TiO₂ nanocrystal and carbon cloth binder-free composite electrode while Fig. 3(b) shows the discharge capacities of a commercial TiO₂ (25 nm in size) that was coated onto copper foil using the traditional slurry-based fabrication method. In comparison, the binder-free composite electrode show better performance than that of the commercial TiO₂ nanoparticles at the same current density, which may be attributed to the unique structure of the binder-free TiO₂. Another reason is due to the small size of the TiO₂ nanoparticles. As the particle sizes decreases below 10 nm, the lithium–TiO₂ intercalation compound starts to behave like a solid solution which allows a higher density of lithium-ions to be packed, increasing the electrode's capacity.²⁹

Fig. 3 (a) Discharge profiles of the composite electrode at different currents with a loading of ∼0.5 mg cm⁻² (b) discharge profiles of commercial TiO₂ with a loading of ∼0.5 mg cm⁻² (c) first charge and discharge profile of the composite electrode (d) rate performance of the composite electrode (e) specific discharge capacity (red) and coulombic efficiency (blue) of the composite electrode over 100 charge and discharge cycles at a loading of ∼0.5 mg cm⁻² and current density of 500 mA g⁻¹.

Fig. 3(c) shows the first charge and discharge curves of the composite electrode at a current density of 100 mA g⁻¹. The curves shows both the Li-ion insertion and extraction potentials, corresponding to the insertion and extraction potential at ∼1.7 V and ∼1.9 V respectively, which matches with results shown in literature.^{19,22,23,30,33} The discharge plateau of TiO₂ (at ∼1.7 V) corresponds to the diffusion-controlled capacity.^30,32,34 In this region, the lithium ions are diffusing into the crystal structure of TiO₂ and the capacity is limited by the solid mass transport of the ions and the diffusion length.^19,35,36 It has been previously shown that as the crystal size decreases, the capacity in this region will increase as the diffusion length of the lithium ions decrease and vice versa.²⁹ The low-sloped, linear region of the discharge curve correspond to the surface-controlled capacity.^30,32,34 In this region, the lithium ions react and adsorb onto the surface/near-surface of the TiO₂ crystals and is limited by the available surface area. When the crystal size decreases, the surface-controlled region will generally also increase due to the increased surface area to volume ratio.³² For higher current densities, the ratio of diffusion- and surface-controlled capacity will decrease since the lithium will have less time to fully permeate the crystal structures.²⁹ Another thing of note is that compared to the discharge curve of the commercial TiO₂ on copper foil at 100 mA g⁻¹ shown in Fig. 3(b) the TiO₂/carbon cloth composite has a significantly higher surface-controlled capacity. This effect may be explained by the increased surface area of the TiO₂ nanocrystals in addition to the increased available surface area of carbon cloth compared to copper foil. Thus, the carbon cloth has a synergistic effect with the nanostructured TiO₂, allowing a greater surface-controlled charge storage capacity. We have also measured the capacity of the neat carbon cloth, shown in Fig. S3,† in order to demonstrate the lithium-ion storage capacity is mostly due to the presence of TiO₂. Fig. 3(d) shows the discharge capacities of the composite electrode at different current densities. The first discharge capacity of the electrode at 100 mA g⁻¹ is approximately 310 mA h g⁻¹, and still remains at 280 mA h g⁻¹ after 10 cycles,^34,37 as the current density increases, the capacity drops due to the aforementioned mechanisms. It can be seen that at each applied current densities, the charge and discharge capacity of the electrode remains relatively stable, revealing the high coulombic efficiency, which indicates the stability of electrode structure during lithium insertion and extraction. When the current returns to the initial value of 100 mA g⁻¹, the reversible capacity returns to 280 mA h g⁻¹, indicating a good rate performance. Fig. 3(e) shows the specific discharge capacity of the composite electrode and its coulombic efficiency over 100 charge and discharge cycles at a current density of 500 mA g⁻¹. We can see that the composite electrode exhibits good cycling stability after 100 cycles, demonstrating a discharge capacity of ∼150 mA h g⁻¹. This is consistent with the discharge capacity exhibited by the electrode during rate performance tests. We also see that the coulombic efficiency of the composite electrode remains ∼100% throughout the 100 cycles of charging and discharging.

Fig. 4 shows the effect of the TiO₂ solution concentration when drop casting (with similar mass loadings) on the capacity of the electrode. As the concentration decreases, the performance of the electrode increases. This may be explained by particle aggregation on the electrode and the low conductivity of TiO₂. As the loading increases, the utilization of the TiO₂ for lithium ion storage decreases.²⁰ Although the particles are able to be fully dissolved in toluene, the high concentrations might allow for aggregation once the solvent evaporates. This result is confirmed by using SEM imaging to probe the surface morphology of the electrodes with similar mass loadings but different casting concentrations as shown in Fig. S4.† Fig. S4(a)–(f)† shows the surface of the composite electrode made with casting concentrations of 2.5, 5 and 10 mg mL⁻¹ respectively. It can be seen that higher casting concentrations leads to more aggregates of TiO₂, resulting in non-uniform thicknesses and cracking. This result is consistent with the performance of the electrodes as shown in Fig. 4. Lower concentrations were not investigated as it would be impractical to fabricate electrodes with higher loadings. However, it can be assumed that lower solution concentrations will increase the performance by ensuring an even distribution of the particles onto the surface of the current collector. The best drop casting concentration (2.5 mg mL⁻¹) was then used for subsequent electrode fabrication.

Fig. 4 Rate performance comparison between different electrodes made from various TiO₂ solution concentrations.

Fig. 5 shows a proof-of-concept flexible half-cell made with the TiO₂/carbon cloth composite electrode. The half-cell demonstrates the flexibility and performance of the composite electrode under flexion. The pouch half-cell was able to power an LED bulb with and without flexion. In addition to the proof-of-concept flexible half-cell, the mechanical stability of the electrode was also tested. Fig. S5(c) and (d)† shows the as-prepared composite electrode before and after 100 cycles of mechanical flexion. Besides a slight physical deformation of the electrode along the bend-axis, the electrode shows no other signs of wear. The mass of the electrode before and after 100 cycles of bending only changed by 0.30%, indicating that the composite electrode has good mechanical stability.

Fig. 5 (a) Flexible pouch half-cell made using the composite anode, (b) flexible pouch half-cell powering an LED bulb under flexion.

4. Conclusions

In this work, TiO₂ nanocrystals with the sizes of about 10 nm were synthesized and drop casted onto carbon cloth to produce a composite flexible electrode material. Importantly, the electrode exhibited a reversible capacity of 270 mA h g⁻¹ at a current density of 100 mA g⁻¹, which is close to the theoretical capacity of TiO₂. The excellent performances are attributed to the unique structure and the small size of TiO₂ nanocrystal. The TiO₂ nanocrystals were uniformly deposited onto the carbon cloth substrate due to the oleic acid capping that helps act as a surfactant to dissolve the crystals in toluene. This is beneficial for reducing the inner resistance and the access of electrolyte to the surface of the active material. Moreover, the small crystal size aids in shortening the lithium-diffusion pathway and also increasing the surface area available to the electrolyte solution. The work presented here may be applied to other metal oxides and other flexible substrates in order to further improve the performance of LIC anodes.

Acknowledgements

This research was financially supported by the Natural Sciences and Engineering Research Council of Canada (NSERC).

References

1. Y. Ma, H. Chang, M. Zhang and Y. Chen, Adv. Mater., 2015, 27, 5296–5308.
2. H. S. Choi and C. R. Park, J. Power Sources, 2014, 259, 1–14.
3. Y. Wang, B. Liu, Q. Li, S. Cartmell, S. Ferrara, Z. D. Deng and J. Xiao, J. Power Sources, 2015, 286, 330–345.
4. A. M. Gaikwad, G. L. Whiting, D. A. Steingart and A. C. Arias, Adv. Mater., 2011, 23, 3251–3255.
5. L. Li, Z. Wu, S. Yuan and X.-B. Zhang, Energy Environ. Sci., 2014, 7, 2101–2122.
6. X. Wang, X. Lu, B. Liu, D. Chen, Y. Tong and G. Shen, Adv. Mater., 2014, 26, 4763–4782.
7. L. Hu and Y. Cui, Energy Environ. Sci., 2012, 5, 6423–6435.
8. L. Hu, H. Wu, F. La Mantia, Y. Yang and Y. Cui, ACS Nano, 2010, 4, 5843–5848.
9. Q. Zheng, Z. Cai, Z. Ma and S. Gong, ACS Appl. Mater. Interfaces, 2015, 7, 3263–3271.
10. M.-S. Balogun, C. Li, Y. Zeng, M. Yu, Q. Wu, M. Wu, X. Lu and Y. Tong, J. Power Sources, 2014, 272, 946–953.
11. B. Liu, J. Zhang, X. Wang, G. Chen, D. Chen, C. Zhou and G. Shen, Nano Lett., 2012, 12, 3005–3011.
12. Y. Tang, Y. Zhang, W. Li, B. Ma and X. Chen, Chem. Soc. Rev., 2015, 44, 5926–5940.
13. H. Zhang, W. Ren and C. Cheng, Nanotechnology, 2015, 26, 274002.
14. G. Lui, J.-Y. Liao, A. Duan, Z. Zhang, M. Fowler and A. Yu, J. Mater. Chem. A, 2013, 1, 12255–12262.
15. H. Zhang, X. Lv, Y. Li, Y. Wang and J. Li, ACS Nano, 2010, 4, 380–386.
16. M. Moradzaman, M. R. Mohammadi and H. Nourizadeh, Mater. Sci. Semicond. Process., 2015, 40, 383–390.
17. A. M. Bakhshayesh and N. Bakhshayesh, J. Colloid Interface Sci., 2015, 460, 18–28.
18. X. Yan, Z. Wang, M. He, Z. Hou, T. Xia, G. Liu and X. Chen, Energy Technol., 2015, 3, 801–814.
19. J. Wei, J.-X. Liu, Z.-Y. Wu, Z.-L. Zhan, J. Shi and K. Xu, J. Nanosci. Nanotechnol., 2015, 15, 5013–5019.
20. M. V. Koudriachova, N. M. Harrison and S. W. de Leeuw, Phys. Rev. Lett., 2001, 86, 1275–1278.
21. H.-Y. Wu, M.-H. Hon, C.-Y. Kuan and I.-C. Leu, Ceram. Int., 2015, 41, 9527–9533.
22. M. Zukalová, M. Kalbáč, L. Kavan, I. Exnar and M. Graetzel, Chem. Mater., 2005, 17, 1248–1255.
23. V. Aravindan, J. Gnanaraj, Y.-S. Lee and S. Madhavi, Chem. Rev., 2014, 114, 11619–11635.
24. Y. Zhang, Z. Jiang, J. Huang, L. Y. Lim, W. Li, J. Deng, D. Gong, Y. Tang, Y. Lai and Z. Chen, RSC Adv., 2015, 5, 79479–79510.
25. H. Wang, C. Guan, X. Wang and H. J. Fan, Small, 2015, 11, 1470–1477.
26. L. Jiang, X. Lu, C. Xie, G. Wan, H. Zhang and T. Youhong, J. Phys. Chem. C, 2015, 119, 3903–3910.
27. H. Yang, C.-K. Lan and J.-G. Duh, J. Power Sources, 2015, 288, 401–408.
28. Y.-S. Hu, L. Kienle, Y.-G. Guo and J. Maier, Adv. Mater., 2006, 18, 1421–1426.
29. M. Wagemaker, W. J. H. Borghols and F. M. Mulder, J. Am. Chem. Soc., 2007, 129, 4323–4327.
30. Z. Chen, Y. Yuan, H. Zhou, X. Wang, Z. Gan, F. Wang and Y. Lu, Adv. Mater., 2014, 26, 339–345.
31. X. H. Wang, C. Guan, L. M. Sun, R. A. Susantyoko, H. J. Fan and Q. Zhang, J. Mater. Chem. A, 2015, 3, 15394–15398.
32. Z. Chen, D. Zhang, X. Wang, X. Jia, F. Wei, H. Li and Y. Lu, Adv. Mater., 2012, 24, 2030–2036.
33. Y. Cai, B. Zhao, J. Wang and Z. Shao, J. Power Sources, 2014, 253, 80–89.
34. E. Madej, E. Ventosa, S. Klink, W. Schuhmann and F. L. Mantia, Phys. Chem. Chem. Phys., 2014, 16, 7939–7945.
35. Y. Cai, B. Zhao, J. Wang and Z. Shao, J. Power Sources, 2014, 253, 80–89.
36. D. Dambournet, I. Belharouak and K. Amine, Chem. Mater., 2010, 22, 1173–1179.
37. G. Zampardi, E. Ventosa, F. L. Mantia and W. Schuhmann, Chem. Commun., 2013, 49, 9347–9349.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra03262a

---