Electrospun Pd-doped mesoporous carbon nano fibres as catalysts for rechargeable Li–O₂ batteries

Abstract

Mesoporous carbon nanofibres doped with palladium nanoparticles (Pd CNFs) are synthesized by electrospinning with subsequent thermal treatment processes and used as electro-catalysts at the oxygen cathode of Li–O₂ batteries. FESEM images show that the spherical Pd nanoparticles (NPs) are homogeneously dispersed on the surface of CNFs and X-ray diffraction (XRD) measurements display a fcc structure of Pd. The surface area of the nanocomposite CNFs is greatly increased with the incorporation of the metal NPs up to 600 m² g⁻¹ and the presence of the metal promotes graphitization of the carbon. Addition of the N-[(aminoethyl)aminopropyl]trimethoxysilane additive in the precursor solution for electrospinning allows the reduction of the Pd NPs particles size, preserving the highly mesoporous N-doped large surface area and graphitic-nitrogen groups of the carbon nanofibres. Incorporating with a Pd/CNFs catalysed cathode, the Li–O₂ battery shows a very low voltage gap of 0.48 V vs. Li⁺/Li between the terminal discharge and charge voltages, as the recharge occurs at a potential underneath 4.0 V vs. Li⁺/Li for about 90 cycles at the curtailed capacity of 200 mA h g⁻¹. The low recharge voltage can relieve parasitic reactions due to the decomposition of electrolyte and favour a longer cycle life.

---

Introduction

At present, global warming, reduction of fossil-fuel supplies, energy production and storage demand the pursuit of renewable energy sources and sustainable storage technologies.¹ Among electrochemical storage and batteries, the Li-ion battery is considered the best performing, having a specific and volumetric energy of around 210 W h kg⁻¹ and 650 W h L⁻¹ respectively. Nonetheless, resizing the Li-ion battery pack for transport applications results in high cost per kW h and unsuitable energy densities.² Unlike Li-ion batteries, Li–air batteries do not exploit the concept of intercalation electrodes as the Li⁺ ions react directly with O₂ in a porous electrode.³ Laboratory scale tests⁴ demonstrated that the oxygen cathode exhibits gravimetric energy 3–5 times higher than that of LiCoO₂ cathode material of the Li-ion cell. In spite of that, the roadmap of the Li–air technology is predicted to be extended to a 20 years window,² because of a large number of unresolved issues.⁵ Substantial efforts have been addressed to develop Li–O₂ batteries with higher performances in terms of capacity and cycle life. Several major issues are responsible for the limited actual capacity and cycle ability. In principal, the stability of the electrolyte solvent^6,7 and the precipitation of lithium oxides discharge products inside the cathode pore structure during cell discharge. The high recharge potentials needed to decompose the insulating lithium peroxide and the parasitic products formed from the electrolyte decomposition result in energy losses for achieving a reversible electrochemical process.⁸ As the oxygen dissolved in the electrolyte is carried through the pore network of the cathode, insoluble products are supposed to narrow such network lowering oxygen diffusivity.⁹ Different approach of multi-scale modelling considered the end of the galvanostatic discharge caused by unavailable active surface area near the air inlet rather than the blockage of oxygen transport.¹⁰ According to Xiao et al.¹¹ the cathode ability to hold lithium oxides depends on the cathode pore system structure and the mesopore volume proved to be very effective in oxide accommodation. Hayashi et al.¹² also confirmed that carbon materials with high surface areas and larger number of mesopores could deliver high discharge capacities. Additional issues lie on the different reaction mechanisms for discharge and for the charge processes that are responsible for the high over-potentials in Li–O₂ cells.¹³ According to Bruce,¹⁴ during discharge the O₂ reduction results in the formation of LiO₂ due to the presence of Li ions from the electrolyte. At the surface of the cathode, LiO₂ can disproportionate to the more stable product Li₂O₂. On charging,¹⁵ Li₂O₂ decomposes directly, in a one-step reaction, to evolve O₂. To date, no commonly accepted consensus on the cathode mechanism for non-aqueous Li–O₂ batteries has been achieved¹⁶ and the literature findings are also discrepant regarding the role of catalysts at the O₂ cathode.⁸ In spite of that, most of the studies showed that the round trip efficiency of the cell increases by incorporation of metal and metal oxides catalysts at the cathode.^15,17–20 These efforts have improved the energy efficiencies of the Li–O₂ cell from 50 to 70% in the last years.²¹ Studies of specific metal clusters on the cathode surface provided evidence that nano-meter sized metals can strongly affect the morphology of the discharge products and that the tailoring of such structures can lead to decrease the charge over-potential and increase the capacity and cycle life of the Li–O₂ cell.^22,23 The above considerations suggest that the optimization of the oxygen electrode is crucial and the selection of an efficient catalyst for Li₂O₂ decomposition at a lower charge potential plays a role to improve the performance of the rechargeable battery.¹⁵ Precious metals are not so wide spread used in Li–air batteries mostly because of their expensive price and limited reserves, in spite of that, they are considered the best catalysts for chemical reactions.¹⁶ Interesting results have been obtained with bi-functional catalysts that combine one metal (Au), which is highly active for the ORR, with another metal (Pt) that is highly active for the OER.²⁴ Nonetheless, it was not demonstrated whether such bi-functional catalyst could lead to good cycling performances.²⁵ Cheaper precious metals, such as palladium nanoparticles, due to the strength of oxygen binding on the Pd surface, have higher intrinsic ORR activities in non-aqueous electrolytes as recent studies well demonstrated.²⁶ Ryu et al.²⁷ incorporated Pd nanoparticles on a PAN nano-fibre membrane and paired it with a carbon nanotube supported Ru nano-particles (NPs) cathode. The as assembled Li–O₂ cell showed a stable cycle ability (60 cycles) with a reduced degree of polarization. Mono-disperse M–Pd alloys NPs supported on reduced graphene oxide displayed capacities of 4407 mA h g⁻¹ at 100 mA g⁻¹ in DMSO based electrolytes.²⁸ Thapa et al.²⁹ incorporated Pd mixed with MnO₂ to prepare a bi-functional catalyst for both the ORR and OER processes. A free-standing honeycomb-like Pd-modified hollow spherical carbon was proposed by the group of Zhang³⁰ displaying long-term cycling ability (100 cycles at a current density of 300 mA g⁻¹ at the curtailed capacity of 1000 mA h g⁻¹). Shao-Horn et al.³¹ reported a refined synthesis of nano-catalysts assembled by M13 virus with manganese oxides, in which the incorporation of 3–5 wt% of Pd NPs enabled the cell to achieve 13 350 mA h g⁻¹ of specific capacity. Lei et al.³² prepared Pd/C through atomic layer deposition (ALD) method. They found that the amount of Pd influenced the discharge capacity of the Li–O₂ cell as the OER activity. As for the carbon support, the use of carbon nano-fibres, CNFs, assures high surface area as well as the high pores volume compared to other carbon based materials.³³ Impressive values of specific capacities of 40 000 mA h g⁻¹ at 500 mA g⁻¹ with vertical aligned CNFs were obtained,³⁴ although the loading of material on the electrode was 6–10 μg, which is rather impractical for battery applications.³⁵ Recently, high capacity values of 20 600 mA h g⁻¹ were obtained using 25 wt% RuO₂ supported on mesoporous CNFs mixed with carbon black KB.³⁶

In the study, different Pd doped mesoporous carbon nanofibres (Pd/CNFs) produced by electrospinning technique were used at the cathode of the Li–O₂ cell in tetraethylene glycol dimethyl ether (TEGDME) based electrolytes. Electrospinning of polyacrylonitrile (PAN) followed by stabilization and carbonization has become a straightforward and convenient route to make CNFs for energy conversion and storage applications.^37,38 Traditionally, the preparation of metal NPs/CNF nano-composites consists of two separate steps: the synthesis of carbon fibres and the subsequent deposition of the metal NPs. Examples include incipient wet impregnation with a Pd precursor,^39,40 chemical reduction of Pd salts⁴¹ or the use of colloidal Pd solutions followed by Pd deposition.⁴² All these methods require multiple steps and in some cases have some limitation for the preparation of highly porous or mesoporous structures. Electrospinning allows engineering fibres architecture by simply adjusting polymer composition⁴³ and the direct metal doping can be achieved during the synthetic process without any additional reductive treatment. Here, we directly synthesized Pd/CNFs by electrospinning of polyacrylonitrile/Pd(acetate)₂ mixture solution followed by thermal treatment processes. The conversion of PAN nanofibres to CNFs was performed at high temperature and involved stabilization and carbonization processes. During such steps, Pd²⁺ in the polymer nano-fibres was reduced to Pd⁰ and aggregated into Pd NPs.⁴⁴ The metal particle size was dependent on amount of Pd(Ac)₂ used and incorporation of complexing silanes in the precursor solution mixture, such as N-[(aminoethyl)aminopropyl]trimethoxysilane (AEAPTS), resulted in highly dispersed small nanoparticles of Pd supported on the CNFs surface.⁴⁵ The homogeneous dispersion of the metal NPs through the CNFs matrix, the increased surface areas and mesoporous structure of the metal/carbon composite, allowed promising electro-catalytic performances at the oxygen cathode.

Experimental

Unless and until mentioned separately, the chemicals used in the study are analytical grade and used as received. Pd(CH₃COO)₂, polyacrylonitrile (PAN, 150 000 mol g⁻¹) and the ligand [3-(2-aminoethylamino)propyl]trimethoxysilane (AEAPTS) were purchased from Sigma Aldrich. Dimethylformamide (DMF) was purchased from Scharlab. Tetraethylene glycol dimethyl ether (TEGDME) was battery grade and used as received from Solvionic, LiClO₄ salt was purchased by Sigma Aldrich. Poly(vinylidenefluoride) (Kynar® 761) binder was purchased from Arkema. N-Methyl-2-pyrrolidone (NMP) solvent was purchased by Sigma Aldrich.

Structural-morphological characterization

Metal content analyses were performed by ICP-MS (Agilent 7500, Agilent Technologies). The morphology of the as prepared samples (Fig. 1) has been analysed by means of a field emission scanning electron microscope (FE-SEM) Merlin (Zeiss). Nitrogen adsorption isotherms at 77 K were recorded with a Nova 2200e equipment (Quantachrome Instruments). High resolution transmission microscopy (HRTEM) measurements were performed using a side entry Jeol JEM 3010 (300 kV) microscope equipped with a LaB6 filament and fitted with X-ray EDS analysis by a Link ISIS 200 detector. For analyses, the powdered samples were deposited on a copper grid, coated with a porous carbon film. All digital micrographs were acquired by an Ultrascan 1000 camera and the images were processed by Gatan digital micrograph. The specific surface area of the samples was calculated by BET method within the relative pressure range of 0.05 to 0.2. The pore size distributions (PSDs) were determined by the BJH method calibrated for cylindrical pores according to the improved KJS method. The X-ray diffractions analysis were carried out using a Philips X'pert MPD powder diffractometer, equipped with Cu Kα radiation. Our analysis were performed with a Solid State PI Xul-ID detector with a number of 255 active channels guarantying a high-resolution detection. Moreover, to improve further the quality of the data acquired, the measurements were carried out at a very slow scan rate of 0.018° s⁻¹. X-ray photoelectron spectroscopy (XPS) measurements were carried out with a Physical Electronics PHI 5800 (USA) multi-technique ESCA system using monochromatic Al Kα X-ray radiation with a beam size of about 400 μm. The survey and narrow spectra were obtained with energy of 187.85 and 209 23.5 eV, respectively. Binding energies calibrated with respect to C 1s at 284.6 eV were accurate within 0.2 eV. The samples were placed in an ultra high vacuum chamber at 2–10 Torr. The morphology of the cathodes was examined using field-emission scanning electron microscopy (FESEM, JEOL-JSM-6700F). The electrical conductivity was measured using the four probe. The set up consist in a four points probe (JANDEL), with a fixed distance(s) of 1 mm, positioned onto the surface material. The material is a typical film of CNFs of 2 × 2 cm² with a thickness around 0.1 mm (the thickness was measured with a digital micrometer). An electrostatic potential between −0.3 V and 0.3 V was applied in the inner contacts meanwhile the electrical current (I) was measured between the outer contacts. The curve I vs. V was measured with a VMP3 potentiostat. In that way, a linear function I vs. V was plotted which slope correspond to the inverse of the resistance value.

Fig. 1 HRSEM images of: (a) Pd5, (b) Pd2.5, (c) Pd2.5A, (d) CNFs.

Preparation of CNFs and Pd/CNFs by electro-spinning

The solutions for the electro-spinning were based on a 10 wt% PAN in DMF, and the different samples were prepared by adding 2.5 mmol and 5 mmol of Pd(CH₃COO)₂ (samples labelled as Pd2.5, Pd5, respectively). Addition of the metal precursors was carried out before the electro-spinning process. When the AEAPTS ligand was used, an amount of 7.5 mmol of AEAPTS was added to the metal precursor solution before mixing with the PAN solution (sample labelled as Pd2.5A). The electro-spinning process was performed using a multi-spinneret with 3 syringes at a flow of 2–3 mL h⁻¹, moving at 120 m s⁻¹, with a distance of 15 cm, a V = 29.9 kV and a collector rod rotating at 300 rpm. The thermal treatment was carried out in a multistage process based on a first step of heating at 1 °C min⁻¹ up to 280 °C, dwelling time of 1 h at 280 °C under air and then a second heating step at 5 °C min⁻¹ up to 1000 °C with a dwelling time of 1 h under N₂.

Preparation of cathodes

The O₂ electrode was prepared as a coating layer over carbon paper gas-diffusion layer (GDL, SIGRACET GDL-24AB, SGL Technologies), hydrophobized substrate with a 5 wt% PTFE loading. The thickness of the GDL was 190 μm and the air permeability was 60 cm³ (cm² s)⁻¹ according to Sigraget data sheets. To prepare cathodes, CNFs and Pd/CNF were mixed with poly(vinylidenefluoride) (Kynar® 761) binder, in the weight ratio of 90 : 10, utilizing a mixer mill (Retsch, MM 400). N-Methyl-2-pyrrolidone (NMP) solvent was then added to the solid mixture in order to obtain a uniform slurry. During the preparation, the slurries with different compositions were tuned with a similar viscosity using NMP solvent. The thickness of the deposited slurry was set at 300 μm. The slurry was coated on top of the micro-porous layer of the GDL using doctor blade technique. This coating was dried at 60 °C overnight to evaporate the NMP solvent. The as-prepared cathodes were further dried in vacuum at 120 °C overnight before testing in Li–O₂ cells. No significant difference in the thickness of the coating layers was observed for various compositions. The coating layer had a total loading of 0.8 mg cm⁻².

Electrochemical tests

A lithium disc (18 × 0.2 mm, Chemetall s.r.l.) was used as the anode. A glass fibre (18 × 0.65 mm, ECC1-01-0012-A/L) saturated in the electrolyte was used as the separator. All the electrolyte formulations comprised 5 wt% of lithium perchlorate (LiClO₄) salt in TEGDME solvent. The Li–O₂ cell was then assembled in an Ar-filled dry glove box (Mbraun Labstar) using an ECC-air electrochemical cell (EL-Cell, GmbH) configuration with openings allowing oxygen to enter and exit through the cathode side. The O₂ flow was set at 3.0 mL min⁻¹ by Smart Thermal Mass Flow Controller (Brooks Instrument). The geometric area of the cells was 2.54 cm². In order to evaluate the discharge capacity, the cells were galvanostatically discharged by an Arbin BT-2000 battery tester at room temperature, from the open circuit voltage (OCV) to 2.25 V vs. Li⁺/Li and recharge up to 4.4 V vs. Li⁺/Li. To investigate the cycle-ability of the cells, galvanostatic time-controlled charge and discharge steps were carried out. The cycling test were limited in time and voltage, with cut-off voltages of 2.25–4.4 V vs. Li⁺/Li and discharge/charge duration of 10 h. Throughout all the electrochemical tests, the cells were continuously purged with dry O₂ at a gas flow rate of 3.0 mL min⁻¹. Prior to each test, the Li–O₂ cell rested 6 h at OCV under oxygen flow.

Results and discussion

The use of palladium acetates as precursors in the electrospun solution allowed the in situ formation of the metal NPs, as during the process, Pd²⁺ is reduced to Pd and aggregated into Pd NPs. It is assumed that the presence of carbon together with the acetate counter ions are responsible for this reduction.⁴⁴

Moreover, the use of AEAPTS ligand improves the distribution of the NPs through the CNF matrix, since the ligand can coordinate Pd²⁺ avoiding NPs agglomeration. Indeed, the average diameter of the metal NPs mainly depends on the type of metal, the reaction conditions and the type of complexing silane in the precursor solution.

In sol–gel processes, the reason of the influence of the precursor is the different exothermicity of the pyrolysis reaction. The resulting local temperature differences promote metal growth to different degrees.⁴⁵ The Pd NPs distribution is shown in the HRSEM images of all the synthesized mesoporous Pd-doped CNFs along with the HRSEM image of the un-doped CNFs (Fig. 1). The as prepared CNFs had diameters ranging from 300 to 500 nm with tens of micrometers in length. Pd NPs were deposited on the surface of CNFs and samples prepared from solutions with less metal content showed smaller Pd NPs than those with higher ones (Pd2.5 vs. Pd5), which is confirmed by the TEM technique in Fig. 2a and b. Addition of AEAPTS in the precursor solution resulted in the formation of Pd aggregates with smaller particles (Fig. 2c). In addition, the Pd aggregates in the Pd2.5A (in Fig. 2c) were formed by loosely packed metal nanoparticles, which could result in a higher density of catalytic sites.

Fig. 2 TEM images of: (a) Pd2.5, (b) Pd5, (c) Pd2.5A.

XRD patterns of the doped CNFs confirmed the presence Pd(0) (Fig. 3). All the catalysts exhibited the characteristic Pd face centred cubic (fcc) structure due to the presence of the Pd (111), Pd (200), Pd (220), Pd (311), Pd (222) planes located at 40.28°, 46.78°, 68.20°, 82.20°, 86.72° in the 2θ axes.⁴⁶ The diffraction peak close to 26° (2θ) is assigned to the diffraction of (002) plane of graphite. This shows that the graphitic structures of CNFs were formed as the consequence of the heat treatment in the presence of Pd NPs (in samples Pd2.5 and Pd2.5A). Inagaki⁴⁷ reported that metal particles can exert a catalytic action on the graphitization of carbon, whereby graphitized or non-graphitized carbon is obtained depending on the ratio between polymer and metal. This may explain the absence of the graphite diffraction peak in sample Pd5, also considering the poorer graphitization ability of Pd compared to other non-precious metals.⁴⁸ The average particle diameters of the samples, calculated by Scherrer equation on the XRD patterns (Fig. 3) were 27 nm (Pd2.5A), 50 nm (Pd2.5) and 63 nm (Pd5) respectively. Determined by ICP AES analysis, the overall Pd percentage in the synthesized samples were 4.78 wt% in Pd2.5A, 7.06 wt% in Pd2.5 and 10.65 wt% in Pd5, respectively (Table 1). Theoretical values reported in Table 1 were calculated considering a 30% yield for the CNFs after the thermal treatment. Although some deviations can be appreciated due to the analytical error, the metal incorporation was almost quantitative. The electrical conductivity of the CNFs was measured using the standard four-probe method and ranged from 0.05 to 0.1 S cm⁻¹. Although the thermal treatment of the pristine PAN nano-fibres produced CNFs with some mesoporosity, the addition Pd(Ac)₂ dramatically influences the porosity and the specific surface area of the carbon composites. The surface structural parameters of the samples, summarized in Table 2, were obtained by N₂ adsorption–desorption measurements. The CNFs possessed a rather small SSA of 16 m² g⁻¹ as well as a low pore volume, whereas the corresponding Pd/CNF samples exhibited larger SSA from 211 m² g⁻¹ (Pd2.5) to 626 m² g⁻¹ (Pd5) with total pore volume ranging from 0.11 cm³ g⁻¹ (Pd2.5A) to 0.31 cm³ g⁻¹ (Pd5). In all the samples, the pore size did not vary with the Pd acetate concentration as the average value of 3.6 nm was observed. Addition of AEAPTS resulted in the partial conversion of the microporosity into mesoporosity (samples Pd2.5 and Pd2.5A, Table 2). XPS (Fig. 4) was employed to analyse the surface properties of the catalysts. The XPS spectra for all the catalysts are referred to a C 1s value of 284.6 eV. The atomic Pd percentage on the surface of Pd2.5 was 0.4%, similar to that of Pd2.5A (0.3%), which suggests analogue location of the Pd particles in both catalysts (Fig. 4a and Table S1 in ESI†). The atomic Pd percentage on the surface of Pd5 was found to be 1.4% consistent to large or agglomerated Pd particles and to the higher amount of Pd. In sample Pd2.5A, the 2.0% atomic Si percentage derived by incomplete removal of the silane additive. The XPS spectrum of Pd 3d of Pd5 sample is shown in Fig. 4c and confirmed the presence of metallic Pd. The Pd 3d peak was de-convoluted and the Pd(0) was located at a binding energy of 335.2 eV and 340.5 eV, 341.3 eV for the Pd 3d_3/2 and Pd 3d_5/2 respectively. The oxidized Pd 3d was located at 335.6 eV, 337.0 eV and 342.2 eV consistent with PdO. The N 1s spectrum was used for analysing the nature of N functionalities and the de-convoluted N 1s spectra of Pd2.5A sample is shown in Fig. S1 in ESI.† Graphitic-nitrogen groups, at about 401 eV, are reported to enhance the ORR activities in metal-free carbons as N atoms reduce the electron density on adjacent C nuclei, resulting in improved electron transfer. All the tested samples (Fig. 4b) showed large fractions of graphitic-N moieties, which are consistent with the temperature of the carbonization process.⁴³ In Fig. S1 in ESI,† the peak at 398.93 eV may either correspond to N-pyridinic (N-6) or most probably to tertiary amine and the shoulder around 402 eV may be assigned to the ammonium peak.⁴⁹ The peak at 401.16 eV should correspond to graphitic-N (N-G) and that at 402.96 eV can be tentatively assigned to some form of oxidised nitrogen, N-oxides (N–X).^50,51 Interestingly, the graphitic-N groups increased with increasing the amount of Pd(Ac)₂ (Pd5 vs. Pd2.5), while the use of AEAPTS results in increased (Pd2.5A vs. Pd2.5 in Fig. 4b) of amines fraction.

Fig. 3 XRD patterns of Pd2.5A, Pd2.5 and Pd5 synthesized materials.

Table 1 Pd content on Pd-doped CNFs obtained by ICP-AES measurements

Sample	Pd2.5	Pd2.5A	Pd5
Measured (wt%)	7.06	4.78	10.65
Theoretical (wt%)	7.3	6.28	15.29

---

Table 2 Properties of investigated CNF and Pd/CNF materials

Sample	CNF	Pd2.5	Pd2.5A	Pd5
Specific surface area (m^2 g^−1)	16.0	211.2	139.5	626.9
Mesoporous	16.0	67.6	98.0	66.4
Microporous	—	143.6	41.5	560.6
Pore size (nm)	3.2	3.7	3.9	3.6
Total pore volume (cm^3 g^−1)	0.05	0.12	0.11	0.31
Mesoporous	0.05	0.058	0.10	0.10
Microporous	—	0.062	0.014	0.21

---

Fig. 4 (a) XPS survey of Pd2.5A, Pd2.5 and Pd5 synthesized materials, (b) converted values of N-species of each sample from the de-convoluted high-resolution N 1s XPS spectra of Pd/CNFs, (c) de-convoluted high-resolution Pd 3d spectra of Pd5.

The electrochemical performances of the Pd/CNFs have been investigated as potential catalysts for the oxygen electrode of the Li–O₂ battery. The main interest was to evaluate how much the porosity of CNFs and the relatively small metal loadings (5–10 wt% of Pd) and particle size of Pd affected the cycling ability of the cell, with the attempt to quantify the impact of the Pd nanoparticles on the morphology of the discharge products. As reference, the CNFs based electrode and the GDL support were also tested in the same conditions. We first discuss the discharge voltage and the gravimetric capacities from Li–O₂ obtained at low discharge current. Fig. 5 shows the discharge/charge curves of the CNFs, Pd5, Pd2.5, Pd2.5A measured between the open circuit potential (OCV) and 2.25 V at a current density of 20 mA g_cathode⁻¹. The Pd2.5 exhibited the highest discharge capacity of 7828 mA h g⁻¹, compared with CNFs (6730 mA h g⁻¹), Pd5 (4941 mA h g⁻¹) and Pd2.5A (3647 mA h g⁻¹). The GDL current collector contributed to a background capacity (not reported in Fig. 5) of 160 mA g⁻¹ at 20 mA g⁻¹ based on the mass of the GDL. The specific capacities reported in Fig. 5 are consistent with previously reported studies using vertical aligned carbon fibres,⁴ although it was also demonstrated that the gravimetric capacities obtained were five times lower if normalized to the weight of discharged electrodes (accounting the Li₂O₂ weight), especially at low current regimes. In this study, Pd apparently has little effect on the discharge capacity compared to that of CNF, with the exception of Pd2.5.

Fig. 5 Full discharged Li–O₂ cells in galvanostatic conditions from OCV to 2.25 V vs. Li⁺/Li at the applied current of 20 mA g_cathode⁻¹.

As concerns the galvanostatic charge profile, in Fig. S2a in ESI,† the CNF electrode exceeded the 4.4 V voltage limit at about 4000 mA h g⁻¹ which corresponds to 58% of coulombic efficiency. Incorporation of about 7 wt% of Pd in the CNF (Pd2.5 sample) enhanced the coulombic efficiency to 95%, resulting in more efficient decomposition of both Li₂O₂ and side products during the charging process in the presence of Pd.

At the applied gravimetric current of 100 mA g_cathode⁻¹, for Pd2.5, (Fig. S2b in the ESI†) the average discharge voltage reduced to 2.6 V vs. Li⁺/Li but the material still retained a high specific capacity of 6674 mA h g⁻¹. The capacity losses at high current rates were attributed to increased cathode passivation and to uneven distribution of Li₂O₂ deposits within the pore volume of the cathode.⁵²

Ex situ FESEM images (Fig. 6) of the discharged cathodes, to 2.25 V vs. Li⁺/Li, clearly revealed the deposition of the crystalline products as toroidal shaped particles on the surface of all the discharged electrodes. Several studies^53,54 reported that Li₂O₂ forms discrete particles with a typical toroid-like shape, which can grow up to about 1 μm of size. Such particular morphology appeared to be more dependant to the true surface-area currents rather than to the type of electrolyte or of carbon used at the cathode.⁵⁵ Nazar et al.⁵⁶ showed that crystalline Li₂O₂ toroids proliferate the cathode surface after discharge at low current densities, while at higher current densities, largely amorphous Li₂O₂ films were formed.

Fig. 6 Post-mortem FESEM images of the cathodes of the Li–O₂ cells after full discharge from OCV to 2.25 V vs. Li⁺/Li at the applied current of 20 mA g_cathode⁻¹ and after recharge up to 4.4 V vs. Li⁺/Li. (a) Discharged CNF cathode, (b) charged CNF cathode, (c) discharged Pd2.5 cathode, (d) charged Pd2.5 cathode.

In this work, we observed for the bare CNFs that the particles of Li₂O₂ entirely covered the surface of the carbon fibres forming a dense compact layer (Fig. 6a). Interestingly, the initial increment of recharge potential in carbon-based cathodes has been related⁵⁷ to the collapse of the dense film formed in discharge by the large-sized particles of insulating Li₂O₂. Ex situ FESEM images of the recharged CNFs cathodes (Fig. 6b) up to 4.4 V vs. Li⁺/Li, revealed incomplete removal of the discharge products supporting the limited recharge ability observed for the CNF electrode (Fig. S2a in ESI†). FESEM images of the discharged Pd2.5-cathode (Fig. 6c) displayed uniformly sized toroids (Li₂O₂) formed through the aggregation of Li₂O₂ nano-crystallites along the CNFs. Ex situ FESEM image of the charged Pd2.5-cathode to 4.4 V vs. Li⁺/Li (Fig. 6d) highlighted that the sidewalls of the CNFs were largely bare, suggesting the full reconversion of the discharge products.

Analysis of the FESEM image (Fig. 7a) of the pristine Pd5 cathode revealed that the use of a larger amount of Pd(Ac)₂ in the PAN/Pd(Ac)₂ mixture resulted in the formation of inner pores and etching the surface of electrospun fibres. This imparted a lower mechanical stability to the fibres, as many appeared broken after the electrode manufacturing. A similar Li₂O₂ morphology (Fig. 7b) to Pd2.5 was observed on Pd5 cathode on discharge and full reconversion of the products after recharge at 4.4 V vs. Li⁺/Li (Fig. 7c) although the sidewalls of the Pd5 were rougher than that of the pristine. Interestingly, with Pd2.5A, Li₂O₂ coated over the Pd2.5A-CNFs with smaller Li₂O₂ particles sizes (Fig. 7d and e). Such Li₂O₂ morphology was considerably different from the previous observed for Pd2.5 and Pd5 discharged cathodes.

Fig. 7 Post-mortem FESEM images of the cathodes of the Li–O₂ cells after full discharge from OCV to 2.25 V vs. Li⁺/Li at the applied current of 20 mA g_cathode⁻¹ and after recharge up to 4.4 V vs. Li⁺/Li. (a) Pristine Pd5 cathode, (b) discharged Pd5 cathode, (c) charged Pd5 cathode. (d) Discharged Pd2.5A cathode and (e) magnification of (d).

XRD patterns (Fig. 8a) of the discharged CNF electrodes until 2.25 V vs. Li⁺/Li, identified Li₂O₂ as the main crystalline phase but not negligible amounts of crystalline Li₂CO₃ were detected. Some LiOH deposition was also noticed. McCloskey et al.⁵⁸ demonstrated that Li₂O₂ induces decomposition of TEGDME through hydrogen abstraction during discharge to form Li formate and –OCH₃-containing organics. Interestingly, Shao-Horn et al.⁵⁹ reported that vertically aligned carbon nanotubes (VACNT) undergo chemical and morphological modifications during discharge. They determined that Li₂CO₃ predominately forms as the result of the reactivity between carbon and Li₂O₂ in discharged electrodes. At high capacities, the formation of Li₂CO₃ was observed to occur predominantly on the carbon surfaces. In our work, the XRD analysis of the discharged Pd/CNF electrodes (Fig. 8b–d) at 2.25 V vs. Li⁺/Li, revealed the presence of crystalline Li₂O₂, LiOH·H₂O and LiOH. Nazar et al.⁵⁴ observed similar LiOH formation with Na_0.44MnO₂ nanowires at the cathode of a Li–O₂ cell with TEGDME-based electrolyte, after discharge at 2 V. They considered the reaction of surface H⁺ within the structure of the catalyst and O₂⁻ produced by reduction of O₂ as possible mechanism of LiOH formation. Evidence of crystalline LiOH formation on discharge was also reported by Amine et al.⁶⁰ with Pd/C cathodes. The Authors also suggested that the presence of Pd on carbon surface leads to better electron transfer for the nucleation and growth of Li₂O₂ and minimize the possible electrolyte decomposition on the carbon defect sites. With our Pd/CNF electrodes, no crystalline Li₂CO₃ was observed, suggesting that Pd/CNF may moderate the reactivity of Li₂O₂ with the carbon surface and the electrolyte and the electrolyte decomposition on carbon.

Fig. 8 XRD patterns of the cathodes of the Li–O₂ cells after full discharge from OCV to 2.25 V vs. Li⁺/Li at the applied current of 20 mA g_cathode⁻¹ and (a) pristine and discharge CNF cathodes, (b) pristine and discharged Pd2.5 cathodes, (c) discharged Pd2.5A cathode, (d) discharged and recharged (up to 4.4 V vs. Li⁺/Li) Pd5 cathodes.

Additionally, recent investigations pointed out that the oxidation of the LiOH occurs at lower charge potential than that of Li₂CO₃, suggesting easier decomposition of such product.⁶¹ This is consistent with the XRD pattern of the full recharged Pd5 cathode, which only highlights the XRD peaks of both Pd and GDL (Fig. 8d). The long term cycling of the cathodes was investigated in galvanostatic condition at the curtailed capacity of 200 mA h g⁻¹, with a round trip cycle duration of 20 h (10 h per discharge) at a current density of 20 mA g_cathode⁻¹. Such set of conditions has been selected by considering that the exposure time of Li₂O₂ discharge product to the electrolyte is a crucial factor in the evaluation of the chemical stability of the deposited discharge products on the surface of cathodes and of the stability of the catalyst. In order to evaluate the contribution of the GDL on the cycling performance the discharge/charge cycles were carried out with no active material at the actual current that was used for all the cycling tests. On the first cycle (Fig. 9), in the presence of Pd (Pd2.5, Pd2.5A and Pd5) the oxidation potential remained below 3.8 V, whereas it increased above 4.25 V in its absence. Remarkably low charge potential of 3.2 V was observed for Pd2.5A. CNF based electrodes demonstrated limited stability (Fig. 10), unable to cycle for more than 20 cycles before irreversibly exceeding the voltage windows of 2.25–4.4 V vs. Li⁺/Li.

Fig. 9 Galvanostatic discharge and charge of the Li–O₂ cells at the curtailed capacity of 200 mA h g⁻¹ at the applied current of 20 mA g_cathode⁻¹ with cathodes: Pd2.5A (black line), Pd2.5 (red line), Pd5 (blue line), CNF (green line) and the GDL (pink line). The voltage vs. capacity plot refers to the 1st cycle.

Fig. 10 Cyclic performance of Li–O₂ cells, at the curtailed capacity of 200 mA h g⁻¹ with catalyzed cathodes: (a) CNFs cathodes; (b) Pd5 cathodes; (c) Pd2.5 cathodes (d) Pd2.5A cathodes; (hollow blue circle)/charge capacities; (solid blue circle)/discharge capacity vs. cycle number. Applied gravimetric current: 20 mA g⁻¹. The voltage vs. capacity profiles of the first 20 cycles are reported in Fig. S3 in ESI.†

Under the same conditions (Fig. 10), the Pd2.5, Pd2.5A and Pd5 demonstrated good cycling profiles for 73, 88 and 61 cycles respectively, corresponding to 50–70 days of cell operation at 100% coulombic efficiency. It is worth noting that the catalyst that contained the lowest amount of Pd and smaller Pd NPs (i.e. Pd2.5A, 4.7 wt% of Pd) was responsible of the higher Li–O₂ cell longevity and of the largest reduction of the charge overvoltage. For Pd2.5A, the difference between the terminal voltage of charge and that of discharge was only 0.48 V at the first cycle (Fig. 9), then the cell recharged underneath 3.5 V for the subsequent 20 cycles (Fig. 10). The terminal charge potential increased but remained below 4.1 V until the 80th cycle. Similar to what previously reported, the growth of Li₂O₂ NPs on Pd sites results in the nano-crystalline Li₂O₂ morphology with small grains that make up the toroids. The charge process is supposed to be facilitated by the small Li₂O₂ grains in the toroids as well as conduction through grain boundaries and electronic contact with the cathode surface.⁶²

The particular morphology of the Li₂O₂ observed in full discharge conditions with Pd2.5A should account for an easier decomposition of the peroxide, resulting in lower over-potentials on the following recharge. Such result indicates that the nucleation of Li₂O₂ strongly depends on the catalytic surface, i.e. Pd particle size, Pd distribution on the CNF and N content. The Pd2.5A, which is responsible of lowering the charge over potential to greater extent, also alleviates concomitant side reactions, reducing accumulation of the carbonate residuals, prolonging cycle duration.

The progressive accumulation of the insulating discharge products on the cathode,⁶⁰ carbonates precipitation in the separator fibres,⁶³ O₂ crossover to the anode with formation of LiOH and carbonates⁶⁴ and possible binder decomposition⁶⁵ leads to limit the cycle performance to about 90 cycles corresponding to 1750 h of cell operation. The combination of the porous catalysed electrode structure and the electrolyte solvent, synergistically, not only determines the chemical nature of the discharge product but also governs the physical size and morphology of it, playing a decisive factor in the recharge ability and cycling performance of the resulting Li–O₂ battery.

Conclusions

Electro-spinning method with the capability to mass produce nano-fibres with good fibre yield open new challenges in the preparation of composite catalysts for cathodes of rechargeable Li–O₂ batteries. In particular, mesoporous carbon nano-fibres doped Pd (Pd/CNFs) can be easily obtained through a simple combination of electro-spinning and thermal process. Interestingly the use of silane-based additive, previously employed in sol–gel process, allows to reduce the Pd NPs particles size, to lower the Pd content and to preserve the mesoporous and large surface area of the CNFs support with high degree of graphitization. The Pd/CNFs catalysed cathode in the Li–O₂ battery showed a very good electro-catalytic activity and stability allowing the cell to perform almost 90 cycles at 100% of coulombic efficiency at the curtailed capacity of 200 mA h g⁻¹.

Acknowledgements

Financial support was provided by the European Union Seventh Framework Programme (FP7/2007–2013) project STABLE (no. 314508). The Authors sincerely thank Dr Vankova Svetoslava for XRD measurements, Dr Salvatore Guastella for XPS, Mr Mauro Raimondo and Dr Maela Manzoli for FESEM and TEM analyses.

Notes and references

1. D. Larcher and J.-M. Tarascon, Nat. Chem., 2015, 7, 19–27.
2. L. Grande, E. Paillard, J. Hassoun, J.-B. Park, Y.-J. Lee, Y.-K. Sun, S. Passerini and B. Scrosati, Adv. Mater., 2015, 27, 784–800.
3. Q. Li, P. Xu, W. Gao, S. Ma, G. Zhang, R. Cao, H.-L. Wang, G. Wu and J. Cho, Adv. Mater., 2014, 26, 1378–1386.
4. R. R. Mitchell, B. M. Gallant, C. V. Thompson and Y. Shao-Horn, Energy Environ. Sci., 2011, 4, 2952–2958.
5. J. Christensen, P. Albertus, R. S. Sanchez-Carrera, T. Lohmann, B. Kozinsky, R. Liedtke, J. Ahmed and A. Kojica, J. Electrochem. Soc., 2012, 159(2), R1–R30.
6. S. A. Freunberger, Y. Chen, N. E. Drewett, L. J. Hardwick, F. Bardé and P. G. Bruce, Angew. Chem., Int. Ed., 2011, 50, 8609–8613.
7. E. M. Erickson, E. Markevich, G. Salitra, D. Sharon, D. Hirshberg, E. De La Llave, I. Shterenberg, A. Rozenman, A. Frimer and D. Aurbach, J. Electrochem. Soc., 2015, 162(14), A2424–A2438.
8. B. D. McCloskey, R. Scheffler, A. Speidel, D. S. Bethune, R. M. Shelby and A. C. Luntz, J. Am. Chem. Soc., 2011, 133, 18038–18041.
9. Y. Wang and S. C. Cho, J. Electrochem. Soc., 2013, 160(10), A1847–A1855.
10. K.-H. Xue, T.-K. Nguyen and A. A. Franco, J. Electrochem. Soc., 2014, 161(8), E3028–E3035.
11. J. Xiao, D. Wang, W. Xu, D. Wang, R. E. Williford, J. Liu and J.-G. Zhang, J. Electrochem. Soc., 2010, 157, A487–A492.
12. M. Hayashi, H. Minowa, M. Takahashi and T. Shodai, Electrochemistry, 2010, 78, 325–328.
13. A. Kraytsberg and Y. Ein-Eli, Nano Energy, 2013, 2, 468–480.
14. S. A. Freunberger, Y. Chen, Z. Peng, J. M. Griffin, L. J. Hardwick, F. Bard, P. Novak and P. G. Bruce, J. Am. Chem. Soc., 2011, 133, 8040–8047.
15. J. Wang, Y. Li and X. Sun, Nano Energy, 2013, 2, 443–467.
16. Z. Ma, X. Yuan, L. Li, Z.-F. Ma, D. P. Wilkinson, L. Zhang and J. Zhang, Energy Environ. Sci., 2015, 8, 2144–2198.
17. Y. Shao, S. Park, J. Xiao, J.-G. Zhang, Y. Wang and J. Liu, ACS Catal., 2012, 2, 844–847.
18. M.-K. Song, S. Park, F. M. Alamgir, J. Cho and M. Liu, Mater. Sci. Eng., R, 2011, 72, 203–252.
19. Y. Lei, J. Lu, X. Luo, T. Wu, P. Du, X. Zhang, Y. Ren, J. Wen, D. J. Miller and J. T. Miller, et al., Nano Lett., 2013, 13, 4182–4189.
20. L. Wang, X. Zhao, Y. H. Lu, M. W. Xu, D. W. Zhang, R. S. Ruoff, K. J. Stevenson and J. B. Goodenough, J. Electrochem. Soc., 2011, 158, A1379–A1382.
21. J. Shui, F. Du, C. Xue, Q. Li and L. Dai, ACS Nano, 2014, 8, 3015–3022.
22. J. Lu, K. C. Lau, Y.-K. Sun, L. A. Curtiss and K. Amine, J. Electrochem. Soc., 2015, 162(14), A2439–A2446.
23. J. Lu, L. Cheng, K. C. Lau, E. Tyo, X. Luo, J. Wen, D. Miller, R. S. Assary, H.-H. Wang, P. Redfern, H. Wu, J.-B. Park, Y.-K. Sun, S. Vajda, K. Amine and L. A. Curtiss, Nat. Commun., 2014, 5, 4895 , 1–8.
24. Y.-C. Lu, Z. Xu, H. A. Gasteiger, S. Chen, K. H. Schifferli and Y. Shao-Horn, J. Am. Chem. Soc., 2010, 132, 12170–12171.
25. F. Cheng and J. Chen, Chem. Soc. Rev., 2012, 41, 2172–2192.
26. Y.-C. Lu, A. Gasteiger and Y. Shao-Horn, J. Am. Chem. Soc., 2011, 133, 19048–19051.
27. W.-H. Ryu, F. S. Gittleson, M. Schwab, T. Goh and A. T. Taylor, Nano Lett., 2015, 15, 434–441.
28. M. Sevin, T. Sener and O. Metin, Int. J. Hydrogen Energy, 2015, 40, 10876–10882.
29. A. K. Thapa, K. Saimen and T. Ishihara, Electrochem. Solid-State Lett., 2010, 13(11), A165–A167.
30. J. Xu, Z.-L. Wang, D. Xu, L.-L. Zhang and X.-B. Zhang, Nat. Commun., 2013, 4, 2438 , 1–10.
31. D. Oh, J. Qi, Y.-C. Lu, Y. Zhang, Y. Shao-Horn and A. M. Belcher, Nat. Commun., 2013, 4, 2756 , 1–8.
32. Y. Lei, J. Lu, X. Luo, T. Wu, P. Du, X. Zhang, Y. Ren, J. Wen, D. J. Miller, J. T. Miller, Y.-K. Sun, J. W. Elam and K. Amine, Nano Lett., 2013, 13, 4182–4189.
33. V. A. Agubra, L. Z. Zuniga, D. Flores, J. Villareal and M. Alcoutlabi, Electrochim. Acta, 2016, 192, 529–550.
34. J. Shui, F. Du, C. Xue, Q. Li and L. Dai, ACS Nano, 2014, 8, 3015–3022.
35. M. Noked, M. A. Schroeder, A. J. Pearse, G. W. Rubloff and S. B. Lee, J. Phys. Chem. Lett., 2016, 7, 211–215.
36. Z. Guo, D. Zhou, H. Liu, X. Dong, S. Yuan, A. Yu, Y. Wang and Y. Xia, J. Power Sources, 2015, 276, 181–188.
37. P. Poudel, L. Zhang, P. Joshi, S. Venkatesan, H. Fong and Q. Qiao, Nanoscale, 2012, 4, 4726–4730.
38. S. Zhang, Z. Lin, L. Ji, Y. Li, G. Xu, L. Xue, S. Li, Y. Lu, O. Troprakci and X. Zhang, J. Mater. Chem., 2012, 22, 14661–14666.
39. J. Zhu, J. Zhou, T. Zhao, X. Zhou, D. Chen and W. Yuan, Appl. Catal., A, 2009, 352, 243–250.
40. E. Díaz, M. León and S. Ordónez, Int. J. Hydrogen Energy, 2010, 35, 4576–4581.
41. W. Yang, S. Yang, J. Guo, G. Sun and Q. Xin, Carbon, 2007, 45, 397–401.
42. Y.-H. Qin, Y.-B. Jia, Y. Jiang, D.-F. Niu, X.-S. Zhang, X.-G. Zhou, L. Niu and W.-K. Yuan, Int. J. Hydrogen Energy, 2012, 37, 7373–7377.
43. G. Su Park, J.-S. Lee, S. T. Kim, S. Park and J. Cho, J. Power Sources, 2013, 243, 267–273.
44. J. S. Huang, D. W. Wang, H. Q. Hou and T. Y. You, Adv. Funct. Mater., 2008, 18, 441–448.
45. S. Martınez, M. Moreno-Manãs, A. Vallribera, U. Schubert, A. Roig and E. Molins, New J. Chem., 2006, 30, 1093–1097.
46. Z. Liu, L. Hong, M. P. Tham, T. H. Lim and H. Jiang, J. Power Sources, 2006, 161, 831–835.
47. M. Inagaki, K. Fujita, Y. Takeuchi, K. Oshida, H. Iwata and H. Konno, Carbon, 2001, 39(6), 921–929.
48. M. Yudasaka, Y. Kasuya, F. Kokai, K. Takahashi, M. Takizawa, S. Bandow and S. Iijima, Appl. Phys. A: Mater. Sci. Process., 2002, 74, 377–385.
49. C. Yuan, W. Chen and L. Yan, J. Mater. Chem., 2012, 22, 7456.
50. Z. Lin, M.-K. Song, Y. Ding, Y. Liu, M. Liu and C.-P. Wong, Phys. Chem. Chem. Phys., 2012, 14, 3381–3387.
51. J. R. Pels, F. Kapteijn, J. A. Moulijn, Q. Zhu and K. M. Thomas, Carbon, 1995, 33, 1641–1653.
52. C. Ó. Laoire, S. Mukerjee, E. J. Plichta, M. A. Hendrickson and K. M. Abraham, J. Electrochem. Soc., 2011, 158(3), A302–A308.
53. B. M. Gallant, R. R. Mitchell, D. G. Kwabi, J. Zhou, L. Zuin, C. V. Thompson and Y. Shao-Horn, J. Phys. Chem. C, 2012, 135, 494–500.
54. J.-H. Lee, R. Black, G. Popov, E. Pomerantseva, F. Nan, G. A. Botton and L. F. Nazar, Energy Environ. Sci., 2012, 5, 9558–9565.
55. Y.-C. Lu, B. M. Gallant, D. G. Kwabi, J. R. Harding, R. R. Mitchell, M. S. Whittingham and Y. Shao-Horn, Energy Environ. Sci., 2013, 6, 750–768.
56. L. Nazar, B. Adams, R. Black, C. Radtke, K. Zaghib and M. Trudeau, Energy Environ. Sci., 2013, 6, 1772–1778.
57. J. Xiao, D. Mei, X. Li, W. Xu, D. Wang, G. L. Graff, W. D. Bennett, Z. Nie, L. V. Saraf, I. A. Aksay, J. Liu and J.-G. Zhang, Nano Lett., 2011, 11(11), 5071–5078.
58. B. D. McCloskey, A. Valery, A. C. Luntz, S. R. Gowda, G. M. Wallraff, J. M. Garcia, T. Mori and L. E. Krupp, J. Phys. Chem. Lett., 2013, 4, 2989–2993.
59. B. M. Gallant, R. R. Mitchell, D. G. Kwabi, J. Zhou, L. Zuin, C. V. Thompson and Y. Shao-Horn, J. Phys. Chem. C, 2012, 116, 20800–20805.
60. Y. Lei, J. Lu, X. Luo, T. Wu, P. Du, X. Zhang, Y. Ren, J. Wen, D. J. Miller, J. T. Miller, Y.-K. Sun, J. W. Elam and K. Amine, Nano Lett., 2013, 13, 4182–4189.
61. T. Zhang and H. S. Zhou, Nat. Commun., 2013, 4, 1817–1824.
62. J. Lu, Y. Lei, K. C. Lau, X. Luo, P. Du, J. Wen, R. S. Assary, U. Das, D. J. Miller, J. W. Elam, H. M. Albishri, D. A. El-Hady, Y.-K. Sun, L. A. Curtiss and K. Amine, Nat. Commun., 2013, 4, 2383 , 1–9.
63. J.-L. Shui, J. S. Okasinski, D. Zhao, J. D. Almer and D.-J. Liu, ChemSusChem, 2012, 5, 2421–2426.
64. R. S. Assary, J. Lu, P. Du, X. Luo, X. Zhang, Y. Ren, L. A. Curtiss and K. Amine, ChemSusChem, 2013, 6, 51–55.
65. H. Geaney and C. O'Dwyer, J. Electrochem. Soc., 2016, 163(2), A43–A49.

---

Footnote

† Electronic supplementary information (ESI) available: Surface functional component obtained from the survey XPS spectra, de-convoluted N 1s spectra for sample Pd2.5A. Gravimetric capacity vs. cell voltage plots of the CNFs and Pd2.5 cathodes in full discharge and recharge; discharge plots at two different applied currents for Pd2.5. Cell voltage profiles vs. capacity of the first 20 cycles of Li–O₂ cells in TEGDME-based electrolyte with: CNF; Pd5; Pd2.5 and Pd2.5A-based cathodes. See DOI: 10.1039/c6ra09721a

---