Ball milling : a green mechanochemical approach for synthesis of nitrogen doped carbon nanoparticles †

ARC Centre of Excellence for Functiona Materials, Deakin University, Waurn Pond chen@deakin.edu.au ARC Centre of Excellence for Electromateria Deakin University, Burwood, Victoria 3125, School of Life and Environmental Sciences, 3216, Australia. E-mail: wenrong@deakin.e 5227 2932 School of Physics, The University of Sydney † Electronic supplementary information fabrication of the catalyst, high-resolutio and TGA of the ball milled graphite. See D Cite this: Nanoscale, 2013, 5, 7970


Introduction
Precious Pt-based materials have been used as an effective electrocatalyst for oxygen reduction reactions (ORRs), 1 yet largescale commercial use has been precluded by high cost, limited supply, and weak durability of platinum.Recent intensive research efforts directed at reducing or replacing Pt-based electrodes in fuel cells have led to the development of new ORR electrocatalysts. 2,3In particular, the state-of-the-art nitrogendoped carbon nanomaterials have been proved to have better resistance to poisoning, 1,4,5 and are relatively cheaper to fabricate 6 than Pt-based catalysts.Carbon nanomaterials exhibit some unusual physical, chemical, and mechanical properties, such as light absorption, mechanical stiffness, and high conductivity. 7Thus, nitrogen-doped carbon materials have been generally accepted as potential candidates for replacing Pt to achieve the commercialization of fuel cell technology.Nevertheless, most of the current methods of synthesizing nitrogen doped carbon materials face some issues like low production yield and use of toxic gas (NH 3 for example).Therefore, the development of a new fabrication method is still required.
Ball milling is a process, in which the moving balls apply their kinetic energy to the milled material, break chemical bonding and produce fresh surfaces by fracturing material particles.The dangling bonds on the newly created surfaces are usually chemically reactive.Also, it has been reported that the impact of a high-energy ball milling process sometimes can create a local high temperature above 1000 C 8 and/or high pressure up to several GPa. 9 Due to this nature of the ball milling process, the process has been used as a mechanochemical synthesis method 8,10 for the fabrication of unique nanostructures with new chemical properties.It has been reported that ball milling can trigger solid-solid reactions to synthesize PbTe alloy from PbO and Te. 114][15] Other than these, ball milling has also been used recently to fabricate a nanostructured material that has unique properties. 16Graphite has a layered structure of aromatic rings.Numerous strong bonds are broken upon disruption of the layers, which produces free valences along the fresh borders.This would open the way for the formation of carbon nanostructures with new functionalities.In the previous work, 17 we have demonstrated that a nanoporous carbon structure can be produced in the graphite samples aer ball milling at ambient temperature.Formation of the nanoporous structure is associated with the disordered graphite structure.The disordered and nanoporous structure is probably fullerene-like in nature.Subsequently, it has also been showed that annealing of a disordered carbon nanostructure at 1400 C produced carbon nanotubes. 18][21] In this article, we report that ball milling can be used as a powerful green method to synthesize nitrogen-doped carbon nanoparticles (NDCPs), which can be used as an electrochemical catalyst for ORRs (Scheme 1 and Fig. S1 †).The resulting carbon nanoparticles exhibited competitive catalytic activity (5.2 mA cm À2 kinetic-limiting current density at À0.5 V compared with 7.6 mA cm À2 on Pt/C reference) and excellent methanol tolerance compared to a commercial Pt/C catalyst.

Materials and methods
Graphite (powder, particle size <20 micron, synthetic) was obtained from Sigma AldrichÒ (Sydney, Australia) and used as the starting material.Nitrogen (99.99% purity) and a nitrogen-15% hydrogen mixture were used as atmospheric gases.

Sample preparation
NDCPs were prepared by ball milling of pristine graphite powders in a high-energy rolling ball mill in the presence of nitrogen gas (300 kPa) at room temperature.The set-up of the ball mill is described elsewhere. 22The volume of the ball mill jar is about 300 mL.In a typical experiment, 2 g of the graphite powders were placed into a milling jar with four Ø 2 cm hardened steel balls.The sample to ball mass ratio is 1 : 132.The milling process lasted for 24 hours with a rotation speed of 150 rpm.The milled samples were then heated in a horizontal tube furnace from room temperature to 700 C at a rate of 25 C min À1 and then annealed for 3 hours.The sample was cooled down to room temperature.N 2 -15% H 2 mixture gas was used as the annealing atmosphere with a ow rate of 0.05 L min À1 .

Instrumentation
The sample structure was characterized using X-ray diffraction (PANalytical X'Pert Pro diffractometer) with Cu Ka radiation (l ¼ 0.15418 nm).The sample morphologies were studied using a scanning electron microscope (SEM, Supra 55VPP) at 3 kV and a working distance of 4.6 mm.The particle size was measured with a dynamic light scattering method (Malvern ZETASIZER).0.5 mg of the samples were added into 5 mL deionized water to make dispersions.The dispersions were ultrasonically treated for 30 min before the test.
Ultraviolet-visible (UV-Vis) spectroscopy was used to characterize the dispersive ability of the samples in water.The dispersions were made by adding 20 mg of samples to a 5 mL volume of deionized water.Then, dispersions were ultrasonically treated for 30 min.Aer that, they were diluted by 50 times to make them adequately transparent for UV-Vis tests.The scans were run from 200 nm to 800 nm with a scan rate of 600 nm min À1 and a scan step of 1 nm.Water and a completely opaque cell were used as 100% and 0% transmittance standards, respectively, to correct the baseline.
X-ray photoelectron spectroscopy (XPS) (model: SPECS-XPS, from SPECS, Germany) was used to analyze the concentrations of elements and binding energies.The XPS system was equipped with a high sensitivity PHOIBOS 150-9 MCD energy analyzer.An Al K-alpha X-ray source (1486.74eV) was used for the analysis.Transmission electron microscopy (TEM) images were taken with a JEOL 2100F instrument.The accelerating voltage used was 200 kV in the experiment.Thermogravimetric analysis (TGA) was done with a NETZSCH STA 409 PC.Samples were heated to 700 C from room temperature with a heating rate of 10 C min À1 .A N 2 -15%H 2 mixture was used as atmosphere.The ow rate was 10 mL min À1 .

Thin lm electrode preparation
The catalyst suspension was prepared by sonication (Soniclean 160T, Soniclean Pty Ltd) of the sample with tetrahydrofuran (THF, 99.9+%, Sigma-Aldrich) for 1 h, resulting in ink with nal concentrations of 5 mg sample per mL.An aliquot of 4 mL of suspension was drop-cast onto a glassy carbon (GC) disk electrode (4 mm diameter, 0.126 cm 2 area, rotating ring-disk electrode (RRDE) Pt ring/GC disk electrode, ALS Co., Ltd).The electrode was previously polished with 0.05 mm alumina slurry (Buehler) on a clean polishing cloth (Buehler), sequentially rinsed with distilled water and acetone, and then dried with lint-free tissue paper.Aer slow evaporation of solvent, 1 mL of NaonÒ solution (5 wt%, Fluka) was drop-cast on the electrode surface and dried slowly overnight inside a closed beaker to attach the catalyst particles onto the glassy carbon disk electrode.Based on the density of NaonÒ solution of $0.874 g cm À3 and the density of the NaonÒ lm of $2.0 g cm À3 , 23 an estimated NaonÒ lm thickness of $1.74 mm was obtained which is reasonably thin to render a negligible O 2 transport resistance within the lm. 24

Electrochemical characterization
Voltammetric experiments were performed with a Biologic VMP2/Z multi-channel potentiostat.Rotating disk electrode (RDE) voltammograms were obtained in 0.1 M potassium hydroxide (KOH) prepared from deionized water and potassium hydroxide pellets (KOH, 99.99%, Sigma-Aldrich) at a scan rate of 20 mV s À1 in either a N 2 or O 2 (ultrahigh purity grade, BOC) saturated atmosphere at room temperature (which corresponds to an electrolyte temperature of $17 AE 1 C).Voltammograms for the reductions were obtained using a three-electrode Scheme 1 Fabrication of nitrogen-doped carbon nanoparticles (NDCPs) and their ORR catalytic properties.
The background capacitive current contribution (obtained from N 2 saturated experiment) is subtracted from each voltammogram curve.Three catalyst lms were prepared for each sample to check the reproducibility of the lm quality and performance; the results of those vary within the experimental error.The current density values were obtained assuming an effective disk surface area of 0.1257 cm 2 (diameter of 4 mm) for all samples.

Results and discussion
The preparation of NDCPs and their ORR catalytic properties are shown in Scheme 1. Ball milled graphite (C 2 ) was fabricated by milling the commercial graphite (C 1 ) in a nitrogen atmosphere.Then, C 2 was annealed to produce NDCPs (C 4 ) with catalytic properties.For comparison purposes, C 1 was also annealed under the same conditions to produce annealed commercial graphite (C 3 ).
Ball milling has been widely used for mechano-chemical synthesis. 8,10We have shown that nitrogen can be effectively doped with materials by ball milling in the presence of nitrogen gas. 25,26It was demonstrated that mechanical energy of ball milling can break the chemical bonds in solid materials and produce dangling bonds, which are very active for chemical reactions. 10,22Simultaneously, gas atoms react with the solid atoms with dangling bonds.SEM images (Fig. 1) show a total breakdown of the graphitic layer structure.Furthermore, XRD patterns (Fig. 2) conrm the crystalline structure breakdown from a signicant widening of the diffraction peaks of graphite. 27These results imply a wide introduction of dangling bonds and the potential to trigger the gas-solid reaction.
The high-resolution SEM images show the different morphologies of four samples.Starting commercial graphite (Fig. 1a, C 1 ) had a ake-like shape.Annealing treatment did not change this morphology, as shown in Fig. 1c, C 3 .However, ball milling treatment has changed the morphology substantially of C 1 .There is no typical ake-like structure in the milled graphite sample (Fig. 1b and d).Instead, the milled samples contain irregular sphere-like structures with a wide distribution of particle sizes.Aer annealing of the milled graphite, abundant edges are found in C 4 , as high resolution TEM images of C 2 and C 4 are shown in Fig. S2.† XRD patterns of the four samples are consistent with the SEM imaging results.The starting (C 1 ) and heated (C 3 ) graphite samples have similar XRD patterns, showing most intensive (002) peaks at $26.5 degrees and some other weak peaks at $45 degrees (100, 101), 55 degrees (004) and 77 degrees (110).However, aer ball milling, the XRD pattern of C 2 shows a much broadened peak of (002), which suggests a structural distortion and the reduction of crystal size.Comparing C 4 with C 2 , the FWHM of peaks becomes narrower aer annealing, indicating that annealing renes the graphitic structure.This structure renement is expected to increase the sample's conductivity. 28PS was used to characterize the element contents of the samples.As expected, in the XPS analysis of sample C 1 , the C 1s peak (for carbon) was the only strong peak found with a very small O 1s peak, corresponding to approximately 2.2% oxygen.Aer ball milling treatment (sample C 2 ), approximately 7.5% oxygen and 6.3% nitrogen were found.The high nitrogen content in the milled sample indicates that the milling process led to nitrogen incorporation into the graphite.The rise in the oxygen level in C 2 might be caused by residual oxygen in the milling chamber or atmosphere contamination before XPS measurement.Both nitrogen and oxygen contents were reduced down to approximately 1.0% for nitrogen and 0.7% for oxygen (sensitivity of the analysis was about 0.1%) (Fig. S3 and S4 †)  aer the annealing treatment (sample C 4 ).As a comparison, direct annealing of commercial graphite in the N 2 -H 2 mixture (sample C 3 ) did not introduce detectable nitrogen using XPS.The possible mechanism of the ball milling in a nitrogen atmosphere to act as an effective process in the synthesis of a Ndoped graphite catalyst is that the mechanical ball milling process introduces active sites such as unpaired bonds during the milling, resulting in nitrogen covalent bonding to the carbon surface. 8PS N1s spectra were analysed to characterize the bonding nature of the nitrogen in the catalysts since different states of nitrogen could result in different catalytic properties. 29The N 1s spectrum of sample C 2 (Fig. 3a) can be tted by 4 components at 398.8 eV, 399.7 eV, 401.1 eV and 403.6 eV, corresponding to the nitrogen in pyridinic (N 1 pyrrolic (N 2 ), graphitic (N 3 ) and nitric oxide (N 4 ) structures. 6,30Aer annealing (sample C 4 , Fig. 3b), these three components remained while the pyrrolic structure (N 2 ) disappeared and the relative amount of the graphitic structure (N 3 ) increased from 43% to 66%.This phenomenon was consistent with previous research that the graphitic structure (N 3 ) had better thermal stability than pyridinic and pyrrolic structures. 29he particle size distribution was measured to characterize how the particle size changed aer annealing (Fig. 4).C 2 had an average particle size as 210 nm; the particle size became 348 nm aer annealing (C 4 ).This phenomenon might be caused by the agglomeration and the renement of the graphitic structure during high temperature annealing, which could be seen from XRD patterns. 31ig.S4 † shows the thermogravimetric analysis (TGA) of the ball milled graphite.The heating process was programmed to be the same as the annealing process.C 1 shows little weight loss during the annealing.Both C 2 and C 4 show a weight loss about 3% below 100 C, which might be contributed by the absorbed moisture.We noticed C 2 has steady mass loss during the annealing while C 4 shows little signicant weight loss aer 100 C.This difference might be due to the reduction of carbon oxide groups and the loss of thermal unstable nitrogen doping of C 2 .
UV-Vis absorption spectroscopy was used to characterize the dispersive and hydrophilic property of samples (Fig. 5).C 1 and C 3 could not form a stable dispersion in water.However, C 2 and C 4 could be dispersed in water, as reected by the presence of a UV-Vis absorption band of the aromatic system in graphite. 32his enhanced dispersive ability and hydrophilicity are likely to be due to the changed surface property and the decreased average particle size created by ball milling. 33Indeed, the introduction of N and O doping 26 has been proved to be able to increase the hydrophilicity.This increased hydrophilicity may be responsible for the improved catalytic property by enabling enhanced mass exchange rates and the stability of the threephase zone. 34,35 Fig. 6 shows ORR in 0.1 M KOH using thin lms of C 1 , C 2 , C 3 and C 4 deposited on the RDE rotated at 1600 rpm.As a control experiment, Fig. 7 was obtained under the same conditions conrming the absence of reduction currents for typical carbons e.g.C 2 and C 4 in the absence of oxygen.The ORR activity is enhanced in the order of C 3 , C 1 , C 2 and C 4 .The ORR enhancement is evident in both the current onset and density e.g. a sample showing more positive onset potential value exhibits a larger current density.C 4 , in particular exhibits a similar onset potential value and higher current density in excess of À0.6 V with respect to Pt/C (40 wt% Pt), the standard benchmark for ORR in conventional fuel cells, implying the remarkable potential of the proposed doping method to synthesize alternative ORR catalysts.This is due to the N doping and the breaking of the graphite basal plane structure.7][38][39] At the same time, it has been shown that the basal-plane aromatic ring structure of graphite is very inert to chemical reactions and electron transfer, [40][41][42][43] which implies that defects and boundary carbon atoms are the real active sites where chemical reactions and catalytic processes take place.As demonstrated, ball milling could result in a large quantity of structural defects and even change the morphology, which would increase the quantity of active catalytic sites.However, it is interesting to compare the RDE results of C 1 , C 2 and C 4 .In fact, the current density of ORR does not change much aer only the ball milling process, while the potential onset shis to be more positive than C 1 .The onset shi is due to the nitrogen doping, while the reason for the low current may be that the damaged structure leads to reduced conductivity.The reduced conductivity causes a limitation of catalytic current and more energy consumption when charge transfer occurred in the material, which could compensate the energy saving brought by the catalytic process.Another possible reason is that although the ball milling induced nitrogen doping, the doped nitrogen is not catalytically active.XPS uses bonding energy to differentiate different nitrogen doping.It is possible that some nitrogen doping or absorbed nitrogen have not any catalytic contribution (which had bonding energy within the tting area) to the extent that the catalytic nitrogen doping might be less than the tting showed.Aer annealing, the catalytic property enhanced even though the nitrogen content was much lower, reduced from 6.3% to 1%.Previous research suggests that the chemical structures of nitrogen doping with active catalytic properties are pyridinic and graphitic nitrogen. 29However, in this case, XPS shows that the content of all kinds of nitrogen doping decreases aer annealing.Thus, in this work, the catalytic property of the sample is not quantitatively related to the nitrogen content.
To determine the ORR characteristics, the relationship between the current density and the rotation rate was examined.The typical ORR current density prole versus rotation rate is shown in Fig. 8 for sample C 4 .Fig. 9 plots the typical inverse of the current density as a function of the inverse of the root square of the rotation rate e.g. the Koutecky-Levich plots. 44,45The slope seems to vary between C 3 /C 1 and C 2 /Pt/C/C 4 .Considering the widely established facts that ORR on Pt/C occurs through a four-electron mechanism whereas conventional carbon through a two-electron mechanism, 44,46 it is most likely that (within the experimental error), ORRs on C 3 and C 1 follow a two-electron mechanism while ORRs on C 2 , Pt/C and C 4 follow a four-electron mechanism pathway.The shi of the mechanism from two-electron to four-electron is consistent with the ORR activity enhancement trend observed above.Also, the kinetic current density (J k ) could be calculated from the y-axis intercept of Koutecky-Levich plots.The J k increased in the order of C 3 , C 1 , C 2 and C 4 , the same with the order of potential onset increase.The J k of C 3 was 4.2 mA cm À2 and 5.2 mA cm À2 for C 4 .C 4 showed a high kinetic current density approaching that of Pt/C (7.5 mA cm À2 ).These results suggest that ball milling has introduced catalytic nitrogen doping in the graphite.Aer ball milling, the mechanism of ORR changed from a less efficient 2-electron mechanism of graphite to a mixture of 2-electron and 4-electron mechanisms, which is due to the nitrogen doping. 29On the other hand, the renement of annealing could also be seen from both the increase of J k and the more preferential 4-electron mechanism of C 4 .The increase of J k may be caused by the rened graphitic structure.
Furthermore, we did not see the change of catalytic current of C 2 compared with commercial graphite, but was signicantly improved aer annealing.Thus, the annealing is considered as a structure rened step of the milled samples.Our results showed ball milling of graphite in nitrogen could distort the graphitic structure and reduce the size of the (0 0 2) plane.As graphite is an anisotropic material, which has signicant anisotropic electroconductivity, the distorted graphitic structure and the tailored (0 0 2) plane may lead to an increased electronic resistance, which reduces the performance and limits the catalytic current.The annealing, in this case, is used to rene the graphitic structure and improve the conductivity.Aer annealing, as our results show, the FWHM of the (0 0 2) peak was smaller, which suggests the renement of the graphitic in-planar structure.In the annealing atmosphere, N 2 -15%H 2 is not compulsory.In fact, the annealed samples in pure N 2 were also tested and proved to have catalytic properties.However, as shown in XPS, the milled samples contain oxygen.In order to reduce the oxygen level, we chose a reducing atmosphere.
In the real fuel cells, the organic compounds like methanol at the anode would penetrate the membrane and reach the Pt cathode, which could change the chemical reaction of oxygen reduction to the oxidation of those organic compounds. 47This phenomenon is called the poisoning of catalyst.The advantage of carbon-based materials becomes apparent in terms of its methanol resistance.The chronoamperometric response was obtained at À0.35 V (Fig. 10).We found that introducing methanol leads to slight deterioration of the ORR current for C 4 .For Pt/C, however, methanol oxidation occurs and causes a total loss of cathodic current.This indicates a much better stability of reaction sites in C 4 than Pt/C in an alkaline environment, of interest for advanced fuel cells.

Conclusions
High-performance ORR electrocatalysts can be synthesized based on nitrogen-doped carbon nanoparticles via ball milling and annealing processes.The nitrogen doped carbon nanostructures exhibit a high activity and excellent tolerance to methanol in comparison to other non-precious metal catalysts.The synthesis strategy demonstrated here exploits novel carbonbased structures, and could represent a general approach to highly efficient electrocatalysts for a range of important electrochemical reactions, which is low-cost, and easy to scale up, providing a feasible method for development of large-scale practical electrocatalysts.

Fig. 1
Fig. 1 Typical SEM images of (a) starting graphite, (b) milled graphite, (c) heated graphite without milling treatment and (d) heated graphite after milling treatment.

Fig. 2
Fig. 2 XRD diffraction patterns of graphite after different treatments.

Fig. 5
Fig. 5 UV-Vis absorption of four samples after ultrasonic treatment.C 1 and C 3 quickly precipitate after ultrasonic treatment and indicate no UV-Vis absorption (inset: C 1 -C 4 from left to right taken after 30 min of ultrasonication).

Fig. 6
Fig. 6 Rotating disk electrode (RDE) voltammograms of C 1 -C 4 and Pt/C reference at 1600 rpm in O 2 saturated 0.1 M KOH at a scan rate of 20 mV s À1 .

Fig. 8
Fig. 8 Rotating disk electrode (RDE) linear sweep voltammograms of C 4 in O 2 saturated 0.1 M KOH with different rotation rates at a scan rate of 20 mV s À1 .

Fig. 7
Fig. 7 RDE voltammograms of C 2 and C 4 in 0.1 M KOH saturated with N 2 or O 2 at a rotating rate of 1600 rpm.