Nitrogen-doped MWCNT/graphene composite materials

in section 4.4, they were selected as precursors for nitrogen-doped ORR catalysts.

This work was inspired by a previous study done in our working group on graphene [126], where it was noted that graphene sheets have the tendency to stack up again during the high-temperature pyrolysis procedure. To avoid this, MWCNTs (which were also already confirmed to have significant activity towards the ORR on their own in our working group [121]) were selected as the spacer material between the graphene sheets. To synthesise N-doped catalysts from these two carbon materials, graphite was first oxidised into GO (section 6.2) and com-mercial MWCNTs were purified (section 6.1). After this, they were mixed with a nitrogen precursor and PVP in liquid phase, sonicated and pyrolysed (section 6.4). During the pyrolysis procedure, the fast decomposition of the nitrogen pre-cursors into gaseous components acts to separate the GO sheets even further and to reduce them into graphene. The precursors used in this work are given in Figure 6.

All of them have a few things in common: they are very nitrogen-rich and de-compose at less than 800 °C. In general, during the pyrolysis process all of these nitrogen sources first form graphitic carbon nitride (g-C3N4), which in turn decomposes at >700 °C to give NH3 and carbon nitride gases (C2N2+, C3N2+, C3N3+) [194–197]. Depending on the specific precursor, by-products such as HNCO, cyanuric acid, triuret, ammelide, ammeline, melamine, melam, melem and melon can also be formed. Using the „flash“ type of pyrolysis, where the sample is quickly entered into the heating zone of the furnace, it is likely that all of these components are present at the same time and react with both the carbon and each other as well. This means that doping with each of the nitrogen sources will give very different results.

SEM images of the 2-NC and NG/NCNT-BR material are shown in Figure 7.

The crumpled sheet morphology characteristic to graphene is clearly visible, with MWCNTs deposited on top of and between the sheets, as intended. The MWCNTs act as spacers to prevent restacking of graphene nanosheets and there-fore increase the availability of the active sites on those sheets. In addition, the MWCNTs themselves are catalytically active. TEM (Figure 7b) confirmed these finding as no larger graphite stacks were visible.

Figure 7. SEM images of 2-NC (a) and NG/NCNT-BR (c) materials and a TEM image of the 2-NC material (b).

To determine elemental composition on the surface of the catalyst materials, as well as the type of nitrogen moieties each of the nitrogen precursors created on the surface, XPS analysis of each of the catalysts was undertaken. The XPS survey spectra and the core-level N1s spectra for the catalysts derived from biuret, carbo-hydrazide and semicarbazide are given in Figure 8.

The XPS peaks at 284.8, 532.1 and 400.0 eV denote the binding energies of C1s, O1s and N1s, respectively. As can be seen, most of the oxygen-containing groups of GO and purified MWCNTs have been either reduced or replaced by nitrogen moieties at the end of the pyrolysis. The presence of an N1s peak around 400 eV indicates the successful incorporation of nitrogen atoms into graphene/MWCNT material. The types of nitrogen species were identified by deconvolution of the core-level N1s spectra. The peaks in the high-resolution N1s spectra in Figure 2d–f can be deconvoluted into four components: pyridine-N-oxide (404.1 eV) [198], quaternary (graphitic) N (401.1 eV), pyrrolic N (400.1 eV) and pyridinic N (398.1 eV). The total nitrogen content in NG/NCNT-BR, NG/NCNT-CH and NG/NCNT-SC was 2.8, 2.5 and 1.7 at.%, respectively. The relative contents of nitrogen moieties are given in Table 2. Interestingly, while the absolute concentration of species changed, the relative contents of nitrogen moieties stayed more or less the same. The XPS spectra corresponding to 1-NC and 2-NC materials are presented in Figure 9.

Table 2. Surface nitrogen contents and speciation of nitrogen for NG/NCNT-BR, NG/NCNT-CH, NG/NCNT-SC, 1-NC and 2-NC.

Catalyst Total N

(at.%) Pyridinic

N % Pyrrolic

N % Quaternary

(graphitic) N % Pyridine-N-oxide %

NG/NCNT-BR 2.8 46 24 21 10

NG/NCNT-CH 2.5 39 26 17 17

NG/NCNT-SC 1.7 35 32 16 17

1-NC 6.0 56 24 12 8

2-NC 4.0 50 26 15 9

Figure 9. XPS survey spectra (a,b) and N1s spectra (c,d) for 1-NC and 2-NC samples, respectively.

In the case of 1-NC and 2-NC, the overall N content of 1-NC was higher (as was thus the content of pyridinic and quaternary nitrogen, which are considered the most important components). The relative quaternary N content was lower than in the case of all of the samples from the first set along with the content of oxidised pyridine N-oxide.

To describe the structure of carbon in the 1-NC and 2-NC samples, Raman spectroscopy of these materials was done in the Institute of Physics, University of Tartu. The resulting Raman spectra are given in Figure 10.

Figure 10. Raman spectra of 1-NC and 2-NC materials.

The most important signals in the Raman spectra of carbon materials are the D-band situated at ~1350 cm⁻¹ and the G-band situated at 1550–1600 cm⁻¹. The D-band arises from a breathing mode vibration of six-member carbon rings [199]

and notes the presence of disordered carbon planes. The G-band arises from the tangential vibrations of carbon atoms in graphite and is characteristic of sp²-hybridised carbons [200]. 1-NC and 2-NC both showed peaks at 1350 and 1580 cm⁻¹. The ID/IG ratio was found to be 0.96 and 1.03 for 1-NC and 2-NC samples, respectively. Slightly larger intensity of the D peak of the sample prepared by pyrolysis of carbon and urea is in accordance with the XPS measurements and refers to a higher content of sp³ carbon in this material. There is also a visible shoulder on the G-peak at ~1604 cm⁻¹ which is associated with defective graphitic structures [201]. Overall, the Raman spectra showed that the 1-NC material was somewhat more disordered than 2-NC.

The RDE method was used as the primary method for testing the ORR activity of the N-doped MWCNT/graphene catalysts. The ORR was studied on GC electrodes modified with 0.1 mg cm^–2 of the selected catalyst, which was deposited from a 1 mg ml^–1 dispersion also containing 0.25% of Tokuyama AS-4 ionomer. The ionomer served as both a dispersing agent and a binder. Figure 11 shows the results on the undoped GO/MWCNT composite and all of the doped materials. The undoped material showed a rather low activity towards the ORR, with an onset potential (Eonset, defined as the potential at –0.05 mA) of –200 mV vs SCE. There is a clear second reduction wave at –0.8 V, which is can be ascribed to quinone-type groups [202] reducing some of the H2O2 further to water. This is supported by the K-L data shown in Figure 12, where the number of electrons transferred per O2 molecule (n) increases from 2 (O2 reduction to H2O2) to 3 (both two- and four-electron electroreduction processes taking place).

The K-L plots (Figure 12) were derived from the ORR polarisation data with the number of electrons transferred per O2 molecule (n) shown in the inset. The n value was calculated from the K-L equation [203]:

= + = − −

. ^{/ / /} , (10)

where I is the measured current, Ik and Id are the kinetic and diffusion-limited currents, respectively, k is the rate constant for O2 reduction (cm s⁻¹), A is the geo-metric electrode area (cm²), F is the Faraday constant (96485 C mol⁻¹), ω is the rotation rate (rad s⁻¹), DO2 is the diffusion coefficient of oxygen (1.9×10⁻⁵ cm² s⁻¹) [204], CO2 is the concentration of oxygen in the bulk (1.2×10⁻⁶ mol cm⁻³) [204]

and ν is the kinematic viscosity of the solution (0.01 cm² s⁻¹) [205]. The values of DO2 and CO2 are given for 0.1 M KOH solution.

Using SC as the dopant shifted the onset potential to –175 mV vs SCE and increased the reduction current values notably. Switching the nitrogen precursor to CH shifted the Eonset potential a further 30 mV to the positive side and increased the reduction current values at a given rotation rate of the electrode, with the ORR becoming diffusion-limited at more negative potentials. Doping with BR had a much more profound effect on the ORR activity than either SC or CH: the onset shifted to –130 mV vs SCE and n was near 4 in the whole potential window studied. The peak at around –0.2 V vs SCE shown at lower rotation speeds is due to the reduction of oxygen trapped in the inner structures of the catalyst layer, indicating that this catalyst has a complex structure with a high specific surface area. Doping with urea yielded a catalyst with moderate activity (1-NC), with onset potential of –132 mV vs SCE, but no clear reduction current plateaux on the RDE polarisation curves and the n value of around 3.5–3.7. DCDA doping (2-NC) was much more successful with an activity similar to NG/NCNT-BR (Eonset of –121 mV), albeit lower reduction currents.

Figure 11. RDE voltammetry curves for oxygen reduction on (a) undoped GO/MWCNT, (b) NG/NCNT-SC, (c) NG/NCNT-CH, (d) NG/NCNT-BR, (e) 1-NC and (f) 2-NC modified GC electrodes in O2-saturated 0.1 M KOH. ν = 10 mV s⁻¹, ω = (1) 360, (2) 610, (3) 960, (4) 1900, (5) 3100 and (6) 4600 rpm. A = 0.2 cm².

Since there is no way to quantitatively measure the number of electrochemically accessible active sites on this type of materials, it is inconclusive exactly what causes the increase in activity. It has been shown, however, both experimentally [102,206] and with DFT (density functional theory) calculations [107,108] that pyridinic and quaternary (graphitic) N increase the ORR activity of carbon materials [207]. It has been proposed that nitrogen atoms can create a high positive charge density on the neighbouring carbon atoms and thus change the adsorption mode of the O2 molecule from end-on adsorption to side-on adsorption [208] in addition to weakening of the O=O bond. Pyridinic nitrogen has been shown to donate a p-electron into the aromatic π-system of the carbon ring, thereby destabilizing the carbon atoms next to it and providing local charge [208,209], has been shown to enhance the second step of the oxygen reduction reaction in acidic media [72] and enable the adsorption of O2 on the carbon atoms next to it in alkaline media. Graphitic nitrogen, which also destabilises the carbon ring it is doped into since it has a higher electronegativity than carbon [210], has been shown to facilitate the chemisorption of oxygen onto the adjacent carbon atom in the first step of the ORR and pyrrolic nitrogen has been correlated with the two-electron reduction of oxygen into hydrogen peroxide as well [72]. Other activity descriptors such as a high surface area (not measured here, but likely quite high due to the high SSA of the starting materials) and defectiveness (visible from Raman spectra) also likely contributed to the high ORR activity of these materials. Compared to results on N-doped graphene [126], which was doped using the same procedure, the separation of graphene layers by addition of MWCNTs was very successful as a means of increasing the ORR activity.

Figure 12. Koutecky–Levich plots for oxygen reduction on (a) undoped GO/MWCNT, (b) NG/NCNT-SC, (c) NG/NCNT-CH, (d) NG/NCNT-BR, (e) 1-NC and (f) 2-NC modified GC electrodes in 0.1 M KOH. E = (★) −0.4, ( ), −0.5, (▶) −0.6, (◀) −0.7, (♦)

−0.8, (▾) −0.9, (▴) −1.0, (●) −1.1 and (■) −1.2 V. Inset shows the potential dependence of n. Data derived from Fig. 11.

Figure 13. Stability of NG/NCNT-BR and 2-NC modified GC electrodes in O2-saturated 0.1 M KOH during 1000 potential cycles with the inset showing LSV results. ν = 10 mV s⁻¹, ω = 960 rpm, inset ν = 100 mV s⁻¹. A = 0.2 cm².

Comparison of the N-doped composite catalysts with a commercial Pt/C catalyst are visible in Figure 14. Both the 2-NC and the NG/NCNT-BR catalyst performed near to the activity of the 20 wt.% Pt/C catalyst from E-TEK, with similar half-wave potentials (E1/2), albeit more negative onset potentials.

The stability of both NG/NCNT-BR and 2-NC were tested during 1000 potential cycles from –1.2 V to 0 V vs SCE (Figure 13). After every 100 cycles an LSV curve and RDE polarisation curve were recorded. The LSV peak current values were nearly identical during the whole test and the ORR polarisation curves showed minimal activity loss (with no shift in the onset potential), indicating that these materials have excellent stability in alkaline media.

For comparison purposes with other catalysts in the literature as well as the materials presented later in this work, the jk values at 0.8 V vs RHE were also calculated. The formula for converting potentials from SCE to RHE used is ERHE = ESCE + 1.008 V, which was determined by calibrating the SCE used. The current values were divided by the electrode area to obtain the ORR current densities. The results are given in Table 3. The 2-NC material obviously has an edge over the other N-doped catalyst, which is why it was chosen for testing in a DMFC.

Table 3. ORR parameters of N-doped MWCNT/graphene composite catalyst materials and Pt/C in 0.1 M KOH.

Catalyst Eonset

(mV vs SCE) E1/2

(mV vs SCE) jk at 0.8 V vs RHE (mA cm^–2)

NG/NCNT-BR –130 –270 2.2

NG/NCNT-SC –170 –364 0.8

NG/NCNT-CH –145 –267 1.9

1-NC –132 –300 1.8

2-NC –121 –260 2.6

20 wt.% Pt/C –105 –236 1.7

After pre-selecting the 2-NC catalyst from the preliminary RDE tests, a full scale DMFC experiment was conducted with this material [II]. Prior to DMFC testing, the methanol tolerance of the 2-NC material was also studied, due to the pos-sibility of methanol crossover from the anode side to the cathode side. This is a common issue when using Pt/C catalysts, because Pt is also highly active towards methanol electro-oxidation, leading to a mixed potential and adsorption of side products such as CO, which poison the electrode [21]. The activity of the 2-NC catalyst was first tested in O2-saturated 0.1 M KOH, after which a 0.1 M KOH methanol solution was added, so that the end concentration of methanol in the electrochemical cell would be 3 M without changing the KOH concentration. The cell was then re-saturated with oxygen and another RDE voltammetry curve was recorded. As seen in Figure 15, the ORR performance of 2-NC does not change significantly (most importantly, there is no methanol oxidation peak) when going from a methanol-free environment to 3 M MeOH solution, so the catalyst is quite resistant to methanol poisoning.

Figure 15. RDE voltammetry curves for oxygen reduction on a 2-NC modified GC electrode in O2-saturated 0.1 M KOH solution with and without 3 M MeOH. ν = 10 mV s⁻¹, ω = 960 rpm.

The 2-NC material was spray-painted as the cathode of an alkaline anion-exchange membrane fuel cell with a FAA3 membrane from Fumatech (more details of the MEA preparation are given in section 6.9) with the results given in Figure 16.

The FAA3 membrane was selected due to its better resistance to methanol cross-over compared to the Tokuyama membranes. The activity was compared with a 60 wt.% Pt/C material. The 2-NC material outperformed even the high-loading Pt/C in the DMFC test, with a maximum power density of 0.72 mW cm⁻² (same as the Pt/C) and an OCV of 0.64 V (compared to 0.44 V for the Pt/C). It is to be noted, however, that the 2-NC material reaches its maximum power density at much higher potentials. The results were also highly competitive with other alkaline DMFCs at the time, including N-doped CNTs [211] and other studies using Pt/C as the cathode [212,213].

This work on nitrogen-doped MWCNT/graphene composites was among the first to demonstrate the activity of metal-free catalysts for alkaline DMFCs. It also showed the success of this doping method (wet mixing of small N-containing molecules with carbon nanomaterials followed by pyrolysis in inert atmosphere) in creating a high relative concentration of pyridinic N on the surface of the material, on which the later improvements made in this work are based. It also showed that DCDA is the most successful of the dopants.