2.1. Physico-Chemical Characterization
Figure 1 shows the X-ray diffraction (XRD) patterns of carbon supported Pt
1Ni
1/KB, Pt
3Ni
2/KB and Pt
2Ni
3/KB as prepared, treated at 900 °C and leached after the thermal treatment. XRD patterns of the Pt
1Ni
1/KB, reported in
Figure 1a show a disordered cubic structure (fcc) for the as-prepared catalyst and a single ordered primitive cubic (L1
2) phase for the alloy treated at 900 °C. The occurrence of the primitive cubic structure in the sample treated at high temperature is evident from the presence of the superlattice reflections, i.e., (001), (110) and (210) and from a better matching with the Joint Committee on Powder Diffraction Standards (JCPDS) card (65-2797) than the JCPDS card (04-0802) related to Pt. The as-prepared catalyst does not show any significant separation between Ni and Pt probably due to the low level of crystallinity, and after the high temperature treatment (900 °C), no separation of metallic Ni phase was observed. A shift towards higher Bragg angles is observed, indicating the formation of the solid solution between Pt and Ni, with a 35 at % of Ni in the alloy (
Table 1). As reported in
Table 1, the crystallite size after the thermal treatment is about 2.7 nm.
The catalyst prepared with an excess of Pt (Pt
3Ni
2/KB), whose XRD patterns are reported in
Figure 1b, shows small crystallites, about 2.3 nm, but also the presence of Ni hydroxide species for the as-prepared catalyst (see the shoulder at about 35° 2θ) that evolve with the formation of a separate metallic Ni phase at 900 °C (not clearly visible, since the Ni peaks are very close to Pt reflections). Thereafter, the effect of leaching was investigated. This post-treatment promoted the dissolution of unalloyed Ni, as proven by the XRD analysis. Nonetheless, a suitable alloying (close to the bulk composition, see
Table 1) was achieved as proven by the shift of fcc reflections.
In the case of the catalyst containing an excess of Ni (Pt
2Ni
3/C), whose XRD patterns are reported in
Figure 1c, the presence of sharp peaks indicating large crystallites of metallic Ni after the thermal treatment (JCPDS card 1-1258) is evident. For the as-prepared catalyst, the presence of Ni hydroxides is clearly evident. After the thermal treatment at 900 °C, the formation of two separate metallic phases occurs, one rich in Ni and the other in Pt. After the leaching post-treatment, only a slightly modification is achieved and the two separate phases are still present. However, a certain degree of alloying is obtained (26% Ni in the alloy). The crystallite size for the Pt phase is larger than 6 nm, as reported in
Table 1.
A morphological analysis of the dispersion of the metallic particles on carbon (
Figure 2) shows an increase of the particle size for the sample with a larger Ni content (average Pt particle size of 7.6 nm). Particle sizes around 4.5 nm are observed for the Pt
1Ni
1 and Pt
3Ni
2 treated at 900 °C. On the other hand, the commercial Pt/C catalyst (E-TEK) exhibits a narrow particle size distribution with average size of 4.7 nm.
A surface characterization of the Pt-Ni samples was carried out by X-ray photoelectron spectroscopy (XPS). There was no significant change in the surface composition of the samples compared to the Pt:Ni ratio of their bulk compositions as evaluated by the energy dispersive X-ray analysis (EDX) analysis (not shown).
Figure 3a shows the Pt 4f and Ni 2p spectra of the Pt
1Ni
1/C sample as prepared. In the case of Pt, the analysis of the photoelectron spectra indicates similar occurrence of metallic and oxidized Pt (2+) on the surface. Regarding Ni 2p, the deconvolution of the main bands shows a prevalence of hydroxide and oxide species; whereas, a very low amount of metallic Ni is present in the surface. The presence of such hydroxide and oxide species in the outermost layers may be due to the oxidation of surface Ni species.
The XPS analysis of the sample containing an excess of Pt (Pt
3Ni
2/C) is reported in
Figure 3b. Based on the quantitative determination reported in
Table 2, it is possible to state that this sample shows a more oxidised Pt on the surface than the Pt
1Ni
1. Furthermore, this sample shows that Ni is, also in this case, largely present as hydroxide and oxide species on the surface with a small amount of zero-valent Ni.
The XPS analysis of the sample containing an excess of Ni (Pt
2Ni
3/C) shows that Pt is essentially occurring on the surface as Pt
2+, whereas the Ni occurs as hydroxide and oxide species (
Figure 3c). A complete dataset of the surface composition and surface oxidation states of the investigated samples is reported in
Table 2.
2.2. Electrochemical Characterization
The electrochemical performance of the PtNi/C cathodic catalysts prepared by the formic acid method and successive thermal and leaching treatments was evaluated in a single cell based on a thin Nafion NR212 membrane. This was selected in order to increase the operating temperature of the PEFC device, reducing simultaneously the relative humidity (R.H.) and pressure to simulate the automotive conditions. However, initially, the cells were conditioned and investigated at 80 °C and full humidification (100% R.H., 3 bar
abs) as reference conditions. The polarization curves under the latter conditions and in the presence of air as the oxidant are reported in
Figure 4a. They show a similar performance for Pt
1Ni
1/C, Pt
3Ni
2/C and the benchmark Pt/C catalyst, in particular, in the low current density (activation) region; whereas, at high current density, the cell equipped with the benchmark catalyst appears slightly better performing than the other catalysts. On the other hand, the behavior of the catalyst rich in Ni and with the presence of large metallic Ni particles was not showing a performance level than the other cathode catalysts investigated (
Figure 4). This was due to the large particle size and the presence of two separate phases (PtNi and metallic Ni). The presence of the Ni phase can be also responsible of the high cell resistance recorded for the cell based on this catalyst, as can be observed in
Table 3.
In order to exacerbate the differences among the curves, oxygen was fed at the cathode and the polarization curves (
Figure 4b) were also recorded under these conditions (80 °C, 100% R.H., 3 bar
abs). The observed trend is exactly the same of that recorded by feeding air at the cathode (see
Figure 4a,b). Since the cell based on the cathode catalyst that was rich in Ni showed a poor performance, this cell was not investigated further; however, the membrane electrode assemblies (MEAs) based on Pt
1Ni
1/C and Pt
3Ni
2/C were tested at 80 °C under low humidification (50% R.H., 1.5 bar
abs) conditions (
Figure 5a) and compared to the one equipped with the commercial Pt/C catalyst.
At low RH, the MEA based on the Pt-rich catalyst (Pt
3Ni
2/C) showed a lower voltage loss in the activation region, compared to the equimolar bimetallic catalyst and benchmark Pt/C. The slightly better behavior of this catalyst could be ascribed to a proper composition or to a lower crystallite size compared to the equimolar PtNi sample. On one hand, the Pt crystallite size of Pt
3Ni
2/KB (2.3 nm,
Table 1) is slightly lower than PtNi/KB (2.7 nm). A smaller particle size means larger electrochemical surface area and more available active sites for the oxygen reduction. However, since negligible differences were observed between PtNi catalysts under full humidification, the different surface characteristics could also play a role in enhancing the catalytic activity of the catalyst. The most relevant characteristic of Pt
3Ni
2/KB is its larger amount of metallic Ni (19.7%) as revealed by XPS, which could contribute to oxygen adsorption on the Pt surface (by geometric and electronic effects). The polarization curves recorded at 95 °C under low R.H. conditions (
Figure 5b) confirm a better catalytic activity of the Pt-rich bimetallic catalyst compared to the other ones. In this case, the PtNi (1:1) also shows a higher catalytic activity than the Pt/C, indicating the beneficial effect of Ni in enhancing the kinetics of the oxygen reduction reaction, in particular under low humidification. The proton availability related to the water content has, of course, an effect on the kinetics. A significant reduction of relative humidity caused a decrease of proton availability and a consequent increase of the activation barrier for the ORR. This aspect could be associated to both membrane and ionomer dry-out, resulting in a lower availability of protons at the catalyst-ionomer electrolyte interface. The presence of Ni, in its oxidized forms, could help to reduce these constraints.
Figure 6 summarizes the cell voltage corrected by the ohmic drop (iR-free) at 0.5 A·cm
−2 for the three most performing catalysts at the different operating conditions used. Although a comparison with the literature is not easy due to the different conditions and materials (also in terms of catalyst loading) used, we can observe that the performance is similar to that recorded in Refs. [
8,
30,
31], where potential values (iR-free) around 0.8 V are reported at 0.5 A·cm
−2. Mani et al. [
8] reported a cell voltage lower than 0.75 V for a PtNi/C cathode-based MEA at 80 °C, 100% RH, H
2/O
2 and Pt loading close to 0.1 mg·cm
−2 at a much lower current density of 0.002 A·cm
−2. Da Silva et al. [
31] employed a higher Pt loading of 0.5 mg·cm
−2 for several PtNi catalysts. The best-performing one achieved a cell voltage of 0.34 V at 1000 A·g
Pt−1, equivalent to 0.5 A·cm
−2 (reported in
Figure 6), but not iR-corrected, at 80 °C and fully humidified H
2/O
2 streams. The series resistance was quite high (0.94 Ω·cm
2) and, after iR correction, the cell voltage is close to 0.81 V. Peng et al. [
30] have recently reported PEFC results for a PtNi catalyst performing 0.8 V at 0.5 A·cm
−2 with 75% RH at 80 °C. Han et al. [
29] obtained very promising results using dealloyed Pt-Ni catalysts. The dealloying method is a useful approach to obtain Pt-rich surface or core-shell structures for bimetallic catalysts [
32]. They showed high performance and durability using a family of dealloyed Pt-Ni catalysts for the ORR [
29,
32]. The behavior at 80 °C and low current density were comparable to that observed in the present work using Pt
3Ni
2/C and Pt
1Ni
1/C catalysts, although 0.2 mg·cm
−2 Pt loading was used in our investigation compared to 0.1 mg·cm
−2 reported in Ref. [
29]. At high current densities, the performance reached by Han et al. [
29] was higher compared to that obtained in the present analysis with our Pt-Ni catalysts; however, it must be taken into account that a thinner membrane was employed in Ref. [
29], which produced a lower cell resistance.
Furthermore, in order to increase the stability of the catalysts, a post-dealloying thermal annealing is usually applied [
29,
32]. This procedure produces an increased Pt surface diffusion rate promoting an improved Pt-skin layer on the bimetallic catalyst [
29,
32]. Furthermore, the dealloying process, in particular starting from high initial non-noble atom concentration, generates porous structure, clearly visible by scanning tunneling electron microscopy (STEM) [
29,
32]. In our approach, the thermal treatment was carried out before the acid leaching procedure, and, in the case of the Pt
1Ni
1/C catalyst, produced a more ordered alloy structure, as observed from the XRD pattern. From TEM images reported in
Figure 2 at low magnification, porous nanoparticles are not detectable, and, although a Pt enrichment of the surface can be envisaged from XPS, the formation of a Pt skin layer should be confirmed by using other techniques. In terms of performance, the formation of a more ordered crystallographic structure for the Pt
1Ni
1/C catalyst did not translate into a better behavior; however, this, together with the possible formation of a Pt skin layer, could be the reason of the improved stability obtained for the cell based on this catalyst (see discussion below). This was also reported in a recently published paper [
33], in which the conversion of PtNi nanoparticles from a disordered solid solution to an ordered intermetallic compound led to an enhanced durability and better ORR activity. In our case, the ORR activity was found to be better for the Pt
3Ni
2 disordered solid solution compared to the ordered Pt
1Ni
1, more likely due to a major presence of Pt in the catalyst.
Accelerated degradation tests (ADTs), i.e., 15,000 step cycles (steps 0.6–1.2 V, cycle time 6 s, H
2–N
2), were carried out for all catalysts at 80 °C, 100% R.H., 1.5 bar
abs. A milder protocol, i.e., 0.6–0.9 V cycling, was first carried out, but the differences in performance before and after the test were found to be almost negligible, as observed in
Figure S1 of the supporting information. After the degradation test, polarization curves were carried out under the same conditions feeding oxygen at the cathode (
Figure 7). From the analysis of the curves recorded before and after the degradation test, it appears that the equimolar bimetallic catalyst (Pt
1Ni
1) shows the best stability since the polarization profile does not change much after the ADT, especially in the low current density region.
The benchmark catalyst shows the largest losses in performance, as can be derived from the histograms reported in
Figure 8, which summarizes the potential losses at two values of current density (of interest for what concerns catalytic activity and practical application) of the various MEAs after the ADT.
Figure 8 shows the cell potential at 200 mA·cm
−2 and 950 mA·cm
−2 for the three MEAs subjected to the ADT. The MEA based on the benchmark Pt catalyst showed the largest potential losses at low current density (8%) after the ADT while the Pt
1Ni
1/C-based MEA was the most stable (2% and 12.5% voltage losses at 200 mA·cm
−2 and 950 mA·cm
−2, respectively). In all cases, the latter MEA presented the highest cell potential after the ADT among the investigated ones both at 200 and 950 mA·cm
−2. Cyclic voltammetry analyses carried out before and after the ADT (not shown) allowed the determination of the electrochemical active surface area (ECSA) that is reported in
Table 4, together with the crystallite size determined by XRD at the BoL and EoL. The values show that all catalysts are affected by significant particle sintering; the crystallite size increases from three to five times after the tests, causing a relevant decrease of ECSA to almost half the initial value. However, the catalyst showing a more ordered crystallographic structure appears less affected by sintering phenomena during the ADT [
33]. It is worth mentioning that the ECSA is higher for the PtNi/C catalysts, related to their smaller particle size as revealed by XRD and TEM analyses. No significant change in the ECSA at the BoL may be attributed to the different content in Ni between PtNi/KB and Pt
3Ni
2/KB. Da Silva et al. reported an increased ECSA when introducing Ni caused by the decrease of Pt–Pt bond distance (geometric effect) and the increase of the 5d-band vacancy (electronic effect) [
31].
To confirm this conjecture, XRD and TEM analyses were carried out after the ADTs; the XRD patterns and the TEM images after the ADTs are reported in
Figure 9. A dramatic increase of crystallite size can be observed in the XRD patterns, together with a significant shift of the peaks towards lower Bragg angles, which is an indication of a dealloying phenomenon. From TEM micrographs, we can derive that a particle sintering occurred together with a leaching of particles due to carbon corrosion (as can be observed in
Figure S2 of the supplementary information). Thus, it appears that this preparation procedure is not suitable to obtain stable catalysts. Recently, Pt-Ni nanocage (PNC) catalysts were synthesized by a solvothermal method and investigated in PEFCs in terms of performance and stability [
30]. The PNC experienced slight agglomeration and some edge loss after cycling (30 K cycles in the range 0.6 V to 1.0 V). However, the majority of the particles retained their cage structure. In our case, large agglomeration and particles losses were observed indicating that the high temperature treatment is not enough to stabilize the PtNi alloy, in particular under these severe testing conditions (0.6–1.2 V cycling). Under milder conditions (0.6–0.9 V cycling) the results envisaged lower stability constraints (see
Figure S1 in the supplementary information).