High-Current Density Durability of Pt/C and PtCo/C Catalysts at Similar Particle Sizes in PEMFCs

Catalyst materials and membrane electrode assembly (MEA) fabrication

Transmission electron microscopy (TEM) of catalysts

Electron-probe microanalysis (EPMA)

Electrochemical measurements and diagnostics

The two catalysts dealt with in this manuscript namely, annealed Pt/HSC-a and PtCo/HSC-a were synthesized via the same method including heat treatment to achieve similar particle sizes and distribution. Table I shows the BOL properties of the two catalysts. Pt/Co atomic ratio of as-received PtCo/HSC-a catalyst is 3.3. Annealed Pt/HSC-a and PtCo/HSC-a catalysts feature an average particle size of 4.3 and 4.5 nm, respectively along with an ECSA value quantified via H_upd measurements in an MEA of 39 to 42 m²/g_Pt. As is well-known because of alloying, PtCo/HSC-a catalyst shows a specific activity of 1308 μA cm⁻² _Pt which is 60% higher due to the alloying effect than the Pt/HSC-a catalyst. A higher mass activity is also observed (0.56 A/mg_Pt for PtCo/HSC-a vs 0.31 A/mg_Pt for annealed Pt/HSC-a).

MEAs containing the annealed Pt/HSC-a and PtCo/HSC-a cathode materials were subjected to the catalyst AST protocol and characterized via various analytical procedure at EOT. Figure 1a shows the particle size distribution for the annealed Pt/HSC-a and PtCo/HSC-a cathode catalysts at BOL (solid lines) and EOT (dashed lines), whereas Fig. 1b shows corresponding bright-field TEM images. Clearly, the particle size distributions for the EOT samples have broadened significantly when compared with the BOL samples. The same trend is observed for the average particle size, which increased to ∼5.5 nm for both the catalysts at EOT. This behavior is expected after 30k cycles due to a variety of possible mechanisms, such as particles migration/coalescence, electrochemical Ostwald ripening and carbon corrosion leading to nanoparticle coarsening. ^22,24 We should point out that the very large particles shown in the bright-field images were excluded from particle size distributions, as they are an exception and thus rarely observed. Higher average particle size and wider distribution observed at EOT is also reflected in the similar magnitude of ECSA loss after 30k cycles as shown in Fig. 2a. Both the catalysts show an EOT ECSA in the range of 26 to 28 m² g⁻¹ that corresponds to ∼33% loss in Pt surface area after 30k cycles. Figure 2b shows the cyclic voltammograms of the Pt and PtCo catalysts wherein the H_upd area of 0.05 to 0.4 V clearly decreases at EOT. Further, the H_upd peaks corresponding to the H-adsorption/desorption on the various facets of Pt nanoparticles reveal the loss of the alloying element Co in the PtCo/HSC-a catalyst. A broad featureless H_upd region is observed for PtCo/HSC-a at BOL in the anodic scan due to the alloying effect, which becomes more Pt-like at EOT as Co is leached from the alloy catalyst nanoparticles.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 3a shows the ORR kinetics of the two catalysts as measured via H₂/O₂ polarization curve. At BOL, PtCo/HSC-a catalyst clearly shows a much improved kinetics as reflected by ∼25 mV lower overpotential at 0.05 A cm⁻² for ORR compared to the annealed Pt/HSC-a. Post-30k durability cycles, the kinetic benefit of the alloy catalyst decreases significantly with PtCo/HSC-a showing only ∼11 mV improvement at 0.05 A cm⁻² compared to annealed Pt/HSC-a. As is well-known, this is due to the leaching of Co from the alloy catalyst that leads to a strain relaxation which was originally responsible for the higher activity. ^7,36 It should be noted that the intrinsic kinetic activity of annealed Pt/HSC-a catalyst marginally increases post-30k durability cycles which is at least partly due to the increase in particle size. ^16,50 This is clearly observed in the specific activity plot (Fig. 3b) wherein PtCo/HSC-a decreases from 1308 ± 55 to 1053 ± 44 μA cm⁻² whereas annealed Pt/HSC-a increases marginally from 797 ± 12 vs 846 ± 52 μA cm⁻². The mass activity does drop for both the catalysts post-30k cycles due to the loss in ECSA that leads to decrease in absolute H₂/O₂ performance normalized to the Pt loading (Fig. 3c).

Zoom In Zoom Out Reset image size

Figure 4a shows the H₂/air polarization performance of annealed Pt/HSC-a and PtCo/HSC-a catalysts at BOL and EOT conditions measured at 0.1 mg_Pt cm⁻² loading up to an operating current density of 2.5 A cm⁻². At BOL, PtCo/HSC-a catalyst provides significant benefit in performance compared to annealed Pt/HSC-a as measured by an improvement of ∼40 to 60 mV in cell voltage at all current densities up to 2.5 A cm⁻². This significant improvement in cell voltage of PtCo/HSC-a is solely brought upon by the positive impact of the alloying element compared to annealed Pt/HSC-a catalyst at BOL. However, the relative performance of the two catalysts depict a drastically different picture after the 30,000 durability cycles. The H₂/air polarization performance of annealed Pt/HSC-a is relatively unchanged whereas PtCo/HSC-a catalyst shows a significant drop. A closer look at the polarization curves in Fig. 4a shows the cell voltage of PtCo is only ∼10 mV better than annealed Pt at EOT conditions up to a current density of 1.0 A cm⁻². Further at current densities greater than 1.5 A cm⁻², a massive drop in cell voltage is observed for PtCo catalyst. At 2.5 A cm⁻², PtCo catalyst has lost ∼200 mV compared to its own BOL performance and is also lower than annealed Pt catalyst by ∼115 mV at EOT conditions. While the loss in cell voltage of PtCo/HSC-a catalyst in the kinetic region can simply be explained by the loss in its specific activity due to the leaching of Co element from the alloy, the loss in high current density performance requires a more closer analysis of the transport properties of the cathode catalyst layer. HFR measured as a function of current density does not show any changes for both the catalyst containing MEAs at EOT conditions indicating that the membrane resistance has not changed measurably. This is expected since the catalyst-specific AST protocol does not cause any structural or compositional changes to the membrane.

Zoom In Zoom Out Reset image size

Various electrochemical diagnostics and ex situ characterization of the MEA were carried out to understand the sources of potential-cycling induced cell voltage losses: i) H_upd measurements to quantify Pt ECSA, ii) CO_ads stripping as a function of RH to understand Pt utilization on carbon support, iii) H₂/N₂ impedance measurement to understand H⁺ transport resistance in CCL, iv) O₂ limiting current density analysis to quantify local-O₂ transport resistance in CCL. Further, TEM-EDS analysis were carried out to quantify Pt/Co atomic ratio in the cathode catalyst and EPMA analysis to quantify metal-cation distribution in the MEA.

Figure 5 shows the Pt utilization measured via CO_ads stripping surface area measurement as a function of % RH (Pt-CO_ads + H₂O → Pt + CO₂ + 2H⁺ + 2eˉ). This is a measurement to investigate the accessibility of Pt to H⁺/H₂O. At a low enough RH of 20%, only the Pt nanoparticles in direct contact with the ionomer are active for CO_ads stripping whereas the Pt nanoparticles present inside the micropores are inactive since they are not in contact with the ionomer. ^15,45 Only at higher RH values of ≥ 60%, Pt nanoparticles inside the micropores become active for CO_ads stripping due to the presence of condensed water. Hence, the utilization at 20% RH could be considered as a measure of the fraction of Pt present on the external surface of the carbon support in direct contact with the ionomer. As can be seen in Fig. 5a, at BOL both annealed Pt/HSC-a and PtCo/HSC-a catalysts show a utilization of ∼70 to 80% which increases to almost to 100% pointing to an improvement in the accessibility of Pt to H⁺/H₂O. It is hypothesized here that two plausible structural changes in the CCL could lead to such an improvement in Pt utilization which are depicted in Fig. 5b: i) Pt nanoparticles in the interior pores of the carbon support undergo dissolution followed by redeposition on the external surface and/or ii) widening of the carbon microporous openings (due to surface carbon corrosion that occurs during catalyst AST potential cycling ⁵¹ ). These two structural changes could enable the apparent improvement in the utilization of Pt nanoparticles.

Zoom In Zoom Out Reset image size

Figure 6 shows the local-O₂ transport resistance measured at BOL and EOT conditions. Figure 6a is a plot of pressure-independent transport resistance as a function of inverse roughness factor for both the catalysts at BOL (solid lines) and EOT (dashed lines). The slope of the plot is a measure of the local-O₂ transport resistance in the CCL which is captured as bar plots in Fig. 6b. ⁸ Several recent studies have shown that the local-O₂ transport resistance in the CCL is directly correlated to the volume of micro- and small meso-pores in the carbon primary particle. ^8–10 These internal pores impose a resistance to O₂ transport and hence hinder the accessibility of Pt to O₂. At BOL, both catalysts exhibit a local-O₂ transport resistance of ∼12 s cm⁻¹ which appears to decrease to ∼8 s cm⁻¹ after 30,000 durability cycles. This is an interesting observation that indicates the resistance towards local-O₂ transport in both the annealed Pt and PtCo-alloy CCL decrease after 30,000 catalyst-AST potential cycles. This result in conjunction with the Pt utilization measured via CO_ads stripping indicates that the possible redeposition of the nanoparticles from the internal pores of carbon support to its external surface and/or the widening of microporous openings could result in the decrease of the local-O₂ transport resistance.

Zoom In Zoom Out Reset image size

Figure 7 shows the H⁺ transport resistance in the CCL measured via H₂/N₂ EIS protocol at BOL and EOT conditions. H⁺ transport resistance remains constant for the annealed Pt catalyst within the error level of the measurement whereas marginal increase is observed for the PtCo-alloy catalyst from 0.14 ± 0.02 to 0.19 ± 0.02 Ω cm². It should be noted that the data in Fig. 7 corresponds to the resistance measurements at an inlet RH of 50% which is reasonably dry enough to capture minor changes in H⁺ transport resistance of the catalyst layer due to various cationic contaminants. At higher RH values (≥75%), the presence of small amounts of contaminants is usually not perceptible via H₂/N₂ EIS measurements. This measurement suggests the presence of contaminants that impedes the overall cationic transport in the bulk ionomer phase of the CCL. It is emphasized that the EIS measurement is only indicative of resistance trends because impedance measurements (AC signal) do not necessarily correspond to fuel cell operation under constant operating conditions. The AC signal ensures that the contaminant cations are mobile whereas under steady state conditions, the contaminant cation flux is equal to zero. ⁵²

Zoom In Zoom Out Reset image size

Figure 8 shows the average Pt content as a function of the PtCo particle sizes for the BOL and the EOT samples, as well as a comparison of the Pt/Co atomic ratio for the pristine powder, the BOL and the EOT samples. For the case of the BOL sample (Fig. 8a), the atomic content of Pt increases as the particle size decreases, whereas for the case of the EOT sample (Fig. 8b), the Pt content remains approximately constant for all particle sizes. Regarding the BOL samples, the smaller particles exhibit thick Pt-rich shells when compared with their size. For larger particles, a significant amount of Pt segregates to the surface to form the shell, leaving the bulk depleted in Pt. ⁵³ During cycling, the smaller particles dissolve and disappear (as shown in Fig. 1a), redepositing Pt on larger particles through Ostwald ripening. As a result, the EOT samples exhibit a rather constant Pt content for all particle sizes. Specifically, at EOT there is an enrichment of Pt atomic content to about 90% irrespective of the nanoparticle size. In addition, when the Pt/Co ratio is compared for the pristine powder and BOL samples, we notice a small enrichment in Pt content from 3.3 to 4.2. This is due to Co²⁺ dissolution during the various stages of MEA fabrication and testing such as catalyst ink milling and break-in. Further, after 30,000 durability cycles, there is a preferable Co dissolution with respect to Pt, where the Co remains in ionomer phase during the whole process, while the dissolved Pt can redeposit in larger particles as mentioned above. ^23,30,53 Pt/Co atomic ratio increases to 10.2 after 30,000 durability cycles indicating significant Pt enrichment of the nanoparticles.

Zoom In Zoom Out Reset image size

Using the Pt/Co atomic ratio from TEM-EDS measurements, simple calculations were carried out to quantify the amount of sulfonic acid sites ion-exchanged with the Co²⁺ cations at BOL and EOT conditions. If all the leached Co²⁺ remains only in the membrane (12 μm thick, 800EW), ∼3% SO₃ ^ˉ sites are calculated to be exchanged with Co²⁺ cations at BOL which increases to 9% at EOT. On the other hand, if all the leached Co²⁺ cations is assumed to remain in the electrode ionomer phase (0.9 I/C, 825EW), then it turns out that 26.5% of the SO₃ ^ˉ sites are exchanged with Co²⁺ cations at BOL which increases to a significantly high value of 63.5% at EOT.

EPMA cross-sectional analysis of the PtCo-alloy catalyst MEA was carried out at BOL and EOT condition to detect the presence of various elements such as Pt, Co, F and S distribution across the MEA. EPMA map scans are shown in Fig. 9a whereas line-scans are shown in Fig. 9b. This measurement allows the detection and quantification of Co²⁺ cations in the membrane. At BOL, most of the Co is present in the CCL with only a small fraction present in the membrane. At EOT, an increase of Co content across the membrane is observed along with a decrease in the CCL.

Zoom In Zoom Out Reset image size

It should be noted that EPMA is an ex situ measurement performed in vacuum which shows the Co²⁺ concentration distribution across the MEA under the absence of any gradients imposed by fuel cell operation. During fuel cell operation, Co²⁺ cations are under the influence of diffusion and migration gradients that transports Co²⁺ cations into the cathode thereby ion-exchanging with the acid sites in the electrode ionomer. As was recently shown by Cai et al., ^32,33 movement of Co²⁺ cations into the cathode increases with various MEA operating conditions such as higher current densities, low RH and temperature. Co²⁺ cation accumulation in the cathode causes depletion of H⁺ thereby leading to i) decreased Pt utilization due to H⁺ inaccessibility, ii) decrease in H⁺ activity (pH effect) and iii) increased H⁺ transport resistances in the CCL. These impacts of Co²⁺ cation contamination is similar to the findings in several studies carried out in the past to investigate the impact of cation contaminants arising from leaching of metallic bipolar plates and external contaminants such as Na⁺ based ions in the cathode air feed. ^34,54

In order to understand the presence of Co²⁺ cations in the membrane, MEAs were fabricated using annealed Pt/HSC-a cathode catalyst and membranes that were pre-doped with 10% Co²⁺ cations. This MEA was subjected to 30,000 durability cycles and the EOT performance was measured. As shown in Fig. 10, the H₂/air polarization performance of the 10% Co²⁺ doped MEA with pure-Pt catalyst shows cell voltage comparable to the cycled PtCo/HSC-a catalyst. Particularly, the increased cell voltage loss profile at current densities greater than 1.5 A cm⁻² is reflective of PtCo/HSC-a catalyst at EOT conditions. This provides a confirmation that the presence of dissolved Co²⁺ cations in the membrane and its accumulation in the CCL is responsible for the significant cell voltage losses of cycled PtCo/HSC-a catalyst at high current densities. It can be noticed in Fig. 10 that the shape of EOT polarization curves at current densities ≥1.75 A cm⁻² is complex. Voltage gap between BOL and EOT curves increase from 1.75 to 2.25 A cm⁻² and tends to form a plateau from 2.25 to 2.50 A cm⁻². The increases in voltage loss from 1.75 to 2.25 A cm⁻² is attributed to the transport of Co²⁺ cations into the CCL. On the other hand, the plateau formation from 2.25 to 2.50 A cm⁻² is likely due to a H₂ pumping phenomena that was observed earlier with cation contaminated MEAs at high current densities. ³⁴

Zoom In Zoom Out Reset image size

A 1D-performance model was used to fit the experimental polarization curves and understand the various durability related cell voltage loss factors (see Appendix for details). Figure 11 shows a good agreement between the experimental and model generated polarization curves up to current densities of 2.5 A cm⁻². Figure 12 shows a breakdown of the voltage losses and gains at 2.0 A cm⁻² arising due to the 30,000 durability cycles from various CCL-specific factors such as changes in ECSA, specific activity, local-O₂ transport resistance and Co²⁺ contamination effect. In the case of the annealed Pt/HSC-a catalyst, cell voltage loss due to the decrease in ECSA appears to be the major factor which is compensated by the cell voltage gains from both specific activity improvement and decrease in local-O₂ transport resistance. On the other hand, for the PtCo/HSC-a catalyst, cell voltage loss due to ECSA decrease continues to be the major factor. A minor contribution from specific activity loss of PtCo catalyst is also observed (first-order effect of Co²⁺ dissolution from the catalyst nanoparticles). A marginal gain from the decrease in local-O₂ transport resistance is observed. Unlike, the annealed Pt catalyst, a major cell voltage loss factor due to the impact of Co²⁺ cation contamination of the electrode ionomer is observed for the PtCo (second-order effect of Co²⁺ dissolution). In summary, annealed Pt and PtCo catalysts show roughly similar magnitude of cell voltage losses due to the decrease in ECSA. However, for annealed Pt catalyst this does not translate to total cell voltage loss due to compensatory effect from improvement in specific activity and decrease in local-O₂ transport resistance. Contrarily for PtCo catalyst, cell voltage losses due to the decrease in ECSA is compounded by the losses arising from the decrease in specific activity and the Co²⁺ contamination effect.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

A direct comparison of the durability of high surface area porous carbon-supported annealed Pt and PtCo catalyst at similar particle sizes have been carried out to understand the factors that cause cell voltage losses at high current densities. Our results clearly show that:

• 1)  
  Both annealed Pt and PtCo catalysts exhibit similar ECSA losses upon 30,000 durability cycles, but PtCo suffers from significantly high cell voltage losses at current densities greater than 1.5 A cm⁻².
• 2)  
  The decrease in specific activity of PtCo catalyst due to Co²⁺ dissolution corresponds to only ∼13 mV in cell voltage losses which can be used to explain the losses in the kinetic region. However, the impact of Co²⁺ contamination of the electrode ionomer is significantly higher leading to ∼52 mV loss at 2.5 A cm⁻². Transport of Co²⁺ contaminant into the cathode during fuel cell operation at high current densities appears to be the predominant reason for large cell voltage losses in the PtCo catalyst containing MEA.
• 3)  
  Changes in Pt location distribution due to electrochemical Ostwald ripening and/or surface carbon corrosion during catalyst-AST durability protocol leads to a decrease in local-O₂ transport resistance that enable ∼20 mV cell voltage gains post 30,000 durability cycles.

Given the secondary effects of the dissolved Co²⁺ cations on the high current density performance, our study clearly points to the need for highly active and durable Pt-alloy catalysts that are resistant to the dissolution of the alloying element.

This work was partially supported by U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy under grants DE-EE0007651 and DE-EE0008821. Paulo Ferreira was partially supported by FCT, through IDMEC, under LAETA, project UIDB/50022/2020. NR would like to thank Dr Srikanth Arisetty for useful discussions. The authors would like to acknowledge David Lindsey, Kenneth Holt, Cristin Keary, Shruti Gondikar, Venkata Yarlagadda and Kathryn Stevick of GM for engineering support.

Appendix. Model development

In our previous paper, ⁸ we described an electrode model that considers catalyst morphology, particularly accounting for ORR taking place at both interior and exterior Pt surfaces. Whereas ORR current at the exterior Pt surface is affected by the local oxygen transport resistance (R_Pt-ionomer) due to Pt-ionomer interaction, an overall effectiveness factor is evaluated based on O₂ transport through micro-/meso-pores of HSC particles for the interior Pt. The effective oxygen permeability in micro-/meso-pores (relating to porosity/tortuosity) was obtained by fitting to the measured polarization curves. In this work, instead of tuning porosity/tortuosity, we have developed a semi-empirical correlation for the overall effectiveness factor of interior Pt in terms of Pt loading (L_Pt), electrochemically active area (ECA), and applied current density (i_app)

$$\begin{eqnarray}&&{\rm{\Theta }}={\rm{f}}\left({{\rm{L}}}_{{\rm{Pt}}},{\rm{ECA}},{{\rm{i}}}_{{\rm{app}}}\right)\end{eqnarray} \tag{ A·1 }$$

utilizing the experimental data (polarization curves, ECSA, specific activity, electrode H⁺ resistance, and R_O2-local) taken from cells built with Pt/HSC catalyst at both BOL and EOT.

Appendix. Impact of Co cation contamination

With the above-obtained overall effectiveness factor for interior Pt, one can further utilize the experimental data taken from cells built with PtCo/HSC catalyst to develop a semi-empirical correlation for the effect of Co²⁺ contamination. As electrode degrades, Co cations leach out of PtCo/HSC catalyst and contaminate ionomer by occupying proton sites. This would result in lowered activity of protons involved in ORR and thus incur a higher kinetic loss. In an extreme scenario, cations are driven towards cathode and effectively deactivate a portion of cathode adjacent to MPL, ⁵⁵ which would cause a large voltage loss at a high current density. Furthermore, lower relative humidity (RH) in cathode catalyst layer leads to more ORR closer to membrane, which would incur additional voltage loss. Therefore, we evaluate the effective Pt loading under the influence of the mole fraction of proton sites occupied by Co cation (x_Co), applied current density (i_app), and cathode RH (RH_cath)

$$\begin{eqnarray}&&{{\rm{L}}}_{{\rm{Pt}}}^{{\rm{eff}}}/{{\rm{L}}}_{{\rm{Pt}}}={\rm{f}}\left({{\rm{x}}}_{{\rm{Co}}},{{\rm{i}}}_{{\rm{app}}},{{\rm{RH}}}_{{\rm{cath}}}\right)\end{eqnarray} \tag{ A·2 }$$

which is then used in Eq. A·1 for calculating the overall effectiveness factor for interior Pt.

Appendix. Effect of ionomer EW

Given I/C and ionomer EW, the volume fraction of ionomer in an electrode can be calculated by, ⁴⁸

$$\begin{eqnarray}&&{\varepsilon }_{{\rm{i}}}=\left({\rm{I}}/{\rm{C}}\right)\cdot \displaystyle \frac{10}{{{\rm{f}}}_{{\rm{t}}\cdot }{{\rm{d}}}_{{\rm{m}},{\rm{dry}}}}\cdot \left(1+\displaystyle \frac{{{\rm{M}}}_{{\rm{w}}}\cdot {{\rm{d}}}_{{\rm{m}},{\rm{dry}}}\cdot \lambda }{{{\rm{d}}}_{{\rm{w}}}\cdot {\rm{EW}}}\right)\end{eqnarray} \tag{ A·3 }$$

where I/C is the weight ratio of ionomer to carbon in the electrode, f_t is 28 ± 2 μm/(mg_C cm⁻² ) for I/C <3/1, d_m,dry is the density of dry ionomer, d_w is the density of water, M_w is the molecular weight of water, and λ is water content depending strongly on RH. According to Eq. A·3, volume fraction of ionomer is proportional to I/C ratio and inversely proportional to EW (due to larger swelling at lower EWs). At a given λ, lower EW and higher I/C lead to a higher volume fraction of ionomer which effectively decreases electrode porosity, thus increasing local oxygen transport resistance and tendency to agglomerate (or equivalently reduced amount of interior Pt). Therefore, one can use the following empirical correlations to account for the effect of ionomer EW

$$\begin{eqnarray}&&{{\rm{R}}}_{{\rm{O}}2-{\rm{local}}}^{{\rm{eff}}}/{{\rm{R}}}_{{\rm{O}}2-{\rm{local}}}={\rm{f}}({\varepsilon }_{{\rm{i}}},{{\rm{i}}}_{{\rm{app}}})\end{eqnarray} \tag{ A·4 }$$

and

$$\begin{eqnarray}&&{{\rm{L}}}_{{\rm{Pt}},\,{\rm{int}}}^{{\rm{eff}}}/{{\rm{L}}}_{{\rm{Pt}},{\rm{int}}}={\rm{f}}({\varepsilon }_{{\rm{i}}},{{\rm{i}}}_{{\rm{app}}})\end{eqnarray} \tag{ A·5 }$$

Here, i_app is included as the impact of O₂ transport resistance which increases with current density. It is worth mentioning that these semi-empirical correlations are supposed to be material specific. However, it is expected that the dependent variables would apply to other material sets, and only empirical coefficients may vary.