Report

Ultralow-loading platinum-cobalt fuel cell catalysts derived from imidazolate frameworks

See allHide authors and affiliations

Science  14 Dec 2018:
Vol. 362, Issue 6420, pp. 1276-1281
DOI: 10.1126/science.aau0630

Combine and conquer

Platinum (Pt)–group metals, which are scarce and expensive, are used for the demanding oxygen reduction reaction (ORR) in hydrogen fuel cells. One competing approach for reducing their use is to create nanoparticles with earth-abundant metals to increase their activity and surface area; another is to replace them with metals such as cobalt (Co) in carbide or nitride sites. Chong et al. thermally activated a Co metal-organic framework compound to create ORR-active Co sites and then grew PtCo alloy nanoparticles on this substrate. The resulting catalyst had high activity and durability, despite its relatively low Pt content.

Science, this issue p. 1276

Abstract

Achieving high catalytic performance with the lowest possible amount of platinum is critical for fuel cell cost reduction. Here we describe a method of preparing highly active yet stable electrocatalysts containing ultralow-loading platinum content by using cobalt or bimetallic cobalt and zinc zeolitic imidazolate frameworks as precursors. Synergistic catalysis between strained platinum-cobalt core-shell nanoparticles over a platinum-group metal (PGM)–free catalytic substrate led to excellent fuel cell performance under 1 atmosphere of O2 or air at both high-voltage and high-current domains. Two catalysts achieved oxygen reduction reaction (ORR) mass activities of 1.08 amperes per milligram of platinum (A mgPt−1) and 1.77 A mgPt−1 and retained 64% and 15% of initial values after 30,000 voltage cycles in a fuel cell. Computational modeling reveals that the interaction between platinum-cobalt nanoparticles and PGM-free sites improves ORR activity and durability.

The oxygen reduction reaction (ORR) is more sluggish in proton-exchange membrane fuel cells (PEMFCs) than hydrogen oxidation and requires three to five times as much platinum (13). The high cost and scarcity of Pt have driven efforts to reduce Pt usage. Recent examples include Pt–transition metal (TM) alloys with distinctive three-dimensional (3D) structures (48). Excellent ORR activity and durability were demonstrated by the rotating disk electrode (RDE) method in oxygen-saturated aqueous electrolyte. Although the RDE approach provides important information about catalytically active sites, it does not fully reflect how the catalysts would perform in operating fuel cell environments of different mass and charge transport limitations (9, 10).

In fuel cells, catalysts in the membrane electrode need to be easily accessible by the reactants, particularly under low fuel cell polarization voltage where a large influx of reactant (O2) must be converted to produce high current density. For a small number of shaped but large crystallites prepared within ultralow Pt loading limitation, there will not be enough crystallites to spread over the electrode surface to encounter all of the O2 before they exit the electrode, resulting in a drop in the fuel cell current. The opposite approach, dispersing Pt to the atomic level, can result in fast Pt dissolution and poor catalytic activity (11). A third approach is to use a platinum-group metal (PGM)–free catalyst, which could eliminate the Pt usage altogether. Such catalysts, generally prepared from earth-abundant elements such as TMs (mostly Fe and Co) embedded in nitrogen-carbon composites (TM-Nx-Cy), have demonstrated promising ORR activity approaching that of Pt (1215).

When prepared from metal-organic frameworks (MOFs) or porous organic polymers as precursors, these catalysts possess densely and uniformly populated active sites throughout the electrode, easily accessible by O2 fluxes (16, 17). The key drawback, however, is their poor stability under PEMFC operations. Unlike Pt catalysts, of which the activity degradation is mainly caused by crystallite dissolution and agglomeration (18), the origin of the PGM-free catalyst deactivation is poorly understood because the nature of the active site is still under debate (19, 20). One possible cause is the oxidative degradation by hydrogen peroxide produced during ORR (21).

If the shortcomings of ultralow-loading Pt and PGM-free catalysts were mutually compensated through a synergistic interaction, Pt usage could be substantially reduced while maintaining excellent activity and durability. We report the design and synthesis of synergistic ORR catalysts containing an ultralow concentration of Pt alloy supported over PGM-free materials, denoted as LP@PF. We used a Co-containing and a Co- and Zn-containing zeolitic imidazolate framework (ZIF, a subgroup of MOFs) as the precursors, which we then thermally activated and catalyzed with Pt to form alloy. The resulting catalysts had very high mass activities (MAs) of 8.64 ± 0.25 A mgPt−1 and 12.36 ± 0.53 A mgPt−1 measured by RDE or 1.08 ± 0.17 A mgPt−1 and 1.77 ± 0.39 A mgPt−1 measured in fuel cells at an internal resistance–corrected (iR-free) voltage of 0.9 V. Both values exceed the U.S. Department of Energy (DOE) target of 0.44 A mgPt−1 (22). The catalysts showed excellent activity in both high-voltage and high–current density domains and good durability in a 30,000 voltage-cycle accelerated stress test (AST) in a fuel cell.

Our catalyst design is based on the following rationales. Pt-Co alloy represents one of the most active ORR catalysts and is currently used in commercial fuel cell vehicles, whereas Co-ZIF–derived PGM-free catalysts have also shown high specific surface areas, densely distributed active sites, and excellent ORR activities in PEMFCs (23). During thermal activation of Co-ZIF, a fraction of Co2+ is reduced to metallic nanocrystallites, whereas other Co ions are converted to atomically dispersed Co-Nx-Cy sites situated nearby. The Co nanocrystallites could serve as the seeds to amalgamate with subsequently added Pt to form alloy nanoparticles (NPs) with a core-shell structure. Close proximity between Pt-Co NPs and Co-Nx-Cy sites could promote synergistic catalysis.

We prepared a monometallic cobalt zeolitic methylimidazolate framework, Co(mIm)2 (also called ZIF-67), and a bimetallic ZIF containing zinc zeolitic methylimidazolate framework, Zn(mIm)2 (also called ZIF-8), coated by ZIF-67 (ZIF-8@ZIF-67). Both ZIF-67 and ZIF-8@ZIF-67 were then thermally activated. A subsequent controlled acid wash formed PGM-free catalyst supports PF-1 and PF-2, respectively. These supports were ORR active by themselves and retained a fraction of metallic cobalt nanocrystallites.

A Pt precursor was subsequently applied to PF-1 and PF-2, followed by in situ reduction in oleylamine and high-temperature annealing under ammonia (NH3) to obtain the final catalysts LP@PF-1 and LP@PF-2. Figure 1A schematically illustrates the characteristics of these catalysts. First, a majority of Pt was converted to Pt-Co NPs that were uniformly dispersed over a substrate of densely populated Co-Nx-Cy sites. Cobalt, however, was found in three different forms. In addition to Pt-Co alloy and Co-Nx-Cy, it also existed as a metal crystallite encapsulated by onion-like graphitic layers [Co@graphene (fig. S1)], which is also often considered catalytically active (13). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images show that Pt-Co NPs are surrounded by an amorphous “particle-free” region, in which individual Co atoms and a trace amount of Pt atoms can be distinguished (Fig. 1, B and C). The energy-dispersive x-ray spectroscopy (EDS) and the electron energy-loss spectroscopy (EELS) analyses identified primarily C, N, and Co2+ in these regions (Fig. 1D and table S1). These compositions represent a typical makeup of PGM-free catalysts (15) with good ORR activity (23). High-resolution transmission electron microscopy (HRTEM) revealed that the Pt-Co NPs had a Pt-Co core and a Pt shell (Fig. 1E and fig. S2 and S3). Apparent ordering of Co and Pt in Pt-Co core with face-centered cubic crystal structures was also observed along the <100> and <110> directions, further supporting the existing of superstructures known to be highly active in catalysis (24).

Fig. 1 LP@PF catalyst structure.

(A) Schematics of LP@PF catalysts, showing coexistence of Pt-Co NPs, Co@graphene, and Co-Nx-Cy PGM-free active sites. (B) A HAADF-STEM image of Pt-Co NPs in LP@PF-1 situated over (C) PGM-free support containing atomically dispersed Co (circled in red) and trace Pt (circled in blue). (D) EELS analysis of the elemental composition of (C). a.u., arbitrary units. (E) HRTEM image of a representative Pt-Co alloy NP with Pt3Co superlattice core and Pt skin partially covered by CoN and CoC terraces.

Lattice contraction led to surface segregation and a highly strained Pt skin of three to four monolayers (fig. S2B), which enhances the ORR activity (25). In many cases, the Pt shell was partially covered by multilayered terraces composed of Co, N, and C and identified as CoN or CoC from their interlayer spacing (Fig. 1E and figs. S2 and S4). The terraces could slow down the dissolution of Pt-Co NPs while keeping the active surface exposed. The Pt:Co ratios of the overall catalysts and the Pt-Co alloy NPs were analyzed by EDS (fig. S5A). The Pt:Co ratios of NP were consistent with alloy compositions of 1:1 in LP@PF-1 and 3:1 in LP@PF-2, respectively, which were further confirmed by x-ray diffraction (XRD) (fig. S6 and table S2). This ratio was substantially lower in bulk catalyst after averaging the contributions from Co-Nx-Cy and Co@graphene sites. The NP sizes were narrowly distributed around average diameters of 5.6 ± 1.6 nm and 5.7± 1.7 nm (fig. S7), and the overall Pt loadings were 2.72 weight % (wt %) and 2.81 wt % for LP@PF-1 and LP@PF-2, respectively. The Brunauer-Emmett-Teller specific surface areas were 343 m2/g for LP@PF-1 and 807 m2/g for LP@PF-2 (fig. S8).

We also investigated the electronic structures of the Pt-Co alloys and the PGM-free catalyst support using x-ray photoelectron spectroscopy (XPS), x-ray absorption near-edge structure (XANES) spectroscopy, and extended x-ray absorption fine structure (EXAFS) spectroscopy. Electron transfer with Co causes a shift in the Pt d-band center energy in Pt-Co alloys, which weakens OHad binding on the Pt surface and thus improved ORR catalytic properties (26). As expected, the Pt XPS shows a ~0.2 eV positive energy shift in LP@PF-1 upon annealing in NH3 (Fig. 2A). Co XPS also showed redistribution to more ionic Co2+ from Co0 (Fig. 2B), with the Co+2:Co peak ratio changing from 1.8 to 2.9 after NH3 treatment, forming additional Co-Nx. The N 1s spectra demonstrated a high content of pyridinic and pyrrolic N embedded in the graphitic matrix with little change after the NH3 treatment (Fig. 2C).

Fig. 2 LP@PF electronic structures.

XPS spectra of LP@PF-1 taken at (A) Pt 4f, (B) Co 2p3/2, and (C) N 1s transitions before (BN) and after NH3 treatment, as well as after a 30,000 voltage cycle AST in a fuel cell. XANES spectra taken at (D) Pt L3-edge and (E) Co K-edge before and after NH3 treatment with metal foils as references.

XANES analysis showed reduction of white line intensity (gray arrow) at the Pt L3 edge, which corroborates electron transfer from Co 3d to Pt 5d orbitals in Pt-Co alloy (27) (Fig. 2D). Alloy formation was further confirmed by a characteristic Pt-Co interaction peak (red arrow) at 11,576 eV (28). Similar changes in XPS and XANES were also observed in LP@PF-2 (fig. S9). The transformation from Pt to Pt-Co alloy was further corroborated by EXAFS (fig. S10A) and XRD (fig. S6). XANES at the Co K-edge was more convoluted because it included the contributions from Pt-Co alloy, metallic Co@graphene clusters, and Co2+ ion embedded in N-decorated C support. After the NH3 treatment, the intensity of the pure Co metal peak at 7013 eV (green arrow) reduced substantially, whereas the peak at 7227 eV (red arrow) grew substantially (Fig. 2E), reflecting hybridized Co 4s and 4p orbitals by Pt in the alloy (28) and conversion of some Co(0) to Co(II)-Nx as corroborated by XPS. EXAFS analysis revealed the loss of Co-Co peak intensity due to a decrease of metallic Co and an increase of alloy formation (fig. S10B). More importantly, it showed an enhancement of peak intensity at Co–N bond distance, indicating the increase of the N-ligated Co2+ population. The atomically dispersed TM ligated by four N atoms in a C matrix has been associated with the active sites for ORR in PGM-free catalysts (15, 20, 29).

We first measured the electrocatalytic ORR activities of LP@PF-1 and LP@PF-2 by the rotating ring-disk electrode (RRDE) at room temperature in an O2-saturated 0.1 M HClO4 solution. For comparison, the PGM-free catalytic substrate PF-2, a commercial Pt/C catalyst (TKK, 46.7 wt % Pt), and an in-house prepared 3 wt % Pt3Co/ZC catalyst were also tested. Pt3Co/ZC was prepared by adding Pt3Co alloy NPs over ZIF-8–derived carbon (ZC). This catalyst is similar in composition and surface property to LP@PF-2, except it lacks Co-Nx-Cy sites. Figure 3A displays the linear sweep voltammetry (LSV) from the kinetic to the diffusion-limiting regions. The halfwave potential E½, a gauge of electrocatalytic activity, increased in the order of PF-2 < Pt3Co/ZC ≤ commercial Pt/C < LP@PF-1 < LP@PF-2, with LP@PF-2 at 0.96 V (table S3). Meanwhile, the electron-transfer number n, calculated from the ring-to-disk current ratios, was 3.99 for both LP@PF-1 and LP@PF-2, suggesting a nearly completed conversion from O2 to H2O instead of H2O2. The Pt MA Tafel plot derived from LSV demonstrated substantially higher values for LP@PF-1 and LP@PF-2 than those of the reference catalysts (Fig. 3B), whereas the specific current density Tafel plot exhibited higher onset potentials and lower slopes (fig. S11). LP@PF-1 and LP@PF-2 delivered high Pt MAs of 8.64 A mgPt−1 and 12.36 A mgPt−1 at 0.9 V versus reversible hydrogen electrode (RHE), respectively, and outperformed the commercial catalyst (Fig. 3C) and some recently reported nanostructured 3D Pt alloy catalysts (table S4) (6, 7).

Fig. 3 Electrocatalytic activity and durability evaluations.

(A) (Bottom) LSVs of different catalysts recorded at a rate of 10 mV s−1 and 1600 rotations per minute in O2-saturated 0.1 M HClO4. j, current density. (Top) Number of transferred electrons (e), n, at different potentials (E). (B) MA Tafel plots derived from (A). (C) Comparison of MAs at 0.9 V versus RHE from (B). (D) H2-O2 fuel cell i-V polarization (solid symbols and lines) and power density (hollow symbols and dashed lines) plots recorded under 1 bar of O2 pressure with cathode Pt loading of 0.033 mgPt cm−2 for LP@PF-1, 0.035 mgPt cm−2 for LP@PF-2, 0.043 mgPt cm−2 for Pt3Co/ZC, and 0.35 mgPt cm−2 for commercial MEA. LP@PF-1, black stars; LP@PF-2, red diamonds; PF-2, green triangles; 3% Pt3Co/ZC, blue spheres; commercial 47% Pt/C from TKK or MEA from BASF, magenta squares. (E) Cathodic MA Tafel plots derived from fuel cell measurement. The green star denotes the U.S. DOE 2025 target. (F) Fuel cell (FC) MAs at 0.9 ViR-free before (solid) and after (hatched) 30,000 voltage cycles, showing that LP@PF catalysts meet or exceed DOE’s 2025 MA targets for before (green dashed line, 0.44 A mgPt−1) and after (red dashed line, 0.264 A mgPt−1 or 40% of the initial value) AST. (G) H2-O2 fuel cell i-V polarizations and power densities after 30 K voltage cycles. (H) H2-air fuel cell performances for the same MEAs containing LP@PF-1 and LP@PF-2 under 1 or 2 bars of pressure. (I) Specific current densities of PF-2, Pt3Co/ZC, LP@PF-2, and the sum of PF-2 and Pt3Co/ZC as a function of iR-free fuel cell voltage measured under 1 bar of H2-O2 pressure. For all fuel cell tests, membrane = Nafion 211, temperature = 80°C, and anode loading = 0.35 mgPt cm−2. For H2-O2 cell PH2 = PO2 = 100 kPa at 100% relative humidity (RH) (back pressure = 50 kPa, absolute pressure = 150 kPa), flow rate = 200 ml min−1. For H2-air cell PH2 = Pair = 100 kPa or 200 kPa at 100% RH, H2 flow rate = 200 ml min−1 and airflow rate = 520 ml min−1 (equivalent of stoichiometries of 1.5/1.8 at 3.5 A cm−2 of the end of polarization).

We further incorporated the LP@PF catalysts in the cathode of the membrane electrode assembly (MEA) and tested their performances in a PEMFC single cell with O2 or air as the cathodic gas feed. The cathodic Pt loading were 0.033 mgPt cm−2 for LP@PF-1 and 0.035 mgPt cm−2 for LP@PF-2, respectively. Figure 3D shows their current-voltage (i-V) polarizations and power density distributions measured under 1 bar of fully humidified O2. For benchmarking, we also tested a MEA with Pt3Co/ZC catalyst with cathodic loading of 0.043 mgPt cm−2, a MEA with PF-2 cathode catalyst, and commercial MEAs with much higher cathodic Pt loadings (Fig. 3D and fig. S12).

Both MEAs with LP@PF-1 and LP@PF-2 displayed higher catalytic activities than the comparative MEAs in the high-voltage region (>0.7 V) in an H2-O2 cell. The MEA with LP@PF-2 cathode catalyst demonstrated higher current density than the commercial MEA through the entire polarization scan, even at 1/10th of the cathodic Pt loading. Its current density continued to increase nearly linearly with polarization voltage, a feature commonly observed in PGM-free fuel cells and characteristically different from conventional MEAs with PGM-only catalysts. Figure 3E shows the Pt MA Tafel plots derived from the internal resistance corrected i-V polarizations (fig. S13) and Pt loading. Again, the LP@PF-1 and LP@PF-2 MEAs showed higher MAs than those of comparative MEAs. The fuel cell–based Pt MAs measured at 0.9 ViR-free are 1.08 A mgPt−1 for LP@PF-1 and 1.77 A mgPt−1 for LP@PF-2, respectively, representing an order of magnitude improvement compared with the commercial MEAs (Fig. 3F and table S5). These values exceed the 2025 target set by DOE (0.44 A mgPt−1 at 0.9 ViR-free for MEA) by factors of approximately two and four and represent record-high ORR activities measured in a PEMFC (22).

The MEAs were also subjected to AST under repeated cell voltage sweeps from 0.6 to 1.0 V according to DOE catalyst stability evaluation protocols (22). Fuel cell polarizations and MAs were measured periodically after designated voltage cycles up to 30,000 (fig. S14). Figure 3G shows fuel cell i-V polarizations and power-density distributions after 30,000 voltage cycles. Although AST caused a substantial activity loss for the commercial MEA, the MEAs with LP@PF-1 and LP@PF-2 cathode catalysts showed improved durability (fig. S15). Especially, the MEA with LP@PF-1 demonstrated the highest durability with its MA retained at 0.672 A mgPt−1 at 0.9 ViR-free (Fig. 3F), or 64% of its initial value. This value surpassed the catalyst durability goal of <40% MA loss after AST set by DOE (22). The MA stability of LP@PF-1 was compared to a state-of-the-art dealloyed PtNi catalyst (30) and showed higher values at both beginning and end of life, although the PtNi MEA demonstrated higher retention of MA at the end of AST. The drop of the fuel cell voltage at current density of 0.8 A cm−2 after 30,000 cycles was 6 mV, well within the DOE target of <30 mV loss. For MEA with LP@PF-2, the MA was reduced to 0.263 A mgPt−1 at 0.9 ViR-free, which is still comparable to the DOE target of 0.264 A mgPt−1 after AST based on 40% loss of the initial activity of 0.44 A mgPt−1. The voltage drop at current density of 0.8 A cm−2 was 47 mV. In addition to voltage cycling, we also tested the MEA durability at constant voltage and found excellent performances with lower decay rates for LP@PF catalysts under both O2 and air compared with the benchmarks (figs. S16 to S18).

Excellent MEA performances by LP@PF-1 and LP@PF-2 were also observed when the fuel cells were tested in H2-air under different stoichiometries (flow rates) and pressures (Fig. 3H and figs. S19 and S20). Both MEAs outperformed commercial MEAs at V > 0.6 V, reaching a current density of 300 ± 10 mA cm−2 at 0.8 V, meeting the DOE target. LP@PF-1 showed slightly better fuel cell performances compared with LP@PF-2 at higher cell voltage but lower current density at low cell potential. We attribute this mainly to the difference in PGM-free substrate structure. PF-1 has a lower surface area but higher PGM-free active site area density and level of graphitization. In addition to high stability, such structure promotes robust synergistic catalysis. PF-2 has higher porosity and surface area, which facilitates the interaction with airflow and, therefore, higher current density near the mass-transport–limited region (9, 23). For comparisons, the MAs and H2-air fuel cell performances of LP@PF MEAs along with representative published reports are provided in table S6.

The morphology, composition, and electronic state of the LP@PF catalysts after AST were analyzed. For example, TEM analysis showed only minor changes in NP size distribution, with the average particle dimension remaining the same within one standard deviation, from 5.6 ± 1.6 nm to 5.7 ± 1.6 nm for LP@PF-1 and from 5.7 ± 1.7 nm to 6.0 ± 1.5 nm for LP@PF-2, respectively (figs. S21 and S22). The HRTEM images also confirmed the retention of Pt-Co core-shell structure beneath the Co-N-C terraces, which likely played an important role in preserving Pt-Co NPs during AST. EDS analysis averaged from multiple samplings showed that the Pt:Co ratios within single NPs after AST were also nearly unchanged for PtCo (44:56) in LP@PF-1 and Pt3Co (74:26) in LP@PF-2, respectively. The Pt:Co ratios in the bulk catalysts, however, increased from 7:93 to 19:81 for LP@PF-1 and 11:89 to 14:86 for LP@PF-2, respectively, presumably owing to the dissolution of a small amount of unalloyed cobalt (fig. S5B). The preservation of Pt-Co alloy structures in both catalysts was further confirmed by XRD of the cathode layers peeled from MEAs after AST (fig. S23). The peeled cathode layers were also investigated by XPS, which revealed an overall Pt:Co ratio of 21:79 in LP@PF-1 and 20:80 in LP@PF-2, in agreement with the EDS measurements (Fig. 2 and fig. S9). Compared to the fresh catalyst, the Pt electronic state in the aged catalyst remained nearly unchanged in the form of alloy. The Co+2:Co ratio also remained unchanged in LP@PF-1 at 2.9 after AST. The most substantial change came from the carbonaceous nitrogens. The pyridinic-N to pyrrolic-N ratio was reduced from 2.7 to 2.2, possibly due to partial conversion of pyridinic- to pyridonic-N shown by the new peak in Fig. 2C. The pyridonic-N was formed by attachment of OH to the carbon atom next to pyridinic-N, which was previously observed in a PGM-free catalyst after ORR (19).

To quantify the synergistic interaction between Pt-Co NPs and PGM-free active sites, we compared the specific current density of a fuel cell containing LP@PF-2 to that from Pt3Co/ZC and PF-2. Figure 3I shows that the specific current density of LP@PF-2 at any given voltage was about twice of the sum of the contributions from Pt3Co/ZC and PF-2. This indicates that the synergistic ORR rate in LP@PF is substantially higher than the simple sum of that from Pt3Co NPs and PGM-free sites. The synergistic catalysis also exhibited improved catalyst stability of LP@PF versus the commercial Pt/C, Pt3Co/ZC, and PGM-free catalysts (Fig. 3G and fig. S14) (23). Such effects were only observed when the Pt-Co alloy NPs were annealed by NH3 over the PGM-free substrate. Because CoN and CoC adlayers were formed during the in situ reduction in NH3, we speculate that they not only protect Pt-Co NPs but also serve as “bridges” in transferring the reaction intermediate H2O2 from PGM-free site to the Pt-Co NPs through a reverse spillover during synergistic catalysis.

To better understand the improved durability of the LP@PF catalyst, we performed density function theory (DFT) calculations for the interface between a Pt3Co NP [represented by strained Pt (111)] and Co-N4 decorated graphene. The calculations determined that the strong interaction of the Pt surface with Co-N4-C sites enhances binding that helps to impede the segregation of the Pt-Co NPs from the support (fig. S24). The simulation also revealed that two or three CoN and CoC adlayers grow preferentially on Pt (100) instead of (111) facets, with formation energies that are more stable than that of the bulk CoN (fig. S25 and table S8). The presence of these adlayers optimizes the exposure of the catalytically more active (111) facet yet reduces Pt dissolution through less stable (100) facets. This result may explain why most alloy particles remain intact after AST. Strong binding of Pt-Co NPs with PGM-free site–mediated surface also generates intimate contact between the two with better charge and reaction intermediate transfers, which are further facilitated through improved hydrophilicity by the adlayer over the Pt surface.

DFT calculations were also carried out to understand the enhanced activity of LP@PF catalysts. We calculated the thermodynamic barriers along two parallel ORR reaction pathways, one over Pt (111) (with 4% strain, near the calculated value for PtCo with a multilayer Pt skin) and another over a Co-N4 active site, through multistep sequential combination of protons and electrons (Fig. 4 and fig. S26). Pt (111) has lower but non-negligible barriers for the reaction steps, including the stabilization of OOH# in step II# and the formation of OH# in step VI#, respectively. Over a Co-N4 site, the kinetic barrier for OOH* formation (step II*) is only <0.1 eV higher than that over the Pt site and is highly facile. The reactions after step II* branch into two concurrent paths, formation of O* and water (step III*) and production of H2O2 (step IV*), with the reaction barrier of the former being less than 0.2 eV. Because H2O2 does not bind to the PGM-free site, it can be released after step IV* and migrate to strained Pt (111) sites in the vicinity, as denoted by the green arrow in Fig. 4. The two pathways over Pt and PGM-free catalytic sites intersect, and the subsequent decomposition of H2O2 over the strained Pt (111) surface is rapid, as it has no thermodynamic barrier. More details on DFT calculations and mechanistic discussion are provided in the supplementary materials.

Fig. 4 Free-energy diagram of the ORR pathways.

The proposed associative reaction coordinates represent the following states: (I) * or # + O2 + 4H+ + 4e, (II) OOH* or OOH# + 3H+ + 3e, (III) O* or O# + H2O + 2H+ + 2e, (IV) * + H2O2 + 2H+ + 2e, (V) 2OH# + 2H+ + 2e, (VI) OH* or OH# + H2O + H+ + e, and (VII) * or # + 2H2O, where * (blue) denotes the binding site on Co-N4 embedded in graphene and # (gray) denotes the binding site on a strained Pt (111) facet. (Inset) Schematics of H2O2 generated over Co-N4 migrating to the strained Pt (111) surface (green arrows), followed by dissociation to OH# and water formation. (Computation was performed at 0.9 V relative to the hydrogen electrode at pH = 1.)

This analysis provides an explanation for our experimentally observed synergistic catalysis over LP@PF with improved activity and durability at both high-voltage (kinetics-limited) and high-current (mass transport–limited) regions of fuel cell polarization. The Pt-Co alloy increases its utilization efficiency by not only performing direct ORR but also facilitating reduction of H2O2 generated from nearby PGM-free sites, leading to the nearly four-electron transfer measured by RRDE and improved catalyst activity observed in the fuel cell test. Because H2O2 is known to corrode TM-based PGM-free sites and porous carbon substrate, its breakdown also helps to preserve the catalyst integrity and durability. Our LP@PF catalysts exhibit improved catalytic activity and durability with lower Pt usage in fuel cells. Remaining challenges include further reducing Pt loading while maintaining synergistic interaction at different fuel cell voltages and catalytic turnover frequencies, better humidity management to ensure effective proton and peroxide transfers over the catalyst surface, as well as improved operation with air.

Supplementary Materials

www.sciencemag.org/content/362/6420/1276/suppl/DC1

Materials and Methods

Figs. S1 to S26

Tables S1 to S8

References (3150)

References and Notes

Acknowledgments: We thank M. S. Ferrandon and D. J. Myers for experimental assistance. Funding: This work was supported by U.S. DOE Fuel Cell Technologies Office through Office of Energy Efficiency and Renewable Energy. The works performed at Argonne National Laboratory’s Center for Nanoscale Materials and Advanced Photo Source, U.S. DOE Office of Science User Facilities, are supported by Office of Science, U.S. DOE under contract DE-AC02-06CH11357. L.C. wishes to thank Argonne National Laboratory for the Maria Goeppert Mayer Fellowship. J.K. and J.G. acknowledge funding through the U.S. DOE, Office of Basic Energy Sciences, Chemical, Biological, and Geosciences Division under DE-SC0010379. J.K. wishes to thank U.S. DOE, Office of Science, Office of Workforce Development for Teachers and Scientists, Office of Science Graduate Student Research (SCGSR) program. The SCGSR program is administered by the Oak Ridge Institute for Science and Education for the DOE under contract number DE‐SC0014664. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a U.S. DOE Office of Science User Facility supported under contract no. DE-AC02-05CH11231. J.Z. and W.D. wish to thank the National Key Research and Development Program of China (2016YFB0701200) and Chinese National Nature Science Foundation (51771112) for support. The views and opinions of the authors expressed herein do not necessarily state or reflect those of the U.S. government or any agency thereof. Neither the U.S. government nor any agency thereof, nor any of their employees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Authors contributions: D.-J.L. and L.C. conceived of the idea and designed the experiments. L.C., J.Z., H.B., W.D., and D.-J.L. participated in the experiments and data analysis. J.W. acquired high-resolution images and performed analysis. J.K., F.G.S., M.C., and J.G. performed computational modeling. L.C., J.K., J.G., M.C., and D.-J.L. prepared the manuscript. Competing interests: The authors declare no competing financial interests. Argonne National Laboratory received U.S. Patent 9,825,308 related to this work. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.

Stay Connected to Science

Navigate This Article