Tuning the activity of Pt alloy electrocatalysts by means of the lanthanide contraction

See allHide authors and affiliations

Science  01 Apr 2016:
Vol. 352, Issue 6281, pp. 73-76
DOI: 10.1126/science.aad8892

A lanthanide boost for platinum

High loadings of precious platinum are needed for automotive fuel cells, because the kinetics of the oxygen reduction reaction (ORR) are relatively slow. Escudero-Escribano et al. studied a series of platinum alloys with lanthanides and alkaline earth elements. When the surfaces were leached to leave pure platinum, they developed compressive strain that boosted the ORR activity—up to a factor of 6 for terbium. Enthalpy effects helped to stabilize these alloys under operating conditions.

Science, this issue p. 73


The high platinum loadings required to compensate for the slow kinetics of the oxygen reduction reaction (ORR) impede the widespread uptake of low-temperature fuel cells in automotive vehicles. We have studied the ORR on eight platinum (Pt)–lanthanide and Pt-alkaline earth electrodes, Pt5M, where M is lanthanum, cerium, samarium, gadolinium, terbium, dysprosium, thulium, or calcium. The materials are among the most active polycrystalline Pt-based catalysts reported, presenting activity enhancement by a factor of 3 to 6 over Pt. The active phase consists of a Pt overlayer formed by acid leaching. The ORR activity versus the bulk lattice parameter follows a high peaked “volcano” relation. We demonstrate how the lanthanide contraction can be used to control strain effects and tune the activity, stability, and reactivity of these materials.

To reduce the Pt loading at the cathode of polymer electrolyte membrane fuel cells (PEMFCs), researchers have intensively studied alloys of Pt with late transition metals such as Ni or Co as oxygen reduction reaction (ORR) electrocatalysts (16). Catalysts exhibiting even greater activity and stability could be designed through the identification of the descriptors that control the performance (710). One single descriptor controls ORR activity, the ΔEOH binding energy, by way of a Sabatier volcano: An ΔEOH ~0.1 eV weaker than Pt(111) yields the optimum value (11). Other indirect descriptors related to ΔEOH include the d-band center (12), the Pt-Pt interatomic distance (4) and the generalized coordination number (8). Stability is a multiparametric challenge, hence requiring several descriptors, such as the alloying energy Ea (11, 13), and dissolution potential (14, 15).

Our earlier studies identified alloys of Pt and rare earths, in particular PtxY and PtxGd, as active and stable catalysts for oxygen reduction, both in the bulk polycrystalline (11, 16) and nanoparticle (NP) form (17, 18). The exceptionally negative Ea of Pt–rare earth alloys should increase their resistance to degradation (11, 13). In contrast, more commonly studied ORR alloys, such as Pt-Ni or Pt-Co, typically degrade in long-term tests via dealloying (19, 20). Nevertheless, new forms of Pt-Ni–based catalysts achieve exceptional activity and stability during short-term accelerated degradation tests (5, 21, 22). Nonetheless, engendering long-term stability in fuel cells (20) may require materials that are inherently less prone to dealloying. Because the rare earth (e.g., Y or Gd) is unstable against dissolution, a Pt overlayer is formed on the surface, as shown on Fig. 1, A and B. We showed that on PtxGd and PtxY NPs, the bulk compressive strain correlated strongly with increased ORR activity; this result suggested that the bulk strain is imposed onto the Pt surface atoms, weakening ΔEOH (23). These observations led us to conjecture that other Pt-lanthanide alloys, exhibiting more optimal levels of compression, would reach the peak of the Sabatier volcano. Here, we show how the decreased radius of the lanthanides with increased filling of the f-shell—i.e., the lanthanide contraction—provides us with a route to engineer such compression. We have systematically studied activity and stability trends of Pt5La, Pt5Ce, Pt5Sm, Pt5Gd, Pt5Tb, Pt5Dy, Pt5Tm, and Pt5Ca, using a combination of experiments and theory to explain our observations.

Fig. 1 Schematic views and electrochemical properties of polycrystalline Pt5M (M = lanthanide or alkaline earth metal) electrocatalysts.

Three-dimensional view of the Pt5M structure (A) during sputter-cleaning and (B) after electrochemistry. (C) Kinetic current density, jk, of Pt5M and Pt at 23°C, 1600 revolutions per minute in O2-saturated 0.1 M HClO4, before and after a stability test consisting of 10,000 cycles between 0.6 and 1.0 V versus RHE at 100 mV s−1. The activity of Pt5Ca after the stability test has been normalized considering the increase of area after the test (see section S3.4 in the supplementary materials). The value normalized by the geometric area (dotted line) is shown for comparison.

We evaluated the electrocatalytic properties of sputter-cleaned polycrystalline Pt5M electrodes by rotating disk electrode (RDE) voltammetry in O2-saturated 0.1 M HClO4. We chose a Pt:M ratio of 5:1 because we could obtain a consistent series of alloys with the same structure, allowing for a systematic investigation. Furthermore, it corresponds to the phase that is most Pt-rich and stable (16). At 0.9 V, Pt5Tb is the most active polycrystalline Pt-based catalyst reported. All of the materials exhibited activity enhancement by a factor of 3 to 6 over pure Pt (see figs. S5 and S6 and table S1 in the supplementary materials). The overall electrocatalytic ranking of ORR activity of polycrystalline Pt alloys is shown in fig. S6: Pt5Tb > Pt5Gd ~ Pt3Y > Pt5Sm > Pt5Ca ~ Pt5Dy > Pt5Tm > Pt5Ce > Pt5Y ~ Pt5La >> Pt (6, 11, 16), demonstrating that these alloys accelerate the ORR more effectively than other polycrystalline Pt alloys. Pt3Co and Pt3Ni alloys prepared this way exhibited enhancement only by a factor of 2 (12, 24). Accelerated stability tests consisting of 10,000 consecutive cycles between 0.6 and 1.0 V versus a reversible hydrogen electrode (RHE) were performed after the initial ORR activity measurements. The electrochemical experiments are summarized in figs. S3 to S11. Figure 1C reports the ORR activities before and after the stability test for all the Pt alloys and pure Pt. Apart from Pt5Ca (which has a lower Ea), all of the Pt-lanthanide alloys retained enhancement by a factor of 3 over pure Pt after the accelerated stability test. Notably, Pt5Gd exhibited a residual activity that was 5 times as great as that of pure Pt.

We characterized the structure and chemical composition of the electrocatalysts by x-ray diffraction (XRD) and x-ray photoelectron spectroscopy (XPS) in order to explain our experimental observations. All of the alloys formed stable intermetallic compounds with a hexagonal structure (figs. S1 and S2 and table S1), in agreement with previous reports (25). The XRD data suggest that the polycrystalline alloys may show different degrees of preferential orientation in the bulk (figs. S1 and S2). However, by presputtering the electrodes, we minimize any differences in surface orientation between the samples under investigation. Moreover, based on Watanabe and co-workers’ in situ scanning tunneling microscopy measurements on sputter-deposited Pt-Fe (26), we expect the acid-leached Pt overlayers to be dominated by (111) terraces, typically the most stable facet termination (27). Most of the elements in the bulk Pt5M alloy form a so-called kagome layer (6, 16) (Fig. 2 and fig. S16), with a nearest-neighbor Pt-Pt distance dPt-Pt = a/2. The lattice parameter a and hence dPt-Pt decreased from left to right in the lanthanide series (Fig. 2).

Fig. 2 Structure of Pt5M.

(A and B) Schematic view of the bulk structure of a Pt5M (illustrated for Pt5Tb), showing Pt5Tb terminated by (A) a Pt and Tb intermixed layer and (B) a Pt kagome layer. Purple spheres represent Tb atoms, and gray spheres represent Pt atoms. Under ORR conditions in an acidic environment, one to two layers of M will be leached out, leaving three to five layers of Pt, as shown on Fig. 1B. (C) Relation between the lattice parameter a of Pt5M measured by XRD (table S1) and the covalent radius of the lanthanide atoms (31). The dotted line shows the linear fit. The upper part of the figure shows the lanthanide contraction across the lanthanide series, the covalent radii decreasing in the same direction as dPt-Pt. The error in a corresponds to the uncertainty in the fit (table S1), whereas the error in the covalent radius corresponds to the estimated standard deviation from (31).

Figure 3, A and B, show the angle-resolved XPS (AR-XPS) depth profiles on sputter-cleaned Pt5Tb before and after electrochemical measurements. For each of the alloys, the Pt to M ratio increased substantially, especially at the most surface-sensitive angles, as shown for Pt5Tb on Fig. 3C, after initial electrochemistry measurements, confirming the formation of a Pt overlayer. The depth profile of Fig. 3B showed that, even after accelerated stability tests, this structure was maintained, demonstrating the stability of these materials upon potential cycling.

Fig. 3 XPS profiles before and after electrochemistry and Pt overlayer thickness as a function of the lattice parameter and activity loss.

(A and B) AR-XPS profiles of polycrystalline Pt5Tb (A) as prepared and (B) after initial ORR activity (solid line) and after stability test (dashed line). (C) Pt to Tb atomic ratios in Pt5Tb from AR-XPS during sputter cleaning, after ORR initial activity, and after stability test. (D) Estimated average thicknesses of the Pt overlayer for Pt5Tm, Pt5Dy, Pt5Tb, Pt5Gd, Pt5Ce, and Pt5La after initial ORR activity and after stability test (taken from Fig. 1C), as a function of a lattice parameter [Pt5Dy is shown as a hollow symbol to demarcate it as an outlier, likely because its as-prepared composition was inconsistent with that of the bulk (fig. S13)]. (E) Percentage of activity loss after stability test as a function of the Pt overlayer thickness. (F) Slab stability represented as dissolution potential versus the strain of the Pt overlayer on Pt5M (from experimental lattice parameter a).

To quantitatively interpret our XPS data (for details, see section S4 of the supplementary materials and figs. S12 to S15), we evaluated the mean Pt overlayer thickness for the alloys. Figure 3D shows how it varied for both the initial ORR activity and after stability tests as a function of the bulk lattice parameter a. The overlayer thickness after the initial testing varied little between the different alloys. However, after stability tests, the mean Pt overlayer generally increased from Pt5La to Pt5Tb [Pt5Ca lies out of the scale; the initial Pt:M ratios for Pt5Ca, Pt5Sm, and Pt5Dy were in fact higher than those expected from the nominal bulk stoichiometry (fig. S13), thus inhibiting a precise calculation of the overlayer thickness]. This difference could explain the anomalous behavior of these alloys, relative to the overall trend. The activity loss also correlates with the thickness of the overlayer (Fig. 3E) (3). Our density functional theory (DFT) calculations on the stability of different Pt overlayers, expressed as dissolution potential, show that the stability decreases as the compressive strain increases (Fig. 3F)—i.e., strain is a stability descriptor. We attribute the apparent thickening of the Pt overlayer with cycling to surface diffusion processes (28); bulk diffusion of lanthanide atoms through the overlayer will be strongly impeded by the strength of Ea (13). The strain-induced destabilization of the Pt overlayer could facilitate surface mobility (28), providing a channel for the dissolution of any residual lanthanide atoms in close vicinity to the surface. In summary, Fig. 3, D to F, shows that the overlayer thickness, activity losses, and thermodynamic stability are all a function of the bulk lattice parameter: Increased strain destabilizes the Pt overlayer and thus accelerates surface diffusion.

Figure 4A is a plot of the ORR activity as a function of the lattice parameter, a and dPt-Pt. Notably, all nine compounds, including the Pt-lanthanides and Pt5Ca, follow the same volcano-type trend, with Pt5Gd and Pt5Tb at the apex. Because ΔEOH is likely correlated with dPt-Pt (3), the most trivial explanation for this trend is that the plot represents a Sabatier volcano: Alloys on the left bind OH too weakly, whereas on the right hand they bind ΔEOH too strongly (as described by the DFT calculations in figs. S17 and S18). Alternatively, beyond a certain level of bulk strain, the overlayer could be unstable, causing the dPt-Pt of the overlayer to relax toward a much lower level of surface strain. On single crystals, the destabilization is manifested as a positive shift in the “reversible” voltammetric peak for OH adsorption (1, 10); however, we do not observe this shift on our polycrystalline materials, presumably because of hysteresis (electrochemical “irreversibility”) or possibly coadsorption of OH and O. Conversely, the lanthanide contraction results in a clear voltammetric shift for the H adsorption region (figs. S3 and S4), plotted on Fig. 4B, which resembles the activity volcano, with Pt5Tb exhibiting the maximum destabilization of adsorbed H. Notably, we also observe a linear relation between the experimental activity and the potential shift in the H adsorption (fig. S7).

Fig. 4 Experimental volcano-type relationships between activity, H adsorption, and Pt-Pt distance.

(A) Kinetic current density at 0.9 V (taken from Fig. 1C) on polycrystalline Pt5M electrocatalysts versus the lattice parameter a of bulk Pt5M (lower axis) and bulk dPt-Pt (upper axis), respectively. The figure shows the kinetic current density, jk, of the alloys after the initial ORR activity (dark gray squares) and after 10,000 cycles of the stability test (colored circles). The dotted and dashed lines represent the experimental trends resulting after initial ORR activity and after stability, respectively. The activity of Pt5Ca after 10,000 cycles has been normalized to account for the increase of area after the stability test. (B) Relation between the potential necessary to adsorb 1/8, 1/6, 1/4, and 1/3 monolayers (ML) of H (UH) from the cyclic voltammograms (CVs) in the H adsorption region in N2-saturated 0.1 M HClO4 on Pt5M (Fig. S5) and dPt-Pt.

Our DFT calculations on strain-activity-reactivity relations (section S5.4) suggest that Pt5Tb, which is the most active electrocatalyst, should exhibit ~3% compression, approaching the optimum OH binding energy of the Sabatier volcano (11). By comparing our activity data and the voltammetric shift in H adsorption to the DFT predictions, we can conjecture that Pt-lanthanide alloys with a shorter dPt-Pt than Pt5Tb form a more relaxed overlayer (figs. S19 to S21]. More generally, our observations suggest that strain effects can only weaken the binding of H and OH to a certain extent. More appreciable destabilization of reaction intermediates can be afforded by ligand effects (1, 10). The implementation of these catalysts in fuel cells will require scalable synthesis methods yielding high surface catalysts. Nonetheless, we have already demonstrated that PtxGd NPs exhibited an outstanding activity of 3.6 Å/mg Pt at 0.9 V RHE in liquid half cells (18, 29) (fig. S6B), only surpassed by Pt3Ni nanoframes (21) and Mo-doped Pt3Ni nanoparticles (22). Careful tuning of the NP composition—for instance, by synthesizing ternary Pt-Gd-Tb alloys, in combination with a judicious choice of annealing treatment (21, 22, 30)—could yield record-breaking catalytic activity and stability over the long term in real devices.

Supplementary Materials

Materials and Methods

Figs. S1 to S21

Table S1

References (3265)

References and Notes

  1. Acknowledgments: The Center for Individual Nanoparticle Functionality is sponsored by the Danish National Research Foundation (DNRF54). We gratefully acknowledge EU FP7’s initiative Fuel Cell and Hydrogen Joint Undertaking’s project CathCat (GA 303492), as well as Danish Strategic Research's project NACORR (12-133817), for funding this work. M.E.-E. is the recipient of a Sapere Aude: DFF-Research Talent grant from the Danish Council for Independent Research. I.E.L.S. is the recipient of the Peabody Visiting Associate Professorship from the Department of Mechanical Engineering at Massachusetts Institute of Technology. We thank C. D. Damsgaard for assistance setting up the XRD measurements and O. Hansen for critically reading the manuscript. The authors declare competing financial interests: Intellectual property pertaining to the materials presented in this Report is protected by three patents (CA2877617-A1, WO2014079462-A1, and CA2767793-A1).
View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article