X-ray Absorption Spectroscopy Investigation of Platinum–Gadolinium Thin Films with Different Stoichiometry for the Oxygen Reduction Reaction

Abstract

Alloys of platinum and lanthanides present a remarkable activity for the oxygen reduction reaction—both in the form of extended surfaces and nanoparticulate catalysts. Co-sputter-deposited thin film catalysts based on platinum and gadolinium show great oxygen reduction activity improvement over pure Pt. The sputter-deposition technique represents a viable and versatile approach for investigating model catalyst systems with different compositions. In this work, co-sputtered Pt₅Gd and Pt_7.5Gd thin films were investigated using X-ray absorption spectroscopy as well as standardized electrochemical techniques. These investigations revealed the importance of forming alloys with specific stoichiometry, supporting the need of forming compressively strained Pt overlayers in order to achieve optimum catalytic performances.

1. Introduction

Hydrogen-fueled energy conversion devices such as low-temperature fuel cells are expected to play a key role in the transition towards a sustainable economy [1]. Pt-based electrocatalysts hold great promise for the oxygen reduction reaction (ORR) in proton exchange membrane fuel cells (PEMFCs) [1]. However, a high loading of 0.2 mg_Pt/cm² is required at the cathode [2], where the ORR takes place. Since the amount of Pt available is very scarce [3], we need to reduce the Pt-loading at the cathode, while enhancing the ORR catalytic activity. Substantial progress [4] has been made in recent years to improve the sluggish ORR kinetics [5]. This rapidly progressing research field has resulted in a wide range of intricate principles for the rational design of Pt-based oxygen electroreduction nanocatalysts [6,7,8,9,10,11].

Elucidating the factors governing ORR is essential to design more efficient electrocatalysts. The ORR activity on Pt-based catalysts follows a Sabatier volcano relationship with the binding energy of the oxygen-containing reaction intermediates, e.g., OH. This has been thoroughly corroborated from extensive density functional theory (DFT) calculations [12,13]. These calculations show that the optimum activity value (the top of the volcano) should be achieved by an electrocatalyst which binds OH ca. 0.-eV weaker than pure Pt [13]. Design principles from a combination of model studies on well-defined electrode surfaces [14,15,16,17,18] and DFT calculations have facilitated the development of tailored Pt-based bimetallic nanocatalysts [19]. Recent strategies for the rational design of enhanced electrocatalysts include tuning the electronic structure by means of Pt–Pt surface strain [6,20,21,22] or ligand effects [16,17], as well as controlling the generalized coordination number [23], the size of nanoparticulate catalysts [24,25], the morphology of the nanoparticles [7,10,19,26] and the utilization of noble metals through single-site optimization [11,27,28,29,30].

In earlier studies, we identified alloys of Pt and rare earth metals such as Y and Gd as active and stable electrocatalysts for ORR in acidic media [9,20,21,22,31,32]. The high stability and activity can be explained by the formation of a thick Pt overlayer under compressive strain—as the rare earth component is vulnerable towards dissolution and tends to leach out from the surface layer, the Pt overlayer protects it against further leaching [33]. Electronic strain effects [34] occur due to the formation of this thick Pt overlayer on top of the bulk Pt–Gd alloy and has been observed to form on both on polycrystalline Pt₅Gd [22] and Gd/Pt(111) single-crystalline electrodes [31,35]. Compressive strain of the Pt overlayer originates from a combination of so-called Kagomé layers stabilized by larger Gd atoms in the bulk Pt–Gd alloy [22,36]. Gd/Pt(111) electrode surfaces show that such strain formation leads to a weakening of the OH binding energy, resulting in improved ORR kinetics [31,35]. Notably, our work on polycrystalline Pt–lanthanide alloys shows that Pt₅Gd electrodes indeed exhibit ORR activities near the top of the Sabatier volcano [19,22]. X-ray diffraction (XRD) measurements showed that all the investigated bulk Pt₅ M (M = lanthanide) alloys exhibited a hexagonal structure [22]. Moreover, we recently investigated the structure of the active phase of Gd/Pt(111) by means of in situ synchrotron grazing incidence X-ray diffraction (GI-XRD) [35]. The compressively strained Pt overlayer forms under open-circuit conditions; the strain relaxes slightly upon repeated electrochemical cycling between 0.6 and 1.0 V vs. the reversible hydrogen electrode (RHE). Interestingly, stability measurements on Gd/Pt(111) cycling between 0.6 and 1.0 V vs. RHE show that the Pt overlayer exhibits a very stable structure when cycling in this potential range [35].

A key understanding of the structure of the active phase from our work on extended Pt–lanthanide alloys enabled the design and development of nanoparticulate Pt–Gd catalysts in the form nanoparticles [21] and thin films [37], both exhibiting enhanced ORR activity over pure Pt. Extended X-ray absorption fine structure (EXAFS) measurements on mass-selected Pt_xGd and Pt_xY nanoparticles showed that their ORR activity increased exponentially with the bulk compressive strain [21].

Investigating Pt–Gd ratios that will lead to a compressive strain of the Pt overlayer that allow the desirable weakening of the OH binding energy would be key to design and develop Pt–Gd catalysts exhibiting optimum values of both ORR activity and stability. Moreover, chemical synthesis of alloys containing metallic Gd has demonstrated exceedingly difficult due to the oxophilic nature of Gd [37]. Physical techniques such as sputtering in ultra-high vacuum (UHV) are very suitable to fabricate Pt-based thin films with different stoichiometry [37,38,39]. We recently investigated Pt, Pt₅Gd and Pt₃Y thin films as catalysts for the ORR [37,38], where we took specific care in ensuring that oxygen had not been incorporated into the thin films and the strained Pt overlayers were formed on top of the alloy surfaces when subjected to electrochemical testing [31,35].

Herein, insights gathered from ex situ X-ray absorption spectroscopy (XAS) measurements on Pt₅Gd and Pt_7.5Gd thin films both as-prepared and after electrochemistry (see Figure 1) will be presented in combination with electrochemical characterization of the co-sputtered Pt_xGd thin films. While our XAS measurements were carried out at Stanford Synchrotron Radiation Lightsource (SSRL), all the fabrication and electrochemical testing was performed at the Technical University of Denmark (DTU).

2. Results and Discussion

Pt_xGd thin films were produced by means of co-sputtering Pt and Gd [37]. Pt_xGd thin films form catalytically active Pt overlayers under electrochemical conditions [37,38]. To investigate the structure of the Pt overlayers, ex situ XAS was employed on 10-nm-thick Pt_xGd samples, both as-prepared and after electrochemical cycling in 0.1-M HClO₄, see Materials and Methods for details.

Our earlier investigations on the formation of Pt and Pt_xGd thin films [37] revealed that thin films of thicknesses <27 nm were prone to pinhole formation during sputtering. Hence, a range of thin films of ~50-nm-thickness was made in parallel with the 10-nm thin films. The 50-nm thin films were the basis for the electrochemical ORR activity and stability measurements, while the 10-nm thin films were the basis for the XAS investigations. Using the small thickness films for XAS was a way of obtaining surface sensitive XAS as thicker 50-nm samples would give information on the bulk structure of the thin films. Hence, 10-nm Pt_7.5Gd and Pt₅Gd co-sputter-deposited thin films were transferred to a N₂-saturated 0.1-M HClO₄ containing three-electrode electrochemical cell (see Figure 1). Pt_xGd thin films were then cycled from 0.6 to 1.0 V vs. RHE until stable CVs were formed (see Figure 2a). By means of CO-stripping measurements and H-adsorption areas from the base CVs, we were able estimate the roughness factor (${\mathsf{\Gamma }}_{\mathrm{CO}/\mathrm{H}ad.}=A_{\mathrm{CO}/\mathrm{H}ad.}/A$) of the thin films, where ($A$) designates the geometric electrode area. Although the diameter of the electrode was 5.0 mm, the outer 0.05 mm around the rim was covered by a mask. However, we neglect this small difference, corresponding to a ~4% smaller sputter area on the electrode (see

Table 1. Fitted extended X-ray absorption fine structure (EXAFS) and electrochemical characterization data for Pt_xGd (10 nm) samples as prepared and after electrochemical conditioning, the latter denoted by “*”. All EXAFS strain data relied on a Pt foil reference measurement [21] in which a relaxed Pt–Pt bond length of 2.775 Å was established, see S.I. EXAFS data are given with ± 0.01 Å. This corresponds to a strain error of about 0.4%.

Parameter	Pt7.5Gd (10 nm)		Pt5Gd (10 nm)
Pt–Pt distance (Å)	2.73	2.75 *	2.69	2.73 *
Pt–Pt strain %	−1.6	−0.8 *	−3.1	−1.6 *
Gd–O distance (Å)	2.39	n.a.*	2.39	2.28 *
Pt-overlayer strain %	n.a.	−0.3 *	n.a.	−0.5 *
$Q_{\mathrm{CO}}$ (μC)	n.a.	162.1 *	n.a.	135.7 *
$Q_{\mathrm{H}ad.}$ (μC)	n.a.	97.1 *	n.a.	79.5 *
${\mathsf{\Gamma }}_{\mathrm{CO}}$	2.4	2.5 *	2.0	2.1 *
${\mathsf{\Gamma }}_{\mathrm{H}ad.}$	2.6	2.7 *	2.1	2.2 *

* after electrochemical cycling until stable CVs formed.

).

Two as-prepared and two electrochemically conditioned 10-nm-thick Pt_7.5Gd and Pt₅Gd thin films were then transferred to SSRL and ex situ EXAFS spectra were obtained for these four samples (see Figure 3).

EXAFS spectra at the Gd L₃ and Pt L₃ line-edges for the four 10-nm Pt_xGd samples were obtained and fitted (see S.I.). Both the electrochemical data and EXAFS results from Figure 2 and Figure 3 are summarized in Table 1. Real-space transforms of EXAFS data can be found in Figures S1 and S2.

The results summarized in Table 1 suggest that, when the Pt_xGd thin films are formed, significant oxygen is present in the surface layer, as observed by the significant Gd–O coordination. This suggests that the oxophilic nature of Gd [37] results in a significantly oxidized surface termination (see S.I.). Moreover, the electrochemical conditioning of the thin films significantly reduces the Gd–O coordination suggesting gadolinium-oxides in the surface dissolve completely during cycling in the acidic electrolyte. Similar observations were made from earlier X-ray photoelectron spectroscopy (XPS) studies [37].

As expected, the nearest Pt–Pt distance from Table 1 increases after electrochemical cycling. The Pt–Pt distance of the pure Pt overlayer lies between that of the Pt_xGd alloy and bulk Pt [36]. Thus, when the thin film is transformed from mostly alloy to part alloy and part overlayer, the average Pt–Pt distance is increased. This is also evident from the raw data in Figure 3a: In reciprocal space, the frequency increases due to longer average distances in real space. After the electrochemical cycling, a compressively strained Pt overlayer is formed from Gd leaching, in agreement with earlier studies of Pt rare-earths and lanthanide alloys [22,35,37,38]. The formation of the overlayer was supported by an additional measurement on the Pt₅Gd sample at a shallower angle (at 0.2° which increases the signal contribution from the surface), in which we observe marginally increased Pt–Pt distances.

Table 1 also shows that relatively high roughness factors are found both from H-adsorption and CO-stripping area measurements. This strongly suggests that extensive pinhole is formed due to the low thickness of 10 nm (see S.I.). Unfortunately, concluding on any compressively strain difference (seen from the Pt–Pt distance) between the Pt_7.5Gd and Pt₅Gd sample is not straightforward, given the extensive roughness. From our earlier work on well-defined extended surfaces [22], we expect Pt₅Gd to exhibit superior ORR activity performance, as its Pt overlayer is more compressively strained than that of Pt_7.5Gd. Indeed, we recently observed that the overlayer strain on polycrystalline Pt₅Gd leads to an activity that approaches the top of the experimental volcano [22]. We note that, according to theoretical calculations [13,32], the OH weakening leading to an optimal ORR activity corresponds to a Pt overlayer under a compressive strain of ca. 3% relative to pure Pt [22].

To further correlate Pt:Gd stoichiometry with ORR activity, 50-nm-thick Pt_7.5Gd and Pt₅Gd thin films were produced and electrochemically characterized. We must note that we carried out the electrochemical measurements on 50-nm thin films as we could not use thinner films due to the presence of pinholes and excess of under-coordinated sites associated (see Figure S3) [37]. Following earlier established electrochemical characterization procedures on Pt₅Gd thin films [37], a minimum of four samples of each composition were tested using standard rotating disk electrode (RDE) measurements in 0.1-M HClO₄ electrolyte using a three-electrode setup, as shown in Figure 1. The base CVs on Pt_xGd (50-nm) thin films before and after accelerated stability tests (see Section 3 Materials and Experimental Methods) were evaluated in conjunction with CO-stripping experiments and CVs under ORR conditions. Representative base CVs and CO oxidation can be seen in Figure 4a,b, respectively.

From the base CVs in Figure 4a, it is noted that features typically associated with H and OH adsorption and desorption [16,40] do not change significantly after accelerated test. This suggests the formation of stable films under the presented electrochemical conditions. Conversely, there is a small shift in the position of the CO stripping peak. In this context, it should be noted that CO can induce Pt surface roughening/smoothening [41,42,43] and general reorganization [44]. The CO-stripping experiments from Figure 4b, presented in

Table 2. Electrochemical data for the 40 nm Pt as-prepared and 50 nm Pt₅Gd and Pt_7.5Gd thin films before and after AST. Specific activities were evaluated at 0.9 V vs. RHE and are shown both using the area derived from CO oxidation and hydrogen adsorption [37].

Parameter	Pt (40 nm) Before AST	Pt5Gd (50 nm) Before AST	Pt5Gd (50 nm) After AST	Pt7.5Gd (50 nm) Before AST	Pt7.5Gd (50 nm) Before AST
$Q_{\mathrm{CO}}$ (μC)	75.4 ± 7	116 ± 25	118 ± 25.5	91 ± 17	93 ± 17
$Q_{\mathrm{H}ad.}$ (μC)	35.3 ± 4	65 ± 16	66 ± 16	55 ± 8	59 ± 9
$A_{\mathrm{CO}}$ (cm2)	0.19 ± 0.04	0.33 ± 0.07	0.34 ± 0.07	0.26 ± 0.05	0.27 ± 0.05
$A_{\mathrm{H}ad.}$ (cm2)	0.18 ± 0.04	0.34 ± 0.08	0.35 ± 0.08	0.29 ± 0.04	0.31 ± 0.04
${\mathsf{\Gamma }}_{\mathrm{CO}}$	1.0 ± 0.1	1.7 ± 0.4	1.8 ± 0.4	1.3 ± 0.3	1.4 ± 0.3
${\mathsf{\Gamma }}_{\mathrm{H}ad.}$	0.9 ± 0.1	1.7 ± 0.4	1.9 ± 0.5	1.5 ± 0.2	1.6 ± 0.2
UCO-peak (mV)	662 ± 16	679 ± 70	672 ± 10	657 ± 13	680 ± 1
jk,CO (mA/cm2)	4.7 ± 0.8	10.0 ± 3.4	9.1 ± 3.1	7.8 ± 0.6	7.3 ± 0.6
jk,H ad. (mA/cm2)	4.9 ± 0.8	9.6 ± 1.9	8.3 ± 1.6	7.2 ± 0.7	6.7 ± 0.7

, reveals that; (i) Quite rough thin film surfaces form. (ii) Pt_7.5Gd thin films are less rough than Pt₅Gd thin films, likely due to the lower degree of Gd dissolution by acid leaching [35]. (iii) The potential of the maximum CO oxidation peak increases after stability tests and the U_CO-peak value attains similar values for both the Pt₅Gd and Pt_7.5Gd samples, suggesting a similar structure is formed after aging.

The ORR activity of the Pt_xGd thin films was also electrochemically evaluated. Representative anodic sweeps during ORR can be seen in Figure 5a. It should be noted that the ORR activities were evaluated using the Koutecky–Levich relation, which identifies the mass-transport corrected kinetic component (j_k) of the oxygen reduction current density (j) taking into account the diffusion limiting current density (j_diff): $j_k^{-1}=j^{-1}-j_{diff}^{-1}$ [45]. The logarithm of the kinetic current density was plotted against the RHE potential to obtain a Tafel plot, as seen in Figure 5b.

The electrochemically active surface areas derived from the CO-stripping charges were utilized to estimate the kinetic current densities in Figure 5b, whereas current densities in Figure 4a,b and Figure 5a were normalized by the geometric electrode area of ca. 0.196 cm² [37]. From the electrochemical characterization (Figure 4 and Figure 5) key parameters such as CO and H-adsorption charge (Q) and associated active electrodes areas (A) and roughness factors ($\mathsf{\Gamma })$ could be extracted. Moreover, kinetic current densities for ORR and CO oxidation peak potential are summarized in Table 2. Table 2 shows that Pt₅Gd (50 nm) and Pt_7.5Gd (50 nm) samples present differences in ORR activity, Pt₅Gd being more active than Pt_7.5Gd, as expected from a higher Pt–Pt compressive overlayer strain [22].

3. Materials and Methods

3.1. Thin Film Fabrication

Pt_xGd and Pt thin films were produced by a (co)sputtering method described elsewhere [37]. A polished and clean glassy-carbon stub (HTW, Ø×h = 5 × 4 mm) was placed in an ultra-high vacuum system. The system base pressure was 10⁻⁹ mbar. Thin films of Pt–Gd alloys were co-sputtered atop using Ar (N5, 50 sccm) and powers of 180 W and 17–20 W for the Pt and Gd targets, respectively (Kurt–Lesker, 99.995%). The film thickness was monitored using a quartz microbalance; the deposition temperature was maintained at 300 °C.

3.2. Electrochemical Conditioning and Characterization of Pt_xGd Co-sputtered Thin Films

The co-sputtered Pt_7.5Gd and Pt₅Gd thin films (10-nm-thick) used for the EXAFS investigation were transferred to a N₂-saturated 0.1 M HClO₄ (Merck, Suprapur^®, 70%) containing three-electrode electrochemical cell (see Figure 1). The samples were immersed into the electrochemical cell under potential control and cycled at 50 mV/s between 0.05 and 1.00 V vs. the RHE until stable cyclic voltammograms (CVs) were obtained. The charge associated with H adsorption ($Q_{\mathrm{H} ad.}$) can be used to estimate the electrochemically active surface area (ECSA) of Pt catalysts [37,46,47,48]. Once stable CVs had formed, CO-stripping measurements were conducted at 10 mV/s [37,49]. The CO-stripping procedure consisted of sample being cycled twice in Ar-saturated electrolyte at 10 mV/s. Then, the electrode was maintained at 0.05 V vs. RHE while bubbling CO for 2 min, saturating the electrode surface. The electrode potential was maintained at 0.05 V vs. RHE for 15 min while bubbling Ar gas purging the remaining CO from the electrolyte. This allowed CO to be adsorbed on the electrode surface, while allowing sufficient time for residual CO to be purged from the cell. The potential was then swept between 0.05 V and 1.00 V vs. RHE at 10 mV/s. The CO oxidation charge ($Q_{\mathrm{CO}}$) derived from the CO-stripping procedure was also used to evaluate the ECSA [44,50,51]. From our earlier work [37,49], the typical CO oxidation charge per area for polycrystalline Pt is ~349 μC/cm². This charge, along with the hydrogen adsorption/desorption value of ~191 μC/cm² [37], was used to estimate the ECSA by CO stripping and hydrogen underpotential deposition (H_UPD), respectively. We must note that, although evaluating the H_UPD area is a standard method of ECSA evaluation on Pt electrodes, H binding may be modified on Pt alloys due to electronic effects such as strain effects [37]. Fifty-nanometer Pt_xGd thin films were used to evaluate electrochemical performance and for direct comparison with the 10-nm samples. We also include reference data of 40-nm-thick pure Pt thin films for comparison. Pt_xGd samples were initially cycled from 0.0 to 1.0 V vs. RHE in N₂-saturated 0.1-M HClO₄ at 200 mV/s and 400 rpm until CVs were stable. The base CVs at 50 mV/s were then gathered at 400 rpm. The ORR activity was evaluated in O₂–saturated electrolyte at 1600 rpm. The samples were subjected to electrochemical impedance spectroscopy (EIS) and Ohmic drop was evaluated at double-layer region at 0.5 V vs. RHE. The Ohmic drop was evaluated by linear regression of the resulting Nyquist plot in the 60 to 400 Hz region [37]. All electrochemical data in this work were presented using a Pt counter electrode and plotted relative to RHE. The RHE potential was evaluated by bubbling H₂ while rotating the working electrode at 1600 rpm and cycling between −0.74 and −0.70 V vs. the Hg_(l)ǀHgSO_4(g) reference electrode (Schott Instruments, 0.6-M K₂SO₄). The value of the RHE potential can be read as the intercept with the current axes [37]. Finally, a base CV in N₂-saturated electrolyte at 50 mV/s and CO-stripping CVs at 10 mV/s were obtained. Accelerated stability tests (AST) were also conducted, consisting of additional 10,000 cycles in O₂-staurated electrolyte between 0.6 and 1.0 V vs. RHE at 100 mV/s without any electrode rotation [52]. After this extended cycling, the electrolyte was changed to a fresh one and the earlier described characterization steps were repeated.

3.3. X-ray Absorption Spectroscopy Measurements

XAS experiments were carried out ex situ at SSRL (Stanford Synchrotron Radiation Lightsource, CA, USA). Both as prepared and electrochemically conditioned samples, were transferred under ambient conditions in a sample holder. EXAFS measurements were carried out with a gracing incidence of 2° (unless otherwise stated) at beam line 11–2 at Stanford Synchrotron Radiation Lightsource (SSRL). Spectra were obtained of both Pt L₃ and Gd L₃ edges and fitted in R-space using a k²-weighing, see Figure 3 and supporting information (S.I.). The spectra were processed and fitted using the software package SIXPack (v1.40, SAMware, Stanford, CA, USA, 2016), using scattering paths from FEFF8 [53].

4. Conclusions

Thin film Pt_xGd electrocatalysts with different stoichiometry were fabricated using a co-sputter-deposition technique. Both 50 nm Pt₅Gd and Pt_7.5Gd thin films are highly active catalysts for ORR. The increased activity compared to pure Pt and the difference in activity between Pt₅Gd and Pt_7.5Gd can be explained by the inherent effect of overlayer strain. This was observed using 10-nm-thick thin film equivalents by EXAFS. We were able to detect compressive strain in the thin films and reveal that the overlayer on Pt₅Gd is more compressively strained, in good agreement with ORR activity measurements on Pt_xGd nanoparticles [21] and Pt₅Gd extended surfaces [22,31,35]. Although ORR experiments were carried out on 50-nm films and EXAFS measurements were performed on 10-nm thin films, the compressive strain values from EXAFS can be used to explain the differences in ORR activity between Pt₅Gd and Pt_7.5Gd thin films. Using EXAFS, it is revealed that the overlayer on Pt₅Gd is more compressively strained than on Pt_7.5Gd. The powerful approach of using EXAFS to screen surface strain and thus optimal ORR response is relevant for a wide degree of Pt-alloy systems. We envisage that such flexible systems combined with powerful synchrotron analysis may open up for new avenues for electrocatalysis research.