The Oxidation of Platinum under Wet Conditions Observed by Electrochemical X-ray Photoelectron Spectroscopy

During the electrochemical reduction of oxygen, platinum catalysts are often (partially) oxidized. While these platinum oxides are thought to play a crucial role in fuel cell degradation, their nature remains unclear. Here, we studied the electrochemical oxidation of Pt nanoparticles using in situ XPS. When the particles were sandwiched between a graphene sheet and a proton exchange membrane that is wetted from the back, a confined electrolyte layer was formed, allowing us to probe the electrocatalyst under wet conditions. We show that the surface oxide formed at the onset of Pt oxidation has a mixed Ptδ+/Pt2+/Pt4+ composition. The formation of this surface oxide is suppressed when a Br-containing membrane is chosen due to adsorption of Br on Pt. Time-resolved measurements show that oxidation is fast for nanoparticles: even bulk PtO2·nH2O growth occurs on the subminute time scale. The fast formation of Pt4+ species in both surface and bulk oxide form suggests that Pt4+-oxides are likely formed (or reduced) even in the transient processes that dominate Pt electrode degradation.


S2
To investigate the cleanliness of the deposition process, Figure S2 shows a comparison of the dark field and bright field TEM images of the same particle. The dark field image is predominantly sensitive to heavy elements (e.g. Pt), whereas the bright field image allows the observation of light elements (e.g. C contamination). The overlay of the two images ( Figure S2c) shows excellent matching, which confirms the absence of a contaminant layer. We should point out, however, that a single layer of small adsobates following air exposure will have likely formed, and would be beyond the detection sensitivity of the analysis.  Figure 1c in the main text. b) Corresponding bright field image. c) Overlay of dark field (red) and bright field (image).
Following the deposition, the FAD55 membranes were soaked in 0.5 M Na2SO4 at room temperature for at least 3 days. The Nafion® samples were kept dry.
Graphene on copper (Graphenea SA) was etched in a 40 g/L (NH4)2S2O8 aqueous solution to remove the copper substrate, leaving a floating graphene sheet on the water surface. After dilution of the solution, the membranes were introduced into the water beneath the graphene sheets. Subsequently, the water level was lowered to transfer the graphene onto the membranes. Following the transfer, the samples were dried in air at room temperature to ensure adhesion of the graphene layer. To prevent the formation of microtears in the FAD55 membranes due to repetitive swelling and drying, the FAD55 membranes were reintroduced to the 0.5 M Na2SO4 solution after approximately 15 minutes of drying (only the surface of the membrane is completely dry under these conditions). Prior to use, the membranes were washed in ultrapure water. Figure S3 shows XPS surveys of graphene-covered Pt on Nafion® (S3a) and on FAD55 (S3b). As expected for the thin, open layers studied here, the Pt layer does not obscure the underlying membrane from XPS observation. For Nafion®, a fluoro-ether with sulfonate functionalization, we observe this in the form of F, O and S peaks. Similarly, N, C, and O can be related to FAD55 in Figure S3b. In addition, FAD55 contains Br. As shown in the main text however, these species only become apparent at high potential (see also Figure 2 in the main text). The only contamination visible in the spectra is a minute Cu LMM Auger peak. This can be attributed to remnants of Cu left after the graphene preparation procedure. We should point out that these remnants do not appear to be the result of insufficient etching. In fact, the Cu substrate is not visible anymore after only 1 hour of etching, whereas the total S3 etching time was 12 hours. Hence, it seems likely that the Cu remnants are encapsulated in defective parts of the graphene. Figure S3: XPS survey spectra of a) 4 nm graphene-covered Pt on Nafion® 117 (corresponding to Figure 5b in the main text) and b) 1 nm graphene-covered Pt on Fumasep® FAD55 (corresponding to Figure 4 in the main text). Both spectra were recorded in the electrochemical cell at OCP. In both cases, the electrolyte was 0.1 M H2SO4.
The graphene layer attenuates the Pt 4f signal due to scattering of the photoelectrons. We established that 300 eV electron kinetic energy is the lower bound for our measurements. Most experiments were conducted with a kinetic energy of roughly 500 eV. Under these conditions the transmission through the graphene is about 40%, estimated using the data at OCP shown in Figure 2 b and c in the main text.
Along with the spectroscopic measurements, we performed an electrochemical characterization of the samples in situ at the beam line. For a proper judgement of the electrochemical data, we should point out that a large fraction of the sample was in contact with the glassy carbon top plate, which contains a hole for spectroscopy (see Figure S4a). Hence, the electrochemical data reflect not only the bare or graphene covered part of the sample, but also the part where the Pt catalyst is sandwiched between the proton exchange membrane and the glassy carbon top plate.
Impedance measurements for both the Nafion® and FAD55 supported samples indicated a contact/solution resistance of 40-50 Ω at OCP. Figures S4b-d show cyclic voltammetry data recorded at the beamline. The Nafion® sample in Figure S4b shows the usual Pt features: surface oxidation onset (OH/O adsorption) at roughly 0.8 V, the oxygen evolution onset at about 1.55 V, the oxide reduction in the cathodic scan at 0.65 V, hydrogen adsorption features between 0.2 V and 0 V and the hydrogen evolution below 0 V. In addition, a large peak is observed at 0.05 V. We attribute this to the oxidation of H2 gas, which may require some time to diffuse away from the glassy carbon-Pt-membrane sandwich. The graphene-covered Pt/FAD55 sample in Figure S4c shows the same features, although the CV is also modified by the electrochemical behavior of the FAD55 membrane itself ( Figure S4d).

S4
XPS was used as a probe for the conductivity in the working electrode. For a grounded electrode, the Pt 0 peak should remain at the same position independent of the applied potential. If electrical contact is lost, the Pt 0 peak will shift to lower binding energy at positive potentials, which easily distinguishes this behavior from Pt oxidation. The same holds for the position of the valence band edge. By checking these two factors, it was established that the electrical contact of all the samples shown in this work was adequate. To test the performance of the system under reaction conditions, we used mass spectrometry to detect O2 generated at on the Pt catalyst at 1.85 V. As shown in Figure S5, our catalyst is clearly able to provide a steady-state conversion, despite the poor quality of platinum as a catalyst for the oxygen evolution reaction. In agreement with this, impedance measurements showed that the resistance between the reference electrode and working electrode was a modest 50 Ohm, implying good mass transport through the membrane. a) Figure S5: Oxygen evolution reaction on 16 nm graphene-covered Pt on FAD55.

S2
Analysis procedure Analysis of the XPS spectra was performed using the CasaXPS 2.3.18 software package. A Shirley background subtraction was used in all cases. The asymmetric Lorentzian LF lineshape was used for fitting of the Pt 4f spectra. While the Doniach-Sunjic lineshape is more physically rigorous, the dampened tail of the LF curve ensures a more unambiguous fit within the limited spectral window that we afforded ourselves in order to limit beam damage. Fitting was performed using 4 doublets: Pt 0 , Pt δ+ , Pt 2+ and Pt 4+ , parameterized as shown in Tables S1-S4. The first two parameters in the LF line shape define the peak asymmetry (equal means symmetric, large difference results in an asymmetric lineshape), the third parameter refers to the width of the Gaussian with which the Lorenzian curve is convoluted and the forth parameter defines the dampening of the curve away from the peak center. The more symmetric lineshape used for the Pt 4+ contribution was chosen not only based on the observed peak shape at high potential, but also based on the physical nature of poorly conducting materials such as PtO2, which have less peak asymmetry due to their limited electron-hole pair excitations.   The ranges for the peak positions were set according to literature data (see literature compilation in the supporting information of Saveleva et al. 1 ). The binding energy of Pt 0 was allowed a +/-0.1 eV variation to account for inaccuracies in setting the beam energy (valence band alignment was regularly performed during the beam time, but was inhibitively inaccurate for the low-coverage samples). For the depth profiling analysis the line shape, peak position and FWHM for Pt 0 and Pt δ+ were determined at open circuit potential, where no oxide was present. The parameters for the Pt 4+ peak were determined by free fitting of the 4 nm thick sample at 1.85 VRHE. In the other datasets, the line shape, peak position and FWHM of Pt 4+ peak constrained to values from the 4 nm case.
For the O K-edge spectra, the energy scale was calibrated using the sharp peak at 537.2 eV in the TEY signal, belonging to the Rydberg state in gas-phase water 2 . The intensities were corrected for the synchrotron ring current and beam line transmission function. After applying an offset to set the preedge region to zero, the spectra were normalized to 1 at 552 eV.

S3 Beam damage effects and prevention measures
The intense X-rays employed in synchrotron measurements can be damaging to materials, in particular oxides and polymers. To investigate the effect of the beam on our samples we compare spectra at the start of beam exposure and after some minutes. As shown in Figure S6a, beam damage effects can be rather severe for the Nafion® samples. The PtO2 is quite strongly reduced after roughly 4 minutes in the beam. In contrast, hardly any PtO2 reduction is observed for the FAD55 sample ( Figure S6b). This rather unexpected difference in behavior underlines the complexity of beam damage phenomena, and the need to study them for individual material combinations. While Pt oxides are stable on FAD55, the drop in the Br peak ( Figure S6b) clearly indicates that the membrane polymer is damaged. To test if this affects the local proton conductivity, we compared the degree of electrochemical PtOx formation on a beam exposed and a fresh spot for both Nafion® and FAD55. For the Nafion® case, we observed that the reactivity is suppressed on the exposed spot, whereas little to no effect was observed for FAD55. Hence, we conclude that while both Nafion® and FAD55 suffer beam damage, only Nafion® loses its proton conductivity as a result of this. Note that Nafion® and FAD55 have similar proton conductivities in their pristine form (as determined by their manufacturers).
The above results make clear that long beam exposure to our samples is undesirable, particularly for the Nafion® samples. For this reason, spectra were either recorded in a single (fast) scan and/or with low beam intensity ( Figure 6a in the main text). Furthermore, each spectrum shown in the main text was recorded on a fresh spot on the surface. Due to the excellent homogeneity of our samples, and our relatively large beam spot size (0.2 mm x 0.1 mm), we do not expect significant variation the behavior of the sample on different spots on the surface. The smooth trends in the Pt 4f and O K-edge spectra confirm this.

Wetting of the Pt nanoparticles
As discussed in the main text, the most important feature that can be used to investigate the wetting of the Pt nanoparticles is the peak at 535 eV in the O K-edge spectra. To determine the contribution of graphene oxide to this peak, Figure S7 shows O K-edge spectra of graphene deposited on a polycrystalline Au/Si substrate. To reduce the noise in the PEY spectrum, an average was taken of 5 spectra recorded at different locations on the surface. The ratio between the 535 eV peak and the peak at 532 eV that accompanies it is close to one. Given the small 532 eV contribution in Figure 2a in the main text, graphene oxide can only have a minor contribution to the observed 535 eV peak. Hence, it predominantly originates from water. It should be noted that the sulfate concentration on the surface was clearly higher for the graphene covered sample, as determined from the S 2p peak area. However, the O K-edge spectrum of sulfate does not show any peaks in the window between 530 eV and 537 eV 3,4 . Some additional information on the wetting of the Pt particles can be obtained from the O 1s spectra. However, the large variety of species (H2Oad, H2Omultilayer, membrane hydrocarbons, graphene oxide and SO4 2-) in a limited spectral window (530-535 eV) make assignments extremely difficult. In addition, attenuation of the substrate, sulfate and water peaks due to the graphene layer hamper the comparison of intensities. Nonetheless, Figure S8 shows that a broadening of the O 1s spectrum occurs as a result of the graphene cover. The reference spectrum of graphene on Au shows that the broadening at higher binding energy is not related to graphene. The O 1s peak of sulfate is found between 532.1 and 532.5 eV 5 , also disqualifying it as an explanation. With a binding energy of 533 eV, multilayer water would explain the broadening at higher binding energy, providing another indication for the high degree of wetting on the Pt samples. To establish the location of the electrochemically created oxide on the 16 nm thick Pt layers, we recorded Pt 4f spectra at several excitation energies. In the probed energy range, 380 eV to 1280 eV, the electron attenuation length in platinum changes from 0.54 nm to 1.16 nm 6 . Hence, the decrease in the Pt 4+ peak at higher excitation energy seen in Figure S9b indicates that the XPS spectra probe an oxide at the outer surface of the catalyst layer (the graphene side). This is clearly not the case for the uncovered sample ( Figure S9a), where the Pt 4+ peak rises with excitation energy. Although some of the rise is related to the decreased spectral resolution at higher excitation energy, this suggests that the oxide is located at the buried Pt-FAD interface.

S6 XPS analysis of the electrolyte species in Nafion
Similar to the case of FAD, the binding energy of species belonging to the Nafion® membrane display shifts resulting from changes in the local electrostatic potential. As shown in Figure S10, the extent of the shifts varies. The C1s peak around 291 eV, which belongs to the fluorinated carbon in Nafion®, shifts 1.05 eV, roughly the expected 1.1 eV. In contrast, the peak at 284 eV, originating from the graphene cover layer, does not show any shift as it is part of the working electrode (which, again, was grounded via the electron analyzer). The O 1s peak shows a mixed response with a shift of 0.43 eV, indicating that the peak contains contributions from both species adsorbed on the working electrode and free electrolyte. For the fluorine signal, the importance of probing depth becomes evident. For the F1s peak, which is very surface sensitive due to the low kinetic energy of the emitted electrons (about 412 eV), a shift of 0.73 eV is observed. In contrast, the F KLL signal is shifted by 1.14 eV. Hence, it appears that the species probed by the F1s peak are predominantly located within the decay length of the electrostatic potential.

Potential-induced redistribution of the membrane in the catalyst layer
The shoulder around 69.3 eV in Figure 2c in the main text originates from Br 3d electrons from the FAD membrane. The rise of the shoulder as a function of potential suggests a redistribution of the membrane's polymer branches in the catalyst layer. Depth profiling ( Figure S9b) confirms that the Br is located at surface at high potentials. Such redistribution can also be observed for Nafion®-supported samples, where we probed the polymer's sulfonate groups with S2p spectra. Again, we observe an increasing peak intensity during anodization (see Figure S11), indicating that the polymer redistribution occurs on this substrate as well. However, in contrast to the case of FAD, the effect appears to be irreversible. Further support for the generality of this phenomenon comes from observations on a phosphoric acid soaked hydrocarbon membrane 7 . In this study, the driving force for the redistribution was assigned to the screening of the electrode potential, and the solvation of the ions required to do this. This interpretation can explain why there is no rise in the Br peak in Figure 2b in the main text. The dry surface for the uncovered sample does not allow for proton and anion conductivity, thus preventing the formation of an electrochemical interface. Hence, there is no interfacial potential to screen and thus no driving force for the membrane redistribution.

Oxidation of the graphene capping layer
Despite its high chemical stability, graphene can be oxidized at sufficiently high potentials. Figure S12 shows C1s spectra recorded along with the experiment in Figure 4 in the main text. The main contributors to the spectra are graphene and the FAD substrate. The graphene is part of the working electrode, which is grounded. Hence, the C1s contribution from graphene should remain at the same position when a potential is applied. The contribution from the FAD substrate will shift however, because it is part of the electrolyte. Taking these considerations into account, only modest chemical changes are observed in the C1s spectra. At 1.85 V one can see a small peak around 288 eV, which suggests the presence of oxidized carbon. Hence, we conclude that only minor graphene oxidation occurs. We should point out however, that the C1s spectra are not a very sensitive for detecting the deterioration of graphene. Hence, there may still be some contribution of graphene oxide to the O Kedge spectra in Figure 4b in the main text. Experiments on blank graphene/FAD confirm that at potentials above 1.2 VRHE some graphene oxide is observable. S12 Figure S12: Evolution of C 1s spectra during the experiment depicted in Figure 4 in the main text.

S9 Bulk Pt oxidation of nanoparticles
To establish whether the oxidation of the single layer nanoparticles went all the way into the core of the particles, we recorded a Pt 4f spectrum at 1.85 V with a beam energy of 1280 eV. With an attenuation length of 1.16 nm, this measurement probes well into the bulk of the nanoparticles. Figure  S13 shows that even for such a more bulk sensitive measurement, there is only a weak Pt 0 contribution, marked by a slight shoulder at 74.3 eV and a mixed Pt 0 / Br 3d peak.