Operating Fuel Cells

Transcription

1 Page 2 of 44 Infrared and X-ray Absorption Spectroscopy of Operating Fuel Cells Ian Kendick and Eugene S. Smotkin* Department of Chemistry and Chemical Biology 360 Huntington Ave, Boston, Massachusetts The first operando cells for infrared and X-ray absorption spectroscopy of fully operating fuel cells are described along with selected data. A new cell, combining the features of the first IR and XAS fuel cells is described. The frequency tuning of CO adsorbed at the Pt-Nafion interface in a membrane electrode assembly showed co-adsorption induced tuning due to the compression/dissipation of CO islands. The attenuated total reflectance spectroscopy of Nafion, polarization modulated infrared reflection spectroscopy of a Nafion/Pt interface and density functional theory calculations of Nafion IR spectra yield a model for Nafion self-assembly that involve the sulfonate and -CF 3 as coadsorbates along with reactive CO. Time dependent X-ray absorption near edge spectroscopy of fuel cell cathodes show that time constants for Pt restructuring with changes can be on the order of hours. 1

2 Page 3 of 44 Physical Chemistry Chemical Physics 1. Introduction Operando fuel cell spectroscopy: The initial experiments Operando IR spectroscopy Operando X-ray absorption spectroscopy Operando IR-XAS Fuel Cell Cell Design Infrared spectroscopy Competitive adsorption of CO onto Pt Mechanically coupled internal coordinates of ionomer vibrational modes Elucidating the ionomer metal interface: X-ray Absorption Spectroscopy Conclusion Introduction That the active state of a catalyst exists only during catalysis 1 is succinct rationale for operando methods of catalyst characterization. The primary challenge to operando spectroscopy is conversion of a practical device into a spectroscopic cell with minimal perturbation of device functionality. Figure 1 schematizes the terminal end-cell of a fuel cell stack. 2 2

3 Page 4 of 44 Figure 1. PEM fuel cell schematic (From ref. 2) Graphite flow-field plates distribute fuel and oxidant to the 5-layer membrane electrode assembly (MEA). An ionomer membrane (e.g., Nafion) supports electrocatalytic layers that contact gas diffusion layers (i.e., porous carbon paper or cloth) that are optimized for reactant transport and electronic conductivity. MEA fabrication methods have been reviewed. 3 Catalyst particles (carbon supported or metal blacks) 4 are dispersed in alcoholic solutions of solubilized ionomer. These inks are deposited onto the gas diffusion layers and hot pressed to the membrane. Alternatively, a catalystcoated-membrane (CCM) can be prepared by immobilizing the membrane on a heated vacuum table (NuVant Systems Inc., Crown Point, IN) for direct ink deposition onto the membrane. Catalytic layers are a complex blend of ionomer, catalyst particles and, sometimes at the cathode, Teflon dispersion. The relative amounts depend on the catalyst composition and device application. The final step of MEA preparation occurs in the operating fuel cell. The hot pressing of the gas diffusion electrode and/or ink drying on the heated vacuum table causes delamination of the ionomer from the catalyst 3

4 Page 5 of 44 Physical Chemistry Chemical Physics particles. The conditioning of the MEA in the operating fuel cell re-wets the catalyst with ionomer and removes the deep oxides that are typical of as-prepared catalysts. 5 There are no intended triple phases. The catalyst active area must be coated with a sub-micron gas permeable, ion conducting layer. 6 A Nafion layer on Pt has been shown to enhance electrocatalysis. 6, 7 Ionomer electrolytes have no mobile ions other than protons or hydroxide ions. Supplemental electrolytes such as aqueous H 2 SO 4 or HClO 4 contribute mobile ions that competitively adsorb onto the surface Supplemental aqueous electrolytes preclude fuel cell operation at the high end of relevant temperatures (e.g o C). Operando spectroscopy requires control of the temperature, flow rate and the humidity of the anode and cathode reactant streams while potential dependent spectra are acquired. Cell component materials must not fluoresce at energies similar to the X-ray edge energies of the catalysts. The uniformity of the polymer electrolyte resistance, governed by the ionomer membrane thickness (e.g., 7 mil for Nafion 117) 11 depends on careful water management and proper flow field design. Whether studying anode or cathode catalysts, the counter electrode can serve as the auxiliary and the reference electrode (counter-reference electrode). 12 Gurau delivered hydrogen to the counter-reference electrode while acquiring liquid feed direct methanol fuel cell anode polarization curves. 13 Pure water can be delivered to the counter-reference electrode: Hydrogen evolution at the counter electrode and the hydrogen ion activity, set by the ionomer equivalent weight and state of hydration, poises 14 the counter-reference electrode. Although the counter-reference is polarized at high currents, the alternative of developing a 3 rd electrode as a reference is more complex than correcting for reference electrode polarization. Fortunately, the exchange current density for the hydrogen electrode is many 4

5 Page 6 of 44 orders of magnitude larger than that for the oxygen reduction reaction. 15 On a practical level, the use of the fuel cell counter-reference affords greater reproducibility between laboratories. This Perspective describes the first fully operando FTIR and X-ray absorption studies of fuel cells. The features of these legacy cells are then combined into a single cell that can accommodate both FTIR and X-ray absorption spectroscopy of operating fuel cells. The operando spectroscopy, complemented with density functional theory (DFT) and polarization modulated infrared reflection absorption spectroscopy (PM-IRRAS) of ionomer-platinum interfaces, suggests a model for selfassembly of Nafion onto a Pt. Time dependent X-ray absorption studies show that catalyst restructuring with potential changes can have time constants of up to hours. 2. Operando fuel cell spectroscopy: The initial experiments 2.1 Operando IR spectroscopy Operando infrared spectroscopy of the catalyst-electrolyte interface is ideal for adsorbate characterization. Adsorbates effect dynamic changes on the catalyst surface such as surface restructuring where surface atoms can be entirely displaced in order to make stronger surfaceadsorbate bonds. 16 These changes can be observed, in operando and in situ spectroscopy, as changes in the vibrational modes of either the surface or the adsorbate atoms. Fan introduced operando fuel cell spectroscopy in , 18 He incorporated a CaF 2 window on a fuel cell flow field, cut a slot into the GDL to expose the catalytic layer, and installed the modified fuel cell into a 1990 s model Harrick Praying Mantis (Pleasantville, NY) diffuse IR accessory. Figure 2 schematizes the cell used to identify methanol oxidation products, including formic acid, in spectra acquired over a range of temperatures from o C with steady state reactant flow to both electrodes. Bo used the same cell 5

6 Page 7 of 44 Physical Chemistry Chemical Physics to measure Stark tuning rates 19 of CO/Pt at 50 o C. 20 The CO/Pt tuning rates, at potentials below and above the current onset potential (0.5 volts vs. NHE), versus the methanol/d 2 O mole ratio were reported (Table 1). Bo observed inverted bipolar peaks and ascribed them to factors related to the nanoscale properties of catalysts incorporated into the MEAs. Figure 3 (left) shows the abnormal bipolar bands due to CO from methanol oxidation. The spectra are referenced to the 0 volt spectrum. The peaks around 2080 cm -1 are from linearly bound CO. The upward high-frequency component of the bipolar subtractively normalized bands corresponds to an increase in the IR reflectivity. The positive going bands indicate an increase in the overall reflectivity coincident with an increase in the absorption by adsorbed CO. There are a number of theories in the literature to explain the inverted peak phenomena Incident IR Integrating sphere Exhaust Nafion membrane D 2, D 2 O CaF 2 PtRu Pt CO, D 2 O inlet Exhaust heating pad Figure 2. Schematic of operando IR cell developed by Fan et al. 6

7 Page 8 of 44 Figure 3. Left: Stark tuning of CO/Pt at fuel cell anode. Right: Stark tuning of CO/Pt vs. methanol/d2o ratio. The cell was operated at 50 C. Taken from reference 20 with permission. Figure 3 (right) shows an abrupt decrease in the Stark tuning slopes at the current onset potential of 0.5 volts. The potential at which the Stark tuning slopes change does not depend on the methanol/d 2 O ratio. The Stark tuning curves continue up to 1 volt because the CO is continuously replenished by the steady state flow of methanol/water. In fact, if the slope of the 0.5 methanol/d 2 O ratio did not change prior to the onset potential, all the curves would intersect at 2084 cm -1, 0.5 volts. The Stark tuning rates before and after the transition point are tabulated (Table 1). The discontinuity in the Stark tuning curves correlates with water activation. Ito et al. 25 report that a change in the CO adlayer structure from (2X2) to 9X 9 at 0.4 volts corresponds to a change in the CO-CO distance 7

8 Page 9 of 44 Physical Chemistry Chemical Physics from 5.6 to 12.2 Å. This decreases dipole-dipole coupling, thus reducing the atop CO stretching frequencies. This potential is about where water activation occurs. Co-adsorbed water injects electrons into the Pt d-band, further reducing ν CO values by increased back-donation into the renormalized CO 2π * band. Figure 4 shows the Stark tuning of adsorbed CO on a fuel cell anode prepared with high surface area Ru at the anode and Johnson Matthey Pt black at the cathode. 2 The results have features similar to Fig. 3 although the transition point is shifted negatively by about 300 mv. Above 300 mv, Ru is an excellent water activator. This is why alloying Ru with Pt enhances the bifunctional mechanism. 26 Similar to the data of Figure 3, the co-adsorbed water reduces the Stark tuning slope. The key point of Figs. 3 and 4, and a comparison of the Stark tuning rates of arc-melted alloys versus nanoscale materials, is that the slopes are not simply a property of the adsorbate-substrate identity. The slopes are strongly affected by co-adsorbates, adsorbate coverage, and particle size. 9 There are Stark tuning slope transitions at the potentials where water becomes activated on Pt and Ru at 0.5 volts and 0.3 volts respectively. It is also noteworthy that Stark tuning slopes CO adsorbed on nanoscale catalysts are always substantially lower in comparison to those obtained on single crystal or arc-melted 10, 27 alloys. 8

9 Page 10 of 44 Figure 4. Stark tuning of CO/Ru on a fuel cell anode of pure Ru operating at 50 C. Table 1 CO/Pt Stark tuning slopes vs. Methanol/D 2 O Ratio Methanol/D 2 O < 0.5 V (cm -1 /V) > 0.5 V (cm -1 /V) Operando X-ray absorption spectroscopy Viswananthan 28 and Stoupin 29 introduced of hydrogen and liquid feed direct methanol fuel cell (DMFC) operando X-ray absorption spectroscopy respectively using the cell of Fig. 5. We simply end-milled a rectangular core out of both flow field blocks and endplates. Palladium was used at the cathode to mitigate interference with the absorption edge of the Pt based anode catalysts. The cored cell design has since been used by a number of workers

10 Page 11 of 44 Physical Chemistry Chemical Physics Figure 5. Operando spectroscopy fuel cell and schematic (from ref. 28) A number of operando XAS studies followed that of Viswananthan with many focused on the oxidation state of the metal components. All found that at relevant fuel cell anode operating potentials, platinum is metallic. The oxidation of CO was studied both on Pt 39, and PtRu, 35, 38, 43, 44 confirming the electronic benefits of Ru as a co-catalyst for CO oxidation. Studies of adsorbed oxygen reduction reaction intermediates on fuel cell catalysts show that adsorption is both potential-dependent and site- 35, specific. Figure 6a shows transmission X-ray absorption spectra of Johnson Matthey PtRu black incorporated into a DMFC. In the Pt L III -edge data (left side), the red line shows the operando Pt XANES at 450 mv, while the green and blue lines are the as-received catalyst (mounted on scotch tape), and Pt foil XANES respectively, also measured in transmission. 10

11 Page 12 of 44 µ(e) (normalized) (a) χ(r) (Å -4 ) (b) Energy (ev) Figure 6. a: Pt LIII-edge XANES: In situ catalyst at 450 mv (red), in situ catalyst at 450 mv from another run (orange), Pt foil (blue), and as-received catalyst (green). b: Pt EXAFS fit (red line) to in situ catalyst data at 450 mv (blue) as a totally metallic environment (only Pt and Ru nearest neighbors). Fit range: 1.5 to 3.1 Å. (from ref. 29) R (Å) The increased white line intensity of the alloyed Pt (red) versus the pure Pt (blue) is due to alloyinduced d-band vacancies, 49 as was observed in XANES from an arc-melted PtRu 80:20 alloy. 5 The large edge intensity of the green curve confirms extensive oxidation of the as-received catalyst. The reduction of this heavily oxidized material is part of the conditioning process. Stoupin et al. observed that the potential-dependent (250, 300, 350, 400, and 450 mv) EXAFS perfectly overlap, confirming that, within the DMFC potential window, the Pt signal is insensitive to potential, similarly to the Pt within supported PtRu catalyst in a H 2 /air fuel cell (i.e. Pt is metallic). Figure 6b shows the excellent first shell fit of the 350 mv EXAFS data (blue) with a model fit (red) simulating a totally metallic environment (fit range 1.5 to 3.1 Å) of mixed Pt and Ru atoms. Figure 7 shows the ex situ Ru K-edge 11

12 Page 13 of 44 Physical Chemistry Chemical Physics XANES of metallic Ru powder (blue), operando Ru spectra at 350 mv (red), as-received PtRu (1:1) catalyst (green), Ru oxide (black), and Ru oxide hydrate (pink). µ(e) (normalized) Energy (ev) Figure 7. Ru K-edge XANES: Operating fuel cell anode at 450 mv (red, and orange another similar experiment), metallic Ru powder (blue), Ru oxide hydrate, (pink), Ru oxide (black), and as-received JM PtRu (1:1) (green) standards mounted on scotch tape. (from ref. 29). The coincidence of the catalyst Ru edge at 350 mv (red) with the metallic Ru powder (blue) confirms that within DMFC operating potentials the Ru is primarily metallic. The deviation of the red from the blue line at 22,120 ev reflects the difference between the nearest-neighbor environments of pure Ru (hexagonal) and Ru in a Pt alloy FCC lattice. The as-received catalyst edge (green) is intermediate in energy between the metallic (blue and red) lines and the oxides of Ru (pink and black) confirming that the as-received catalyst is substantially oxidized. The near edge data of the asreceived catalyst is intermediate between the saddle point of the Ru oxide and the metallic catalyst line 12

13 Page 14 of 44 (red), again confirming a substantially oxidized as-received catalyst. As mentioned before, the final step of catalyst synthesis is within the operating fuel cell , 50 A number of operando x-ray absorption studies have followed with many focused on the oxidation state of the metal components. All found that at relevant fuel cell anode operating potentials, platinum is metallic. The oxidation of CO was studied both on Pt 39, 30 and PtRu 35,38, confirming the electronic benefits of Ru as a co-catalyst for CO oxidation. Studies of adsorbed oxygen reduction reaction intermediates on fuel cell catalysts show that adsorption is both potential-dependent and site- 35, specific. 3. Operando IR-XAS Fuel Cell. 3.1 Cell Design. The Viswananthan cell was designed for transmission X-ray absorption measurements. Fluorescence measurements provide better signal-to-noise because interference due to cell components is minimized. The most recent cell design (Fig. 8) combines features of the Viswananthan cell 28 with the specular reflectance infrared cell reported by Fan et al. 18 The IR-XAS cell enables operando XAS (in transmission and fluorescence) and specular reflectance FTIR 20, 51, 52 spectroscopy. The top flow field accommodates a CaF 2 window for IR reflectance studies. A pin-style upper flow field optimizes flow distribution around the CaF 2 window inset. The CaF 2 window can be removed when the working electrode of interest is an air-breathing electrode. 13

14 Page 15 of 44 Physical Chemistry Chemical Physics Figure 8. Left: IR-XAS cell with DE9 connector for electrodes, heater cartridge and thermister. Right: 1) X-ray transmission, 2) Reflectance IR or fluorescence X-ray, 3) CaF 2 window housing, 4) Teflon gasket, 5) gas outlet insert, 6) slider assembly, 7) thermocouple/heater cartridge port, 8) lower flow field, 9) MEA, 10) upper flow field, 11) top plate. The IR-XAS cell slides into a Pike Technologies (Watertown, WI) diffuse reflectance accessory (with a modified front plate) that is accommodated by most commercial FTIR instruments (Fig. 9). 14

15 Page 16 of 44 Figure 9. Cell installed in Pike Technologies (Watertown, WI) diffuse reflectance accessory in a Bruker Optics (Billerica, MA) Vertex 70 FTIR Spectrometer. A slot under the slider assembly ensures precise positioning of the cell under the Diffuse-IR accessory integrating mirrors. Figure 10 shows background CV (solid) and the CO stripping wave (dotted) obtained at 50 o C. The CO stripping wave (10 mv/sec) extends from mv. The potential dependent spectra of CO adsorbed at 100 mv vs. NHE are shown in Figure Current (ma) Figure 10. Cyclic voltammetry (10mV/s) of fuel cell working electrode (5-cm 2 geometric) under humidified nitrogen at 50 C (solid). Humidified H 2 (50 sccm) at the counter-reference. CO stripping wave (dashed). 3.2 Infrared spectroscopy Potential (mv) Competitive adsorption of CO onto Pt The corresponding Stark tuning plots for CO adsorbed at 100, 200, 300 and 400 mv (Fig. 12). show the effects of co-adsorption on the CO stretching frequencies. These results are similar to those of 15

16 Page 17 of 44 Physical Chemistry Chemical Physics Stamenkovic et al. studying CO on Pt(111) in 0.5 M H 2 SO 4. 8 Stamenkovic correlated the complex Stark tuning of CO ads stretching frequencies (ν CO ) to the compression/dissipation of CO ads islands: A linear Stark tuning region with a subtle ν CO blue shift from the extrapolated linear region was followed by a precipitous drop in ν CO, followed by an upturn. The potential where ν CO precipitously drops (E onset ) correlates with CO ads oxidation. Stamenkovic identified the co-adsorbate as HSO - 3 originating from the 0.5 M H 2 SO 4 electrolyte used in their study. Kendrick, using no supplemental electrolyte (i.e., operando conditions), identified co-adsorbates as the Nafion sulfonate exchange group and the side chain CF 3 group. CO ads oxidation, induced by OH - adsorption, diminishes dipole-dipole coupling and thus precipitously decreases ν CO. The upturn is ascribed to increased adsorption of sulfonate species relative to OH -, which reestablishes repulsive dipole interactions that compress CO ads islands and increases ν CO. The Stark tuning rates of this work, 6.8 ± 0.6 cm -1 /mv, are within the range of 2.5 cm -1 /V to 18.3 cm -1 /V of previous operando Stark tuning studies of direct methanol fuel cells at potentials negative of 0.5 V vs. NHE

17 Page 18 of mv 450 mv 400 mv 350 mv 300 mv 250 mv 200 mv 150 mv 100 mv Figure 11. Potential-dependent spectra of CO on Pt; Adsorption potential of 100 mv operating at 50 C Wavenumber (cm -1 ) 17

18 Page 19 of 44 Physical Chemistry Chemical Physics Wavenumber (cm -1 ) Wavenumber (cm -1 ) mv mv mv Potential (mv) Wavenumber (cm -1 ) Wavenumber (cm -1 ) mv Potential (mv) Potential (mv) Potential (mv) Figure 12. Stark tuning plots of linearly adsorbed CO adsorbed on Pt at 50 C. Figure 13 shows how E onset dependends on E ads. At 400 mv, E onset and E ads coincide because at higher potentials CO ads favors adsorption sites with higher enthalpies of adsorption (e.g., steps and kinks) 53, 54 and there is no thermodynamic force driving migration to sites of lower H ads. At an E ads of 300 mv, the E onset exceeds E ads by 60 mv because the distribution of adsorption sites are more heterogeneous and there is a driving force for CO ads on terraces to migrate to steps, kinks and ad atoms. Thus the difference between E onset on E ads would be expected to increase as E ads is decreased. At E ads of 200 mv and lower, the E onset is leveled to 320 mv 18

19 Page 20 of 44 Figure 13. CO electro-oxidation onset potentials versus adsorption potentials 19

20 Page 21 of 44 Physical Chemistry Chemical Physics Stark tuning plots of CO / Pt obtained at 30 C, 50 C and 70 C (Fig. 14) are similar to those of figure 12. Wavenumber (cm -1 ) Wavenumber (cm -1 ) Wavenumber (cm -1 ) Figure 14. Operando CO/Pt Stark tuning Potential (mv) C 50 C 70 C Potential (mv) Potential (mv) 20

21 Page 22 of 44 Figure 15 shows that E onset decreases linearly (slope = cm -1 /K) with temperature. The relationship of this linear variation to the kinetics of the inner sphere processes 55 will be addressed in future work Slope: R 2 : Potential (mv) Temperature ( C) Figure 15. The potential at which CO begins to oxidize as a function of temperature. Figures 12 and 14 show an adsorption process despite the lack of mobile ions typical of aqueous sulfuric acid. The operando spectroscopy suggests a need for elucidation of Nafion functional groups responsible for the complex Stark tuning plots Mechanically coupled internal coordinates of ionomer vibrational modes. Understanding the competitive adsorption processes, elucidated by Stark tuning plots, motivated an 21

22 Page 23 of 44 Physical Chemistry Chemical Physics FTIR analysis of the Nafion/Pt interface. Such a study required a reliable set of Nafion infrared band assignments. A brief review of the literature showed that the assignments of key infrared bands were incorrect. Nafion and relevant derivatives are schematized (Scheme 1). Scheme 1: Structure of Nafion and its derivatives A discussion of selected spectra (Fig. 16) highlights the problematic assignments. The attenuated total reflectance (ATR) spectra of hydrated Nafion (a) and the short-side-chain ionomer (b) focus on the ~1060 cm -1 and a multiplet that includes a shoulder at 995 cm -1 and two peaks ~983 and ~970 cm - 1, hereafter referred to as and, respectively. The ν hf and ν lf have been conventionally assigned to ether groups in proximity to the backbone and the sulfonate group, respectively. Cable et al. 56 associated to the ether linkage closest to the sulfonate group because of its enhanced sensitivity to ion exchange and the fact that the persists in the Dow short-side chain ionomer spectrum. The short-side chain ionomer has only one ether group, positioned adjacent to the sulfonate group. The sulfonyl fluoride precursor was also compared to Nafion. In the sulfonyl fluoride spectrum (c), and the ~1060 cm -1 are absent. These observations were reconciled by invoking solvation effects as responsible for the sensitivity of to ion exchange: In hydrated Nafion, the sulfonate group is 22

23 Page 24 of 44 embedded in an aqueous phase. The assumption has been that the ether group closest to the sulfonate group is subject to solvation as well and thus sensitive to ion exchange. The, which is essentially insensitive to ion exchange, was attributed to the ether link distant from the sulfonate group. These assumptions were used to rationalize the observation that the 1060 cm -1 and peaks shifted together, and were assigned to -SO 3 - and ether link modes, respectively. Byun et al. 57 also reported the same loss of the upon substitution of the sulfonic acid group for a sulfonyl imide (spectrum f), and assigned as did Cable et al. (Fig. 16e,f). Figure 16. (a) Nafion-H and (b) short-side-chain ionomer (c) sulfonyl fluoride and (d) Nafion-H, (e) H + form of Nafion and (f) sulfonyl imide. Spectra a-d from Cable et al. 56, spectra e and adapted from Byun et al. 57 Webber et al. explain why the assignment of bands to a single functional group (e.g., 1060 cm -1 and ) as has been done for several decades precludes proper analysis of the spectra. 58 The consideration of the mechanical coupling of the internal coordinates of the -SO 3 - and its near-neighbor COC 23

24 Page 25 of 44 Physical Chemistry Chemical Physics provides an explanation of the concurrent shifts of ~1060 cm -1 and peaks without the need for invoking solvation of the ether link adjacent to the sulfonate group. Transmission infrared spectra of Nafion 112 were obtained on a Bruker Vertex 80V Spectrometer (Bruker Optics Inc, Billerica, MA) under dry air or vacuum. The transmission spectra (Fig. 17) show a concurrent loss of intensity of 1062 cm -1 and due to dehydration of the membrane, simultaneous with evolution of peaks at 1415 and 908 cm -1. Figure 17. Transmission IR spectra of Nafion 112 showing the evolution of 1415 cm -1 and 908 cm -1 bands upon dehydration. We attribute the transition of the dehydrated (red) to the hydrated (blue) spectrum to a change in the point group symmetry of the sulfonic acid group. The following density functional theory (DFT) calculations show that as the proton dissociates from the sulfonic acid group (e.g., with hydration), the local point group symmetry changes from C 1 to C 3V. Unrestricted DFT 59, 60 with the hybrid X3LYP 61 functional was used for geometry optimization and calculations of the normal mode frequencies and corresponding IR spectra of triflic acid, the Nafion 24

25 Page 26 of 44 side chain, and the entire 55-atom Nafion repeat unit (Fig. 18: left). The DFT calculated spectrum of the repeat unit provides 159 normal mode frequencies and intensities. Figure 19 shows the theoretically derived peak positions and intensities (black lines) superimposed upon the attenuated total reflectance (ATR) spectrum (red line) of hydrated Nafion. Figure 18. Left: Nafion repeat unit (Red= O, White= H, Grey= C, Purple= S, Green= F). Right: Segment and atom labeling for Nafion 25

26 Page 27 of 44 Physical Chemistry Chemical Physics Figure 19. DFT calculated normal modes (black lines) and Nafion ATR spectrum (red). Although hydrated Nafion has C 1 symmetry overall, it has regions of local symmetry namely the - SO 3 (C 3V ) and the ether groups (C 2V ). Maestro (Schrodinger Inc., Portland, OR) converts Jaguar output files to vibrational mode animations. The full animations of selected calculated modes are in the Supporting Information of Kendrick et al. 9 - Snapshots of the CF 3 SO 3 symmetric stretch (ν s (A 1 )) and the CF 3 OCF 3 asymmetric (ν as (B 2 )) and rocking modes (ρ r (B 2 )) and associated frequencies are shown (Fig. 20, top row). Hereafter the ν s (A 1 ), ν as (B 2 ), and ρ r (B 2 ) modes are referred to as pure modes. The equilibrium positions and vibrational mode extrema of the repeat unit modes corresponding to the bands at 969 cm - 1 ( ) and 1062 cm -1 (Fig. 17) are shown in Figure 20. The animations enable visualization of how the pure mode internal coordinates mechanically couple to yield repeat unit modes. The center row snapshots show a 984 cm -1 mode that results from the coupling of the ν s (A 1 ), ν as (B 2 ), and ρ r (B 2 ) with the dominant mode being the ν s (A 1 ). The full animations show that the dominant pure mode of the 26

27 Page 28 of cm -1 peak is actually the CF 3 OCF 3 ν (as) (B 2 ) mode with a much weaker contribution from the ν s (A 1 ) of triflic acid. The 1059 cm -1 is primarily a ν as (B 2 ) mode mechanically coupled to the internal - coordinates of the ν s (A 1 ) of the -SO 3 group. The key point is that the ether link nearest the exchange group has internal coordinates that are mechanically coupled to the ν s (A 1 ) mode: The 1062 cm -1 and peaks cannot be purely ascribed to the -SO - 3 and COC modes, respectively. Figure 20. Maestro animation snapshots of DFT calculated modes of the full side chain and backbone: Top row: Small molecule pure modes; Middle row: 984 cm -1 ; equilibrium positions in center panel. Bottom row: 1059 cm -1 ; equilibrium positions in center panel. Extrema at left and right panels of center panels. In fact,, conventionally assigned as an ether mode, derives from the triflic acid - SO - 3 ν s (A 1 ) mode with a calculated average of 974 cm

28 Page 29 of 44 Physical Chemistry Chemical Physics The above analysis obviates the need to invoke ether link solvation for analysis of the Figure 16 spectra. The, and the 1060 cm -1 shift concurrently with ion exchange and/or dehydration because the internal coordinates of the ~1060 cm -1 and peaks are mechanically coupled. Consider the spectra of Figure 16 in the context of mechanically coupled coordinates. The is not in the shortside-chain spectra (b) because was backbone the ether link that was not mechanically coupled to the -SO - 3. In the short-side-chain derivative, the remaining is now the backbone ether link mechanically coupled to the -SO - 3 group (i.e., the functional groups responsible for the mechanically coupled ν s (A 1 ), ν as (B 2 ), and ρ r (B 2 ) are now in close proximity to the backbone). In the sulfonyl fluoride spectra (c), the C 3V symmetry of the -SO - 3 is lost and both the 1060 and the vanish (similar to the effect of dehydration). The sulfonyl imide spectra (f) behaves similarly to that of the sulfonyl fluoride. The peak at 1069 cm -1 was confirmed by Korzeniewski to be an asymmetric S-N-S stretch. The ~1060 cm -1 and peaks result from the mechanical coupling of the internal coordinates of - - SO 3 and the COC pure modes. The 1059 cm -1 mode is dominated by an ether link mode. The calculated mode at 984 cm -1 -, a major contributor to the peak, is dominated by the -SO 3 ν s (A 1 ) mode. The consideration of mechanically coupled internal coordinates is essential for the analysis of infrared spectra of ionomers and their interfaces Elucidating the ionomer metal interface: The aggregate of operando spectroscopy, attenuated total reflectance spectroscopy of Nafion 117, PM-IRRAS of Nafion spin-coated onto Pt and DFT calculated Nafion spectra suggest a model for the Pt-Nafion interface the includes the Nafion-CF 3 group as an important co-adsorbate at the ionomer-pt interface. 28

29 Page 30 of 44 PM-IRRAS enhances (relative to the ATR) vibrational modes of functional groups ordered by the Pt surface. Figure 21 shows the ATR spectrum (red), the PM-IRRAS spectra of Nafion-H/Pt interface (grey line), Nafion-Li/Pt interface of Li + exchanged Nafion (blue line) and 6 selected (from the 159 calculated) DFT calculated frequencies and intensities. Figure 21. Theoretical and experimental spectra. ATR of hydrated Nafion (red); PM-IRRAS of Nafion-H on Pt (grey); PM-IRRAS of Nafion-Li on Pt (blue); Selected DFT peaks (black lines 1-6). Figure 18 (right) shows the Nafion structure with functional groups labeled for ease of discussion. The low-frequency ATR band (Fig. 21) at 971 cm -1 (corresponding to theoretical 984 cm -1 ; line-1) and the 1056 cm -1 band (corresponding to theoretical 1059 cm -1 ; line-3) have recently been thoroughly assigned by Webber et al. 58 Animations of the DFT calculated internal coordinates reveal that the observed 1056 cm -1 and 971 cm -1 peaks both have internal coordinates resulting from the mechanical coupling of the adjacent the sulfonate and COC (A) ether link. 58 Thus these peaks shift concertedly with changes in the sulfonate environment. Consider the ATR and PM-IRRAS spectra of the protonic form of Nafion (Fig. 21, grey line). The 1056 cm -1 and 971 cm -1 peaks concertedly shift to higher 29

30 Page 31 of 44 Physical Chemistry Chemical Physics frequencies in the PM-IRRAS because of the interaction of the sulfonate functional group with the Pt surface. A similar effect is observed with Li + exchange of the adsorbed Nafion (blue line). A convention for correlating PM-IRRAS enhanced peaks to the calculated DFT peaks would enable identification of functional groups ordered by the Pt surface: The association of observed PM-IRRAS 58, 62 peaks with DFT peaks, assigned by visualization of mechanically coupled internal coordinates, provides the basis for such a convention. Normal mode coordinate animations (generated by Maestro from DFT output files) explicitly show how neighbor functional groups (called out in figure 18) are mechanically coupled. The calculated internal coordinates are viewed in the context of calculated normal modes of relevant small molecules (e.g., triflic acid, CF 3 OCF 3, 10-carbon CF 2 backbone, etc.) referred to as pure modes as discussed in the previous section. The pure modes serve as the basiselements for assigning DFT calculated normal modes associated with observed peaks. Figure 22 shows the assignments of the 6 selected DFT peaks and snapshots of the corresponding Maestro animations. The atoms contributing to the dominating motion (black circles) and the next most significant atom motions (dotted circles) comprise pure modes that form the basis for the assignments. An alternate strategy for determining the dominant mode is to consider the contribution to the potential energy surface on an atom by atom basis. 63 While this may change the selection of the dominant mode, it does not alter what pure modes contribute to the assignments. The correlation of the DFT to PM-IRRAS peaks (Fig. 21) and the resulting assignments in terms of the mechanically coupled modes are tabulated in Table 2. 30

31 Page 32 of 44 Figure 22. Normal mode coordinate animation snapshots of the Nafion side-chain anion and backbone fragment. Left and right views are extrema positions of the vibrational mode. Functional groups associated with the dominant internal coordinates and next most significant motions are designated by solid and dotted boundary lines respectively. 31

32 Page 33 of 44 Physical Chemistry Chemical Physics Wavenumber (cm -1 ) PM-IRRAS DFT Pure Mode Components SO 3- ν s + COC(A) ν as + COC(B)ρ r CF 2 ω (BB def ) + COC(B)δ s COC(A) ν as + SO 3- ν s CF 2 δ s (BB stre ) + CF 2 (BB def )ρ r + COC(A)ω CF 3 ν as + COC(A)δ s + COC(B)δ s CF 2 δ s (BB def ) Symmetric stretch, ν s ; Asymmetric stretch, ν as Wagging, ω; Scissoring, δ s ; Twisting, τ; Rocking, ρ r Backbone deformation, BB def ; Side-chain deformation, SC def ; Backbone Stretching, BB stre Table 2. PM-IRRAS and DFT IR adsorption peaks and assignments. The pure mode peak assignments (Table 2) elucidate functional groups ordered by the Pt surface. Animations of the pure modes and the internal coordinates of the 6 selected peaks are Supporting Information as.avi files of Kendrick et al. 9 The rational for the key functional group assignments (Table 2) is supported by the overlap of the DFT calculated peak positions with the PM-IRRAS peaks. Consider the DFT and PM-IRRAS peaks in the contexts of the bulk-nafion ATR and the report by Cable et al. 56 that the 1056 cm -1 and 971 cm -1 peaks shift with alterations of the sulfonate group environment. The bulk ATR peak at 1056 cm -1 (red), the PM-IRRAS peak of protonated Nafion adsorbed on Pt (grey line) at 1061 cm -1 and the PM-IRRAS peak of lithiated Nafion adsorbed on Pt (blue line) at 1077 cm -1 (Fig. 21) confirm that Pt surface atoms induce frequency shifts, as do extentof-hydration 58 and ion exchange of Nafion. 56 Thus the PM-IRRAS enhances bulk-nafion-modes that are shifted due to functional group interactions with Pt. Less explicit than the 1056 cm -1 peak, are 32

33 Page 34 of 44 PM-IRRAS peaks derived from bulk-nafion-modes that are convoluted within the Nafion ATR broad envelop region ( cm -1 ), in particular the 1164 and 1260 cm -1 PM-IRRAS peaks. The animation of the theoretical peak at 1254 cm -1 (Fig. 21, line-5), associated with PM-IRRAS peaks at 1260 cm -1 (blue and grey lines), suggests that the -CF 3 internal coordinates dominate that normal mode. The insensitivity of the 1201 cm -1 peak, to ion exchange, suggests that the internal coordinates are essentially uncoupled to the sulfonate group. The 1260 cm -1 intensity is over 10-fold greater than that the cluster of peaks (i.e., associated with theoretical lines 1 and 2) that are mechanically coupled to the sulfonate pure mode: The -CF 3 is a co-adsorbate of comparable importance to the sulfonate group in the self-assembly of Nafion onto Pt. This is supported by the repeat-unit-atom atomic charges obtained by Mulliken population 64 analysis. Table 3 shows the average charges of the backbone, side chain, -CF 3 fluorine atoms and the sulfonate oxygen atoms. The charges for chemically equivalent atoms (e.g.,-cf 3 fluorine and sulfonate oxygen atoms) differ because the calculations correspond to the lowest energy Newman projections where the atomic environments are different for chemically equivalent atoms because of the absence of symmetry in the full molecule. The chemically equivalent atoms have smaller charge standard deviations as would be expected. The average charge of the -CF 3 fluorine atoms are about 18% that of the sulfonate oxygens. Segment Backbone (F) Side-chain (F) CF 3 (F) Sulfonate (O) 13 atoms 8 atoms 3 atoms 3 atoms Avg. Partial Charge Standard Deviation

34 Page 35 of 44 Physical Chemistry Chemical Physics Table 3. Average partial charges of selected Nafion segments Gaussian 03 Viewer (Gaussian, Wallingford, CT) was used to construct a flexible (i.e., enables dihedral angle rotation while maintaining the native functional group bond angles) 2-repeat unit model. The -CF 3 and -SO - 3 groups, oriented with planes defined by their fluorine and oxygen atoms parallel to a Pt surface, effect ordering of the -CF 2 - backbone segments with respect to the Pt surface (Fig. 23). Thus, peaks due to the backbone -CF 2 - would be expected in the PM-IRRAS: The peak at 1164 cm -1 is associated with the theoretical peak (line-4) at 1168 cm -1. The line-4 animation shows that -CF 2 - backbone internal coordinates dominate the 1168 cm -1 mode, supporting the suggestion of ordered -CF 2 - groups. The numbers (yellow) associate DFT calculated IR peaks (line-1 6, Fig. 21) - and associated PM-IRRAS peaks with regions of order induced by the -CF 3 and -SO 3 functional group adsorbates. 2,4, Figure 23. Gaussian 03 Viewer Nafion-Pt interface model. Oxygen (red), Sulfur (yellow), Fluorine (light blue), Carbon (grey), Pt (dark blue). 34

35 Page 36 of 44 Thus the aggregate of the Stark tuning data of Figure 12 and 14, the PM-IRRAS and DFT calculations support Figure 23 as a model for the self-assembly of Nafion onto Pt. The details of exactly how adsorbed-cf 3 functional groups influence the operando Stark tuning curves are not yet established. The low density of functional group adsorption sites, relative to the number of backbone 6, 65 -CF 2 - groups suggests an explanation as to why Nafion is observed to enhance electrode processes. The anchoring functional groups represent about 25% of the footprint of the adsorption model. Neutron reflectivity experiments may provide information concerning water within free volume between the metal surface and the Nafion backbone X-ray Absorption Spectroscopy. Figure 24 shows the cell as an air breathing fuel cell at the MRCAT beam line at Argonne National Laboratory Advanced Photon Source. The source beam (arrow) passes through the cathode gas diffusion layer via the beveled (158º) cell top plate. Fluorescence is detected by the Lytle detector. Figure 24. IR-XAS cell at MRCAT beam line, Argonne National Laboratory. Figure 25 shows an IR-XAS cell polarization curve (40 mv/min). Humidified H 2 (50 sccm) and air 35

36 Page 37 of 44 Physical Chemistry Chemical Physics (250 sccm) were delivered to the counter and working electrode respectively at 50 C Potential (mv) Current (A) Figure 25. IR-XAS fuel cell polarization curve. Figure 26 shows the subtractively normalized operando X-ray absorption near edge spectra (XANES) of the as-prepared membrane electrode assembly (dry MEA) prepared from carbon supported Johnson Matthey Pt before installation into the fuel cell, and versus potential in the IR-XAS fuel cell operating at 50 o C. The cathode gas-diffusion-layer was exposed to the beam-line hutch air while humidified hydrogen was delivered to the Pd counter-reference electrode. After a one-hour conditioning period 67 the reference XANES (to which other spectra were normalized) was obtained at 0 volts vs. NHE. The ex-situ dry MEA spectrum shows that the Pt is highly oxidized before exposure to the fuel cell environment, consistent with the results of Stoupin et al. 29 At open circuit voltage, the white line intensity is still substantially more intense than spectra at 900 mv. 36

37 Page 38 of 44 Figure 26. Pt LIII-edge XANES of a Pt air-breathing cathode of the IR-XAS cell subtractively normalized to XANES at 0 volts vs. Pd hydrogen anode. Left: time dependent data at 530 mv; Right: potential dependent steady state data. Figure 26-left shows the time-dependent Pt XANES (at 530 mv), each obtained as an average of three consecutive scans. The two peaks at 11553eV and 11578eV decrease with time as Pt is reduced. The large time constant (on the order of hours) cannot be attributed solely to reduction of chemisorbed oxygen. The reduction of sub-surface oxygen, native to the crystallite core structure, may be responsible for the long time-constant. Stoupin et al. have shown that in the case of PtRu, the oxide phase extends deep into the particle core. 29 A large Pt restructuring time-constant explains the better 37

38 Page 39 of 44 Physical Chemistry Chemical Physics performance obtained when acquiring cathode polarization curves from low to high potentials. 68 It also explains the hysteresis often observed over a fuel cell voltammetric cycle. Figure 27 shows the Ni-edge time-dependent data, using the same experimental conditions as those on the left, from a PtNi/C cathode catalyst. This PtNi (1:1) was provided by ETEK (Somerset, NJ). Jia et al. performed EXAFS fits of the same data, 69 and attributed the shorter time constant for the Ni edge transient to the predominance of the Ni within the metal core: The surface, dominated by Pt, has only small amounts of Ni available for adsorption/reduction of chemisorbed oxygen. The de-alloying of Pt alloy catalyst is thoroughly discussed by Strasser and co-workers. 70 The data of Figure 26 and 27 suggest that a broad window of times constants is required for understanding time-dependent phenomena. The time constants associated with the XAS of this work range from minutes to hours. Tada et al., 71 using transmission studies done on a Viswananthan-type cored cell reported a method for obtaining 1s time-resolved full EXAFS spectra from a fuel cell being cycled between 0.4 and 1.0 V. They have observed separate time constants for charging and discharging, as well as the formation and dissociation of surface Pt-O bonds. Even shorter time constants will be possible using dispersive EXAFS techniques and lifetime studies using catalysts kept in operation for hundreds of hours will become possible as operando fuel cells become more available. 38

39 Page 40 of 44 Figure 27. Time-dependent XANES at 530 mv of the Ni edge of an air breathing PtNi cathode catalyst subtractively normalized to 0 volts. 4. Conclusion The salient features of the operando IR and X-ray spectroscopy cells of Fan and Viswananthan respectively have been integrated into the design of an IR-XAS cell that enables acquisition of X-ray absorption, and IR reflectance spectra of fuel cell catalysts and adsorbates with flowing fuel and oxidant streams at controlled temperatures and potentials. We obtained complex frequency-tuning rates for CO on Johnson Matthey Pt catalysts incorporated into an operating membrane electrode assembly as a function of fuel cell operating temperature and the potential at which the CO was dosed into the fuel cell. Similar studies by Stamenkovic et al. of CO on Pt(111) suggest that the complex tuning is due to compression/dissipation of CO islands by competitive adsorption for electrolyte anions and reactive OH. The absence of mobile electrolyte anions in an operating fuel cell suggest that 39

40 Page 41 of 44 Physical Chemistry Chemical Physics the Nafion sulfonate exchange groups plays a similar role as did the sulfonate of the Stamenkovic study. The PM-IRRAS of Nafion-Pt interfaces, and ATR spectroscopy of Nafion, correlated with DFT calculated normal mode frequencies assigned by visualization of normal mode animations suggest that the Nafion sulfonate and -CF3 groups co-adsorb during the self-assembly of Nafion onto Pt. The FTIR analysis required development of an IR band assignment method that uses normal mode animations for peak assignments. The general strategy for linking theoretical peaks to experimental bands is to systematically alter the exchange group environment and analyze the spectra in terms of mechanically coupled internal coordinates. Time-dependent subtractively normalized XANES show that potential dependent Pt restructuring can have time constants on the order of hours. The time constants for metals that are partially de-alloyed from the catalysts (e.g., Ni in PtNi) are shorter. Work is in progress to install a benchtop IR source and detector at the MRCAT beamline to enable simultaneous FTIR-X-ray studies of operating fuel cells. References 1. H. Topsoe, Journal of Catalysis, 2003, 216, E. S. Smotkin, in In-Situ Spectroscopic Studies of Adsorption at the Electrode and Electrocatalysis, eds. S.-G. Sun, P. A. Christensen and A. Wieckowski, Elsevier, Oxford, UK, 1st edn., 2007, pp S. Gottesfeld and T. A. Zawodzinski, in Polymer Electrolyte Fuel Cells, eds. R. C. Alkire and D. M. K. a. C. W. T. H. Gerischer, Wiley-VCH, 1997, vol. 5, pp L. Liu, C. Pu, B. Viswanathan, Q. Fan, R. Liu and E. S. Smotkin, Electrochimica Acta, 1998, 43, R. Viswanathan, G. Hou, R. Liu, S. R. Bare, F. Modica, G. Mickelson, C. U. Segre, N. Leyarovska and E. S. Smotkin, Journal of Physical Chemistry B, 2002, 106, L. Liu, R. Viswanathan, R. X. Liu and E. S. Smotkin, Electrochem. Solid State Lett., 1998, 1,

41 Page 42 of L. Ploense, M. Salazar, B. Gurau and E. S. Smotkin, Solid State Ionics, 2000, , V. Stamenkovic, K. C. Chou, G. A. Somorjai, P. N. Ross and N. M. Markovic, Journal of Physical Chemistry B, 2005, 109, I. Kendrick, D. Kumari, A. Yakaboski, N. Dimakis and E. S. Smotkin, J. Am. Chem. Soc., 2010, 132, E. A. Lewis, I. Kendrick, Q. Jia, C. Grice, C. U. Segre and E. S. Smotkin, Electrochimica Acta. 11. The first two digits are the equivalent weight of Nafion divided by 100; the last digit is the thickness of Nafion in thousandths of an inch. 12. H. Rivera, J. S. Lawton, D. E. Budil and E. S. Smotkin, Journal of Physical Chemistry B, 2008, 112, B. Gurau and E. Smotkin, J. Power Sources, 2002, 112, A. Bard and L. Faulkner, Electrochemical Methods: Fundamentals and Applications, John WIley & Sons, A. J. Bard, Electrochemical methods : fundamentals and applications / Allen J. Bard, Larry R. Faulkner, Wiley, New York :, G. A. Somorjai, Introduction to Surface Chemistry and Catalysis, John Wiley & Sons, Inc., New York, NY, Q. Fan, C. Pu, K. L. Lay and E. S. Smotkin, Journal of the Electrochemical Society, 1996, 143, L21-L Q. B. Fan, C. Pu and E. S. Smotkin, Journal Of The Electrochemical Society, 1996, 143, S. C. Chang and M. J. Weaver, Journal of Physical Chemistry, 1991, 95, A. L. Bo, S. Sanicharane, B. Sompalli, Q. B. Fan, B. Gurau, R. X. Liu and E. S. Smotkin, Journal Of Physical Chemistry B, 2000, 104, R. Ortiz, A. Cuesta, O. P. Marquez, J. Marquez, J. A. Mendez and C. Gutierrez, Journal of Electroanalytical Chemistry, 1999, 465, T. Iwasita-Vielstich, in Advances in Electrochemical Science and Engineering, Wiley-VCH Verlag GmbH, 2008, pp P. A. Christensen, A. Hamnett and S. A. Weeks, Journal of Electroanalytical Chemistry and Interfacial Electrochemistry, 1988, 250, S.-G. Sun, Journal of Electroanalytical Chemistry, 2002, 529, K. Yoshimi, M.-B. Song and M. Ito, Surface Science, 1996, 368, M. Watanabe and S. Motoo, Journal of Electroanalytical Chemistry and Interfacial Electrochemistry, 1975, 60, L. Liu, R. Viswanathan, R. Liu and E. S. Smotkin, Electrochemical and Solid-State Letters, 1998, 1, R. Viswanathan, R. Liu and E. S. Smotkin, Review of Scientific Instruments, 2002, 73, S. Stoupin, E.-H. Chung, S. Chattopadhyay, C. U. Segre and E. S. Smotkin, Journal of Physical Chemistry B, 2006, 110, A. E. Russell and A. Rose, Chemical Reviews, 2004, 104,