Ts

ND

a

ARRAA

KOXIR

1

1

prao

smeOc

H

O

p

h0

Catalysis Today 283 (2017) 11–26

Contents lists available at ScienceDirect

Catalysis Today

j our na l ho me page: www.elsev ier .com/ locate /ca t tod

wenty years of operando IR, X-ray absorption, and Ramanpectroscopy: Direct methanol and hydrogen fuel cells

eili Loupe, Jonathan Doan, Eugene S. Smotkin ∗

epartment of Chemistry and Chemical Biology, Northeastern University, Boston MA 02115, United States

r t i c l e i n f o

rticle history:eceived 14 December 2015eceived in revised form 1 June 2016ccepted 4 June 2016vailable online 20 June 2016

eywords:perando spectroscopy-ray absorption spectroscopy

a b s t r a c t

Mixed metal catalysts in polymer electrolyte fuel cell membrane electrode assemblies were character-ized by operando infrared, X-ray absorption and Raman spectroscopy. Windowed direct methanol andhydrogen-air fuel cell assemblies allow for spectral acquisition with controlled potential, temperatureand flowing reactant streams to the graphite flow fields. Johnson Matthey Pt and PtRu (1:1), in-houseprepared PtRu (1:1 and 3:1), PtRuOs (65:25:10), PtNi (1:1), and a melamine-synthesized FeNC catalystwere studied. The near invariant core structure of phase segregated alloy catalysts over practical fuelcell potentials enable a simultaneous fit of metal component EXAFS to quantitatively characterize corestructures: JM PtRu is a face centered cubic lattice with 50% of the Ru in an amorphous phase. Stark tun-

nfrared spectroscopyaman spectroscopy

ing of COads as a reaction intermediate and as a surface structure probe elucidates potential dependentco-adsorption of Nafion functional groups on Pt. Operando Raman spectroscopic tracking of membranehydration at a fuel cathode is facilitated by symmetry-based group mode assignments of Nafion IR bands.Fully hydrated Nafion sulfonate groups have C3V local symmetry. Dehydrated sulfonic acid groups haveC1 local symmetry. C3V and C1 IR bands coexist at intermediate states of membrane hydration.

© 2016 Elsevier B.V. All rights reserved.

. Introduction

.1. Polymer electrolyte membrane fuel cells

Polymer electrolyte membrane (PEM) fuel cells are ideal forortable, traction and stationary power. This perspective summa-izes our studies, since 1996, of direct methanol fuel cell (DMFC)nd hydrogen-air fuel cell electrode layers, with an emphasis onperando spectroscopy [1–14].

Fig. 1 schematizes a fuel cell and the electrocatalytic layertructure. The membrane electrode assembly (MEA) is a Nafionembrane sandwiched between anode and cathode catalytic lay-

rs. A fuel (e.g., MeOH or H2) is oxidized at the anode and air (or2) is reduced at the cathode. The hydrogen-air fuel cell anode andathode half reactions are:

2 → 2H+ + 2e− (1)

− +

2 + 4e + 4H → 2H2O (2)

Catalytic layers are formed by deposition of catalyst inks uponorous carbon paper/cloth gas diffusion layer (GDL) surfaces. The

∗ Corresponding author.E-mail address: e.smotkin@neu.edu (E.S. Smotkin).

ttp://dx.doi.org/10.1016/j.cattod.2016.06.012920-5861/© 2016 Elsevier B.V. All rights reserved.

GDLs are both current collectors and diffusers for gaseous or reac-tants between the flow fields and catalytic layers. Teflon dispersionin the cathode layer rejects water from the oxygen reduction reac-tion (ORR). The solubilized Nafion component of catalyst inks [15]imparts proton conductivity throughout the catalytic layer. Cat-alyzed GDLs are hot pressed onto Nafion to yield a 5-layer MEA.Catalyst inks can also be directly applied to the membrane to yielda catalyst coated membrane (CCM). MEA fabrication methods arereviewed by Gottesfeld [16].

Ideally, PEM fuel cells will tolerate direct oxidation of liquidfuels (e.g., MeOH) and/or H2 derived from hydrocarbon reforma-tion (with typically 50–100 ppm CO). In either case, adsorbed CO(COads) blocks active catalytic sites for MeOH or H2 oxidation. Pthas been the standard anode catalyst with the best performanceof all pure metals [18]. The best-known CO tolerant catalysts arePt-based mixed metal catalysts [3].

1.2. DMFC and hydrogen-air fuel cell catalysis

Condensed MeOH has five times the energy density of com-pressed H2 [19]. Unlike H2 fuel cells, DMFCs can leverage anexisting liquid fuel infrastructure. Improved catalysts for the MeOHhalf-reaction (Eq. (3)) would obviate the need for the reformer,

12 N. Loupe et al. / Catalysis Today 283 (2017) 11–26

brane electrode assembly. Inset shows cathode catalytic layer structure.A

wf

C

oGonc

P

P

P

P

P

M

P

HPttst(

mDiiapfec

tia

Fig. 1. Polymer electrolyte membrane fuel cell with a 5-layer memdapted from Ref. [17].

ater-gas-shift reactor and preferential oxidation units requiredor production of pure H2.

H3OH + H2O → CO2 + 6H+ + 6e− (3)

Gilman proposed the mechanism for methanol electro-xidation on Pt [20]. “Bifunctional theory” was coined to applyilman’s reaction pair concept to Pt-rich alloy catalysts for MeOHxidation [21]. The following is a modification of Gilman’s mecha-ism by Frelink et al [22] and Watanabe and Motoo [23] (alloyingomponent, or promoter metal, is represented as M):

t + CH3OH → Pt-(CH3OH)ads (4)

t-(CH3OH)act → Pt-(CH3O)act + H+ + e− (5a)

t-(CH3O)act → Pt-(CH2O)act + H+ + e− (5b)

t-(CH2O)act → Pt-(CHO)act + H+ + e− (5c)

t-(CHO)act → Pt-(CO)act + H+ + e− (5d)

+ H2O → M-(H2O)act (6)

t-(CO)ads + M-(H2O)act → Pt + M + CO2 + 2H+ + 2e− (7)

Within practical DMFC anode potentials (<300 mV vs. Reversibleydrogen Electrode (RHE)) the rate determining step (RDS) on puret is water activation (Eq. (6)). Deuterium isotope studies confirmhat at 350–400 mV, the RDS on PtRu shifts from water activation tohe first C H activation step (Eq. (5a)) [24] Ru, as a promotor metal,hifts water activation by over 200 mV vs. Pt (Fig. 2). The state-of-he-art DMFC catalyst is Johnson Matthey (JM) unsupported PtRu1:1).

Pt alloying also benefits reformate hydrogen fuel cells. Refor-ate hydrogen electrocatalysis is distinctly different from that ofMFCs, where CO is a reaction intermediate that must be oxidized

f all six electrons in the MeOH oxidation half reaction are to be real-zed (Eq. (3)). When reformate H2 is used, CO cumulatively poisonsctive sites. Since the hydrogen anode potential is never sufficientlyositive to oxidize CO, the more important issue to reformate H2uel cells is the ligand effect: The reduction of the CO adsorptionnthalpy by alloying Pt with promoter metals improves hydrogenompetitive adsorption on active sites [22,25].

Frelink [22] presents the ligand effect as an alloy electronic effecthat impacts the Pt-C bond strength, while Nørskov [25] explains itn terms of d-band center shifting. Our DFT results attribute the lig-nd effects to a d-band dispersion mechanism [26] With DMFCs the

Fig. 2. Steady-state MeOH polarization curves; JM Pt and JM PtRu (1:1).Adapted from Ref. [5].

bifunctional mechanism is essential because, in contrast to refor-mate fuel cells, the CO must be electrooxidized.

1.3. Stark tuning of COads

The variation of vibrational band frequencies with electrodepotential is Stark tuning [27,28]. In a model describing the interac-tion between the CO molecule and Pt d states, Nørskov describesthe bonding in two steps (Fig. 3) [25,29]. In the first step, the 2�*and the 5� MOs are shifted down in energy and broadened due tocoupling with the Pt s,p electrons. The renormalized CO orbitals arethen mixed with the Pt d-band resulting in the splitting of the 2�*MO into Pt-C antibonding and bonding orbitals.

Thus as the electrode potential is increased, the Pt-C �-bondingorbital (derived in part from the CO 2�* molecular orbital) is depop-ulated. This increases the C O bond order concomitant with thestretching frequency (i.e., Stark tuning). A second factor is dipole-

dipole coupling where stretching frequencies increase with surfacecoverage, or compression of existing coverage by co-adsorbates.This phenomena is due to increased spatial dipole-dipole interac-tions via coupling through metal electrons [30–34]. Oxidation of

N. Loupe et al. / Catalysis Tod

Fig. 3. Schematic illustration of the two-step model for CO adsorption on a Pt metalsurface.

R

Cttaoi

1

((t[(mMbgcrc

vmo[5oaIcsm[

(bwn

fWaspbi

the dual path mechanism proposed by Herrero [56], which is a CO

eprinted with permission from Ref. [29].

Oads results in decompression of the adsorbate layer and a reduc-ion of the stretching frequencies. The sensitivity of COads layerso co-adsorbates makes COads an excellent probe for adsorbatest the cathode and anode of fuel cells. This work exemplifies howperando Stark tuning can be used to characterize electrocatalyticnterfaces.

.4. Promoter metal (M) selection

Water activators (Eq. (6)) require M-O diatomic bond energiesD◦

M-O) similar to D◦Pt-C (≈590 kJ/mol) [3]. Fig. 4a shows D◦

M-Osolid symbols) near D◦

Pt-C (dashed line). Viewing left to right, nearhe D◦

Pt-C line, identifies V, Mo, Re, Os, Ru, and Sn as M candidates3]. Although D◦

Os-O is closest to D◦Pt-C, the PtOs phase diagram

Fig. 4b) shows an Os solubility limit of 20 mol%. The bifunctionalechanism (Eq. (7)) suggests the need for high surface M mole%.etal component diluents, that increase the overall M mole%, are

est selected from M candidates (e.g., Ru). The PtRuOs phase dia-ram, derived from PtRu and PtOs phase diagrams, maps out aompositional phase space that minimizes excessive phase seg-egation (Fig. 4b hatched region). This is the rational for ternaryompositions such as PtRuOs [3,35–41].

The lack of any rational basis for quaternary compositions moti-ated development of unbiased, high speed combinatorial searchethodologies. Reddington et al. developed and used such method-

logies to discover the quaternary PtRuOsIr DMFC anode catalyst35] Ir is a C H activator (i.e., Pt substitute) that can promote eqs.a-d of the bifunctional mechanism. This suggested the inclusionf elements with D◦

M-C similar D◦Pt-C (Fig. 4a) [42] to identify Pt

lternatives (Fig. 4, open symbols). The Ley diagram then identifiesr and Os as co-catalysts for C H activation. It is noteworthy that Osan serve as both a water activator and a C H activator. This workhows the benefits of a feedback loop between irrational searchethodologies (combinatorial methods) and fundamental studies

43–46].Fig. 4 suggests that a mischmetal fuel cell catalyst strategy

e.g., Ovshinsky NiMH battery anodes [47]) using M componentsetween the 5th and 12th periodic groups with Do

M-C and DoM-O

ithin ±50 kJ/mol of DoPt-C may be worth pursuing using combi-

atorial methods.The initial mechanistic research on mixed metal direct methanol

uel cell catalysts relied on polished arc melted alloys [3,36,48–50].e initiated rotating disc electrode studies of mixed metal cat-

lysts inks [38,51], and then transitioned to operando fuel cellpectroscopy for characterization of reaction intermediates, com-

etitive adsorption, and interfacial structure. This work focuses onulk and surface properties of Pt and mixed metal catalytic layers

ncorporated into operating MEAs.

ay 283 (2017) 11–26 13

1.5. Operando spectroscopy

Because the active state of a catalyst exists only during cataly-sis, operando spectroscopy is required to bridge the gap betweensurface science and practical catalysis [52]. The term “operandospectroscopy” was broadly introduced by Weckhuysen [53] as aform of in-situ spectroscopy. We qualify operando as spectroscopyof practical devices (e.g., fuel cells). Our primary challenge was con-version of a practical fuel cell, requiring controlled potential andflowing reactant streams, into a spectroscopy cell with minimalperturbation of functionality during acquisition of X-ray, IR andRaman spectra.

Fig. 5 summarizes the evolution of fuel cell operando spec-troscopy in our lab. The operando IR cell designed by Fan [1](1996), was primarily used for study of DMFC electrode lay-ers. The Viswanathan XAS cell [8] (2002) was used for hydrogenand liquid feed direct methanol fuel cells. The Lewis cell [10](2009), combining IR and XAS capabilities of the previous operandocells, was modified by Kendrick [14] for operando Raman micro-spectroscopy. These operando cells are commercially available(NuVant Systems Inc., Crown Point, IN). Further information,including the fuel cell type and catalyst compositions, are in Table 1.

2. Direct methanol fuel cells

2.1. Operando fuel cell IR spectroscopy on Pt, PtRu, and PtRuOsDMFC anodes

The Fan IR cell probed DMFC reaction intermediates on unsup-ported JM Pt and JM Pt-Ru (1:1) anode electrocatalysts. The cellhas a CaF2 window (25 × 2 mm) and a slot in the GDL for expo-sure of the anode catalytic surface to the IR beam. Catalytic inkswere decal-transferred onto opposite faces of the Nafion mem-brane as described [16]. A carbon cloth GDL (4.5 cm2) is used as thecathode current collector through which O2 (40 sccm) is delivered.A carrier gas (N2, 40 sccm) passed through a controlled tempera-ture humidifier filled with aqueous methanol for fuel delivery. ARHE reference electrode was contacted to the Nafion membranefor potential control. The spectral range of 700–4000 cm−1 wasinvestigated by signal averaging 300 scans at a resolution of 8 cm−1.

Fig. 6 shows I–V data and spectra of Pt (a,c) and PtRu (b,d)in the vapor fed DMFC. As expected, the CO2 bands at 2365 and2331 cm−1 increase with anode potential. On Pt, formic formationstarts at 0.3 V. The bands at 1335 and 1290 cm−1 are formic acidmodes, while those at 1194 and 1164 cm−1 are methylformate C-O-C modes. The bands at 1766 and 1745 cm−1 are methylformateCO stretching modes. Bridge-bound and linear-bound CO stretch-ing modes are at 1918–1951 cm−1 and 2050 cm−1 respectively. Thebridge-bound CO mode intensities increase with potential. On PtRu,the 1648 cm−1 band may be formate adsorbates [54,55]. In contrastto Pt, there is no evidence of bridge-bound CO on PtRu. The inten-sities of atop CO bands on PtRu are much smaller than those on Ptblack. The bifunctional mechanism is facilitated by Ru (Eq. (6)), thusreducing the CO coverage on Pt. On PtRu, the formation of formicacid and esters occur at potentials above 0.6 V, suggesting that PtRuis more selective for CO than Pt within the DMFC potential range.

The above data was the first evidence that formic acid is formedin a non-CO pathway for MeOH electro-oxidation on Pt understeady state fuel cell conditions, and that alloying Ru into Pt shutsdown the formic acid pathway at potentials within the DMFCpotential range. A formic acid pathway is to be differentiated from

free pathway accessible by high scan rates that preclude the for-mation of a poisoning CO layer. Fig. 6 shows that under steadystate conditions, formic acid and ester formation initiates at 0.3 V

14 N. Loupe et al. / Catalysis Today 283 (2017) 11–26

Fig. 4. (a) Metal-X bond dissociation energy (Do) vs. periodic group at 298◦ K. Dashed line is DoPt-C. Solid symbols Do

M-O; Open symbols DoM-C (b) PtRuOs phase diagram

approximated from PtRu and PtOs binary phase diagrams. PtRuOs compositional phase space with 5 selected compositions (solid dots) in the FCC (�) single phase regime(hatched area).

Table 1List of cell designs and descriptions.

Fuel cell type Cell design Spectroscopy type Catalyst composition Reference

DMFC Fan ReflectanceIR

JM Pt, JM PtRu(1:1) [1]JM Pt, JM PtRu(1:1) [2]JM Pt, PtRu(1:1),a PtRuOs(65:25:10)a [3]JM Pt [4]JM Pt, JM PtRu(1:1) [5]

Viswanathan TransmissionXAS

JM PtRu(1:1) [6]PtRu(3:1)a [7]

Hydrogen-air

Viswanathan TransmissionXAS

JM PtRu(1:1) [8]JM PtRu(1:1) [9]

Lewis TransmissionXAS/Reflectance IR

JM Pt [10]JM Pt, ETEK PtNi (1:1) [11]

ReflectanceIR

JM Pt [12]JM Pt [13]

os

(9i2awpmPoolo

uIbsPadXdcnp

Kendrick Micro Raman

a Indicates in-house prepared catalysts.

n Pt and 0.6 V on PtRu. Fig. 6 also shows that alloy catalysts haveubstantially less affinity for bridge bonded CO.

Fig. 7 (Fan IR cell) are (a) JM Pt, (b) in-house prepared PtRu1:1) and c) in-house prepared PtRuOs (65:25:10) anode catalysts at0 ◦C. Fig. 7d shows I–V data from 5 cm2 active area MEAs installed

n standard liquid feed DMFCs. Linear-bound COads on Pt appear at083 cm−1: Bridge-bound COads on Pt appear at 1851 cm−1. On PtRund PtRuOs, linearly bound COads bands (2075–2078 cm−1) appearhile bridge bound COads bands are absent. The PtRuOs catalyst,articularly at higher current densities (i.e., short circuit), exhibitsuch less absorption intensity in the CO region than either PtRu or

t. On both the Pt anode and the PtRu anode, bands characteristicf COads persist up to the fuel cell short circuit, but do not appearn the ternary catalyst at 0.1 V. These results correlate with theiquid feed DMFC data of Fig. 7d showing enhanced PtRuOs MeOHxidation activity.

The benefit of Os substitution into PtRu has been demonstratedsing COads data, MeOH electro-oxidation data, and operando

R spectroscopy. CO tolerance and MeOH oxidation studies onoth single-phase arc-melted electrodes and on dispersed highurface area catalysts incorporated into full DMFCs suggest thattRuOs (65:25:10) is superior to PtRu (1:1) catalysts for DMFCnodes. These rankings pertain to catalysts prepared by borohy-ride reduction of metal salts. Finally, although characterization byRD confirms single-phase FCC lattices as the product of borohy-

ride reduction of the noble metal salts, the surfaces require furtherharacterization. A later discussion will show that alloy compo-ents often phase segregate from the Pt FCC lattice as an amorphoushase. Although Os toxicity precluded third party testing of PtRuOs,

Barton FeNC catalysta [14]

the high performance of PtRuOs and PtRuOsIr validated Ley [3]diagram selection of H2O and C H activators.

2.2. Stark tuning of COads on Pt and PtRu DMFC anodes

Fig. 8a shows subtractively normalized potential-dependentspectra of COads on a Pt MEA obtained (Fan cell) [4]. Fig. 7b showsoperando MEA Stark tuning curves vs. the MeOH:D2O ratio deliv-ered to the anode. D2O was used to shift interfering water bands.At all ratios there is an abrupt decrease in the Stark tuning slopesat 0.5 V, which defines a transition potential that is independent ofthe MeOH:D2O ratio. The Stark tuning lines persist to 1 V becausethe CO is continuously replenished by the steady state flow to theanode. If the slope of the 0.5 MeOH:D2O ratio did not change priorto the onset potential, all the curves would intersect at 2084 cm−1,0.5 V. Fig. 7b shows that at 0.5 V, MeOH oxidation is facile on nanos-tructured Pt incorporated into a DMFC anode. The most importantobservation here is that Pt activates water (Eq. (6)) at potentialssubstantially negative of the potential at which Pt OH is formed.This is why Eq. (6) shows H2Oact rather than M OH (or Pt OH withpure Pt) in the bifunctional mechanism.

Stark tuning (Fig. 8b) can be explained in the context ofNørskov’s two-step CO adsorption model. As the MeOH:D2O ratioincreases, the renormalized CO 2�* MO density of states (DOS)increases. There are more 2�* orbitals competing for d-band elec-

trons causing the stretching frequencies to become less sensitiveto changes in the extent overlap of the d-band with the renormal-ized CO orbitals. For example, at MeOH:D2O = 0.5, the Stark tuningrate (STR) is 7 times larger than when MeOH:D2O = 4, at potentials

N. Loupe et al. / Catalysis Today 283 (2017) 11–26 15

Fig. 5. Evolution of fuel cell operando spectroscopy: Fan IR Cell, Visw

Table 2Fan IR cell: Pt STRs versus MeOH:D2O ratio [4].

MeOH:D2O Ratio 0.5 1 2 4

nScMr

sr(t

similar features to Fig. 8b, although the transition point is shifted

STR (cm−1/V) < 0.5 V 18.3 14.3 6.54 2.5STR (cm−1/V) > 0.5 V 8.2 9.2 3.3 1.1

egative of transition point. At the highest MeOH:D2O ratio, theTR is anomalously low. At low ratios, when the coverage is lower,hanges in the electrode Fermi level have a greater effect on the 2�*O DOS. Thus STR slopes are negatively correlated to MeOH:D2O

atios.The discontinuity in the STR slopes before and after the tran-

ition point (Table 2) correlate to H O activation. Ito et al. [57]

2eport that a change in the CO adlayer structure from (2 × 2) to√

19 × √19) at 0.4 V corresponds to an increase in the CO-CO dis-

ance from 5.6 to 12.2 Å. This decreases dipole-dipole coupling and

anathan XAS Cell, Lewis IR-XAS Cell, and Kendrick Raman Cell.

reduces atop CO stretching frequencies where water activationoccurs.

Fig. 8b shows distinct changes in STR slopes that correlates toH2O activation, not the formation of Pt OH. Thus, MEA Pt and PtRuoxidize CO at potentials negative of Pt OH formation (Fig. 2).

Simultaneous fit of EXAFS data suggests that 50% of Ru in a JMPtRu (1:1) catalyst (discussed later) is phased segregated from thePt FCC lattice. Fig. 9a shows operando STR of COads on JM PtRu (1:1).Our arc melted alloy studies showed that the band at 2020 cm−1 isCOads on pure Ru [37]. Fig. 9a contains spectra of COads on the phasesegregated PtRu (1:1) catalyst.

Fig. 9b shows the STR of COads on an anode prepared with highsurface area Ru and JM Pt at the cathode [17]. The results have

negatively by about 200 mV. Ru is an excellent M component (Eq.(6)) because it activates water 200 mV positive of that of Pt [21].STRs are not simply a property of the adsorbate-substrate iden-

16 N. Loupe et al. / Catalysis Today 283 (2017) 11–26

Fig. 6. Fan IR cell. Steady-state MeOH I–V data (25 ◦C). (a) JM Pt. (b) JM PtRu (1:1). (c) Anode spectra corresponding to a. (d) Anode spectra corresponding to b.Reprinted from Ref. [1].

Fig. 7. Fan IR cell. MeOH anode spectra (90 ◦C). (a) JM Pt, (b) PtRu (1:1), and (c) PtRuOs (65:25:10) versus cell potential. (d) MeOH polarization curves in 5 cm2 fuel cell.Anode: 0.5 M MeOH; 25 ml/min. Cathode: Dry O2, 400 ml/min, 10 psig.

Reprinted from Ref. [3].

Fig. 8. Fan IR cell. (a) Potential-dependent spectra of COads (50 ◦C); JM Pt. (b) Stark tuning of COads/Pt vs. MeOH:D2O ratio.Reprinted from Ref. [4].

N. Loupe et al. / Catalysis Today 283 (2017) 11–26 17

; JM PtA

tp0lc

2

[nfmTbpcd

lAfifaDRugcnfaocof

2

cdtsbsvp

Fig. 9. Fan IR cell. (a) Potential-dependent spectra of COads

dapted from Ref. [5].

ity. They are affected by co-adsorbates, adsorbate coverage, andarticle size [12]. STR transitions occur on Pt and Ru at 0.5 V and.3 V, respectively. It is also noteworthy that STRs of nanoscale cata-

ysts are always substantially lower than those obtained from singlerystals or arc-melted alloys [11,36].

.3. Operando XAS spectroscopy

The Viswananthan XAS cell [8,9] differs from that of Herron58] and similar in-situ [59] designs. The operando cell includeson-invasive cut-outs on both end plates and flow fields (Fig. 5) to

acilitate X-ray transmission through the cell. No alterations wereade to the flow field. No liquid electrolyte additives are required.

he absence of auxiliary electrolytes eliminates co-adsorbates (e.g.,isulfate, sulfate) that can block catalytic sites [44,50,60]. Althougherchlorate ions are non-adsorbing [61], steady state currentsause migration and buildup of mobile perchlorate ions within theouble layer of the anode catalytic surface.

The Viswananthan cell was first used at Materials Research Col-aborative Access Team (MRCAT) beamline, Sector 10-ID, of the ANLdvanced Photon Source in 2002 [8,9]. Anode and cathode humidi-ers, a load unit, heated transfer lines and mass flow controllers and

uel cell were installed in the hutch. The cell was operated with H2nd reformate simulant. The same cell was also used for liquid feedMFCs [6,62]. The cell design was adapted for use by Principi [63],oth [64], Tada [65], Russel [66] and Gilbert [67]. Viswananthansed Pd/C at the cathode to mitigate interference with Pt edge ener-ies of mixed metal anode catalysts. The more recent Lewis IR-XASell mitigates such interference by use of glancing angle XAS [68]. Aumber of operando XAS [63,64,69–76] studies followed which

ocused on the metal component oxidation states. All found thatt relevant fuel cell anode operating potentials, Pt is metallic. Thexidation of CO was studied both on Pt [73], and PtRu [49,70,72,77],onfirming the electronic benefits of Ru as a promoter metal for COxidation (DMFCs) or improved competitiveness for H2 (reformateuel cells).

.4. Phase segregation of PtRu catalysts

Fig. 10a shows the anode (red), full cell (black) and calculatedathode polarization curves obtained at 35 ◦C with 0.01 M MeOHelivered to the anode. The current onset at 380 mV vs. Pd/H2 washe basis for selection of the 250 mV–550 mV range for operandotudies. Fig. 10b shows absorption transition events due to CO2

ubble formation (Eq. (7)) in the anode flow field. The red linehows transition to lower absorption at 21,870 eV because a voidolume (bubble) is expelled. The flow fields have a single condensedhase throughout the remainder of the red line. The green line

Ru (MeOH:D2O = 1:1). (b) Stark tuning of COads/Ru (50 ◦C).

is another data set that shows initiation of a bubble at 22,200 eV(green line). Bubble formation is complete at 22,750 eV (green line).The absorption intensity increases because of the lower cross sec-tion of the void volume. Such transition events reduce the amountof extended X-ray absorption fine structure (EXAFS) data availablefor signal averaging. CO2 phase-out from the MeOH fuel stream waseliminated by application of a 120 Torr back pressure at the anodeoutlet.

Fig. 11a is the ex-situ Ru K-edge XANES of Ru powder (blue),operando Ru edge of the PtRu catalyst at 450 mV (red), as-receivedJM PtRu (1:1) catalyst (green), Ru oxide (black) and Ru oxidehydrate (pink). The coincidence of the catalyst Ru edge at 450 mVwith the metallic Ru powder confirms that within the DMFC poten-tial window the Ru is primarily metallic. The deviation of the redfrom the blue line at 22,120 eV reflects the difference between thenear neighbor environments of pure Ru (hexagonal) versus Ru in aPtRu FCC lattice.

In comparison to the operando catalyst, the as-received cata-lyst edge (green) is intermediate in energy between the metallic(blue & red) lines and the oxides of Ru (pink & black). The extentto which the as-received catalyst is oxidized and additional detailsare addressed by simultaneous fitting of the Pt and Ru EXAFS. ThePt LIII edge data of the same as-received PtRu catalyst (mountedon scotch tape), at 450 mV in DMFC mode, and the Pt foil refer-ence were measured in transmission mode (Fig. 11b). The alloyedPt had an increased white line intensity (green) versus the pure Ptfoil (Blue) due to alloy-induced d-band vacancies, as was observedin our XANES from an arc-melted PtRu (80:20) alloy [8], and byprevious researchers [61]. The large edge intensity confirms exten-sive oxidation of the as-received catalyst. The potential-dependent(250 mV, 300 mV, 350 mV, 400 mV, 450 mV) EXAFS had near per-fect overlap confirming that, within the fuel cell operating potentialrange, the Pt is insensitive to potential (as expected), similarlyto the Pt within supported PtRu catalyst in a H2/air fuel cell [8].Fuel cell polarization losses are primarily due to the four-electron-reduction of O2 (∼=400 mV) with little polarization at the H2 anode.This enables use of a H2 anode as a reference electrode.

Least squares fits of the PtRu operando XANES (Fig. 11c), withRu powder and Ru oxide-hydrate standards at all potentials rele-vant to DMFC anodes (250 mV–450 mV) showed that the fractionof Ru oxide-hydrate required for the best fit in pure water, 0.1 Mor 2 M MeOH within experimental error, is invariant. Since there isminimal change in the state of oxidation of the Pt and Ru through-out the DMFC operating potential window, we pooled the data and

simultaneously fit the Pt and Ru EXAFS. The result were structuralparameters obtained from MEA1 in remarkable agreement withthe MEA2 parameters (Table 3) with the exception of Ru O Debye-Waller factor which was positive for MEA2. The ratio of the atomic

18 N. Loupe et al. / Catalysis Today 283 (2017) 11–26

Fig. 10. Viswananthan XAS cell. (a) 0.01 M MeOH (1 ml/min), 35 ◦C. Fuel cell performance with air (100 sccm) at counter electrode (black). Anode polarization curve with4% H2 balanced N2 to counter electrode (red). Cathode polarization curve (i.e., full cell + anode curve) (blue). (b) Absorption transition events resulting from CO2 bubbleformation/expulsion.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Reprinted from Ref. [6].

Fig. 11. Viswananthan XAS cell. (a) Ru K-edge: various metals/catalysts (labeled). (b) Pt LIII-edge: Operando XANES (JM PtRu, 450 mV) MEA1 (red), MEA2 (orange). Pt foil(blue). As-received JM PtRu (green). (c) Least squares fits of in-situ XANES with Ru powder and Ru oxide-hydrate standards showing fractions of Ru oxide-hydrate versuspotential: pure H2O (blue), 0.1 M MeOH (green), and 2 M MeOH (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the webversion of this article.)

Reprinted from Ref. [6].

Table 3Catalyst structural parameters from Pt and Ru EXAFS for three different samples: MEA1 in 0.1 M MeOH at all potentials fit simultaneously; MEA2 in 0.1 M MeOH at 450 mVanode potential (improved EXAFS background, similarly prepared MEA); and as-received catalyst (same structural model plus Pt O bond). Uncertainties can be found inRef. [6].

EXAFS parameter notation MEA1 MEA2 As-received catalyst

Ru around Pt Fraction* yRu 0.27 0.30 0.25Pt around Ru Fraction yPt 0.55 0.49 0.57Ru Ru bond DWF**, (Å2) �2

Ru-Ru 0.0037 0.0045 0.0042Ru Pt bond DWF, (Å2) �2

RuPt 0.0038 0.0034 0.0038Pt Pt bond DWF, (Å2) �2

Pt-Pt 0.0057 0.0047 0.0064Ru Ru bond length, (Å) RRu-Ru 2.649 2.652 2.66Ru Pt bond length, (Å) RPt-Ru 2.698 2.695 2.70Pt Pt bond length, (Å) RPt-Pt 2.717 2.716 2.71Total number of first shell atoms Ntot 8.2 8.2 5.5Ru O bond DWF, (Å2) �2

Ru O −0.003 0.01(fixed) 0.0057Ru O bond length, (Å) RRuO 1.86 1.88 1.94Ru O coordination number NRuO 0.3 0.4 2.6Pt O bond DWF, (Å2) �2

PtO N/A N/A 0.0049Pt O bond length, (Å) RPt O N/A N/A 1.93

N

R

fstip

Pt O coordination number NPtO

eprinted from Ref. [6].

ractions ratio yRu = NPtRu/Ntot (fraction of Ru atoms around Pt), and

imilarly yPt (fraction of Pt atoms around Ru) where Ntot is the fit-

ed total number of near neighbours, provides the ratio of Ru to Ptn the FCC lattice. That ratio, 0.49, suggests that 50% of the Ru ishase segregated from the FCC lattice. The absence of observed Ru

/A N/A 1.5

metal bands in the XRD suggests that the phase segregated Ru hasno long range order (amorphous). The fit value of yRu = 0.27 ± 0.02 is

less than 0.34 (FCC phase for nominal Ru composition) suggests thatthe EXAFS model of the alloy FCC lattice is not totally disordered.

N. Loupe et al. / Catalysis Today 283 (2017) 11–26 19

Fig. 12. FCC lattice spacing versus Pt mole%. High surface area catalysts (red circles)and arc-melted alloys (black squares). Red lines suggest that high surface area PtRu(1:1) is phase segregated with a PtRu alloy phase of 76% Pt. (For interpretation of thero

R

2

asfi(ipiiafihmt

FaiGsoPktro

MRtafivtXano

Table 4Summary: EXAFS of operando and as received JM PtRu (1:1).

Operando Catalyst (MEA avg.) As Received Catalyst

Ru Oxidation 15% 58%N 8.2 5.5[Ru]/[Pt] 0.50 0.44Pt O bonds None Present

eferences to colour in this figure legend, the reader is referred to the web versionf this article.)

eprinted from Ref. [44].

.5. Comparison of operando PtRu to as-received PtRu

The XAS data were reconciled with ex situ XRD (Fig. 12) and XRFnalysis of the raw catalyst to further confirm the extent of phaseegregation suggested by the EXAFS analysis. The XRF analysis ofve grams of JM PtRu confirmed the Pt:Ru mole ratio to be 1.04:1∼1:1) [6]. A lattice parameter analysis of several fuel cell catalysts,ncluding the JM catalyst [78] showed that the JM PtRu (1:1) latticearameter us larger than that of arc-melted PtRu of the same nom-

nal composition. This confirms that the alloy phase of the catalysts richer in Pt than what would be expected from a 1:1 compositions confirmed by XRF. Application of a Vegard law line (Fig. 12) con-rms that the JM catalyst is phase segregated with an alloy phaseaving a ratio of Pt:Ru of 1.89:1 [78]. This is in remarkable agree-ent with the 2.04:1 determined by the simultaneous analysis of

he Pt and Ru EXAFS data.Our finding that half the Ru is phase segregated from the PtRu

CC lattice correlates with the Gasteiger et al. study of arc meltedlloys [48]. They report an optimum 33 mol% Ru at 60 ◦C on pol-shed arc-melted Pt alloy surfaces. We operate our DMFCs at 60 ◦C.asteiger’s finding of 2:1 Pt:Ru fits the JM PtRu 1:1 nominal compo-ition since 50% of the Ru is phase segregated. The alloy compositionf the ideal polished arc melted alloy and the alloy phase of JMtRu catalyst are roughly the same. This is the first work, to ournowledge, where Ru and Pt EXAFS data were fit simultaneouslyo characterize the bulk structure of a nanostructured catalyst withesults that are consistent with a lattice parameter and XRF analysisf the same catalyst.

The 1.87 Å Ru O bond length (Table 3) was reproducible fromEA1 to MEA2 and shorter than the 2.02 Å reported for RuO2 or

uO2·H2O [79]. This is not unexpected as our analysis shows thathe Ru is primarily metallic in the JM catalyst during fuel cell oper-tion. A 2.02 Å Ru O distance implies no Ru-Ru bonds. This differsrom Ru incorporated in either an FCC PtRu lattice or a primar-ly metallic amorphous Ru phase. Table 3 suggests that the 1.87 Åalue fits in a systematic trend. The as-received catalyst Ru O dis-ance is 1.94 Å, much closer to the 2.02 Å literature value [79]. The

ANES data shows that the as-received catalyst is highly oxidized,nd can be fit with a similar model yielding an overall metal coordi-ation number (Ntot) of 5.5, as well as a significantly larger numberf Ru O (2.6) and Pt O (1.5) bonds averaged over all atoms in

Pt O bonds ∼0.24 avg. ∼2.8 avg.

Adapted from Ref. [6].

the catalyst. The XANES also shows that upon introduction of thefuel, the JM PtRu catalyst is reduced to an essentially metallic state(Fig. 11a). This is a structural transformation from the as-receivedcatalyst (with oxygen deep in the FCC nanoparticles and phasedout Ru) to an essentially metallic operational catalyst that has asmall amount of residual oxygen either at the surface of the clus-ters and/or in the amorphous phase. The trend is thus 2.02 Å Ru Obond lengths (fully oxidized Ru), 1.94 Å Ru O bond lengths (sub-stantially oxidized Ru within the as-received catalyst) and 1.87 ÅRu O bond lengths (operating JM catalyst) which is not similar toa fully oxidized phase.

Table 4 summarizing Table 3, succinctly points out the need foroperando analysis of fuel cell catalysts. The aggregate of the XASresults of Table 4, and lattice parameter analysis of PtRu shows thathalf the Ru is phased out of the alloy phase and is likely amorphous.This bulk phase analysis is a small piece of a complex questionconcerning processes at surfaces.

3. Hydrogen - air fuel cells

3.1. Operando IR-XAS cell

The Lewis IR-XAS cell combines features of the Viswananthan[9] and Fan cells (Fig. 5) to enable transition of operando specularreflectance FTIR spectroscopy to XAS (transmission and fluores-cence) without requiring MEA extraction. A fluorescence XAS beamis not attenuated by the lower graphite flow field. The IR beamaccesses the catalytic layer through a CaF2 window [4,80] posi-tioned above a slot in the MEA GDL. The Lewis IR-XAS cell interfacesto a potentiostat through a 9-pin connector on the slider hous-ing. The slot on the bottom side of the slider housing preciselypositions the window within a Pike diffuse reflectance accessory.The catalytic layer facing the lower graphite flow field serves asboth a H2 reference and counter electrode when charged withH2. The low CO oxidation currents do not measurably polarize thecounter/reference electrode. This strategy of shorting the counterand reference leads is discussed by Gurau et al. [81] XANES wereobtained at the MRCAT.

3.2. Time-dependent XANES of hydrogen - air cathodes

Operando spectroscopy was carried out with humidified airand H2 delivered to the Pd anode and Pt cathode, respectively.After an initial one-hour conditioning period [5], the fuel cell wasstepped from open circuit voltage to 530 mV. Time-dependentXANES were sequentially acquired at two minute intervals. Thespectra were subtractively normalized to the reference spectrum(obtained at 0 V). Fig. 13a shows subtractively normalized XANESof a Pt cathode catalytic layer with the Lewis cell operating as an airbreathing fuel cell at 50 ◦C. The bands at 11,553 eV and 11,578 eVdecrease with time, as the Pt is reduced, over a period of hours.

The large time constant cannot be attributed solely to reduction ofchemisorbed oxygen. The sub-surface oxygen, native to the crystal-lite core structure, may be responsible for the long time-constant.The large restructuring time-constant for Pt explains why better

20 N. Loupe et al. / Catalysis Today 283 (2017) 11–26

Fig. 13. Lewis IR-XAS cell. Time-dependent XANES at 530 mV subtractively normalized to 0 V. (a) Pt edge of air-breathing Pt cathode. (b) Ni edge of an air breathing PtNicathode catalyst.

Reprinted from Ref. [11].

F orkin( tra of

A

pfos

(ioNca

msobcow

Ustt

ig. 14. Lewis IR-XAS cell. (a) Polarization curve. (b) Cyclic voltammetry (CV). W50 sccm). Background CV (solid). CO stripping wave (dashed). (c) Stark tuning spec

dapted from Ref. [11].

erformance obtained when acquiring cathode polarization curvesrom low to high potentials [82]. Oxides are undesirable at the cath-de and it takes time to deplete sub-surface oxygen reservoir whencanning from high to low potentials.

Fig. 13b shows the Ni-edge time-dependent data of a PtNi1:1)/C cathode catalyst (ETEK, Somerset, NJ), using the same exper-mental conditions as was used for the Pt catalyst. An EXAFS fitf Fig. 13b data [83] attributes the shorter time constant for thei edge transient to the predominance of the Ni within the metalore: The surface, dominated by Pt, has only small amounts of Nivailable for adsorption/reduction of chemisorbed oxygen.

The time constants associated with the above XAS range frominutes (PtNi) to hours (Pt). Tada et al. [65], using transmission

tudies done on a Viswananthan-type cell, reported a method forbtaining 1 s time-resolved full EXAFS spectra from a fuel cell cycledetween 0.4 and 1.0 V. They observed separate time constants forharging and discharging, as well as the formation and dissociationf surface Pt O bonds. Even shorter time constants will be possiblehile using dispersive EXAFS techniques.

XAS probes the core and surface of nanostructured catalysts.

nfortunately, the contribution of surface bonds to XAS models is

mall relative to that of the bulk. It is often difficult, or impossibleo draw concrete conclusion about the surface. Our concerns abouthe use of Janin cluster models for surface analysis by subtractively

g electrode: Humidified N2 (50 ◦C). Counter-reference electrode: Humidified H2

COads/Pt (100 mV adsorption potential).

normalized XAS included signal/noise issues and the inability ofJanin cluster models to manifest ligand effects [10].

3.3. Stark tuning of COads in hydrogen fuel cells

Fig. 14a is a Lewis cell polarization curve (0.04 mV/min, 50 ◦C).Humidified H2 (50 sccm) and air (250 sccm) were delivered to thecounter/reference and working electrodes respectively. Fig. 14b isthe background CV (solid) and the CO stripping wave (dashed)obtained at 50 ◦C. The CO stripping wave, acquired at 10 mV/sec,extends from about 600–900 mV. In Fig. 14c, the potential-dependent IR spectra (Stark tuning) of CO adsorbed at 100 mV vs.RHE are shown.

Stark tuning data for CO dosed at 100, 200, 300 and 400 mVvs. RHE (Fig. 15a) was obtained at 50 ◦C using the Lewis IR-XAFcell. The potential was set to 100 mV prior to acquisition of foursignal-averaged (250 scans) spectra, at 50 mV increments, until theCO vibrational bands were no longer observable. The STRs of thiswork are typically about 7 cm−1/mV. STR for MEA incorporated

Pt are lower than those reported for polished polycrystalline Pt(e.g., 23.8 cm−1/V–27.5 cm−1/V) [37,84]. The differences are due tounique surface conditions provided by the operando environment(e.g., Pt in the MEA is Nafion-coated, Pt is multifaceted, etc.).

N. Loupe et al. / Catalysis Today 283 (2017) 11–26 21

Fig. 15. Lewis IR-XAS cell. (a) Stark tuning of linearly bound COads/Pt MEA versus potential, 50 ◦C. (b) CO electro-oxidation onset potential versus adsorption potential, withm tationv

A

dcwa�bntE2p

3

3iHfmsCaiuctwspiaPise

3

(

odels of COads coverage. Atom color code: C: black, O: red, Pt: grey. (For interpreersion of this article.)

dapted from Ref. [11].

Fig. 15b shows how the oxidation onset potential (Eonset)epends on adsorption potential (Eads). At 400 mV, Eonset and Eadsoincide because at higher potentials COads favors adsorption sitesith higher enthalpies of adsorption (e.g., steps and kinks) [85,86]

nd there is no thermodynamic force driving migration to lowerHads sites. At Eads of 300 mV, the Eonset exceeds Eads by 60 mV

ecause the distribution of adsorption sites are more heteroge-eous and there is a driving force for COads on terraces to migrateo steps, kinks and adatoms. Thus the difference between Eonset onads would be expected to increase as Eads is decreased. At Eads of00 mV and lower, the Eonset is leveled to 320 mV. Site specificity isotential-dependent.

.4. Temperature-dependent COads on Pt in hydrogen fuel cells

Fig. 16 shows operando STR plots of linearly bound COads/Pt at0, 50, and 70 ◦C obtained using the Lewis cell. The data are sim-

lar to those of Stamenkovic et al. on single crystal Pt in dilute2SO4 [87]. However, Fig. 16 data is obtained from an operating

uel cell: There is no aqueous sulfuric acid electrolyte and thus noobile anions. At all temperatures, a linear region (below 200 mV)

egues to a subtle �CO blue shift. We attribute the blue shift toOads compression due to repulsive dipole interactions with co-dsorbed Nafion exchange site sulfonate groups [12]. This processs illustrated by the transition of Fig. 16 panel 1 to panel 2. Thepturn is followed by a precipitous drop (i.e., reduced dipole-dipoleoupling) caused by COads oxidation induced by H2Oact absorp-ion (Fig. 16, panel 3). A final upturn becomes more pronouncedith higher temperatures. This is attributed to further adsorption of

ulfonate exchange sites (as well as Nafion–CF3 groups) that com-ress COads islands and reestablishes dipole-dipole coupling that

ncreases �CO (Fig. 16, panel 4). It is possible that CO crowding islso induced by activated water. The self-assembly of Nafion ont, reported by Kendrick et al. [12], using polarization modulatednfrared reflection absorption spectroscopy of a Nafion/Pt interface,uggests that Nafion–CF3 groups co- adsorb as well the sulfonatexchange groups.

.5. Potential-dependent spectra of hydrogen - air Pt cathodes

The Pt MEA cathode spectra obtained with the Lewis cell1.2 V–0 V vs. RHE counter/reference) show the appearance of

of the references to colour in this figure legend, the reader is referred to the web

Nafion IR bands at 0.9 V (Fig. 17a). Fig. 17b shows the I–V curve(black) superimposed upon the potential-dependent interfero-gram amplitudes (blue). The increased amplitude corresponds toincreased Pt surface reflectivity. At potentials where Pt is passi-vated with oxide (1.2–1 V vs. H2 anode) there is no change vs. thereference spectrum at 1.2 V. At the ORR onset-current a 2% step-increase in Pt reflectivity coincides with the emergence of NafionIR bands. The step-change in reflectivity coincides with exposure ofoxide-free Pt to O2. The reflectivity maximizes just after the currentonset and then slightly diminishes with decreasing potential. The1060 cm−1 and the envelope region (1100–1300 cm−1) persist tothe short circuit current. The 1060 cm−1, a group mode associatedwith a dissociated Nafion exchange site with a C3V local symmetry[12,88–90] has been assigned as COC �as, SO3

− �s.The 1060 cm−1C3V mode is sensitive to state-of-hydration. The

emergence of C3V modes coincides with H2O formation and the stepincrease in the Nafion/Pt interface reflectance.

3.6. Operando Raman confocal micro-spectroscopy

Operando Raman spectroscopy of solid oxide fuel cells has beenpreviously reported [91–93]. Because PEM fuel cells are chargedwith humidified reactant streams, condensed water layers (>5 �m)can obliterate IR bands. Raman spectroscopy is routinely used toprobe such buried interfaces. Although there are previous reports ofwater management/transport in PEM fuel cells [94–99], these ref-erences provided no potential-dependent data. We used operandoRaman confocal micro-spectroscopy to study the MEA cathode of ahydrogen fuel cell vs. cell potential. Through density functional the-ory (DFT) calculated normal mode analysis of the Nafion side chain,we correlated Nafion vibrational group modes to fuel cell cathodepotentials. Water formation at the cathode directly impacts theNafion ion exchange site local symmetry (vide infra).

Confocal micro-spectroscopy enabled depth profiling of theoperando cell from the center of the MEA, through the catalyticlayer, up to, and including, the flow field region. This low resolu-tion profiling provided guidance as to the proper positioning of thelaser focal point for study of the catalytic layer.

The upper flow field of the Raman micro-spectroscopy cell(Fig. 2) has a recessed disc area that accommodates a fused quartzwindow (GE 124, General Electric) positioned above the work-ing electrode. 11 The lower flow field delivers H2 to the counter

22 N. Loupe et al. / Catalysis Today 283 (2017) 11–26

Fig. 16. Lewis IR-XAS cell. Panels 1 - 4: Model for Stark tuning of COads/Pt. C: black, O: red, S: yellow, F: aqua, Pt: grey. Bottom: Stark tuning of COads/Pt MEA at 30 ◦C (left)50 ◦C (middle) 70 ◦C (right). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Adapted from Ref. [12].

F b) Intp

R

eehiFta

ig. 17. Lewis IR-XAS cell. (a) Potential-dependent spectra of a Pt cathode (N2). (olarization curve under O2.

eprinted from Ref. [13].

lectrode layer enabling its use as both the counter and referencelectrode as described by Gurau et al. [81] The cell is equipped witheater elements and a thermistor for temperature control. The cell

s positioned at the working distance of the microscope objective.

ig. 18 is a polarization curve obtained with H2 and O2 delivered tohe anode and cathode respectively. The current onset is observedt 750 mV.

erferogram amplitude as a function of potential superimposed over an operando

Fig. 19a shows a dimensioned cross sectional schematic of theMEA loaded within the operando cell. The laser focal point wasvaried along the z-axis until the pure Nafion spectrum attained amaximum intensity (black spectrum). This focal point position is

referred to as +0 �m. Additional spectra are obtained (Fig. 19b) asthe focal point height is progressively increased from the referencepoint, through the membrane, catalytic layer, gas diffusion layer,

sis Today 283 (2017) 11–26 23

aczdta

bTapE[glOot

3

(psfve

Oed(s[1tgN[ca

Fig. 18. Kendrick Raman cell. Polarization curve: Pt anode (H2), Fe-Nx/C cathode

Fw

R

N. Loupe et al. / Cataly

nd graphite flow field. The focal point position cannot be directlyorrelated to a position within the fuel cell assembly because the

resolution (23 �m), confocal volume (33 �m3), and complexitiesue to the various physical layer characteristics (diffraction, scat-ering, etc.). The complex optics for even simpler membrane layersre discussed by Peng et al [94].

The green spectrum (100 �m) shows the initial emergence of aand for gaseous O2 within the porous catalytic layer at 1556 cm−1.he intensity of this band increases as the focal point is lifted upnd through the fuel cell flow field. Bands related to the carbon sup-ort also emerge at 100 �m. The two prominent features are the2g band 1600 cm−1and the disorder band (D band) at 1360 cm−1

100]. The D band arises from an A1g breathing mode due to edges ofraphene planes on clusters smaller than 200 Å [100]. The catalyticayer has thin layer Nafion originating from the catalyst ink [15].ur interest is a region within the catalytic layer that simultane-usly displays both gaseous O2 and Nafion bands (+350 �m abovehe reference point).

.7. Potential-dependent nafion Raman and IR spectra

The polarization curve, with the focal point set at + 350 �m,Fig. 18) shows a current onset at 750 mV [101]. Fig. 20 comparesotential-dependent Raman spectra (left) with transmission IRpectra of fully dehydrated (blue), partially hydrated (purple), andully hydrated (red) Nafion. Visualization of normal mode eigen-ector animations resulted in the assignments categorized by thexchange site local symmetry (Table 5).

At cathode potentials positive of 800 mV (Fig. 20) there is noRR. The Raman spectra in the ORR inactive region (>800 mV)xhibit a band at 910 cm−1 also present in the spectrum of dehy-rated Nafion (blue). This band corresponds to a group mode786* cm−1) that is dominated by a dehydrated Nafion exchangeite in the sulfonic acid form having no local symmetry (C1)12,88–90]. At potentials negative of 700 mV, Raman bands at066 cm−1 and 969 cm−1 emerge. These bands, also observed inhe hydrated Nafion IR spectrum (red), are the C3V,HF and C3V,LFroup modes, respectively that are dominated by a dissociated

afion exchange site (sulfonate form) with C3V local symmetry

12,90,102–106]. These IR group modes involve the mechanicallyoupled ether link and ion exchange site and thus cannot bessigned as single-functional-group modes [12,90,102–106].

ig. 19. Kendrick Raman cell. (a) Schematic of MEA installed in the operando spectroscopyith O2 flow.

eprinted from Ref. [14].

(O2).

Reprinted from Ref. [14].

3.8. Catalyst Raman bands (560–631 cm−1)

Fig. 20a shows low intensity bands at 564 cm−1, 600 cm−1,and 635 cm−1. To facilitate analysis of these bands, Fig. 21 showsoperando spectra with cathode catalytic layers of: Fe-Nx/C with O2flow (left), C/(no Fe-Nx/C) with O2 flow (middle), and Fe-Nx/C withN2 flow (right). JM Pt/C was used in the anode catalytic layer. Sincethe band frequencies are not potential dependent, they were aver-aged to improve signal to noise. “Average” spectra are shown inseparate panels below their respective source panels.

The 564 cm−1 band is attributed to Fe-O2 stretching modes[107–110] and present at all cell potentials under O2 flow (left).Under N2 flow (right), the band is non-discernable from the back-ground. This is consistent with the presence of Fe in the catalyticlayer. There is a total absence of intensity at 564 cm−1 with a purecarbon cathode under O2 flow (middle).

Buzgar et al. [111] report C3V binding of transition metals tosulfates in the wavenumber region 600 cm−1 to 635 cm−1. Hester

and Krishnan [112] confirmed C3V binding of divalent metals to drymolten sulfates. We have reported divalent ions binding with C3Vsymmetry to sulfonate oxygens in dehydrated Nafion [88–90]. The

fuel cell. (b) Confocal Raman microscope depth profiling spectra of Fe-Nx/C cathode

24 N. Loupe et al. / Catalysis Today 283 (2017) 11–26

Fig. 20. Kendrick Raman cell. (a) Potential-dependent spectra of Fe-Nx/C cathode catalyst (cathode: O2; anode: H2). (b) Transmission spectra of Nafion 212. Fully dehydrated(top), partially dehydrated (middle), fully hydrated (bottom). Group theory labels refer to exchange site local symmetry.

Reprinted from Ref. [14].

Table 5Group mode assignments, DFT calculated normal modes, Transmission IR, and Raman bands for hydrated and dehydrated Nafion.

Local symmetry Group mode assignment DFT (cm−1) Transmission (cm−1) Raman (cm−1)

Hydrated Nafion CF3 �u, COC-B �s, COC-A �s 738* 730BB 883* 806C3v,LF SO3

−1 �s, COC-A �as 983* 969 969C3v,HF COC-A �as, SO3

−1 �s 1059* 1061 1066Dehydrated Nafion C1,LF CF3 �u, COC-B �r, COC-A �r 731* 910 730

SO3H �s, COC-A �s 786* 910C1,HF COC(B) , BB 820* 1414 806

S brella

hrcBtwbcabsSl

SO3H �as, COC-A �as 1405*

ymmetric stretching, �s; Asymmetric stretching, �as; Wagging, ; Bending, �s; Um

igh performance of fuel cell of this study, along with the aboveeferences, suggests that other metal-oxygen interactions may beontributing to vibrational spectral features in Figs. 20 and 21 .rogan et al. [113] assigned bands around 550 cm−1 as Pt O vibra-ional modes. We recently reported on the direct interaction of Ptith the Nafion sulfonate group [12] and the emergence of Nafion

ands at similar operando spectroscopy fuel cell potentials in Ptathode fuel cells [13]. The possibility of a Pt contaminant in thes-received Fe- Nx/C catalyst prompted us to request a Pt analysisy Robertson Microlit Laboratories, Ledgewood, NJ. The analysishowed 13.4 ppm Pt in the as-received catalyst.1 The absence of

tark tuning of the 564 cm−1 band suggests that the Fe-Nx/C cata-yst does not participate in ORR.

1 The original vial with remaining catalyst was sent for ICP-MS analysis.

bending, �u; Rocking, �r; Backbone, BB; Starred values (*) are calculated.

4. Conclusion

Operando IR, Raman and X-ray absorption spectroscopy wasused to study direct methanol and hydrogen/air fuel cells. The evo-lution of operando cell designs and basis for methodologies areprovided. Pt catalysts and a non-PGM iron based catalyst were stud-ied at fuel cell cathodes over an extended period, as well as Pt-basedmixed metal catalysts at anodes. The membrane electrode assem-bly catalysts, catalyst-ionomer interfaces and bulk ionomer phaseswere subjected to operando spectroscopic studies, complementedwith XRD lattice parameter analyses and density functional theoryvibrational normal mode analysis. Some of the salient findings aresummarized.

(1) The best DMFC catalyst, Johnson Matthey unsupported PtRu(1:1), has 50% of its Ru phase segregated from the Pt alloy FCClattice. Within the DMFC operating potential window, Ru and Pt

N. Loupe et al. / Catalysis Tod

Fe-Nx/C cathodecatalyst opera�ngunderO2

C (no Fe) cathodecatalyst opera�ngund erO2

Fe-Nx/C cathodecatalystop era�ngunderN2

631 cm-1

600 cm-1

564 cm-1 631 cm-1

600 cm-1

564 cm-1 631 cm-1

600 cm-1

564 cm-1

0 mV

100 mV

200 mV

300 mV

400 mV

500 mV

600 mV

700 mV

800 mV

900 mV

1000 mV

1100 mV

Average

Fig. 21. Kendrick Raman cell. Potential-dependent Raman spectra (650–550 cm−1)of the Fe-Nx/C cathode catalyst operating under O2 (left), under N2 (middle), andu

A

(

(

(

H. Kim, S. Thomas, A. Wieckowski, J. Phys. Chem. B 104 (2000) 3518–3531.

nder O2 after Fe removal (right).

dapted from Ref. [14].

are essentially metallic. IR studies of COads on JM PtRu show lin-ear bound CO on Pt atoms and phase segregated Ru. In general,alloy catalysts are phase segregated.

2) The Ley diagram overlays metal-carbon and metal-oxygenbinary bond energies (vs. periodic group) that are in proximityto the Pt-C bond energy of 590 kJ/mol. Pt-C Ley diagrams iden-tify Mo, Re, Os, Ru and Sn as water activators and Pt, Ir, Rh andOs as C H activators for bifunctional electrocatalysts. Goddardconfirmed by DFT that Os is both a water and C H activator.

3) The linear bound CO stretching frequencies on catalyst surfaceatoms vary with the membrane electrode assembly potential(Stark tuning). The Stark tuning slopes have transition pointsthat correlate to water activation. Activated water is the bifunc-tional mechanism reactant, not M-OH (e.g., Ru OH or Pt OH).CO oxidation initiates at the Stark tuning transition point.On Ru, the Stark tuning slopes transition from 23 cm−1/V to7 cm−1/V.

4) Operando confocal Raman micro-spectroscopy tracked wateraccumulation in the ionomer phase of H2-air fuel cell mem-brane electrode assemblies as the cathode potential varied from1.1 to 0 V vs. RHE. Course depth profiling of an entire fuel cellassembly, including the flow field grooves, catalytic layers andthe Nafion membrane, enabled selection of the appropriate

focal point for study of the electrode layer. The innovation isassignment of vibrational normal modes in terms of the localsymmetry of the Nafion exchange site using DFT calculated nor-mal mode analysis. When Nafion is dehydrated, the exchange

ay 283 (2017) 11–26 25

site is in the sulfonic acid form with no local symmetry (C1).As the ionomer hydrates, the exchange site dissociates to a sul-fonate form with C3V symmetry. Operando Raman spectroscopyof the fuel cell tracks the gradual transition of vibrational bandsfrom C1 modes to C3V modes as the cathode potential is reduced,corresponding to ionomer hydration as a result of the oxygenreduction reaction.

Funding sources

Funding was provided by the Army Research Office grantsW911NF-12-1-0346 and W911NF-14-1-0365.

References

[1] Q.B. Fan, C. Pu, E.S. Smotkin, J. Electrochem. Soc. 143 (1996) 3053–3057.[2] Q.B. Fan, C. Pu, K.L. Lay, E.S. Smotkin, J. Electrochem. Soc. 143 (1996)

L21–L23.[3] K.L. Ley, R. Liu, C. Pu, Q. Fan, N. Leyarovska, C. Segre, E.S. Smotkin, J.

Electrochem. Soc. 144 (1997) 1543–1548.[4] A.L. Bo, S. Sanicharane, B. Sompalli, Q.B. Fan, B. Gurau, R.X. Liu, E.S. Smotkin,

J. Phys. Chem. B 104 (2000) 7377–7381.[5] S. Sanicharane, A. Bo, B. Sompalli, B. Gurau, E.S. Smotkin, J. Electrochem. Soc.

149 (2002) A554–A557.[6] S. Stoupin, E.H. Chung, S. Chattopadhyay, C.U. Segre, E.S. Smotkin, J. Phys.

Chem. B 110 (2006) 9932–9938.[7] S. Stoupin, H. Rivera, Z. Li, C.U. Segre, C. Korzeniewski, D.J. Casadonte Jr., H.

Inoue, E.S. Smotkin, Phys. Chem. Chem. Phys. 10 (2008) 6430–6437.[8] R. Viswanathan, G. Hou, R. Liu, S.R. Bare, F. Modica, G. Mickelson, C.U. Segre,

N. Leyarovska, E.S. Smotkin, J. Phys. Chem. B 106 (2002) 3458–3465.[9] R. Viswanathan, R. Liu, E.S. Smotkin, Rev. Sci. Instrum. 73 (2002) 2124–2127.

[10] E.A. Lewis, C.U. Segre, E.S. Smotkin, Electrochim. Acta 54 (2009) 7181–7185.[11] E.A. Lewis, I. Kendrick, Q. Jia, C. Grice, C.U. Segre, E.S. Smotkin, Electrochim.

Acta 56 (2011) 8827–8832.[12] I. Kendrick, D. Kumari, A. Yakaboski, N. Dimakis, E.S. Smotkin, J. Am. Chem.

Soc. 132 (2010) 17611–17616.[13] I. Kendrick, E.S. Smotkin, Int. J. Hydrogen Energy 39 (2014) 2751–2755.[14] I. Kendrick, J. Fore, J. Doan, N. Loupe, A. Vong, N. Dimakis, M. Diem, E.S.

Smotkin, J. Electrochem. Soc. 163 (2016) H3152–H3159.[15] M.S. Wilson, S. Gottesfeld, J. Appl. Electrochem. 22 (1992) 1.[16] S. Gottesfeld, T.A. Zawodzinski, Advances in Electrochemical Science and

Engineering, Wiley-VCH, 1997.[17] E.S. Gottesfeld, in: S.-G. Sun, P.A. Christensen, A. Wieckowski (Eds.), In

in-situ Spectroscopic Studies of Adsorption at the Electrode andElectrocatalysis, 1st edn., Elsevier Oxford, UK, 2007, pp. 247–272, Ch. 8.

[18] A. Marinkas, F. Arena, J. Mitzel, G.M. Prinz, A. Heinzel, V. Peinecke, H. Natter,Carbon 58 (2013) 139–150.

[19] H. Rivera, J.S. Lawton, D.E. Budil, E.S. Smotkin, ECS Meeting Abstracts 802(2008), 890–890.

[20] S. Gilman, J. Phys. Chem. 68 (1964) 70.[21] M. Watanabe, S. Motoo, J. Electroanal. Chem. Interfacial Electrochem. 60

(1975) 275–283.[22] T. Frelink, W. Visscher, J.A.R. van Veen, Surf. Sci. 335 (1995) 353–360.[23] E. Watanabe, S. Nishimura, H. Ogoshi, Z. Yoshida, Tetrahedron 31 (1975)

1385–1390.[24] H.-W. Lei, S. Suh, B. Gurau, B. Workie, R. Liu, E.S. Smotkin, Electrochim. Acta

47 (2002) 2913–2919.[25] J.K. Norskov, T. Bligaard, A. Logadottir, S. Bahn, L.B. Hansen, M. Bollinger, H.

Bengaard, B. Hammer, Z. Slivancanin, M. Mavrikakis, Y. Xu, S. Dahl, C.J.H.Jacobsen, J. Catal. 209 (2002) 275.

[26] N. Dimakis, H. Iddir, R.R. Diaz-Morales, R. Liu, G. Bunker, E.-H. Chung, E.S.Smotkin, J. Phys. Chem. B 109 (2005) 1839–1848.

[27] I. Villegas, M.J. Weaver, J. Phys. Chem. B 101 (1997) 5842–5852.[28] D.K. Lambert, J. Chem. Phys. 89 (1988) 3847–3860.[29] B. Hammer, O.H. Nielsen, J.K. Norskov, Catal. Lett. 46 (1997) 31–35.[30] R. Hammaker, S. Francis, R. Eischens, Spectrochim. Acta 21 (1965)

1295–1309.[31] M. Scheffler, Surf. Sci. 81 (1979) 562–570.[32] G. Mahan, A. Lucas, J. Chem. Phys. 68 (1978) 1344–1348.[33] M. Moskovits, J. Hülse, Surf. Sci. 78 (1978) 397–418.[34] P. Hollins, J. Pritchard, Surf. Sci. 89 (1979) 486–495.[35] E. Reddington, A. Sapienza, B. Gurau, R. Viswanathan, S. Sarangapani, E.S.

Smotkin, T.E. Mallouk, Science (Washington, D.C.) 280 (1998) 1735–1737.[36] L. Liu, R. Viswanathan, R. Liu, E.S. Smotkin, Electrochem. Solid State Lett. 1

(1998) 123–125.[37] R.X. Liu, H. Iddir, Q.B. Fan, G.Y. Hou, A.L. Bo, K.L. Ley, E.S. Smotkin, Y.E. Sung,

[38] R.X. Liu, K. Triantafillou, L. Liu, C. Pu, C. Smith, E.S. Smotkin, J. Electrochem.Soc. 144 (1997) L148–L150.

[39] N. Dimakis, N.E. Navarro, E.S. Smotkin, J. Chem. Phys. (2013) 138.[40] N. Dimakis, T. Mion, E.S. Smotkin, J. Phys. Chem. C 116 (2012) 21447–21458.

2 sis Tod

Chem. Soc. 100 (1978) 6083–6088.[111] N. Buzgar, A. Buzatu, I.V. Sanislav, An. Stiint. U. Al. I-Mat 55 (2009) 5–23.

6 N. Loupe et al. / Cataly

[41] N. Dimakis, M. Cowan, G. Hanson, E.S. Smotkin, J. Phys. Chem. C 113 (2009)18730–18739.

[42] E.S. Smotkin, R.R. Diaz-Morales, Annu. Rev. Mater. Res. 33 (2003) 557–579.[43] J. Kua, W.A. Goddard III, J. Am. Chem. Soc. 121 (1999) 10928–10941.[44] B. Gurau, R. Viswanathan, R.X. Liu, T.J. Lafrenz, K.L. Ley, E.S. Smotkin, E.

Reddington, A. Sapienza, B.C. Chan, T.E. Mallouk, S. Sarangapani, J. Phys.Chem. B 102 (1998) 9997–10003.

[45] N. Dimakis, F.A. Flor, N.E. Navarro, A. Salgado, E.S. Smotkin, J. Phys. Chem. C(2016), http://dx.doi.org/10.1021/acs.jpcc.6b02086.

[46] T.E. Mallouk, B.C. Chan, E. Reddington, A. Sapienza, G. Chen, E.S. Smotkin, B.Gurau, R. Viswanathan, R. Liu, U.S. Patent No. 6,284,402. Washington, DC:U.S. Patent and Trademark Office, 2001.

[47] K. Sapru, K. Hong, M.A. Fetcenko, S. Venkatesan, U.S. Patent No. 4,551,400.Washington, DC: U.S. Patent and Trademark Office, 1985.

[48] H.A. Gasteiger, J. Electrochem. Soc. 1994 (1795) 141.[49] A.E. Russell, S. Maniguet, R.J. Mathew, J. Yao, M.A. Roberts, D. Thompsett, J.

Power Sources 96 (2001) 226–232.[50] S. Mukerjee, J. McBreen, J. Electroanal. Chem. 448 (1998) 163–171.[51] R. Liu, K. Triantafillou, L. Liu, C. Pu, C. Smith, E.S. Smotkin, J. Electrochem. Soc.

144 (1997), 2942–2942.[52] H. Topsoe, J. Catal. 216 (2003) 155–164.[53] B.M. Weckhuysen, Chem. Commun. (2002) 97–110.[54] T. Nagayasu, C. Yoshioka, K. Imamura, K. Nakanishi, J. Colloid Interface Sci.

279 (2004) 296–306.[55] T. Nagayasu, K. Imamura, K. Nakanishi, J. Colloid Interface Sci. 286 (2005)

462–470.[56] E. Herrero, W. Chrzanowski, A. Wieckowski, J. Phys. Chem. 99 (1995)

10423–10424.[57] K. Yoshimi, M.-B. Song, M. Ito, Surf. Sci. 368 (1996) 389–395.[58] M.E. Herron, S.E. Doyle, S. Pizzini, K.J. Roberts, J. Robinson, G. Hards, F.C.

Walsh, J. Electroanal. Chem. 324 (1992) 243–258.[59] J. McBreen, W. O’Grady, K. Pandya, R. Hoffman, D. Sayers, Langmuir 3 (1987)

428–433.[60] B.C. Schardt, J.L. Stickney, D.A. Stern, A. Wieckowski, D.C. Zapien, A.T.

Hubbard, Surf. Sci. 175 (1986) 520–534.[61] S. Mukerjee, S. Srinivasan, M.P. Soriaga, J. McBreen, J. Phys. Chem. 99 (1995)

4577–4589.[62] S. Stoupin, E.A. Lewis, D.E. Ramaker, C.U. Segre, E.S. Smotkin, ECS Meeting

Abstracts 801 (2008), 1155–1155.[63] E. Principi, A. Di Cicco, A. Witkowska, R. Marassi, J. Synchrotron Radiat. 14

(2007) 276–281.[64] C. Roth, N. Martz, T. Buhrmester, J. Scherer, H. Fuess, Phys. Chem. Chem.

Phys. 4 (2002) 3555–3557.[65] M. Tada, S. Murata, T. Asakoka, K. Hiroshima, K. Okumura, H. Tanida, T.

Uruga, H. Nakanishi, S. Matsumoto, Y. Inada, M. Nomura, Y. Iwasawa,Angew. Chem. Int. Ed. 46 (2007) 4310–4315.

[66] A.E. Russell, A. Rose, Chem. Rev. 104 (2004) 4613–4635.[67] J.A. Gilbert, N.N. Kariuki, X.P. Wang, A.J. Kropf, K. Yu, D.J. Groom, P.J. Ferreira,

D. Morgan, D.J. Myers, Electrochim. Acta 173 (2015) 223–234.[68] O. Petrova, C. Kulp, M.W.E. van den Berg, K.V. Klementiev, B. Otto, H. Otto, M.

Lopez, M. Bron, W. Gruenert, Rev. Sci. Instrum. 82 (2011).[69] E. Principi, A. Witkowska, S. Dsoke, R. Marassi, A. Di Cicco, Phys. Chem.

Chem. Phys. 11 (2009) 9987–9995.[70] C. Roth, N. Benker, T. Buhrmester, M. Mazurek, M. Loster, H. Fuess, D.C.

Koningsberger, D.E. Ramaker, J. Am. Chem. Soc. 127 (2005) 14607–14615.[71] R.J.K. Wiltshire, C.R. King, A. Rose, P.P. Wells, M.P. Hogarth, D. Thompsett,

A.E. Russell, Electrochim. Acta 50 (2005) 5208–5217.[72] F.J. Scott, C. Roth, D.E. Ramaker, J. Phys. Chem. C 111 (2007) 11403–11413.[73] S. Maniguet, R.J. Mathew, A.E. Russell, J. Phys. Chem. B 104 (2000)

1998–2004.[74] J. Singh, M. Tromp, O.V. Safonova, P. Glatzel, J.A. van Bokhoven, Catal. Today

145 (2009) 300–306.[75] D.E. Ramaker, D.C. Koningsberger, Phys. Chem. Chem. Phys. 12 (2010)

5514–5534.

ay 283 (2017) 11–26

[76] V. Croze, F. Ettingshausen, J. Melke, M. Soehn, D. Stuermer, C. Roth, J. Appl.Electrochem. 40 (2010) 877–883.

[77] F.J. Scott, S. Mukerjee, D.E. Ramaker, J. Electrochem. Soc. 154 (2007)A396–A406.

[78] R. Diaz-Morales, R. Liu, E. Fachini, G. Chen, C. Segre, A. Martinez, C. Cabrera,E. Smotkin, J. Electrochem. Soc. 151 (2004) A1314–A1318.

[79] Y. Mo, M.R. Antonio, D.A. Scherson, J. Phys. Chem. B 104 (2000) 9777–9779.[80] FTIR studies require a slot cut into the GDL to enable beam access to the

catalytic surface.[81] B. Gurau, E. Smotkin, J. Power Sources 112 (2002) 339–352.[82] V.S. Murthi, R.C. Urian, S. Mukerjee, J. Phys. Chem. B 108 (2004)

11011–11023.[83] Q. Jia, E.A. Lewis, C. Grice, E.S. Smotkin, C.U. Segre, J. Phys. Conf. Ser. 190

(2009) 12157.[84] G.Q. Lu, A. Lagutchev, D.D. Dlott, A. Wieckowski, Surf. Sci. 585 (2005) 3–16.[85] C.S. Kim, C. Korzeniewski, Anal. Chem. 69 (1997) 2349–2353.[86] J. Yoshinobu, N. Tsukahara, F. Yasui, K. Mukai, Y. Yamashita, Phys. Rev. Lett.

90 (2003) 248301.[87] V. Stamenkovic, K.C. Chou, G.A. Somorjai, P.N. Ross, N.M. Markovic, J. Phys.

Chem. B 109 (2005) 678–680.[88] M. Webber, N. Dimakis, D. Kumari, M. Fuccillo, E.S. Smotkin,

Macromolecules 43 (2010) 5500–5502.[89] J. Doan, N.E. Navarro, D. Kumari, K. Anderson, E. Kingston, C. Johnson, A.

Vong, N. Dimakis, E.S. Smotkin, Polymer 73 (2015) 34–41.[90] I. Kendrick, A. Yakaboski, E. Kingston, J. Doan, N. Dimakis, E.S. Smotkin, J.

Polym. Sci. Pt. B: Polym. Phys. 51 (2013) 1329–1334.[91] R.C. Maher, L.F. Cohen, J. Phys. Chem. A 112 (2008) 1497–1501.[92] M.B. Pomfret, J.C. Owrutsky, R.A. Walker, Anal. Chem. 79 (2007) 2367–2372.[93] H.W. Abernathy, E. Koep, C. Compson, Z. Cheng, M.L. Liu, J. Phys. Chem. C

112 (2008) 13299–13303.[94] Z. Peng, P. Huguet, S. Deabate, A. Morin, A.K. Sutor, J. Raman Spectrosc. 44

(2013) 321–328.[95] P. Huguet, A. Morin, G. Gebel, S. Deabate, A.K. Sutor, Z. Peng, Electrochem.

Commun. 13 (2011) 418–422.[96] Y. Tabuchi, R. Ito, S. Tsushima, S. Hirai, J. Power Sources 196 (2011) 652–658.[97] M. Hara, J. Inukai, K. Miyatake, H. Uchida, M. Watanabe, Electrochim. Acta 58

(2011) 449–455.[98] H. Matic, A. Lundblad, G. Lindbergh, P. Jacobsson, Electrochem. Solid State

Lett. 8 (2005) A5–A7.[99] W.A. Rigdon, X.Y. Huang, J. Power Sources 272 (2014) 845–859.

[100] F. Jaouen, F. Charreteur, J.P. Dodelet, J. Electrochem. Soc. 153 (2006)A689–A698.

[101] I. Kendrick, J. Monahan, M. Diem, E.S. Smotkin, ECS Meeting Abstracts, 2012,MA2012-01, 1537.

[102] J. Doan, E. Kingston, I. Kendrick, K. Anderson, N. Dimakis, P. Knauth, M.L. DiVona, E.S. Smotkin, Polymer 55 (2014) 4671–4676.

[103] K.M. Cable, K.A. Mauritz, R.B. Moore, J. Polym. Sci. B: Polym. Phys. 33 (1995)1065–1072.

[104] D.S. Warren, A.J. McQuillan, J. Phys. Chem. B 112 (2008) 10535–10543.[105] C.K. Byun, I. Sharif, D.D. DesMarteau, S.E. Creager, C. Korzeniewski, J. Phys.

Chem. B 113 (2009) 6299–6304.[106] Y. Okamoto, Chem. Phys. Lett. 389 (2004) 64–67.[107] T.K. Das, M. Couture, Y. Ouellet, M. Guertin, D.L. Rousseau, Proc. Natl. Acad.

Sci. U. S. A. 98 (2001) 479–484.[108] R. Holze, Electrochim. Acta 33 (1988) 1619–1627.[109] J.M. Burke, J.R. Kincaid, T.G. Spiro, J. Am. Chem. Soc. 100 (1978) 6077–6083.[110] J.M. Burke, J.R. Kincaid, S. Peters, R.R. Gagne, J.P. Collman, T.G. Spiro, J. Am.

[112] R. Hester, K. Krishnan, J. Chem. Phys. 49 (1968) 4356–4360.[113] M.S. Brogan, T.J. Dines, J.A. Cairns, J. Chem. Soc. Faraday Trans. 90 (1994)

1461–1466.