Pt current collectors arti ﬁ cially boosting praseodymium doped ceria oxygen surface exchange coe ﬃ cients †

The chemical oxygen surface exchange coe ﬃ cient ( k chem ) values used to quantify and rank oxygen evolution reaction (OER) and oxygen reduction reaction (ORR) catalyst performance for high-temperature, oxygen-exchange-enabled devices (such as Solid Oxide Fuel Cells, Solid Oxide Electrolysis Cells, oxygen sensors, etc. ) are often determined electrically, with the aid of precious metal current collectors. However, the curvature relaxation ( k R ) and Time-of-Flight Secondary Ion Mass Spectroscopy (ToF-SIMS) analyses performed here on Pulsed Laser Deposited thin ﬁ lms of the oxygen exchange catalyst Pr 0.1 Ce 0.9 O 2 (cid:1) x (PCO) show that unpolarized platinum current collectors dramatically improve the k chem of Si-contaminated PCO by reducing the Si concentration at the PCO surface and/or di ﬀ using into the PCO; even for PCO thin ﬁ lms only exposed to mild temperatures of 500 (cid:3) C. This suggests that precious metal current collectors are likely responsible for some of the large k chem variation reported in the literature for “ identical ” materials tested under “ identical ” conditions.


Introduction
Solid Oxide Fuel Cells (SOFCs) are solid state chemical to electrical energy conversion devices that exhibit high power density, high energy density, high energy conversion efficiency, and fuel exibility. [1][2][3][4] In addition, SOFCs can be used in reverse as Solid Oxide Electrolysis Cells (SOECs) to store energy, produce chemicals, and/or generate fuels. 4-7 Unfortunately, high system costs, high operating temperatures, and high degradation rates have complicated the commercial deployment of these Solid Oxide Cells (SOCs). 4 With time however, it seems likely that improved materials will help lower these commercialization barriers. Since today's "State-of-the-Art" SOCs are typically limited by poor oxygen ion transport (1) through the bulk of the solid electrolyte, (2) across the solid electrolyte-solid electrode interface, and/or (3) across the solid electrode-gas phase interface, 8 it seems likely that future material sets will feature increased bulk oxygen ion conductivities, reduced interfacial resistivities, and/or increased chemical oxygen surface exchange coefficients.
The chemical oxygen surface exchange coefficient, k chem , is commonly dened as: where J is the oxygen ux across a surface, C o is the oxygen concentration just inside the material, and C s is the oxygen concentration on the surface required to maintain equilibrium with the surrounding atmosphere. 9 As such, k chem can be regarded as a "materials property" that can be used to rank a surface's ability to facilitate oxygen transport between a material's interior and the surrounding environment. For today's most common Mixed Ionic Electronic Conducting (MIEC) oxygen exchange catalysts, all of which have electronic transference numbers close to 1, this ranking can be done through 1) a direct k chem comparison (assuming that the materials have similar lattice oxygen concentrations that vary similarly with oxygen partial pressure (p O2 ), as is oen the case as shown in Fig. S1 of the ESI †), or (2) through its official conversion into a near-open-circuit oxygen surface exchange resistance, R s , via the equation: where k B is Boltzman's constant, T is the temperature in Kelvin, e is the elementary charge of an electron, C o is the oxygen ion concentration just inside the material, and k o is the oxide ion surface exchange rate constant (which is oen approximated as the electrically-determined oxygen surface exchange coefficient, k q (ref. 10)) dened as: Here, k o was assumed to be zk q , as commonly done in the literature 10 102 Bold studies were performed on thin films, italic studies were performed on loose or partially sintered powders, and unformatted studies were performed on bulk pellets with relative densities >92%.
where g o is the thermodynamic factor for oxygen dened as: [11][12][13][14][15] A material's oxygen surface exchange properties (i.e. a material's k chem , k o , k q , R s etc.) can also be combined with SOC structure-property-performance models [16][17][18] to predict SOC electrode polarization resistances [19][20][21] and/or identify optimal SOC electrode microstructures. 22 Other devices (such as solar thermochemical cells, 23 oxide memristors, 24 oxygen separation membranes, 25 catalytic converters, 26 oxygen sensors, 27 etc.) can also have their performance and/or efficiency impacted by the oxygen surface exchange properties of the materials within them.
However, before the scientic community can effectively engineer oxygen surface exchange coefficients for improved device performance, it must rst gure out how to reliably measure them. Specically, Fig. 1 shows that, even for the most common MIEC materials, huge k chem , k o and R s variations exists in literature for "identical" materials under "identical" conditions. For instance, in Fig. 1 there is a >1000 times variation in the extrapolated 700 C Pr 0.1 Ce 0.9 O 2Àx (PCO) k chem values, a >10 000 times variation in the measured 650 C La 0.6 Sr 0.4 -FeO 3Àx (LSF) k chem values, and a >100 000 times variation in the measured 500 C Ba 0.6 Sr 0.4 Co 0.8 Fe 0.2 O 3Àx (BSCF) k chem values.
The reasons for these oxygen surface exchange property variations are manifold. For instance, previous studies have shown that systematic differences in lm crystallinity, 28,29 grain size, 30 surface orientation, 31-34 surface chemistry, [35][36][37][38][39] and/or surface gas phase adsorbates 34,40,41 can all impact a material's measured oxygen surface exchange properties. Experimental complications such as poor gas phase mixing, 42 noninstantaneous atmospheric switching times, 43 and/or large amounts of sample oxygen release that can undesirably alter the atmospheric p O2 (ref. 44) can also produce k chem variations.
However, despite the fact that precious metal surface additions are well known to improve the performance of the MIEC materials used as SOC electrodes, oxygen sensors, vehicular catalytic converters etc., [45][46][47][48] an examination of how current collectors (especially the precious metal current collectors commonly used in the existing oxygen surface exchange literature 36,40,[49][50][51][52][53][54][55][56][57][58][59][60][61][62] ), affect oxygen surface exchange properties is less clear. This uncertainty is perpetuated by the fact that (1) today's most-common k chem , k o , k q, and R s measurement techniques (i.e. Electrical Conductivity Relaxation (ECR) and Electrochemical Impedance Spectroscopy (EIS)) require the electrical polarization of a current collector, and (2) many ECR and EIS studies do not include descriptions of the current collector microstructure, materials, thickness, pattern geometry, and/or electrical-polarization used to perform the oxygen surface exchange measurements. This is problematic because electrically-polarized current collectors could alter oxygen exchange through (1) the enhanced oxygen exchange properties some precious metals have been shown to catalyze at metal-MIEC-air interfaces, 47,[63][64][65] (2) the chemical gettering of surfacesegregated MIEC ions/impurities that some current collecting materials are known to possess, 66 (3) current collector coefficient of thermal expansion (CTE) mismatch stress induced alterations in the point defect and/or surface chemistry that some mechano-chemically-active MIECs are known to exhibit, 67,68 (4) electric-eld induced alterations in the point defect, surface chemistry, and/or metal-MIEC-air triple-phaseboundary widths that are known to occur in some MIECs, 63,68-70 etc.
Although a few studies to the contrary exist, 71 many more studies have found that precious metal surface decoration can signicantly alter a material's measured oxygen surface exchange properties. For instance, ECR measurements by Egger and Sitte 72 demonstrated that a thin layer of silver increased the 600 C k chem of La 2 NiO 4 by an order of magnitude, and Zhang et al. 32 observed similar ECR k chem improvements for lanthanum strontium manganate decorated with platinum or palladium nanoparticles. EIS and oxygen isotope exchange depth proling experiments by Sahibzada et al. 73 showed that palladium surface decoration decreased the 700 C k o of La 0.6 -Sr 0.4 Fe 0.8 Co 0.2 O 3Àx (LSFC) by an order of magnitude but improved the electrochemical performance of LSFC at lower temperatures. Similarly, microelectrode EIS measurements performed by Riedl et al. 74 showed that platinum surface decoration either increased or decreased the R s of La 0.6 Sr 0.4 -FeO 3Àx (LSF) by up to $100 times depending on the p O 2 . Further, Ma and Nicholas 75 found that the 600 C PCO k chem values obtained from current-collector-free curvature relaxation (kR) and current-collector-free optical transmission relaxation (OTR) experiments were identical, but more than 10-100 times lower than those obtained from k chem techniques utilizing precious metals. Similarly, simultaneous $300-600 C measurements by Perry and coworkers 59,76 on strontium titanium ferrite thin lms showed that roughly identical k chem values were obtained from OTR and in-plane ECR measurements utilizing current collectors that only covered a small portion of the MIEC surface, but that $10 times higher k chem values were obtained via through-sample EIS measurements utilizing current collectors covering a large portion of the MIEC surface.
Despite these reports, precious metal current collectors are still being actively used to measure the oxygen surface exchange properties of MIEC materials. 36,[49][50][51][52][53][54][55][56][57][58][59][60] Hence, the rst objective of this work was to dramatically demonstrate the role that even minute amounts of unpolarized precious metals can have on the k chem of MIEC materials. The second objective of this work was to go beyond the aforementioned past studies and identify any local chemistry changes likely responsible for the observed k chem alteration. These objectives were achieved by pairing (1) the remarkable ability of the curvature relaxation technique to obtain direct k chem measurements on PCO thin lms that were either uncovered, partially Pt-covered, or completely Pt-covered, with (2) Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) analyses of the sample chemistry before and aer curvature relaxation.

Experimental methods
Here, PCO thin lms were produced by Pulsed Laser Deposition (PLD), optionally coated with one of the platinum current collector geometries shown in Fig. 2, and tested via curvature relaxation (kR).

Sample fabrication
Prior to thin lm deposition, 25.4 mm diameter, one-sidepolished, 200-micron-thick, (100)-oriented (Y 2 O 3 ) 0.095 (ZrO 2 ) 0.905 (YSZ) single crystal substrates (Crystec GmbH, Berlin, Germany) were air-annealed at 1450 C for 20 h utilizing 5 C min À1 nominal heating and cooling rates to remove any residual stress within them. This wafer annealing schedule was chosen because it had previously been shown capable of reducing the curvature changes observed upon heating these YSZ wafers from 25 to 700 C to less than 0.005 m À1 (so that residual stress changes would not obscure the $0.002 m À1 curvature changes caused by lm reduction or oxidation at a single, constant temperature). 75 In addition, $94% dense Pr 0.1 Ce 0.9 O 2Àd PLD targets were produced from PCO powder fabricated via the glycine nitrate combustion (GNC) method. 77 This GNC process was conducted by rst dissolving 99.9% pure cerium nitrate (Strem Chemicals, Newburyport MA), 99.9% pure praseodymium nitrate (Strem Chemicals, Newburyport MA), and 99% pure glycine (Millipore Sigma, Burlington, MA) in 18.2 MU water (Millipore Sigma, Burlington, MA) at a 1 : 1 glycine : nitrate molar ratio using a Teon coated stir bar (Fischer Scientic, Pittsburgh, PA) in a Pyrex beaker (Fischer Scientic, Pittsburgh, PA). The resulting solutions were then ignited over a hotplate in a stainless-steel reaction vessel (Polar Ware, Kiel, WI). To remove unreacted glycine, the resulting powder was calcined in a 99.8% pure alumina crucible (CoorsTek, Golden, CO) at 1000 C in air for 1 hour using 5 C min À1 nominal heating and cooling rates. The calcined powder was then uniaxially compacted to $63 MPa in a 38 mm stainless-steel die (MTI, Richmond, CA) and sintered at 1450 C for 20 h in air using nominal 3 C min À1 heating and 10 C min À1 cooling rates. The resulting sintered target was then ground down to 25 mm diameter and 2.5 mm thick using 240 grit SiC sandpaper and bonded with silicone to a $2.5 mm thick, 25 mm diameter copper backing.
PCO PLD thin lms were produced from the aforementioned target by depositing $265 nm of PCO onto 700 C-preheated YSZ wafers within 30 mTorr of oxygen using the PLD/MBE 2300 system (PVD Products, Inc) at the Northwestern University PLD User Facility. PLD was conducted using a 100 mm target-to-substrate distance and 15 000, 10 Hz, 200 mJ, 248 nm ΚrF excimer laser pulses. All the PCOjYSZ samples reported here were (1) cooled to room temperature at 10 C min À1 upon the completion of PCO deposition, (2) produced consecutively during a single PLD session, and (3) stored within individual polypropylene wafer containers when not undergoing subsequent processing or characterization. Aer thin lm deposition, each PCOjYSZ sample had its oxygen vacancy defect concentration re-equilibrated with air via an hour-long 1100 C hold utilizing nominal 3 C min À1 heating and cooling rates.
To produce patterned platinum current collectors covering $2.7% of the PCO geometric surface area (such as those for the "partially Pt-covered PCOjYSZ samples shown on the right-hand side of Fig. 2) some of the 1100 C re-equilibrated PCOjYSZ samples were subjected to an acetone rinse, a methanol rinse, a deionized water rinse, nitrogen gas drying, a 5-10 minute excursion to 115 C, 10 seconds of 700 rpm rotation in a spin coater, and nally 30 seconds of 3000 rpm rotation in a spin coater. Next, the exposed PCO surfaces of these samples were covered with Photoresist S1813 (Microchem Corp, Westborough, MA) by slowing the PCOjYSZ sample rotation to 700 rpm, dripping $50 mL of photoresist solution onto the PCOjYSZ sample, removing the PCOjYSZ sample from the spin coater, and baking the spin-coated sample for 1 min at 115 C. Current collector patterns were then introduced into the photoresist via UV curing in a Karl Suss MJB3 mask aligner (SUSS MicroTec SE, Garching, Germany) using a positive mask (Photoscience Inc, Lexington, KY) and 274 watts of 250 nm light for 90 seconds. The UV-exposed portion of the photoresist was then removed by exposing it to photoresist developer MF319 (Microchem Corp, Westborough, MA) for 45 seconds, rinsing it with deionized water, and then baking it in air at 115 C for 5 min. The samples were then placed in an Axiss PVD System (Kurt J. Lesker, Jefferson Hills, PA), and the total chamber pressure was pumped down to 10 À6 torr. Aer backlling with Ar gas until the total chamber pressure was 20 mTorr, and aer initiating 30 rpm of sample rotation, platinum (from a 99.95% pure Pt target) was sputtered onto the sample surface for 1 minute using 200 W of power. Aer removal from the sputtering chamber, these samples were dipped in acetone and then rinsed in isopropanol to remove the photoresist.
To produce "completely" Pt-covered PCOjYSZ samples such as those shown in the le side of Fig. 2 (i.e. those where 100% of the geometric PCO surface area was covered with 89% dense 78 Pt lms), some PCOjYSZ samples were placed in the aforementioned Axiss PVD system and subjected to the same platinum sputtering conditions used for the partially Pt-covered PCOjYSZ samples. However, these completely Pt-covered samples were not subject to the solvent cleaning, photoresist application, UV curing, or photoresist removal processes needed for the partially Pt-covered samples.
To make sure the kR signal of the completely Pt-covered PCOjYSZ samples came from oxygen exchange into/out of the PCO (instead of the Pt), a bare YSZ wafer was completely covered with platinum using the same procedures as those used for the completely Pt-covered PCOjYSZ samples.

Oxygen surface exchange measurements
Here, thin lm PCO oxygen surface exchange coefficient measurements were performed using the kR method described, and derived, previously in the literature. 75,[79][80][81][82] In brief, the curvature relaxation techniques determines k chem from the PCO volume changes accompanying small, isothermal, p O 2 -induced changes in the oxygen nonstoichiometry (Dd) that occur in PCO above $400 C in air via the reaction: 75,83-85 Assuming that as this mechano-chemically-active defect reaction proceeds (1) the oxygen-exchange-active thin lm remains well bonded to an oxygen-exchange-inactive substrate, (2) the lm and substrate only deform in a linear elastic manner, (3) the lm is more than $500 times thinner than the substrate but is thick enough to ensure that its volume changes bend the wafer in a detectable manner, (4) the lm thickness is at least $500 times lower than the lm materials characteristic thickness (L C ¼ D chem /k chem where D chem is the chemical diffusion coefficient) to ensure that the sample response is only controlled by surface exchange and not partially or fully controlled by oxygen diffusion in the bulk of the lm, and (5) the p O 2 step size is small enough to ensure that (a) the oxygen surface exchange process remains linear, and (b) the resulting change in oxygen nonstoichiometry (Dd) is small enough to ensure that the lm strain (3 C ) can be described by: where a C is the chemical expansion coefficient, the PCO oxygen surface exchange coefficient can be mathematically extracted from how quickly the PCOjYSZ sample curvature equilibrates to a new p O 2 by tting the observed curvature relaxation to the following solution to Fick's second law: 9 where k is the instantaneous sample curvature, k 0 is the sample curvature before the relaxation, k N is the sample curvature aer the relaxation, t is time, and h f is the lm thickness. In addition to being a non-contact, in situ thin lm k chem measurement technique that works with, or without, overlying current collecting layers, the kR technique has the benet that no specic lm or substrate materials property values are required to obtain accurate thin lm k chem values via eqn (7) (so long as the lm and substrate have isotropic in-plane mechanical properties, as was the case here).
To determine if multiple processes with different time constants (or different rates of the same process in different parts of the sample) were active, all curvature data was replotted to search for slope changes in ln vs. time plots during each relaxation, as has been done previously. 31,79,86-88 Fig. 3 shows a schematic of the controlled atmosphere multibeam optical stress sensor setup used to perform the kR measurements reported here (see Nicholas 81 for the exact dimensions). In preparation for kR experiments, each PCOjYSZ sample was placed atop the central quartz support tube shown in Fig. 3, covered with the controlled-atmosphere gaspreheating manifold shown in Fig. 3, heated to 500 C at 5 C min À1 , and held at 500 C for 1 hour in air to equilibrate and homogenize the PCO oxygen vacancy concentration. Then, the uncovered, partially Pt-covered, and completely Pt-covered PCOjYSZ samples were measured in 25 C increments from 675 to 725 C, 500 to 725 C, and 500 to 425 C, respectively, aer holding for at least 30 minutes at each temperature to promote thermal equilibrium. Curvature relaxations were triggered by switching the p O 2 around each sample from 100 sscm of 0.21 (21% O 2 -79% Ar, i.e. synthetic air) to 100 sccm 0.021 (2.1% O 2 -97.9% Ar, i.e. 10 times diluted synthetic air) using a four-way valve. Since the same synthetic gas mixtures were used to test all the samples, atmospheric impurity content differences should not have to contributed to the observed differences in sample behavior. As shown in Fig. S2-S4 of the ESI, † each sample was allowed to re-equilibrate at each p O 2 for at least 5 times the characteristic relaxation time (to allow for reliable kR tting, as explained in den Otter et al.). 89 Only k chem values that were not reactor ush time limited (i.e. those where switching between two 100 sccm gas ows or between two 300 sccm gas ows yielded the same k chem values) were reported here.

Additional sample characterization
To evaluate PCO thin lm phase purity, X-Ray Diffraction (XRD) was performed in room temperature air from 20 to 80 with a 0.01 step size and a 1 s per step dwell time using unltered Cu k a radiation from a SmartLab diffractometer (Rigaku, The Woodlands, TX) operated at 44 kV and 40 mA with 10 mm-wide detector and source slits.
To evaluate PCO thin lm surface composition, X-Ray Photoelectron Spectroscopy (XPS) was performed at room temperature and 10 À9 torr on uncovered PCOjYSZ samples aer (1) post-deposition re-equilibration in air and (2) aer curvature relaxation. XPS was performed using the Al Ka X-ray radiation inside a Phi 5600 XPS (PerkinElmer, Waltham, MA). XPS survey scans were collected with a step size of 0.4 eV and a pass energy of 187.5 eV. XPS peak positions were calibrated with the carbon 1s peak at 284.8 eV. Due to concerns that the vacuum encountered during XPS analysis might alter the PCO oxygen vacancy concentration, the samples used for XPS analyses were not used for kR experiments.
To evaluate lm thickness, Scanning Electron Microscopy (SEM) was performed on fractured PCOjYSZ samples coated with $5 nm of Ti and imaged using a TESCAN MIRA3 Field Emission SEM (Tescan, Brno, Czechia).
To evaluate PCO thin lm bulk composition, PCOjYSZ samples were analyzed via ToF-SIMS depth proling conducted by EAG Labs (East Windsor, NJ, USA). The absolute concentrations of Zr, Y, and O in the lms were calculated from the relative concentrations produced by SIMS by setting the SIMSdetected concentrations in the bulk of the YSZ substrate equal to those found in (Y 2 O 3 ) 0.095 (ZrO 2 ) 0.905 . Likewise, the absolute concentrations of Pr and Ce were calculated from the measured SIMS proles by offsetting each Pr and Ce SIMS prole until the Pr and Ce concentrations in the interior of the lm equaled those found in Pr 0.1 Ce 0.9 O 2 . Likewise again, the Pt concentration proles were offset until those in the Pt current collector equaled those found in dense Pt metal. In contrast, a single Ta 2 O 5 standard with a known amount of Si was used to estimate the Si concentrations in all the PCO lms, assuming that the SIMS matrix effects in Ta 2 O 5 and PCO were similar. Even if this assumption was incorrect and as a result introduced errors in the absolute Si concentration values, the fact that the same Si calibration, SIMS collection settings, and SIMS equipment were used for all the samples suggests that any relative Si concentration differences between the samples, and within each sample, should be valid. Fig. 4 shows representative XRD scans indicating that only crystalline, uorite-structured PCO and crystalline, uoritestructured YSZ peaks were present in the PCOjYSZ samples. Consistent with previous literature reports, 75,[90][91][92][93][94][95] all the PCO lms here exhibited (100) preferred orientation, consistent with the (100) oriented YSZ substrates they were grown upon. Fig. 5 shows representative XPS scans indicating that the sample fabrication procedures used here produced "clean" PCO surfaces containing only Ce, Pr and O. No Si, Zr, Y or any other element was found in any of the XPS scans, even for uncovered PCOjYSZ samples kR tested to 725 C in a Si-containing kR test rig known to vapor deposit Si on samples held for extended times at and above 600 C. 96 This suggests that any surface impurities in the PCOjYSZ samples here were present in amounts below the $1 atomic percent XPS detection limit. 97 Fig   YSZ and ceria under the conditions used here, 99 a h f /h S < 0.002, a h f < 0.0001 Â L C (based on the reported 450-725 C PCO L C values, 90 D chem activation energies 100,101 and k chem activation energies 51,75,90,102 ), and the reproducible curvature behavior with p O 2 switching for all the PCOjYSZ samples tested here and shown in Fig. S5 of the ESI, † ensured that all the assumptions needed to ensure the validity of eqn (7) were met. Fig. 6 shows that covering either $2.7% or 100% of the PCO surface with unpolarized Pt current collectors signicantly boosted the PCO oxygen surface exchange coefficient. Further, uncovered PCOjYSZ samples measured at the beginning (green circles) and end (red triangles) of the k chem measurement series yielded identical k chem values, attesting to the high reproducibility of the sample fabrication and measurement procedures used here. Interestingly, compared to uncovered PCOjYSZ samples from Ma and Nicholas 75 that were prepared in a different PLD chamber, grown at a lower temperature, and (unlike the present samples) alkaline etched to remove surfacesegregated impurities, the uncovered PCOjYSZ samples here had lower k chem values. However, the $0.6 eV activation energies observed for the present study's uncovered PCOjYSZ samples were similar to those measured above 500 C in the curvature relaxation measurements on uncovered PCO PLD thin lms from Ma and Nicholas, 75 the optical transmission relaxation measurements on uncovered PCO PLD thin lms from Chen et al., 102 the electrochemical impedance spectroscopy measurements on Ag-paste-covered PCO PLD thin lms from Chen et al., 90 and the optical transmission relaxation measurements on uncovered PCO PLD thin lms from Nicollet et al. 51 This suggests that the same rate-determining step for oxygen incorporation into PCO was active above $500 C for all the PCO samples here and in the literature, even its rate was faster in some samples versus others.

Results and discussion
In contrast, Fig. 6 shows that at temperatures less than $500 C, the completely Pt-covered PCOjYSZ samples (blue diamonds) had activation energies essentially double those of the uncoated PCOjYSZ samples measured above $500 C. Similarly, although not mentioned by the authors, high sub-500 C PCO activation energies of $1.3 eV can also be seen in the optical transmission spectroscopy data of Chen et al. 102 on uncovered (i.e. Pt-free) PCO PLD thin lms that transitioned to k chem activation energies of $0.5 eV above $500 C. (Unfortunately, the exceeding-fast, low-temperature oxygen exchange kinetics of the completely Pt-covered PCOjYSZ samples meant that reactor ush time limitations prevented them from being measured above 500 C, when their k chem > 1 Â 10 À6 cm s À1 ). The fact that this previously unrecognized k chem activation energy transition was seen in Chen et al.'s 102 samples without platinum current collectors indicated that it is a general characteristic of PCO, and is not related to the absence or presence of precious metal current collectors. The observation of a higher activation energy at lower temperatures is the opposite of (1) what is commonly observed in other doped ceria compositions, 103 and (2) what might be expected based on the idea that lower temperatures would de-activate higher activation energy barrier pathways for oxygen exchange into/out of a static Fig. 7 kR behavior of a completely Pt-covered PCOjYSZ sample at 475 C compared to a partially Pt-covered PCOjYSZ sample at 725 C and an uncovered PCOjYSZ sample at 725 C. The dashed lines are eqn (7) fits to the data. structure. Hence, it is likely that the $500 C transition to a higher k chem activation energy upon cooling below 500 C is related to a change in the rate-determining step for oxygen exchange brought about by the increased difficulty in exchanging oxygen as the thin lm PCO Pr 3+ /Pr 4+ ratio and oxygen nonstocichometry rapidly trend toward zero below $500 C. 75 Regardless of the exact reason for this newly-recognized 500 C activation energy change, Fig. 6 makes it clear that completely covering the PCO surface with sputtered Pt dramatically boosts the measured k chem . The fact that sputtered Pt did not reduce the measured k chem is consistent with the thin nature of the sputtered Pt current collectors, the low grain boundary and/or surface-path resistivities observed previously for oxygen transport in sputtered Pt lms, 64,104 and the likelypercolated $11% porosity of the sputtered Pt lms 78 facilitating gaseous oxygen transport to the PCO surface. Fig. 6 also shows that the partially Pt-covered PCOjYSZ samples exhibited k chem values between those of the uncovered and completely Pt-covered PCOjYSZ samples. With only $2.7% of the PCO surface covered by Pt, the observed k chem enhancements were modest and hence the partially Pt-covered PCOjYSZ samples could only be measured above 500 C, where they displayed activation energies similar to those measured previously. Fig. 7 provides a graphical summary of just how much faster oxygen exchange was in the completely Pt-covered PCOjYSZ samples than in either the uncovered or partially Pt-covered PCOjYSZ samples. Specically, even though the temperature for the completely Pt-covered PCOjYSZ curvature relaxation is 250 C lower than the uncovered or partially Pt-covered PCO-jYSZ samples in Fig. 6, the time required to complete the oxygen exchange process is $2 times less. Fig. 8 suggests that the platinum current collectors used here improved the PCO k chem by removing Si from the PCO surface and/or diffusing into the PCO. Specically Fig. 8a and b show that the uncovered PCO lms had approximately 1 Si atom for every 100 Pr atoms in their bulk and had enriched amounts of Si at their air-PCO and PCO-YSZ interfaces. This Si contamination probably resulted from the silica-based glassware used to produce the PCO powder utilized in PLD target fabrication (this contamination could likely have been avoided through the use of polyethylene beakers but was retained here to illustrate the interaction common processing and environmental contaminants can have with PCO platinum current collectors). Si enrichment at the air-PCO interface may help explain the low k chem values observed for the uncovered PCOjYSZ samples, since surface-segregated Si impurities are well-known for inhibiting oxygen surface exchange and bulk oxygen transport in PCO and a variety of other ceria-based compositions. 96,105-109 Fig. 8a and b also show a slight enhancement of Zr and Ce at the PCO surface, but since those elements occurred at concentrations 10 to 100 times less than the Si concentration, Si was assumed to be responsible for the low k chem values observed in the uncovered PCOjYSZ samples. Fig. 8a and b also show interdiffusion of Pr, Ce, Zr, and Y between the lm and substrate, and diffusion of Si from the lm into the substrate, the majority of which presumably happened during the 1 hour, 1100 C re-equilibration in air. Interestingly, a comparison of Fig. 8a and b shows that although the bulk Si level in the lm and near the exposed surface roughly doubled (presumably due to silica vaporization from the silica kR test rig known to occur at and above $600 C) 96,105 and some of the bulk Si migrated to the PCOjYSZ interface, no other signicant changes in the sample chemistry, or chemical distribution, occurred during kR testing up to 725 C. Fig. 8c suggests that the completely Pt-covered PCOjYSZ current collectors applied here cleaned the air-PCO interface by serving as a chemical getter for the surface-segregated Si, Zr and Ce present on uncoated lms. The ability of Pt current collectors to act as Si getters to improve the PCO k chem is not completely surprising since (1) Ma and Nicholas 96 recently quantied the relationship between the PCO Si surface content and k chem , and (2) Zhao et al. 105 found that La or Sm PCO surface additions improved k chem by gettering surface Si. However, it is also possible that the near-surface dips in the Si, Zr, and Ce proles could be artifacts caused by the SIMS front transitioning from the metal to the oxide.
In addition to suggesting that Pt cleared Si from the PCO surface, Fig. 8c shows that signicant quantities of Pt diffused into the PCO lm (with the Pt concentration exceeding the Pr doping level for more than 50 nm into the PCO lm). This is somewhat surprising because the completely Pt-covered PCO-jYSZ samples only saw temperatures up to 500 C during kR testing. However, this behavior is consistent with prior reports that platinum can diffuse into ceria at room temperature, 110 especially when ceria experiences reducing conditions, 111 as likely occurred during the Pt current collector sputtering process.
Whether from the removal of Si from the PCO surface, Pt diffusion into the ceria, or some of the other mechanisms proposed in the literature, 63,111,112 the results here make it clear that completely covering a signicant portion of a material's surface with Pt current-collectors can led to k chem values signicantly different from those the material would display on its own. As such, the small k chem enhancements observed here for the partially Pt-covered samples in Fig. 6 were assumed to be due to the portion of the PCO close to the Pt current collectors behaving like the completely Pt-covered PCOjYSZ samples, while those portions of the PCO further away from the Pt current collectors behaved more like the uncoated PCOjYSZ samples. Fig. 9 compares the curvature response of a completely Ptcovered PCOjYSZ sample to a completely Pt-covered YSZ wafer. The fact that essentially no curvature response was observed in the completely Pt-covered YSZ wafer with p O 2 switching conrms that the observed k chem enhancements were not the result of oxygen exchange into/out of the bulk of the Pt, but instead were related to Pt current collectors inuencing oxygen exchange into/out of the PCO.

Conclusions
Despite many studies suggesting that precious metal current collectors and/or impurities can impact the performance of oxygen-exchange-enabled devices, [45][46][47][48] precious metal current collectors are still routinely used to transport electronic species into/out of oxygen exchange materials in Electrical Conductivity Relaxation, 36,[49][50][51][52][53][54][55][56][57][58] and Electrical Impedance Spectroscopy 59,60 oxygen exchange experiments. The results here demonstrate that precious metal current collectors are not necessarily inert (i.e. they can diffuse into and/or chemically clean the surface of the oxygen exchange material of interest), can unexpectedly alter the measured k chem values, and, hence, should be treated with caution when used for oxygen surface exchange measurements. This is especially true when performing oxygen exchange measurements on either bulk or thin lm samples where a signicant fraction of the oxygen-exchange-active surface is in close proximity to a precious metal current collector (such as ECR k chem measurements on materials with low electronic resistivities that require the close placement of interdigitated electrodes to get measurable electronic conductivities, experiments performed with ne-grained colloidal precious metal pastes and/or precious metal thin lms applied over most of the oxygen-exchanging surface, etc.). Further, if precious metal current collectors dramatically improve the oxygen exchange kinetics of the neighboring/underlying MIEC, articially high k chem values could also be recorded for electrical (ECR, EIS, etc.) k chem measurements performed using small precious metal current collectors placed a signicant distance apart (since the precious-metal-activated portions of the MIEC surface would act as "major leaks" for oxygen incorporation into the bulk of the material and hence would "drown-out" the response from slower oxygen incorporation along the bare MIEC surfaces). For all these reasons, it is likely that precious metal current collectors are responsible for a signicant portion of the k chem variation reported in the literature for "identical" materials tested under "identical" conditions.

Author contributions
Y. Ma performed the thin lm deposition and characterization, T. E. Burye performed the k chem literature review, and J. D. Nicholas conceptualized and led the write-up of this work.

Conflicts of interest
There are no conicts to declare.