Probing Grain-Boundary Chemistry and Electronic Structure in Proton

Oct 3, 2016 - Nikolay A. Danilov , Julia G. Lyagaeva , Dmitry A. Medvedev , Anatoly K. Demin , Panagiotis Tsiakaras. Electrochimica Acta 2018 284, 551...
1 downloads 0 Views 3MB Size
Download PDF
Letter pubs.acs.org/NanoLett

Probing Grain-Boundary Chemistry and Electronic Structure in Proton-Conducting Oxides by Atom Probe Tomography Daniel R. Clark,*,†,‡ Huayang Zhu,§ David R. Diercks,‡ Sandrine Ricote,§ Robert J. Kee,§ Ali Almansoori,∥ Brian P. Gorman,‡ and Ryan P. O’Hayre*,†,‡ †

Renewable Energy Materials Research Science and Engineering Center (REMRSEC), Colorado School of Mines, 1500 Illinois St., Golden, Colorado 80401, United States ‡ Center for Advanced Ceramics, Metallurgical and Materials Engineering, Colorado School of Mines, Golden, Colorado 80401, United States § Department of Mechanical Engineering, Colorado School of Mines, 1500 Illinois St., Golden, Colorado 80401, United States ∥ Department of Chemical Engineering, The Petroleum Institute, Post Office Box 2533, Abu Dhabi, United Arab Emirates S Supporting Information *

ABSTRACT: A laser-assisted atom-probe-tomographic (LAAPT) method has been developed and applied to measure and characterize the three-dimensional atomic and electronic nanostructure at an yttrium-doped barium zirconate (BaZr0.9Y0.1O3−δ, BZY10) grain boundary. Proton-conducting perovskites, such as BZY10, are attracting intense interest for a variety of energy conversion applications. However, their implementation has been hindered, in part, because of high grain-boundary (GB) resistance that is attributed to a positive GB space-charge layer (SCL). In this study, LAAPT is used to analyze BZY10 GB chemistry in three dimensions with subnanometer resolution. From this analysis, maps of the charge density and electrostatic potential arising at the GBs are derived, revealing for the first time direct chemical evidence that a positive SCL indeed exists at these GBs. These maps reveal new insights on the inhomogeneity of the SCL region and produce an average GB potential barrier of approximately 580 mV, agreeing with previous indirect electrochemical measurements. KEYWORDS: Atom probe tomography, proton-conducting ceramics, grain boundary characterization, space-charge layer

P

technique is here applied to the specific case of BZY10, the analysis method is quite general and can be applied to examine the electronic properties of other materials interfaces as well. The physical and chemical reasons for high grain-boundary (GB) resistance in barium zirconates (or other ion-conductors) are not well-understood. Hypotheses include increased barium depletion at the GB possibly forming ZrO2 under extreme conditions,5,6 increased disorder in the GB region,4 dopant/ impurity segregation,13−16 or intrinsic charge at the GB due to preferential oxygen segregation.7−12,17−22 Recent density-functional-theory (DFT) studies also predict positively charged oxygen-vacancy segregation at GB interfaces due to differences in defect formation energies between the bulk and the GB.19−22 Direct experimental measurements of atomic-scale GB structure and chemistry contribute greatly to validating or refuting alternative hypotheses. The LAAPT approach provides nanometer-scale quantification of cation composition and oxygen stoichiometry. These small relative changes in stoichiometry cannot generally be quantified with alternative

roton-conducting perovskite (ABO3) ceramics, of which yttrium-doped barium zirconate (BaZr0.9Y0.1O3−δ, BZY10) is a prototypical example, are poised to play transformative roles in intermediate-temperature (300−600 °C) energyconversion technologies such as fuel cells, electrolyzers, and membrane reactors.1−3 However, high grain-boundary resistance (as much as 1000 times greater than the bulk)4 can significantly reduce total conductivity, impeding practical and commercial deployment in applications that require high proton conductivity. The present study is concerned with developing a fundamental atomic-level understanding of this grain-boundary behavior. To this end, laser-assisted atom probe tomography (LAAPT) is used to interrogate the grainboundary region of BZY10 in three dimensions with nanometer spatial resolution and parts-per-million chemical sensitivity. The application of computational modeling to the LAAPTderived data, via the solution of a three-dimensional Poisson− Boltzmann equation, further enables characterization of the electrostatic-potential fields in the grain-boundary region. This novel LAAPT approach is very different from typical methods to assess the electronic properties of grain boundaries such as electrochemical impedance spectroscopy,4−9 direct-current-bias techniques, 10,11 or reduction experiments.12 While the © XXXX American Chemical Society

Received: July 14, 2016 Revised: September 7, 2016

A

DOI: 10.1021/acs.nanolett.6b02918 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 1. Concentration contour maps of BZY10 grain boundary collected by laser-assisted atom probe tomography. (a) TEM micrograph of BZY10 samples pre- and post-LAAPT. (b) 3D reconstruction showing the Y, YO, and YO2 ionic species (not all atoms are shown for simplicity) and the region of interest (with average charge density contour map) generated from the contour maps in c. (c) Concentration maps (at %) of the grain boundary for (from left to right) barium, zirconium, yttrium, and oxygen. Red corresponds to regions of higher concentration, and blue corresponds to lower concentrations (see scale bar).

LAAPT was performed in a Cameca LEAP 4000X Si system equipped with a UV laser (neodymium:yttrium vanadate laser, frequency tripled to 355 nm wavelength) and a spot size less than 1 μm at peak full-width half max (fwhm). The laser energy was 10 pJ with a pulse frequency of 625 kHz and a specimen base temperature of 40 K. Data analysis was performed using Cameca’s IVAS v.3.6.6 software. The specimen reconstruction was generated using the tip profile method (more detail found in the Supporting Information). For compositional analysis, the highest three peaks were manually ranged at full-width tenth maximum (FWTM) of the peak (the intensity of each peak was determined, and the upper and lower bounds were ranged at 10% of the maximum counts). These peaks were Ba2+, ZrO+2, and O2+. The other peaks in the mass spectra were ranged to just above background levels. Example mass spectra are shown in Supporting Information, Figure S1. The peak at 17 Da OH+ was ignored and assumed to be adsorbed water, along with other hydrogen and water signals (1, 2, 3, 18, 19, and 33 Da). Although proton-conducting oxides contain mobile protonic (hydrogen) defects, their concentration is negligible compared to external hydrogenic contamination signals from water in the chamber and on the specimen tip and holder and hydrogen permeating through the stainless steel vacuum chamber. Oxide stoichiometry was determined using the isotopic decomposition of peaks method to account for peak overlap. Overlapping peaks included 8 9 Y + + + / 9 0 Zr + + + , 1 6 O 2 + / 9 0 Zr 1 6 O + + + , 89 16 2+ 90 16 2+ Y O / Zr O , and 89Y16O+/90Zr16O+. Software-based background correction was accomplished using the local mass based correction algorithm. See Supporting Information for additional information on material synthesis and atom probe data analysis and calculating SCLs from literature. APT-Quantification of Local Grain-Boundary Chemistry. As it is one of the most well-studied and technologically relevant proton-conducting oxides, 10 mol % yttrium-doped barium zirconate, or BZY10, was chosen for this study. Bulk polycrystalline BZY10 samples were synthesized using the inexpensive, commercially relevant solid-state reactive sintering

techniques such as aberration-corrected analytical scanning transmission electron microscopy (STEM). LAAPT uses positively biased sharp needle-shaped specimens. Elemental or molecular ions are field-evaporated from the specimen’s surface during a laser pulse. The LAAPT instrument measures the vaporized species’ times-of-flight and the spatial positions where they strike a microchannel plate.23 Using a series of successive pulses, a volume of material is removed, from which a three-dimensional assembly of the atomic identities and positions can be reconstructed. Until recently, atom probe tomography (APT) was restricted to highly conductive materials (usually metals). However, LAAPT enables characterization of insulators and semiconductors (i.e., materials with conductivities below 100 S cm−1). Thus, directly reconstructing GBs at the nanoscale, including oxygen nonstoichiometry, is now possible for oxides such as BZY10. The present study is the first to report three-dimensional atomic and electrostatic-potential fields within the GB region of a proton-conducting ceramic. Laser-Assisted Atom Probe Tomography. Atom probe specimens were prepared from bulk samples with a dual-beam FEI Helios NanoLab 600i FIB using a standard wedge lift-out technique and annular milling patterns.24 The specimens were mounted onto a copper 200 mesh grid, which had been cut in half and each post prethinned in the dual-beam FIB.25 This grid was mounted in a TEM/APT compatible holder that allowed for analysis on both the TEM for imaging and the atom probe for elemental characterization.25 TEM micrographs are in an all-beam condition and were collected in a Phillips CM200 operating at 200 keV. Iterations between the FIB (sharpening the tip) and the TEM were made in order to position grain boundaries 1400 °C) of the BZY10 specimen and were “frozen” during the subsequent hydration at 600 °C RT.28 The mobile defect concentrations within the GB region can be evaluated from eqs 1 and 2 as ⎛ z Fϕ(x) ⎞ [Xk](x) = [Xk∞]exp⎜ − k ⎟ ⎝ RT ⎠

three-dimensional Poisson−Boltzmann equation is solved over the entire domain, using partial charge densities as measured by LAAPT. For more information on the treatment of the raw data and the measurement, see Supporting Information. Figure 2a projects the results of the average two-dimensional charge-density distribution perpendicular to the GB, obtained

(3)

where ϕ(x) = Φ(x) − Φ∞ is the local electrostatic potential difference within the GB region. The electrostatic potential in the bulk grain Φ∞ is an arbitrary reference potential. Both the mobile and the immobile species contribute to the net local charge density ρ, which can be expressed as

ρ = ρim +

∑ zkF[Xk]

(4)

k

where ρim is the local charge density due to the immobile charge species. Considering the charges from the single mobile defect species (OH•O) present in BZY10 under the sample preparation conditions (zOH•O = 1), the net local charge density can therefore be expressed as ρ = ρim + F[OH•O]

Figure 2. Charge density and grain boundary potential for BZY10. Map of average charge density (C/m3) with lines indicating positions of scans across the grain boundary for solving the Poisson−Boltzmann equation. Red corresponds to regions of higher positive charge density, and blue corresponds to a lower negative charge density (see scale bar).

(5)

Because the immobile charge species (i.e., YZr ′ and V•• O ) can be measured spatially by LAAPT, only the mobile protonic charge density contribution is unknown. The challenge is thus to evaluate (or at least estimate) the distribution of the mobile protonic carriers, and from that compute the total chargedensity map, and hence the electrostatic potential map. The local charge density ρ is related to the electrostaticpotential distribution ϕ by the Gauss equation as ∇·ε0εr∇ϕ = −ρ

via the analysis discussed above. This result directly confirms, for the first time, the positive GB space-charge zone predicted for BZY10 based on EIS measurements and DFT calculations.4−9,16,19−22 Furthermore, Figure 2 reveals that the charge distribution is highly nonuniform. To gain greater insight into the spatial variation of the GB charge density, a series of line scans (the orange lines in Figure 2) were taken across the GB. These line scans were subsequently converted to potential profiles via numerical solution of eq 7 (discussed later). It should be noted that the overall bulk, as shown in Figure 2, was measured to have a negative charge density. This is due to an inherent collection problem of Ba which evaporates more easily in the LAAPT experiment, causing some loss of measured atoms due to evaporation between collection pulses. The Ba deficit was corrected for by imposing charge neutrality on the overall material and is discussed in further detail in the Supporting Information. To further illustrate the spatial variation in GB charge density, two-dimensional projections of the charge density were extracted along the plane of the GB at different planes approaching the feature (Figure 3). The charge density varies significantly at the nanometer-scale both along the plane of the GB and across the GB. This finding directly contrasts with current analytical models of the SCL that assume a monotonic and smooth GB potential barrier. The effective barrier confronted by a proton as it migrates across this highly heterogeneous GB region likely represents a nonlinear average of the high and low charge density regions modified by the tortuosity of the charge density landscape. Continuing the approach discussed above, the charge density data in Figure 2 are converted into electrostatic potential using

(6)

where ε0 is the vacuum dielectric constant and εr is the relative dielectric constant. For BZY10, the relative dielectric constant εr is taken to be 45.0 (see Supporting Information). Combining eqs 3 and 5 makes clear that the charge-density map ρ(x) is a function of the charged-defect concentrations [Xk](x), and hence the electrostatic-potential map ϕ(x). Thus, the Poisson− Boltzmann equation emerges from eq 6 as ⎡ ∇·ε0εr∇ϕ = −⎢ρim + ⎢⎣

⎛ zkFϕ ⎞⎤ ⎟⎥ RT ⎠⎥⎦

∑ zkF[Xk∞]exp⎜⎝− k

(7)

Given a suitable domain with appropriate boundary conditions, this three-dimensional nonlinear elliptic partial differential equation can be solved computationally for the electrostatic-potential distribution ϕ(x). The computational domain is defined to encompass the GB region and extend into the bulk grains. Far from the GB, the electrostatic potential vanishes. Once the electrostatic-potential distribution is found, the unmeasured defect concentrations can be recovered from eq 3. Typical models of GB behavior divide the domain into a core region, a SCL, and a bulk crystal. The models typically seek to resolve the SCL. The present model is different inasmuch as there is no explicit separation of the core and SCLs. Rather, the D

DOI: 10.1021/acs.nanolett.6b02918 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 3. Projections of charge density along the plane of the grain boundary. Vertical cuts a−e representing projections of charge density (C/m3) along the plane of the grain boundary (defined as peak charge density). The upper left shows the charge density of the grain boundary (shown in Figure 2), and the orange lines (a−e) represent 1 nm apart cuts perpendicular to the grain boundary. These are projected in the right showing areas of high (red) and low (blue) charge density (50 nm × 20 nm panels).

the Poisson−Boltzmann equation (eq 7). The results of this analysis for the one-dimensional line scans across the GB designated in Figure 2 are summarized in Figure 4, which enables statistical quantification of the average GB SCL barrier height and standard deviation for this GB. Figure 4b shows an average GB potential value of 580 mV, with a 1σ standard deviation of ±470 mV (calculated from the peak values of each line scan). As discussed previously, a simple average of the GB

potential barriers derived from line scans across the GB is an imperfect description of the effective proton transport barrier given the significant nonmonotonic, three-dimensional variation in charge density shown in Figures 2 and 3. Nevertheless, this simple average GB potential barrier height of 580 mV is consistent with previous literature estimates derived from electrical measurements and/or DFT calculations, which typically fall in the range of 400−700 mV for BZY10.5−12,28 The approximate 5−7 nm (at 2σ) thickness of the positively charged SCL region (δgb) is also in good agreement with previous estimates.5−9,11,12,19,22 A comparison of the SCL values obtained in this work versus previous studies using alternative techniques is provided in Supporting Information, Table S1. Figure 4c illustrates the resulting severe depletion of protons in the grain-boundary region, also predicted from the SCL theory. As was discussed earlier in the context of the apparent width of the oxygen-ion depletion region, we note again that the GB is curved in the XY direction as well as the YZ direction, which likely causes apparent broadening of the GB region. The large variation in the SCL properties observed across even a single GB as revealed by this LAAPT study is important to note. Some of the charge density variation can be attributed to statistical uncertaintyfor example, the APT detector counts approximately 40% of the ion species, sometimes producing random spikes in certain ions and leading, e.g., to a variation of approximately 0.7 at. % for oxygen. (The uncertainty per contour point is discussed further in Supporting Information.) However, even when accounting for variations due to statistical uncertainty (which lead to an approximate uncertainty of 1% of the charge density-discussed further in Supporting Information), the results of Figures 2−4 clearly suggest that SCL potential is spatially heterogeneous and can vary significantly (by a factor of 2 or more) over short (nmscale) distances. In fact, we observe that the space-charge potential at the GB reduces almost to zero in a few locations, although it never reverses polarity. SCL estimates obtained from AC impedance or DC bias techniques typically represent an ensemble average across many grain boundaries. This LAAPT-based study suggests that the classic “brick layer” model for normalizing GB conductivity with grain size is not always accurate39 and motivates potential extensions to standard approaches for modeling GB resistance. For example, a series of two-dimensional charge density maps

Figure 4. Grain-boundary potential and proton distribution for BZY10. (a) Potentials (V) across the grain boundary for each individual line scan shown in Figure 2. (b) Average potential (V, black) and standard deviation (gray) of the potential (V) across the grain boundary. (c) Proton charge density (C/m3) across the grain boundary, excluding line scans that included zero values (outside of the tip, lines 1, 2, 10, and 11). E

DOI: 10.1021/acs.nanolett.6b02918 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Foundation, Protonic Capital Funds, and the Office of Naval Research Grant No. N00014-08-1-0539. The atom probe instrument used in this research was supported under NSF Award No. DMR-1040456.

parallel to the GB, such as the one shown in Figure 3, could be used to seed a parallel resistor network model to calculate effective GB resistance. Although more protons will clearly flow through regions of lower charge density, the relative area and spatial tortuosity of these regions will also strongly influence the overall effective resistance of the GB. The LAAPT results also reveal that the GB possesses distinct curvature which may create local lattice distortions or varying coincident sites along the boundary, ultimately affecting the nanoscale charge density. This finding suggests new avenues for DFT studies, which generally assume a flat interface with constant charge density.19−22 Finally, future LAAPT configurations, when combined with in situ TEM, might enable the exciting possibility to study the influence of GB orientation on GB chemistry and SCL behavior. Laser-assisted atom-probe-tomography (LAAPT) provides the unique opportunity to reconstruct grain-boundary chemistry in three dimensions with subnanometer spatial resolution. This information provides the basis to extract the local charge density and electrostatic-potential distribution in the grainboundary region, enabling the direct visualization of these properties. Using LAAPT, yttrium and oxygen vacancy accumulation and zirconium depletion at a grain boundary are quantified for the first time in the archetypical protonconducting material BZY10. The oxygen vacancies and yttrium dopant, which segregate the grain boundary and are trapped after sintering, cause a positively charged potential barrier, impeding proton transport. Based on subsequent analysis of the LAAPT data, we calculate an average space-charge potential barrier, taking into account mobile proton charge carriers, of approximately 580 mV and an average space charge layer thickness of 5−7 nm (at 2σ) for a single BZY10 grain boundary. These values are in close agreement with previous results from electrical measurements and DFT calculations.4−11,16,19−22 Importantly, however, the charge density is observed to be nonuniformly distributed both across and along the plane of the grain boundary. LAAPT proves to be a useful technique that can be expanded to give further understanding on the effects of processing, dopants, and impurities on grainboundary chemistry and local charge/potential.

■

■

(1) Takahashi, T.; Iwahara, H. Energy Convers. 1971, 11 (3), 105− 111. (2) Iwahara, H.; Esaka, T.; Uchida, H.; Maeda, N. Solid State Ionics 1981, 3−4, 359−363. (3) Kreuer, K. D. Annu. Rev. Mater. Res. 2003, 33 (1), 333−359. (4) Iguchi, F.; Sata, N.; Tsurui, T.; Yugami, H. Solid State Ionics 2007, 178 (7−10), 691−695. (5) Babilo, P.; Uda, T.; Haile, S. M. J. Mater. Res. 2007, 22 (5), 1322−1330. (6) Iguchi, F.; Tsurui, T.; Sata, N.; Nagao, Y.; Yugami, H. Solid State Ionics 2009, 180 (6−8), 563−568. (7) Iguchi, F.; Sata, N.; Yugami, H. J. Mater. Chem. 2010, 20 (30), 6265−6270. (8) Kjølseth, C.; Fjeld, H.; Prytz, Ø.; Dahl, P. I.; Estournès, C.; Haugsrud, R.; Norby, T. Solid State Ionics 2010, 181 (5), 268−275. (9) Chen, C.-T.; Danel, C. E.; Kim, S. J. Mater. Chem. 2011, 21 (14), 5435−5442. (10) Iguchi, F.; Chen, C.-T.; Yugami, H.; Kim, S. J. Mater. Chem. 2011, 21 (41), 16517. (11) Shirpour, M.; Merkle, R.; Lin, C. T.; Maier, J. Phys. Chem. Chem. Phys. 2012, 14 (2), 730−740. (12) Shirpour, M.; Merkle, R.; Maier, J. Solid State Ionics 2012, 216, 1−5. (13) Kingery, W. D. J. Am. Ceram. Soc. 1974, 57 (2), 74−83. (14) Yan, M. F.; Cannon, R. M.; Bowen, H. K. J. Appl. Phys. 1983, 54 (2), 764. (15) Desu, S. B.; Payne, D. A. J. Am. Ceram. Soc. 1990, 73 (11), 3398−3406. (16) Shirpour, M.; Rahmati, B.; Sigle, W.; van Aken, P. A.; Merkle, R.; Maier, J. J. Phys. Chem. C 2012, 116 (3), 2453−2461. (17) Jamnik, J.; Maier, J.; Pejovnik, S. Solid State Ionics 1995, 75 (0), 51−58. (18) Shirpour, M. PhD Thesis, University of Stuttgart, 2011. (19) Helgee, E. E.; Lindman, A.; Wahnström, G. Fuel Cells 2013, 13 (1), 19−28. (20) Tauer, T.; O’Hayre, R.; Medlin, J. W. J. Mater. Chem. A 2013, 1 (8), 2840−2846. (21) Kim, J.-S.; Yang, J.-H.; Kim, B.-K.; Kim, Y.-C. Solid State Ionics 2015, 275, 19−22. (22) Lindman, A.; Helgee, E. E.; Wahnström, G. Solid State Ionics 2013, 252, 121−125. (23) Kelly, T. F.; Miller, M. K. Rev. Sci. Instrum. 2007, 78 (3), 031101. (24) Thompson, K.; Lawrence, D.; Larson, D. J.; Olson, J. D.; Kelly, T. F.; Gorman, B. Ultramicroscopy 2007, 107 (2), 131−139. (25) Gorman, B. P.; Diercks, D.; Salmon, N.; Stach, E.; Amador, G.; Hartfield, C. Microsc. Today 2008, 16 (4), 42−47. (26) Tong, J. H.; Clark, D.; Hoban, M.; O’Hayre, R. Solid State Ionics 2010, 181 (11-12), 496−503. (27) Ricote, S.; Bonanos, N. Determination of the dielectric constant of different BZY10; available upon request, 2010. (28) Ricote, S.; Bonanos, N.; Manerbino, A.; Sullivan, N. P.; Coors, W. G. J. Mater. Chem. A 2014, 2 (38), 16107−16115. (29) Zhu, H.; Ricote, S.; Coors, W. G.; Kee, R. J. Faraday Discuss. 2015, 182, 49−74. (30) Krivanek, O. L.; Chisholm, M. F.; Nicolosi, V.; Pennycook, T. J.; Corbin, G. J.; Dellby, N.; Murfitt, M. F.; Own, C. S.; Szilagyi, Z. S.; Oxley, M. P.; Pantelides, S. T.; Pennycook, S. J. Nature 2010, 464 (7288), 571−574. (31) Jia, C. L.; Lentzen, M.; Urban, K. Science (Washington, DC, U. S.) 2003, 299, 870−873.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.6b02918. Material synthesis and atom probe data analysis and calculating SCLs from literature (PDF)

■

REFERENCES

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected] (D.R.C.). *E-mail: [email protected] (R.P.O). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work has been supported by the National Science Foundation MRSEC program under Grant No. DMR0820518, the National Science Foundation Graduate Research Fellowship under Grant No. DGE-1057607, The Petroleum Institute in Abu Dhabi, UAE, the Colorado School of Mines F

DOI: 10.1021/acs.nanolett.6b02918 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters (32) An, J.; Park, J. S.; Koh, A. L.; Lee, H. B.; Jung, H. J.; Schoonman, J.; Sinclair, R.; Gür, T. M.; Prinz, F. B. Sci. Rep. 2013, 3, 2680. (33) van Duin, A. C. T.; Merinov, B. V.; Han, S. S.; Dorso, C. O.; Goddard, W. A. J. Phys. Chem. A 2008, 112 (45), 11414−11422. (34) Cahn, J. W.; Hilliard, J. E. J. Chem. Phys. 1958, 28 (2), 258−267. (35) Cahn, J. W. J. Chem. Phys. 1977, 66 (8), 3667. (36) Mebane, D. S.; De Souza, R. Energy Environ. Sci. 2015, 4 (8), 1166−1169. (37) Diercks, D. R.; Tong, J.; Zhu, H.; Kee, R.; Baure, G.; Nino, J. C.; O’Hayre, R.; Gorman, B. P. J. Mater. Chem. A 2016, 4 (14), 5167− 5175. (38) Stokes, S. J.; Islam, M. S. J. Mater. Chem. 2010, 20, 6258−6264. (39) Beekmans, N. M.; Heyne, L. Electrochim. Acta 1976, 21 (4), 303−310.

G

DOI: 10.1021/acs.nanolett.6b02918 Nano Lett. XXXX, XXX, XXX−XXX