Imaging heterogeneity and transport of degraded Nafion membranes

Abstract

Accelerated aging experiments of Nafion® 212 (N212) membranes were carried out with Fenton's reagent under a series of degradation durations. Morphological changes and appearance of surface defects on degraded membranes were characterized by scanning electron microscopy (SEM). X-ray photoelectron spectroscopy (XPS) mapping was performed to investigate heterogeneity in chemical composition “on” and “off” surface defects. In addition, scanning ion conductance microscopy-scanning electrochemical microscopy (SICM-SECM) was used to map the heterogeneous permeability of degraded membranes to a cationic redox probe.

---

Polymer electrolyte membrane fuel cells (PEMFCs) represent an important component of present and future energy schemes. A key component of the PEMFC, the polymer electrolyte membrane (PEM), i.e., perfluorosulfonic acid (PFSA) ionomer membrane,^1,2 polyetheretherketone membrane (PEEK)^3,4 and polybenzimidazole (PBI) fuel cell membrane,^3,5,6 serves as an ion conductor, gas barrier and mechanical support; all factors which play a critical role in PEMFC operation. Despite the fact that durability of PEMs has been improved dramatically over past decades, membrane degradation and failure continues to present a critical challenge to PEMFC lifetime.^2,7,8

PFSA ionomer membranes, such as Nafion® (Dupont™), are the most widely-used electrolyte membranes for PEMFCs. The structure of the ionomer is shown in Scheme 1.^2,7 When cast in thin films, the polytetrafluoroethylene (PTFE) backbone provides mechanical stability, and fluorous-ether side chains bearing sulfonic acid groups provide proton conductivity. Under extended use, the electrolyte membrane in PEMFCs can be exposed to extensive oxidative stress as a result of reactive intermediates formed in the membrane electrode assembly. These reactive intermediates attack the polymer membrane, leading to chain scission, subsequent membrane thinning, and formation of surface defects such as bubbles/tears.⁹ Fuel cells cannot operate efficiently if even a small amount crossover occurs, and as defects develop and expand within the polymer membrane, failure of the cell will occur. Degradation of Nafion membranes has been studied extensively,^1,2,7–27 and the primary cause for chemical degradation during fuel cell operation is accepted as the attack of hydroxyl (·OH) and peroxyl (·OOH) radical species. Fenton's reaction (Scheme 2)²⁸ has been employed widely as a source of hydroxyl radicals in accelerated degradation tests of membrane durability.

Scheme 1 Structure of Nafion.

Scheme 2 Fenton's reaction.

Previous studies have also explored the roughened texture and appearance of defects (bubbles/tears) of degraded Nafion films, however quantitation of the heterogeneous nature of changes in the chemistry and conductance, relative to the surface defects has not been confirmed.^{1,2,19,22–24} Herein, accelerated degradation of Nafion® 212 (N212) membranes has been carried out with Fenton's reagent, to produce oxidative degradation in reasonable time scales (compared to degradation that occurs after thousands of hours under regular fuel cell operating conditions). Characterization of degraded N212 membrane samples was performed by spectroscopy, microscopy and electrochemical methods. To examine morphological changes after degradation, SEM analysis was carried out. In our studies, special attention is paid to analyze heterogeneity of the membrane and degradation processes. To investigate heterogeneity of chemical composition in degraded membrane samples, XPS mapping was performed. To study heterogeneous local permeability of N212 membranes, SICM-SECM was employed. In combination, results reveal heterogeneity in both chemical composition and transport properties of degraded membranes.

An ion exchange method reported by Inaba and Kinumoto²⁵ was used in membrane aging experiments. Briefly, pre-hydrated N212 membrane samples were soaked in 0.1 M FeSO₄ at 70 °C for 2 h. Membrane samples were then thoroughly rinsed with de-ionized water and placed into vials that contained 50 mL of 30% H₂O₂, and placed in a 70 °C water bath. Membrane samples were removed at 3 h, 6 h, 24 h and 48 h degradation time and further conditioned in 0.5 M H₂SO₄ at 70 °C for 2 h to remove any residual iron.

To analyze changes in membrane morphology after degradation, scanning electron microscopy (FE-SEM, FEI Quanta-FEG, Hillsboro, OR) was used. Images from SEM analysis are shown in Fig. 1a–d. Surface was roughened by the presence of defects, such as bubbles and tears (as indicated by arrows in Fig. 1d) for all degradation conditions. Overall, defect size increased with duration of exposure to peroxide solution. Analysis of defect diameter was performed by ImageJ (National Institute of Health), with obtained diameter distributions shown in histograms of Fig. 1e–h. In addition, a summary for both the average and the maximum diameter of surface defects for all the degradation durations is shown in Table 1. After 3 h degradation (Fig. 1e), the diameter of defects ranged from hundreds of nm to ∼5 μm, and an average diameter of 2.6 μm was measured (Table 1). For 6 h degradation (Fig. 1f), the average diameter of defects was 4.1 μm (Table 1), with ∼70% of defects ranging from hundreds of nm to 5 μm, and the remaining ∼30% were between 5 μm and 15 μm. Degradation for 24 h led to further increase in defect size, with ∼50% of the defect diameters between 10 to 15 μm (Fig. 1g), and ∼50% that range from 2 to 5 μm. Finally, for 48 h (Fig. 1h), ∼70% of defects were between 5 to 15 μm, and the diameter of the remaining ∼30% ranged from 15 to 35 μm. With no surprise, based on the above size analysis, longer degradation duration led to larger defect sizes, which indicates more severe degradation.

Fig. 1 Scanning electron microscopy (SEM) images of the N212 membrane degraded for (a) 3 h; (b) 6 h; (c) 24 h and (d) 48 h. Inset: 4× zoomed-in images for the corresponding degradation conditions (note: contrast was adjusted for the zoomed-in images for better visualization). Histograms of defect size distribution for (e) 3 h; (f) 6 h; (g) 24 h and (h) 48 h.

Table 1 Average and maximum diameters for all degradation conditions. n is the number of defects measured for each degradation conditions

Duration	Average diameter (μm)	Maximum diameter (μm)
3 h	2.60 ± 0.97 (n = 1383)	5.26
6 h	4.07 ± 2.63 (n = 946)	15.93
24 h	9.07 ± 4.86 (n = 223)	19.36
48 h	12.51 ± 5.31 (n = 186)	36.29

---

XPS has been used in previous studies to examine chemical changes of intact and degraded membranes.^11,12,21 However, to our knowledge, no detailed investigation of heterogeneity in chemical composition of degraded membranes has been performed. Herein, XPS mapping was carried out to reveal differences in chemical structures “on” and “off” surface defects.

Earlier versions of Nafion® were proposed to degrade via hydroxyl radical attack on carboxylic acid (–COOH) groups at the terminus of the polymer backbone.⁷ Decomposition of the (CF₂)_n backbone follows the general unzipping mechanism proposed by Curtin et al.⁷ To combat this degradation mechanism, postfluorination of the polymer backbone termini results in “chemically-stabilized” grades of Nafion, e.g., Nafion® 212, in which the concentration of terminal –COOH groups was decreased to negligible levels. In these stabilized N212 membranes, the backbone unzipping degradation pathway is effectively eliminated.^{26,27,29–31} However, membrane degradation persists even if no carboxylic groups are present, which occurs from alternative degradation mechanisms, for instance, via polymer side chains. Ghassemzadeh observed significant side chain degradation via solid state ¹⁹F NMR spectroscopy and proposed a possible side chain degradation mechanism, as shown in Schemes 3 ²⁶ and 4.²⁷ Direct attack of hydroxyl radicals on the α-OCF₂ bond (labelled in Scheme 3) initiates the side chain degradation (Scheme 3), and hydroxyl radicals continue to attack further up the side chain, as indicated by step 1, 2 and 3 in Scheme 4.^26,27 Ishimoto et al. reported that the α-OCF₂ is more vulnerable to radical attack compared to ether groups that bridge the main chain and side chain (β-OCF₂) on the basis of density functional theory (DFT) calculations.¹⁷ The relative reactivity of α-OCF₂ (k₂) and β-OCF₂ (k₁) with hydroxyl radicals was theoretically analyzed by k₂/k₁ (the ratio of rate constants for the corresponding reactions). As k₂/k₁ was calculated to be 4.0 × 10³, cleavage of the α-OCF₂ was proposed to be the major pathway for side chain degradation.¹⁷ In Scheme 3, side chain degradation leads to direct loss of – the ionic SO₃⁻ head groups, as well as –CF₂–CF₂– segments. Further attack of the side chain leads to additional loss of –CF_n groups (Scheme 4).

Scheme 3 Initiation of side chain degradation.

Scheme 4 Progression of side chain degradation.

In Fig. 2a, the averaged C 1s XPS spectra collected “on” surface defects (red) and “off” surface defects (green) are shown. Of note, to determine the binding energies of other XPS peaks, spectra were referenced to the binding energy (E_b) of carbon at 284.8 eV. Contributions to the signal for all C 1s species include –CF₂–, –OCF–, –OCF₂–, –CF– and –CF₃– groups.¹⁶ Spatially selected spectra for etched membranes displayed two peaks in the C 1s region, at E_b = 292.2 eV and at E_b = 284.8 eV. The peak intensity at 292.2 eV decreased “on” surface defects compared to that “off” the defects, while the intensity of the peak at 284.8 eV was higher “on” defects. “Off” defects, the ratio of these two peaks (C 1s : R_292/284) was calculated to be 2.2. While “on” surface defects, C 1s : R_292/284 decreased to 0.6. In Fig. 2b, a ratio map obtained by dividing the map of C 1s at 292.2 eV and at 284.8 eV is shown, with the corresponding scanning X-ray induced secondary electron image (SXI) of the area of interest shown in the inset. This demonstrates existence of chemical heterogeneity in degraded membranes. Furthermore, averaged F 1s spectra along with the corresponding F 1s XPS map are shown in Fig. S1.† A drastic decrease in peak intensity at E_b = 688.0 eV was observed “on” defects, which again, indicates a larger fluorine loss as a result of more severe degradation “on” defects, as compared to that “off” defects.

Fig. 2 (a) X-ray photoelectron spectroscopy (XPS) spectra for C 1s “on” surface defects (red) and “off” surface defects (green); (b) ratio map optained by dividing the C 1s XPS map at 292.2 eV and the C 1s XPS map at 284.8 eV; inset: scanning X-ray induced secondary electron image (SXI) of the area under study (c) XPS spectra for O 1s “on” surface defects (red) and “off” surface defects (green).

In Fig. 2c, the averaged O 1s spectra acquired “on” surface defects (red) and “off” defects (green) are displayed. In Nafion, oxygen exists in two functionalities; as sulfonic acid groups and as ether groups. The O 1s region showed distint changes “on” and “off” defects as well. “Off” defects, the peaks indicated at E_b = 535.0 eV and at E_b = 533.0 eV (Fig. 2c, green) yield a ratio, O 1s : R_535/533 of 1.3. “On” surface defects, (Fig. 2c, red), a net decrease for O 1s : R_535/533 to 0.3, was observed, which indicates significant heterogeneity in the chemical composition of oxygen for the two regions of film.

Taken in total, our results obtained from XPS mapping studies reveal heterogeneous chemical composition in degraded N212 membranes, and demonstrate the membrane was more severely degraded “on” surface defects compare to that “off” defects. In particular, with the degradation mechanism proposed, these chemical changes result in loss of charged, anionic sulfonate groups, which are responsible for imparting the proton selectivity to Nafion. The net result of these defects results in physical and chemical changes to the membrane, which includes surface defects (bubbles/tears) that may lead to crossover and ultimately failure of membrane operations. Membrane permeability, a key determinant of overall membrane performance, might be affected by formation of bubbles/tears as well.

As confirmed by XPS mapping, chemistry of degradation was heterogeneous at surface defects, which poses questions of heterogeneity in membrane transport properties. In previous studies, SECM has proven to be a well-suited tool to study transport properties in polymer membranes.^32–38 Powerful as it is, conventional SECM lacks reliable probe-sample distance control, and the probe is kept at a constant height during scanning. As a result, any variation in surface topography will result in changes in probe-sample distance, which will complicate the SECM measurements/analyses. Combination of SECM with SICM enables simultaneous collection of both the electrochemical activity and the surface topography.^39–42 Herein, to examine heterogeneity in transport, SICM-SECM was used to correlate membrane topography to local permeability of redox mediators. Fig. 3a shows a typical SICM-SECM probe utilized in experiments described here, which consists of an open nanopore (100–150 nm in radius) and a gold crescent (250 nm in thickness) electrode. A detailed probe fabrication procedure has been described in previous reports, and a description is provided in ESI.†^42–44

Fig. 3 (a) SEM image of an end-on view of the SICM-SECM probe. (b) Schematic diagram of the SICM-SECM setup. (c) Cyclic voltammogram (CV) acquired with the AuE for Ru(NH₃)₆³⁺/Ru(NH₃)₆²⁺ reduction at a scan rate of 0.1 V s⁻¹; (d) SICM topography image of two nanopores and a larger feature. (e) SECM faradaic current responses of the two nanopores seen in (d).

For SICM-SECM experiments, a commercial SICM (Park Systems XE-Bio SICM/AFM, Suwon, South Korea) in combination with a potentiostat (Chem-clamp, Dagan Corporation, Minneapolis, MN) was used. A schematic diagram of the SICM-SECM setup is depicted in Fig. 3b. Ion current between the Ag/AgCl electrode inside the nanopipette (PE, pipette electrode) and another Ag/AgCl electrode in the bath solution (RE, reference electrode) was employed as the feedback for probe-sample distance control. A crescent-shaped gold electrode (AuE) was used to acquire electrochemical signal. To investigate membrane permeability, a perfusion cell was used and the membrane of interest was mounted between the top and bottom chamber. A concentration gradient was established across the membrane by filling the top chamber with 0.1 M KCl and the bottom chamber with 20 mM ruthenium hexamine (Ru(NH₃)₆³⁺, a cationic redox mediator) and 0.1 M KCl, such that the Ru(NH₃)₆³⁺ molecules diffused from bottom chamber across the membrane, to the top chamber. Membrane permeability to Ru(NH₃)₆³⁺ can be determined by monitoring the electrochemical signal (reduction of Ru(NH₃)₆³⁺ to Ru(NH₃)₆²⁺) with the AuE. To determine the potential to bias the AuE at, cyclic voltammograms (CVs) were recorded in bulk solution (100 mM KCl and 5 mM Ru(NH₃)₆³⁺), shown in Fig. 3c. For the reduction of Ru(NH₃)₆³⁺/Ru(NH₃)₆²⁺, a steady-state current was observed at −0.5 V (Fig. 3c). To obtain both topography and electrochemical images, potential of the PE and the AuE were set at +0.1 V and −0.5 V, respectively. We have chosen Ru(NH₃)₆³⁺ as a redox probe in an effort to examine the properties of the membrane towards cation transport. An important point to note is that in most functional Nafion films, H⁺ is the cationic species transported, which is smaller, diffuses faster, and can be transported by all-together different mechanisms than Ru(NH₃)₆³⁺. Thus, our measurements are only illustrative of transport for Ru(NH₃)₆³⁺, and are not directly comparable to H⁺, but do allow inference of aspects of heterogeneity in the degradation process.

To validate SICM-SECM measurements, topographic (Fig. 3d) and electrochemical (Fig. 3e) images of two nanopores in a polyimide (PI) porous membrane were acquired. The nanoporous membrane was prepared as described in previous reports.⁴⁵ In the SICM topographic image (Fig. 3d), two nanopores (denoted by arrows) with an average diameter of 0.44 μm and a larger feature (denoted by a dashed circle) with a diameter of ∼1.85 μm were observed. For the electrochemical image (Fig. 3e), only the two nanopores showed faradaic current responses (originating from the diffusion of Ru(NH₃)₆³⁺ molecules across the membrane), with a nominal current of ∼65 pA and an average full width at half maximum (FWHM) of 1.05 μm. No faradaic current response was observed from the large feature, which suggests that the feature was only a variation in topography without any discernable permeability to Ru(NH₃)₆³⁺ molecules.

We applied SICM-SECM to interrogate heterogeneity in permeability of the degraded N212 membranes. Imaging degraded membranes was enabled by approach-retract scanning (ARS) mode,⁴⁶ which is ideal for imaging samples with significant surface roughness. In Fig. 4, representative SICM-SECM images for 3 h (a), 6 h (b) and 12 h (c) degradations are shown. In Fig. 4a, after 3 h degradation, surface defects were observed in the SICM topography, and an average size of 2.18 ± 0.49 μm (n = 10) was measured. However, no distinct faradaic current response over features was observed, which suggests that although defects can be seen on the membrane surface, most defects do not penetrate the membrane or alter the net permeability to Ru(NH₃)₆³⁺. Longer duration of exposure (6 h) resulted in defects with an average size of 4.48 ± 1.14 μm (n = 11) (Fig. 4b). Not only did the defects increase in size, but some of the defects began to show faradaic current responses. In both the topography and electrochemical images, arrows were added to correlate defects that exhibited faradaic current responses. Two important findings are present here. First, every spot with faradaic current response in the SECM image can be correlated with a surface defect in the SICM image. This suggests that leakage of degraded membrane starts at surface defects, which is consistent with the XPS mapping results, as defect sites experienced more severe degradation compared to non-defect sites. Second, electrochemical signals were only observed on ∼50% of surface defects, which confirms existence of heterogeneity in permeability of degraded membranes. As exposure duration further increased to 12 h, the size of the defects increased to 6.90 ± 1.76 μm (Fig. 4c). The amplitude of the maximum faradaic current increased from ∼50 pA (6 h) to ∼75 pA (12 h), which indicates a larger increase in membrane permeability. Moreover, with longer degradation duration, the number of surface defects with distinguishable faradaic current responses also increased; nearly all surface defects were permeable to Ru(NH₃)₆³⁺ molecules. An increase in background current magnitude was also observed, which is further direct evidence that the membrane became leakier upon longer degradation durations. Herein, SICM-SECM allows the correlation of membrane leakage with surface defects for the first time. In the future, this technique could be further extended to investigate change in membrane performance, in terms of transport property change, along with the chemical degradation process, in hopes to provide some mitigation strategies to increase membrane durability.

Fig. 4 SICM-SECM topographic (left) and faradaic current (right) images of N212 membrane degraded for (a) 3 h, (b) 6 h, (c) 12 h. Arrows were added to indicate bubbles/tears with distinct faradaic current responses for better correlation to topographic images.

Conclusions

In conclusion, we investigated the physical and chemical composition, as well as local permeability change of the degraded membrane samples. From the SEM analyses, surface defects such as bubbles/tears were observed for all degradation conditions. XPS mapping experiments demonstrated the local chemical structure difference “on” and “off” surface defects. In addition, SICM-SECM enabled correlation between the membrane morphology and the local permeability of Ru(NH₃)₆³⁺, and provides a promising platform to study heterogeneity in permeability for technologically important membranes.

Acknowledgements

Electronic Instrument Services (EIS) and Mechanical Instrument Services (MIS) at Indiana University are gratefully acknowledged. The authors appreciate XPS and scanning electron microscope use provided by the IU Nanoscale Characterization Facility. Access to XPS was provided by NSF Award DMR MRI-1126394. Assistance from Dr Yaroslav Losovyj is greatly appreciated for XPS training and discussion. Acknowledgment is given to Park Instruments for SICM use. Financial support provided by Indiana University is also acknowledged.

References

1. A. C. Fernandes and E. A. Ticianelli, J. Power Sources, 2009, 193, 547–554.
2. K. A. Mauritz and R. B. Moore, Chem. Rev., 2004, 104, 4535–4586.
3. D. J. Jones and J. Rozière, J. Membr. Sci., 2001, 185, 41–58.
4. S. J. Zaidi, S. Mikhailenko, G. Robertson, M. Guiver and S. Kaliaguine, J. Membr. Sci., 2000, 173, 17–34.
5. L. Xiao, H. Zhang, E. Scanlon, L. S. Ramanathan, E.-W. Choe, D. Rogers, T. Apple and B. C. Benicewicz, Chem. Mater., 2005, 17, 5328–5333.
6. Q. Li, J. O. Jensen, R. F. Savinell and N. J. Bjerrum, Prog. Polym. Sci., 2009, 34, 449–477.
7. D. E. Curtin, R. D. Lousenberg, T. J. Henry, P. C. Tangeman and M. E. Tisack, J. Power Sources, 2004, 131, 41–48.
8. V. O. Mittal, H. R. Kunz and J. M. Fenton, J. Electrochem. Soc., 2007, 154, B652–B656.
9. L. Gubler, S. M. Dockheer and W. H. Koppenol, J. Electrochem. Soc., 2011, 158, B755–B769.
10. A. Bosnjakovic and S. Schlick, J. Phys. Chem. B, 2004, 108, 4332–4337.
11. C. Chen and T. F. Fuller, J. Electrochem. Soc., 2009, 156, B1218–B1224.
12. C. Chen, G. Levitin, D. W. Hess and T. F. Fuller, J. Power Sources, 2007, 169, 288–295.
13. F. D. Coms, ECS Trans., 2008, 16, 235–255.
14. M. Gummalla, V. Atrazhev, D. Condit, N. Cipollini, T. Madden, N. Kuzminyh, D. Weiss and S. Burlatsky, J. Electrochem. Soc., 2010, 157, B1542–B1548.
15. J. Healy, C. Hayden, T. Xie, K. Olson, R. Waldo, M. Brundage, H. Gasteiger and J. Abbott, Fuel Cells, 2005, 5, 302–308.
16. C. Huang, K. Seng Tan, J. Lin and K. Lee Tan, Chem. Phys. Lett., 2003, 371, 80–85.
17. T. Ishimoto, T. Ogura and M. Koyama, ECS Trans., 2011, 35, 1–6.
18. M. Kitazawa, A. Nosaka and Y. Nosaka, J. Appl. Electrochem., 2008, 38, 491–496.
19. S. Kundu, L. C. Simon and M. W. Fowler, Polym. Degrad. Stab., 2008, 93, 214–224.
20. T. Madden, D. Weiss, N. Cipollini, D. Condit, M. Gummalla, S. Burlatsky and V. Atrazhev, J. Electrochem. Soc., 2009, 156, B657–B662.
21. M. Schulze, M. Lorenz, N. Wagner and E. Gülzow, Fresenius' J. Anal. Chem., 1999, 365, 106–113.
22. H. Tang, S. Peikang, S. P. Jiang, F. Wang and M. Pan, J. Power Sources, 2007, 170, 85–92.
23. R. Lin, E. Gülzow, M. Schulze and K. Friedrich, J. Electrochem. Soc., 2011, 158, B11–B17.
24. N. Macauley, A. S. Alavijeh, M. Watson, J. Kolodziej, M. Lauritzen, S. Knights, G. Wang and E. Kjeang, J. Electrochem. Soc., 2015, 162, F98–F107.
25. T. Kinumoto, M. Inaba, Y. Nakayama, K. Ogata, R. Umebayashi, A. Tasaka, Y. Iriyama, T. Abe and Z. Ogumi, J. Power Sources, 2006, 158, 1222–1228.
26. L. Ghassemzadeh and S. Holdcroft, J. Am. Chem. Soc., 2013, 135, 8181–8184.
27. L. Ghassemzadeh, T. J. Peckham, T. Weissbach, X. Luo and S. Holdcroft, J. Am. Chem. Soc., 2013, 135, 15923–15932.
28. C. Walling, Acc. Chem. Res., 1975, 8, 125–131.
29. T. H. Yu, Y. Sha, W.-G. Liu, B. V. Merinov, P. Shirvanian and W. A. Goddard, J. Am. Chem. Soc., 2011, 133, 19857–19863.
30. L. Ghassemzadeh, K.-D. Kreuer, J. Maier and K. Müller, J. Phys. Chem. C, 2010, 114, 14635–14645.
31. L. Ghassemzadeh, K. D. Kreuer, J. Maier and K. Müller, J. Power Sources, 2011, 196, 2490–2497.
32. P. Bertoncello, I. Ciani, F. Li and P. R. Unwin, Langmuir, 2006, 22, 10380–10388.
33. A. Kishi, M. Inoue and M. Umeda, J. Phys. Chem. C, 2009, 114, 1110–1116.
34. M. Parthasarathy and V. K. Pillai, J. Chem. Sci., 2009, 121, 719–725.
35. R. Cornut and C. Lefrou, J. Electroanal. Chem., 2008, 623, 197–203.
36. O. D. Uitto and H. S. White, Anal. Chem., 2001, 73, 533–539.
37. B. D. Bath, R. D. Lee, H. S. White and E. R. Scott, Anal. Chem., 1998, 70, 1047–1058.
38. M. Pyo and A. J. Bard, Electrochim. Acta, 1997, 42, 3077–3083.
39. Y. Takahashi, A. I. Shevchuk, P. Novak, Y. Zhang, N. Ebejer, J. V. Macpherson, P. R. Unwin, A. J. Pollard, D. Roy and C. A. Clifford, Angew. Chem., Int. Ed., 2011, 50, 9638–9642.
40. Y. Takahashi, A. I. Shevchuk, P. Novak, Y. Murakami, H. Shiku, Y. E. Korchev and T. Matsue, J. Am. Chem. Soc., 2010, 132, 10118–10126.
41. D. J. Comstock, J. W. Elam, M. J. Pellin and M. C. Hersam, Anal. Chem., 2010, 82, 1270–1276.
42. R. Thakar, A. E. Weber, C. A. Morris and L. A. Baker, Analyst, 2013, 138, 5973–5982.
43. W. Shi, N. Sa, R. Thakar and L. A. Baker, Analyst, 2015, 140, 4835–4842.
44. C. A. Morris, C. C. Chen and L. A. Baker, Analyst, 2012, 137, 2933–2938.
45. C. C. Chen, Y. Zhou and L. A. Baker, ACS Nano, 2011, 5, 8404–8411.
46. T. Ushiki, M. Nakajima, M. Choi, S.-J. Cho and F. Iwata, Micron, 2012, 43, 1390–1398.

---

Footnote

† Electronic supplementary information (ESI) available: Additional experimental details, XPS maps are included. See DOI: 10.1039/c5ra20291d

---