Probing Charge Transfer at Surfaces Using Graphene Transistors

pubs.acs.org/NanoLett

Probing Charge Transfer at Surfaces Using Graphene Transistors Pierre L. Levesque,*,†,‡ Shadi S. Sabri,†,§ Carla M. Aguirre,†,| Jonathan Guillemette,†,§ Mohamed Siaj,†,⊥ Patrick Desjardins,†,| Thomas Szkopek,*,†,§ and Richard Martel*,†,‡ †

Regroupement Que´be´cois sur les mate´riaux de pointe, ‡ De´partement de Chimie, Universite´ de Montre´al, Montre´al, Que´bec H3C 3J7, Canada, § Department of Electrical and Computer Engineering, McGill University, Montre´al, Que´bec H3A 2A7, Canada, | De´partement de Ge´nie Physique, E´cole Polytechnique de Montre´al, Montre´al, Que´bec H3C 3A7, Canada, and ⊥ De´partement de Chimie, Universite´ du Que´bec a` Montre´al, Montre´al, Que´bec H3C 3P8, Canada ABSTRACT Graphene field effect transistors (FETs) are extremely sensitive to gas exposure. Charge transfer doping of graphene FETs by atmospheric gas is ubiquitous but not yet understood. We have used graphene FETs to probe minute changes in electrochemical potential during high-purity gas exposure experiments. Our study shows quantitatively that electrochemistry involving adsorbed water, graphene, and the substrate is responsible for doping. We not only identify the water/oxygen redox couple as the underlying mechanism but also capture the kinetics of this reaction. The graphene FET is highlighted here as an extremely sensitive potentiometer for probing electrochemical reactions at interfaces, arising from the unique density of states of graphene. This work establishes a fundamental basis on which new electrochemical nanoprobes and gas sensors can be developed with graphene. KEYWORDS Electron transport, graphene, carbon nanotubes, charge transfer, redox.

T

p-doping of diamond surfaces.7 We present further experiments demonstrating how graphene FETs can be used to monitor redox chemistry at dielectric interfaces. The dopingdependent electrical resistance of graphene serves here as a natural Fermi level potentiometer that allows one to quantitatively probe the kinetics of electron transfer reactions. The graphene-SiO2 FETs were prepared on a 285 nm thermal SiO2 on n++ Si substrate. For the grapheneparylene FET, the substrates consisted of 168 nm of parylene-C deposited by chemical vapor deposition atop of 94 nm thermal SiO2 on n++ Si. In both cases, the substrate layer thicknesses were chosen for high optical contrast per graphene layer under λ ) 550 nm illumination, with a Fresnel thin-film model predicting 15% (13%) reflection contrast per graphene layer on the oxide (parylene) substrates.8 As a first step, a standard photolithography was used to define alignment marks on the substrate, followed by metal evaporation and lift-off. Graphene was then mechanically transferred onto the prepared substrate by exfoliation of natural graphite using adhesive tape. The graphene flakes were then located and identified as monolayers using a quantitative optical reflection microscopy system.8 Standard electron beam lithography and lift-off techniques with PMMA as a resist were finally used to fabricate the contact electrodes to the graphene layer. The electrodes are made of Ti/Pd (0.5 nm/ 40 nm) or Cr/Au (10 nm/110 nm). All electrical transport characteristics were measured at room temperature with a source-drain (S-D) bias voltage of Vds ) 10 mV in a probe station coupled to a semiconductor parameter analyzer (Agilent 1500B). For vacuum and controlled atmosphere experiments, a vacuum probe station (Lake-

he impact of adsorbates on graphene FETs has been extensively explored in order to control doping, improve mobility, or develop new sensor applications. For example, intentional exposure of a back-gated graphene on SiO2 FET (graphene-SiO2 FET) to potassium (K) under ultrahigh vacuum resulted in the expected n-type (donor) doping and a reduction in mobility, consistent with charged impurity scattering.1 Donor doping has also been observed in graphene-SiO2 FETs exposed to dilute NH3 vapor,2-4 dilute CO vapor,2 vapors of 1,5-naphthalenediamine and 9,10-dimethylanthracene, and liquid-phase poly(ethylene imine) (PEI).5 On the contrary, acceptor (p-type) doping has been observed in graphene-SiO2 FETs exposed to dilute H2O and NO2 vapors,2 vapors of tetrasodium 1,3,6,8-pyrenetetrasulfonic acid (TPA) and 9,10-dibromoanthracene,6 and liquid-phase 4-bromobenzenediazonium tetrafluoroborate.5 Although some of these results are easily understood, such as the propensity of K to ionize to K+ and to thereby act as an electron donor, proposed mechanisms behind doping of graphene by small molecules, H2O or air, are less straightforward and have not been fully tested. We have thus performed experiments using graphene as a probe to uncover the specific mechanism behind gas doping. An electron transfer reaction involving the O2/H2O redox couple was found responsible for air doping, highlighting a gas doping mechanism that has not been considered to date in electronic devices. Such a mechanism is responsible for

* Corresponding authors: Pierre L. Levesque, [email protected]; Thomas Szkopek, [email protected]; Richard Martel, [email protected]. Received for review: 8/25/2010 Published on Web: 12/09/2010 © 2011 American Chemical Society

132

DOI: 10.1021/nl103015w | Nano Lett. 2011, 11, 132–137

FIGURE 1. Air effect on a graphene field-effect transistor (FET) with different substrates: (A) graphene on SiO2 and (B) graphene on parylene; red lines, transfer characteristics of graphene FETs measured in vacuum at room temperature after a 4 h anneal at 400 K; cyan lines characteristics measured after 30 min in air. The arrows indicate the direction of the gate (G) voltage sweep. The aspect ratios (length/width) are 0.45 and 0.8 for the SiO2 and parylene devices, respectively.

shore Desert Cryogenics model FWP6) with a base pressure below 5 × 10-7 Torr was used. This chamber was further equipped with a gas handling manifold. Gases and liquids were initially expanded in the manifold and were subsequently introduced into the vacuum probe station through a precision leak valve. The manifold consisted of two gas or vapor sources connected through a bypass and a precision-leak valve to a small expansion chamber. The absolute pressure in the probe station was measured with two capacitance manometers covering a pressure range between 10-5 and 1000 Torr. Deionized water (Milli-Q) was degassed by repeated freeze-thaw cycles under vacuum prior to experiment. Oxygen and nitrogen (Praxair Canada, Inc., UHP grade stated purity 99.994% and 99.999%, respectively) were used without further purification. Overall nine graphene FETs were extensively tested with different gas exposures, including four FETs with graphene on parylene (2/Pd, 2/Au) and five FETs with graphene on SiO2 (4/Pd, 1/Au). The same general behavior was observed for the different batches of grapheneparylene FETs and of graphene-SiO2 FETs, independently of the metal used. We investigated gas exposure effects on back-gated graphene on SiO2 FETs and back-gated graphene on parylene FETs. The substrates were chosen because of their different chemical nature, i.e., SiO2 is hydrophilic and parylene is hydrophobic, to isolate the important role of water in gas experiments. That is, only the SiO2 substrate will accumulate a full layer of water on its surface under atmospheric conditions.9 The transfer characteristics of graphene-SiO2 FETs and graphene-parylene FETs taken after annealing in vacuum at 400 K for 4 h and subsequent exposure for 30 min to air are shown in panels A and B of Figure 1, respectively. The striking result is that only the graphene-SiO2 FET shows p-doping by ambient atmospheric exposure (Figure 1A), as evidenced by the shift of the conductivity minimum (known as the © 2011 American Chemical Society

Dirac point or neutrality point, VDirac) by roughly 14 V, corresponding to a change in carrier density of ∼1 × 1012 cm-2. The p-doping by air and reduction of p-doping following vacuum annealing of graphene-SiO2 FETs are consistent with previously reported experiments.2,3,10,11 In the case of a high density of strongly (e.g., covalent) bonded adsorbates to the graphene, VDirac is not welldefined,12 but the highly reversible nature of gas adsorption precludes strong bonding from playing an important role in our experiments. VDirac of a graphene transfer characteristic thus specifies the carrier doping density and Fermi level. Experiments (not shown) have established that the initial position of VDirac for both parylene and oxide FETs in vacuum depends mainly upon the device processing and annealing time in vacuum. No visible doping by air was observed for all graphene-parylene FETs, while all graphene-SiO2 FETs presented a VDirac shift toward positive gate voltage (Vg), the final value of which depended mostly on the exposure time. Explanations of p-doping of graphene have invoked impurities,13 air adsorbates including water,2 silanol groups at the surface of the dielectrics,14 or metal contacts.15,16 Our experiments (Figure 1) show that there is a direct correlation between air exposure of graphene and the nature of the dielectric used. Direct adsorption of molecular species on the graphene layer can thus be discarded as being responsible alone for p-doping of graphene. A variation in contact metal work function that would dope graphene can also be excluded because the same metal stack was used in both devices. We therefore conclude that p-doping is intimately related to the nature of the dielectric used, consistent with previous studies of carbon nanotube FETs,17 graphene,18,19 and organic semiconductor FETs.20 The nature of the dielectric alone cannot explain why VDirac (and thus doping of the graphene layer) can be tuned using vacuum annealing or air exposure. To identify the 133

DOI: 10.1021/nl103015w | Nano Lett. 2011, 11, 132-–137

FIGURE 2. Gas exposure. Transfer characteristics of graphene FETs in different atmospheres: (A) and (C), SiO2-FETs; (B) and (D), paryleneFETs measured in vacuum after a 4 h anneal at 400 K (orange) and exposed to atmospheric gases at pressure corresponding to their partial pressure in air: 160 Torr O2 (black), 15 Torr H2O (green), and 160 Torr O2 + 15 Torr H2O (blue).

solvated O2 to the nearby graphene electrode, which sets the conditions for the redox reaction

key molecular species in air that are involved in this tunable doping and to shed light on the mechanism, the graphene FETs were exposed to controlled gas doses. The FETs were inserted into a high vacuum chamber with a base pressure below 5 × 10-7 Torr and subsequently annealed at 400 K for 4 h to set the neutrality point, VDirac, near 0 V. The FETs were then exposed at room temperature to the main atmospheric constituents at pressures corresponding to their partial pressure in air: N2 (600 Torr), O2 (160 Torr), H2O (15 Torr, i.e. ∼50% relative humidity at 23 °C). As presented in Figure 2, exposure to O2 (and N2 but not shown) produced virtually no shift in VDirac in both graphene-parylene FETs and grapheneSiO2 FETs. In the case of the graphene-SiO2 FETs, surprisingly, water shifted VDirac toward negative Vg (ndoping). Only the combined exposure to O2 and H2O reproduced the air effect. In the case of the grapheneparylene FET, there is negligible shift in neutrality point. The similarity between the observed doping effects and past work on CNT FETs17 supports charge transfer doping mediated by the O2/water redox couple as the key doping mechanism. As described in the first report on O2/water mediated surface charge transfer to explain p-doping of diamond,7 water fulfils the important role of supplying © 2011 American Chemical Society

ox

O2(aq) + 4H+ + 4e-(graphene) y\z 2H2O

(1)

red

at slightly acidic pH. The Fermi level or redox potential, Eredox, of the electrons associated with this reaction at equilibrium under atmospheric conditions is, according to the Nernst equation, at about -5.3 eV relative to vacuum level (see Supporting Information). The Fermi level of graphene (∼4.6 eV)21 lies above the electrochemical potential of the solution providing a strong driving force to shift the graphene Fermi level well into its valence band (i.e., p-doping). A convenient way to model such a charge transfer process is through the Marcus-Gerischer (MG) theory. MG theory has been successfully used to describe charge transfer in electrochemical reactions between semiconductor (or metal) and redox systems.22-25 We apply MG theory here to explain FET operation in the presence of a solution. Electrons will be transferred from the occupied states of the graphene to the unoccupied states of the O2/water redox 134

DOI: 10.1021/nl103015w | Nano Lett. 2011, 11, 132-–137

Crucially, this negative charge stabilizes a net positive charge (hole) in the graphene layer, explaining the positive shift of VDirac in Figure 1 and Figure 2. When water alone is introduced into the vacuum probe station, only traces of O2 remain and Dox decreases by a factor of 1010 at 10-7 Torr. The solution Fermi level Eredox shifts up and drives the electrons from the occupied levels in the solution toward the unoccupied states of the graphene layer. This process consumes reactive oxygen species and generates H+, which fixes a net positive charge on the SiO2 surface. This is consistent with an effective n-doping of the graphene layer, as observed by the water exposure results in Figure 2C. The reverse reaction toward neutral O2 will also be favored in presence of water, if the concentration of intermediates (O2•-, H2O2, •OH) is significant. Thus, the doping induced by pure water (free of gas) will depend primarily upon the past history of the surface (e.g., applied Vg bias in air) and on the chemical species present. The rate of charge transfer doping during gas exposure is also well described by a simple MG model. Figure 4A presents the time evolution of the transfer characteristics for a SiO2-based graphene FET while exposed to the simulated air conditions (O2 + H2O). The dominant effect is a nonlinear shift with time of VDirac toward positive Vg (i.e., p-doping), illustrated in Figure 4B. The observed VDirac (right axis) is directly proportional to the doping n (left axis) at the surface of graphene and allows the Fermi level of graphene to be determined at any time. The time evolution of the charge doping density (dn/dt) is proportional to the net rate of electron transfer through the redox system. The electron transfer rate from the graphene layer to the solution will be proportional to the overlap between the occupied states of graphene and the unoccupied states of the solution as depicted in Figure 3. The DOS of the electrode is much larger than that of the solution due to the low areal density of solvated species, so that the transfer rate is24,25

FIGURE 3. Electron-transfer mechanism within the Marcus-Gerischer theory: right, schematic of the water/oxygen redox couple density of states (DOS) for an equivalent concentration of oxidizing and reducing species; left, comparison with the graphene DOS. The arrow indicates the direction of the charge transfer reaction.

system, as shown in Figure 3. The MG theory also treats the reverse reaction (i.e., graphene reduction) and considers that molecules or ions in the solution have discrete energy levels that fluctuate in time, owing to the polarization of the solvent. For a single electron transfer, ox + e- h red, of an oxidizing (reducing) agent ox (red), which may also be thought of as an electron acceptor (donor), the electron energy level probability distribution Wox(red) around a mean energy Eox(red) is

[

Wox(red)(E) ) W0 exp -

(E - Eox(red))2 4kBTλ

]

(2)

where λ is the reorganization energy of the solvent and W0 o ) (4kBTλ)-1/2 is a normalization factor. By definition, Eredox o o ≡ (1/2)(Eox + Ered), Ered ≡ Eredox - λ, and Eox ≡ Eredox + λ. Finally, the electronic density of states (DOS) is related to the concentration of molecules or ions in the solution, being defined as Dox ) coxWox and Dred ) credWred for the unoccupied and the occupied states, respectively. The Fermi level of the solution Eredox is the energy where the DOS of the reducing and the oxidizing species are equal: Dox(Eredox) ) Dred(Eredox). As illustrated in Figure 3, electrons will flow from the graphene layer to the unoccupied levels of the solution until equilibrium is reached, i.e., when the electrochemical potentials of the graphene and the solution are aligned. The O2/water redox reactions necessitate the transfer of 4e- in sequence, involving up to four intermediate reactions.26 These reactions generate highly reactive species such as the superoxide anion (O2•-), peroxide (H2O2), and hydroxyl radical (•OH). These strong Brønsted bases are prone to form “traps” at the SiO2 surface that fix a net negative charge. © 2011 American Chemical Society

dn ) coxνκel dt

∫-∞∞ f(E)Wox(E) dE

(3)

where f(E) is the Fermi-Dirac distribution, Wox(E) is the distribution of the oxidizing species (unoccupied states) of eq 2, ν is a frequency factor, and κel is the electronic tunnelling probability. From eq 3, dn/dt depends on the Fermi level of the graphene sheet EF, which is related to the charge density n by

n)

∫EE g(E) dE F

0

(4)

where g(E) ) 2|E - Eig|/πp2νF2 is the graphene DOS, Eig is the intrinsic graphene Fermi level, E is the initial doping level and νF is the Fermi velocity. This model assumes that the graphene/SiO2 interface presents an unpinned Fermi level and that the first step of 135

DOI: 10.1021/nl103015w | Nano Lett. 2011, 11, 132-–137

concentration of 2 × 1017 molecules cm-3 and a water layer thickness of 0.5 nm (approximately two layers). Water adlayer formation of similar thickness between graphene and a hydrophilic surface has been recently observed.28 Taking ν to be a typical vibrational frequency (1011-1013 s-1) gives a tunneling probability κel of the order of 10-4-10-6. Such tunneling rates are indeed expected for a molecule separated by a small distance from a surface.29 Therefore, the good agreement between our theoretical model and kinetic data supports the proposed mechanism of charge transfer and also demonstrates that the key elements of the doping process have been captured. It also suggests that possible pinning of the graphene Fermi level by the SiO2 interface is negligible in comparison to the DOS associated to the oxygen in the water layer. The present study shows the intimate relation between the SiO2 substrate and the effective p-doping of graphene in atmospheric conditions. It also suggests that graphene based gas sensors can take advantage of a combination of a substrate having a sensitive functional coating to minimize background response to water and oxygen. Most interestingly, our work demonstrates that graphene FETs are sensitive surface probes. The atomic thickness of graphene provides exquisite electrochemical coupling to molecular adsorbates, its mechanical strength permits relatively large substrate areas to be probed, while the conductance minimum at the neutrality point provides a natural reference to electrically read-out the local electrochemical potential. In the case of nanoscale systems, where cathodic/anodic currents may be exceedingly small, the graphene provides increased measurement sensitivity well below the attoampere.

FIGURE 4. Air doping kinetics. (A) Doping states taken from 1.7 × 1012 cm-2 to zero during exposure to air within 10 h. Circles and squares show the current at a density of 1 × 1012 cm-2 for holes and electrons, respectively. (B) The net charge density at the surface (left axis) is estimated during air exposure (circle) using VDirac (right axis). The errors in VDirac and time are (1 V and less than 2 min, respectively.ThedopingkineticssimulatedwiththeMarcus-Gerischer model are shown by the red curve. The net charge density is estimated using, n ) (εε0/eds)∆VDirac where ε ) 3.9ε0 and ds ) 285 nm are the gate dielectric (SiO2) permittivity and thickness, respectively.

Acknowledgment. The authors thank F. Lapointe for useful discussions and P. Gagnon for artistic and technical support. Financial support from the Natural Sciences and Engineering Research Council of Canada, the Canada Research Chair program, the Canadian Institute for Advanced Research, and the Fonds que´be´cois de recherche sur la nature et les technologies are acknowledged. This work was also made possible by the McGill University MicroFabrication Facility and the Central Facilities at Polytechnique/UdeM.

the reaction (O2 + e- f O2-) is the limiting step controlling the kinetics.27 Moreover, it is assumed that most of the transferred charges are rapidly stabilized at the surface so that reverse reactions are negligible. A comparison between the theoretical charge transfer versus time and the experimental data is given in Figure 4B. The cathodic current is extremely small (∼1 aA), but indirect measurement of this current via the graphene neutrality point is simple and powerful. Equation 3 was integrated numerically by fitting only the value of the parameter product coxνκel. Excellent agreement was found with experiment using known values of the limiting redox reaction O2/O2- for Wox (Eredox ) -4.11 eV and λ ) 1 eV).27 The best-fit of the parameter product is (4.8 ( 0.1) × 1017 cm-2 s-1, consistent with the expected values for the three individual parameters in the product. For instance, the O2 concentration at the surface is estimated as cox ∼ 1 × 1010 molecules cm-2 from a volumetric O2 © 2011 American Chemical Society

Supporting Information Available. Additional details and discussions regarding hysteresis and minimum conductance, electron chemical potential, and mobility. This material is available free of charge via the Internet at http:// pubs.acs.org. REFERENCES AND NOTES (1) (2) (3) (4)

136

Chen, J.-H.; Jang, C.; Adam, S.; Fuhrer, M.; Williams, E.; Ishigami, M. Nat. Phys. 2008, 4, 377–381. Schedin, F.; Geim, A. K.; Morozov, S. V.; Hill, E. W.; Blake, P.; Katsnelson, M. I.; Novoselov, K. S. Nat. Mater. 2007, 6, 652–655. Lohmann, T.; von Klitzing, K.; Smet, J. H. Nano Lett. 2009, 9, 1973–1979. Romero, H. E.; Joshi, P.; Gupta, A. K.; Gutierrez, H. R.; Cole, M. W.; Tadigadapa, S. A.; Eklund, P. C. Nanotechnology 2009, 20, 245501. DOI: 10.1021/nl103015w | Nano Lett. 2011, 11, 132-–137

(5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17)

Farmer, D. B.; Golizadeh-Mojarad, R.; Perebeinos, V.; Lin, Y. M.; Tulevski, G. S.; Tsang, J. C.; Avouris, P. Nano Lett. 2009, 9, 388– 392. Dong, X. C.; Fu, D. L.; Fang, W. J.; Shi, Y. M.; Chen, P.; Li, L. J. Small 2009, 5, 1422–1426. Chakrapani, V.; Angus, J. C.; Anderson, A. B.; Wolter, S. D.; Stoner, B. R.; Sumanasekera, G. U. Science 2007, 318, 1424–1430. Gaskell, P. E.; Skulason, H. S.; Rodenchuk, C.; Szkopek, T. Appl. Phys. Lett. 2009, 94, 143101. Zhuravlev, L. T. Colloid Surf., A 2000, 173, 1–38. Dan, Y. P.; Lu, Y.; Kybert, N. J.; Luo, Z. T.; Johnson, A. T. C. Nano Lett. 2009, 9, 1472–1475. Romero, H. E.; Shen, N.; Joshi, P.; Gutierrez, H. R.; Tadigadapa, S. A.; Sofo, J. O.; Eklund, P. C. ACS Nano 2008, 2, 2037–2044. Robinson, J. P.; Schomerus, H.; Oroszlany, L.; Fal’ko, V. I. Phys. Rev. Lett. 2008, 101, 196803. Ishigami, M.; Chen, J. H.; Cullen, W. G.; Fuhrer, M. S.; Williams, E. D. Nano Lett. 2007, 7, 1643–1648. Hong, X.; Zou, K.; Zhu, J. Phys. Rev. B 2009, 80. Huard, B.; Stander, N.; Sulpizio, J. A.; Goldhaber-Gordon, D. Phys. Rev. B 2008, 78, 121402. Xia, F. N.; Mueller, T.; Golizadeh-Mojarad, R.; Freitag, M.; Lin, Y. M.; Tsang, J.; Perebeinos, V.; Avouris, P. Nano Lett. 2009, 9, 1039–1044. Aguirre, C. M.; Levesque, P. L.; Paillet, M.; Lapointe, F.; St-Antoine, B. C.; Desjardins, P.; Martel, R. Adv. Mater. 2009, 21, 3087–3091.

© 2011 American Chemical Society

(18) Sabri, S. S.; Levesque, P. L.; Aguirre, C. M.; Guillemette, J.; Martel, R.; Szkopek, T. Appl. Phys. Lett. 2009, 95, 242104. (19) Lafkioti, M.; Krauss, B.; Lohmann, T.; Zschieschang, U.; Klauk, H.; Klitzing, K. v.; Smet, J. H. Nano Lett. 2010, 10, 1149–1153. (20) Chua, L. L.; Zaumseil, J.; Chang, J. F.; Ou, E. C. W.; Ho, P. K. H.; Sirringhaus, H.; Friend, R. H. Nature 2005, 434, 194–199. (21) Barone, V.; Hod, O.; Scuseria, G. Nano Lett. 2006, 6, 2748–2754. (22) Barbara, P. F.; Meyer, T. J.; Ratner, M. A. J. Phys. Chem. 1996, 100, 13148–13168. (23) Nozik, A. J.; Memming, R. J. Phys. Chem. 1996, 100, 13061– 13078. (24) Memming, R. Electron Transfer Theories. In Semiconductor Electrochemistry; Wiley-VCH: Weinheim, 2001; pp 112-150. (25) Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and Applications, 2nd ed.; John Wiley & Sons: New York, 2001; p 836. (26) Sawyer, D. Oxygen Chemistry; Oxford University Press: New York, 1991. (27) Ignaczak, A.; Schmickler, W.; Bartenschlager, S. J. Electroanal. Chem. 2006, 586, 297–307. (28) Xu, K.; Cao, P.; Heath, J. R. Science 2010, 329, 1188–1191. (29) Adams, D. M.; Brus, L.; Chidsey, C. E. D.; Creager, S.; Creutz, C.; Kagan, C. R.; Kamat, P. V.; Lieberman, M.; Lindsay, S.; Marcus, R. A.; Metzger, R. M.; Michel-Beyerle, M. E.; Miller, J. R.; Newton, M. D.; Rolison, D. R.; Sankey, O.; Schanze, K. S.; Yardley, J.; Zhu, X. Y. J. Phys. Chem. B 2003, 107, 6668–6697.

137

DOI: 10.1021/nl103015w | Nano Lett. 2011, 11, 132-–137