Electrochemistry on a Localized Surface Plasmon Resonance Sensor

Dec 18, 2009 - The peak position, height, and radius of curvature were determined by fitting the spectrum ...... Merriam , J. B. Geophysics 2007, 72, ...
1 downloads 0 Views 5MB Size
Download PDF
pubs.acs.org/Langmuir © 2009 American Chemical Society

Electrochemistry on a Localized Surface Plasmon Resonance Sensor Takumi Sannomiya,† Harald Dermutz,† Christian Hafner,‡ Janos V€or€os,† and Andreas B. Dahlin*,† †

Laboratory of Biosensors and Bioelectronics and ‡Laboratory for Electromagnetic Fields and Microwave Electronics, Department of Information Technology and Electrical Engineering, ETH Z€ urich, Z€ urich, Switzerland Received November 9, 2009. Revised Manuscript Received November 30, 2009

The optical signal of a localized surface plasmon resonance (LSPR)-based sensor combined with electrochemistry was investigated. Gold nanoparticles were immobilized on an indium tin oxide (ITO) substrate, which functioned as working electrode. Using cyclic voltammetry synchronized with LSPR sensing, surface reactions on gold were detected both electrically and optically. In the capacitive charging regime, optical signals linear to the applied potential were detected. Gold was found to be dissolved above the oxidation potential and partially redeposited during the reduction, which changed size and conformation of the gold nanoparticles. In kinetic measurements, slower potential establishment was observed at lower salt concentrations. Simulations by multiple multipole program (MMP) suggested the formation of a lossy layer by combination of charge depletion of gold and negative ion adsorption even below the reaction potential. We consider the results presented here of importance for any future sensors based on combined plasmonics and electrochemistry.

Introduction Localized surface plasmon resonance (LSPR)-based sensors have been extensively investigated due to their simple yet sensitive nature.1 The intrinsically small scale of LSPR sensors is a great advantage to be integrated into portable and disposable sensing chips or to achieve extreme high throughput. In LSPR sensing the confined electric field gives high sensitivity only at the vicinity of the surface of the plasmonic structure. Such optical resonances, which are sensitive to the structure and surrounding environment, originate from the dispersive dielectric constant of the metal. Nanostructures made of noble metals, such as gold or silver, are known to exhibit LSPR for UV-vis-IR wavelengths. The electrically conductive property of such LSPR structures offers a possibility to combine the sensing system with electrochemistry. Similar to LSPR biosensing, electrochemical reactions at the surface of the plasmonic electrode can be sensed as a resonance shift when a change of refractive index occurs.2 Indeed, electrochemistry in combination with conventional SPR (propagating plasmons in thin films) can be used to detect surface sensitive reactions optically,3,4 while cyclic voltammetry (CV) provides complementary information. Formation of surface oxide film of noble metals at anodic potential has been observed also by ellipsometry and quartz crystal microbalance.5 Although there *Corresponding author. E-mail: [email protected]. (1) Anker, J. N.; Hall, W. P.; Lyandres, O.; Shah, N. C.; Zhao, J.; Van Duyne, R. P. Nat. Mater. 2008, 7, 442–453. (2) Wanga, T. J.; Lin, W. S. Appl. Phys. Lett. 2006, 89, 173903. (3) Zhang, N.; Schweiss, R.; Zonga, Y.; Knoll, W. Electrochim. Acta 2007, 52, 2869–2875. (4) Jiang, X.; Cao, Z.; Tang, H.; Tan, L.; Xie, Q.; Yao, S. Electrochem. Commun. 2008, 10, 1235–1237. (5) Xia, S. J.; Birss, V. I. J. Electroanal. Chem. 2001, 500, 562–573. (6) Leroux, Y.; Lacroix, J. C.; Fave, C.; Trippe, G.; Felidj, N.; Aubard, J.; Hohenau, A.; Krenn, J. R. ACS Nano 2008, 2, 728–732. (7) Miyazaki, T.; Hasegawa, R.; Yamaguchi, H.; Oh-oka, H.; Nagato, H.; Amemiya, I.; Uchikoga, S. J. Phys. Chem. C 2009, 113, 8484–8490. (8) Novo, C.; Funston, A. M.; Gooding, A. K.; Mulvaney, P. J. Am. Chem. Soc. 2009, 131, 14664–14666. (9) Sardar, R.; Funston, A. M.; Mulvaney, P.; Murray, R. W. Langmuir 2009, DOI: 10.1021/la9019475. (10) Mulvaney, P. Langmuir 1996, 12, 788–800.

Langmuir 2010, 26(10), 7619–7626

have been some studies combining LSPR sensors with electrochemistry,6-11 in situ combination with CV has not been reported. Besides the explicit electrochemical reaction, which can be observed as a reaction peak in CV, LSPR changes that occur in the potential range of capacitive charging have been reported.11 Similar results measured by other optical methods can be found in the literature.12,13 However, the origin of the optical signal from an applied potential is still not fully understood.14 In this study, we combine LSPR sensing with electrochemistry to investigate the influence of electrochemical reactions and the electric double layer formation on the spectrum. In detail, we conduct synchronized CV and LSPR spectroscopy on gold nanoparticles immobilized on an indium tin oxide (ITO) substrate. Electrochemical reactions of gold are observed using the resonance shift (plasmon energy), height (plasmon magnitude), and radius of curvature changes (plasmon lifetime) as functions of the applied potential sweeps. The formation of the electric double layer and the associated optical responses are further discussed by modeling the system by simulating the spectrum with the aid of the multiple multipole program (MMP). Charge formation dynamics was also optically studied at different salt concentrations.

Methods Experimental Section. 50 nm gold colloid particles (British Biocell) were immobilized on the ITO-coated glass substrate, previously coated with poly(L-lysine) (PLL) for electrostatic binding, which was removed by subsequent oxygen plasma treatment, resulting in the physical contact of gold particles and ITO layer.15 The substrate was mounted in an electrochemical optical flow cell that has platinum counter electrode and chloridized silver wire as a semistable reference electrode. 100 mM sodium chloride solution was prepared from ultrapure water (Milli-Q, Millipore Corp.) and was used for all experiments unless stated otherwise. (11) Ung, T.; Giersig, M.; Dunstan, D.; Mulvaney, P. Langmuir 1997, 13, 1773–1782. (12) Horkans, J.; Cahan, B. D.; Yeager, E. Surf. Sci. 1974, 46, 1–23. (13) Stankowski, S.; Ramsden, J. J. J.; Phys, D. Appl. Phys. 2002, 35, 299–302. (14) Bazant, M. Z.; Thornton, K.; Ajdari, A. Phys. Rev. E 2004, 70, 021506. (15) Sannomiya, T.; Hafner, Ch.; Voros, J. Opt. Lett. 2009, 34, 2009–2011.

Published on Web 12/18/2009

DOI: 10.1021/la9042342

7619

Article

Sannomiya et al.

Figure 1. (a) Schematic illustration of the experimental setup. (b) Extinction spectrum and (c) SEM image of the 50 nm gold particles immobilized on an ITO substrate. All experiments were done at room temperature with stagnant liquid. The transmission spectra were recorded by an Ocean Optics spectrometer (USB2000) and SpectraPro 2150 (PIXIS 400, Princeton Instruments) using halogen lamp illumination and fiberoptic readout. The recorded data were evaluated by a custom-made program. The peak position, height, and radius of curvature were determined by fitting the spectrum to a parabolic function. The peak radius of curvature corresponds to the broadening of the peak (larger radius of curvature means broader peak). Cyclic voltammetry was conducted after stabilizing the open-circuit potential using IPS Jaissle PGU10 V-1A-IMP-S potentiostat (Jaissle Elektronik GmbH, Germany). The scanning speed was set to be 10 mV/s. The schematic drawing of the experimental setup, a scanning electron microscopy (SEM) image of the gold particle-coated ITO, and the extinction spectrum of the sample in water solution are shown in Figure 1. Simulation. The calculation of the extinction cross section was performed by MMP using the MaX-1 software package, now replaced by the open source software OpenMax available from ETH Z€ urich.16,17 Spherical particles with different coating layers were modeled. For isolated spherical particles, MMP gives accuracy of more than 10 digits as MMP approach is based on analytical Mie scattering theory.17 The dielectric constant of the aqueous solution was assumed to be nondispersive as 1.777. Measured data were used for the dielectric constant of gold.18

Results Cyclic Voltammetry. Figure 2 shows the cyclic voltammogram with the scan range up to 500 mV and synchronized spectral responses of peak shift, peak height change, and peak radius of curvature change. The flow cell was filled with 100 mM NaCl in pure water under stagnant conditions. In the CV curve no reaction was observed except hydrogen formation at negative potential. The peak position and the radius of curvature showed linear increase while the peak height showed decrease, meaning peak broadening while red-shifting. We attribute the slight deviation from the linearity at the negative applied potential to hydrogen generation. The CV curve and corresponding spectral responses for an identical experiment with scan range up to 800 mV are shown in Figure 3. In the CV curve, a small peak started to appear at around 600 mV in the forward scan and at 350 mV in the backward scan. These peaks correspond to the gold chloride formation and to its reduction at the surface of the gold particles, (16) Hafner, Ch. MaX-1: A Visual Electromagnetics Platform; Wiley: Chichester, UK, 1998. (17) Hafner, Ch. Phys. Status Solidi B 2007, 244, 3435–3447. (18) Johnson, P.; Christy, R. Phys. Rev. B 1972, 6, 4370–4379. (19) Hu, Z.; Ritzdorf, T. J. Electrochem. Soc. 2007, 154, D543–D549.

7620 DOI: 10.1021/la9042342

respectively.19 The broad peak around 150 mV at forward scan can be explained by reoxidation of the adsorbed hydrogen from the previously applied negative voltage. The LSPR peak shift and radius of curvature plots showed hysteresis above 300 mV (Figure 3b,d). The positive deviation (negative for peak height) from the linearity occurs at around 600 mV, which matches the gold chloride formation in the CV curve. At a potential of 350 mV in the backward scan, the spectral response came back to the original linear trace, which corresponds well to the reduction of gold chloride in the CV curve. On the other hand, hysteresis was not clearly observed in the peak height plot, which might be due to the unstable signal of the illumination (peak position and curvature are not affected by a total intensity change for all wavelengths). The spectral response by gold chloridation and reduction seems to be reversible after the voltage scan cycles. CV scans up to 1500 mV exhibited a large peak above 900 mV as shown in Figure 4, which can be assigned to the oxidation of gold.12,19 In the backward scan the oxidized gold was reduced at around 300-500 mV. In the synchronized peak shift plot, some wavy responses were seen above 900 mV in the first forward scan. Below 600 mV in the backward scan, a large increase in the peak shift was observed, which kept increasing until the scan flipped and reached around 200 mV in the forward scan. The peak height plot shows a large signal decrease above 900 mV in the forward scan and a signal increase below 500 mV in the backward scan. The peak height keeps decreasing after each cycle, indicating an irreversible process. The behavior in the radius of curvature plot is basically opposite to the peak height response and signal irreversibility is also clearly observed. Considering all the results, it is indicated that the gold is dissolved into the solution during the oxidation above 900 mV and a part of the gold ions are redeposited below the reduction potential (see below). Kinetics and Salt Concentration. To investigate the kinetics of the process, we performed measurements with different salt concentrations using the linear part of signal response below 500 mV (Figure 2). The kinetic signal response was recorded during the stepwise potential application from 0 V to 250 mV and back. Figure 5a,b shows the peak shift kinetics at different salt concentrations. Although 0.1 mM concentration (or below) shows unstable signal, which is probably due to the reduction (chloride ion release) of the silver chloride reference electrode at low chloride ion concentration, there is a tendency that lower salt concentration gives a slightly lower magnitude of the signal change with a slower kinetic response, which is true for both potential increase and decrease. The time constants of potential decrease (250 mV to 0 V) were smaller than those of potential increase (0 V to 250 mV) at all the concentrations. The peak shifts and the time constants are summarized in Figure 5c,d. The radius of curvature signal showed identical response (see Figure S1 in the Supporting Information).

Discussion Chloridation and Oxidation of Gold. To confirm the reversibility in the gold chloride formation when scanning up to 800 mV and the irreversibility of gold oxidation when scanning up to 1500 mV, we recorded SEM images after the electrochemical measurements (Figure 6). The SEM image after the scan up to 800 mV did not show apparent difference from the original state (compare with Figure 1). This agrees well with the reversible spectral response in Figure 3. Therefore, it can be concluded that the chloridation at the potential up to 800 mV probably occurred only at the outmost layer of the gold particle without losing the gold atoms into the solution. Similar hysteresis has been observed Langmuir 2010, 26(10), 7619–7626

Sannomiya et al.

Article

Figure 2. (a) Cyclic voltammetry, (b) synchronized peak position, (c) synchronized peak height, and (d) synchronized peak radius of curvature plots with the scan from -200 up to 500 mV in 100 mM sodium chloride solution. First three scans are shown as different colors. The scan speed was set to be 10 mV/s.

Figure 3. (a) Cyclic voltammetry, (b) synchronized peak position, (c) synchronized peak height, and (d) synchronized peak radius of curvature plots with the scan from -200 up to 800 mV in 100 mM sodium chloride solution. First three scans are shown as different colors. The scan speed was set to be 10 mV/s.

for oxide film formation without the presence of chloride ions.12 On the other hand, the SEM image after the scan up to 1500 mV exhibits particles with different sizes and shapes as well as aggregates of particles (see also Supporting Information). The irreversibility in the spectral measurements in Figure 4 arises from the dissolution of gold during the oxidation process12,19 and the partial redeposition of gold onto the particles during the reduction process, resulting in diverse particle sizes and shapes. It is also noted that the smaller particles become chemically less stable Langmuir 2010, 26(10), 7619–7626

because of the higher surface energy (similar to the Ostwald ripening process). During the gold dissolution process, the physical binding of the particles to ITO could be loosened, allowing short migration of the particles to form aggregates. These aggregates are physically connected probably by redeposition of gold during reduction. The particles detached from the substrate lose the applied potential which stops the oxidation (and reduction of oxidized gold ions takes place instead), thus resulting in the deposition of gold on the detached gold particles, which also DOI: 10.1021/la9042342

7621

Article

Sannomiya et al.

Figure 4. (a) Cyclic voltammetry, (b) synchronized peak position, (c) synchronized peak height, and (d) synchronized peak radius of curvature plots with the scan from -200 up to 1500 mV in 100 mM sodium chloride solution. First three scans are shown as different colors. The scan speed was set to be 10 mV/s.

Figure 5. Kinetic peak shift plots after stepwise potential application at different salt concentrations: (a) the potential was applied from 0 V to 250 mV and (b) from 250 mV to 0 V at the time 0. To eliminate the variation of different samples and positions, the magnitudes are normalized to the measurement at 1 M concentration which was recorded at the same spot. (c) Summarized peak shift plot at saturation and (d) time constant plot for potential application of 250 mV from 0 V and 0 V from 250 mV as functions of concentration.

enhances the wide size distribution. These structural changes, including the nucleation and growth of small particles, resulted in complicated and irreversible spectral responses during the cyclic voltammetry scans up to þ1500 mV. By chloridation at þ600 mV during the scan up to þ800 mV (Figure 3), we observed resonance damping, which is essentially the lowering of peak height and the increase of radius of 7622 DOI: 10.1021/la9042342

curvature (peak broadening). This is similar to the silver iodine formation in the study by Mulvaney.10 Although we could not find optical data of gold chloride, we assume it has absorbing (lossy) material properties similar to a gold oxide layer.5,12 Electric Double Layer and Spectral Response. According to the DLVO theory, the applied potential on the surface forms electric double layer in solution, which consists of the Stern layer Langmuir 2010, 26(10), 7619–7626

Sannomiya et al.

Article

Figure 6. SEM images after the cyclic voltammetry scan with the range of (a) -200 to 800 mV and (b) -200 to 1500 mV.

at the closest vicinity of the surface and the diffusive (Gouy-Chapman) layer in the solution phase.20,21 We assume that the linear spectral response in Figure 2 is due to the formation of the electric double layer. A chloride ion has higher polarizability than a water molecule, which presumably induces a red shift of the resonance when it is attracted to the gold surface at positive potentials, in analogy to protein adsorption in biosensing.1 The concentration dependence of the potential-induced peak shift in Figure 5c indeed implies the formation of an electric double layer; the higher salt concentrations showed more peak shift than the lower salt concentrations due to more ions attracted to the gold surface. This argument holds both for the diffusive layer and the associated Stern layer. Another possibility is the electron depletion inside the gold which compensates for the charge of the attracted ions. The electron depletion inside the gold particles close to their surface will also cause LSPR shifts because the dielectric constant of the metal is dependent on the electron density.10 At the anodic potential, complex formation, such as metal halide and metal oxide monolayer films, seems also possible even below the reaction potential.12 For these atomic layers, simple Maxwell’s equation with defined boundaries may not be entirely applicable. However, simulations using idealized thin films would still give an idea of the optical response. Here we simulate only free particles as the effect of the substrate is not very critical, which improves the computational speed and accuracy (see Supporting Information for inhomogeneous models). According to our model calculation, the influence on the LSPR signal from the diffusive layer in the solution phase is very small (estimated peak shift of ∼0.1 nm) by a reasonable effective value for the surface potential (1 mM) by the other circuit resistances and the Stern layer capacitance, which in this case are much larger than solution resistance and diffusive layer capacitance. However, the interpretation of the observed kinetics is not so intuitive when the molecular details of the adsorption-desorption process are considered because the discharging process is also slower for lower ion concentration; i.e., the adsorbed negative ions stay longer when the bulk ion concentration is lower. As described above, this is actually more related to the potential (field) distribution in the whole system than just simple Langmuir type ad/desorption. The adsorbed negative ions at the positively charged surface cannot simply desorb when the potential is released. The pairs of the negative ions and positive surface holes can only be decoupled when the charge in the metal is effectively screened, which requires the arrival of positive counterions. DOI: 10.1021/la9042342

7625

Article

A complete explanation could in principle be achieved by introducing an extensive model that solves the coupled PoissonBoltzmann equations for the entire flow-cell system, but this is considered beyond the scope of the present work. Another possibility for slower response at lower concentration is the slower formation of Au-HCO3- surface species, which are probably the dominant surface complexes at lower salt concentration, while Au-Cl is dominantly formed at high salt concentration. As the potential drop by the ion attached to the surface is heavily dependent on the distance from the surface, formation of such surface species could retard the potential establishment.

Conclusion We have demonstrated synchronized cyclic voltammetry and LSPR sensing for gold colloids in contact with a NaCl solution. The chloridation and oxidation of gold were captured as changes in peak shift, height, and radius of curvature. Above the chloridation potential, the optical signal showed a hysteresis loop while the response was linear to the applied potential at low potentials. Above the oxidation potential, dissolution of gold was observed as an irreversible process, which was confirmed by SEM images after the electrochemical cycles. Applying a small step potential allowed for kinetic measurements showing smaller and slower optical response at lower salt concentrations. The optical response in the linear regime was investigated by comparing the experimental spectrum with simulations based on different layer models. The damping effect could be explained only by a lossy layer model, which suggests the formation of an optically lossy layer even well below the reaction potential. The models considering only diffusion layer or only Stern layer of chloride ions did not fully agree with experimental results. In other words, the local RI increase from adsorbed Cl- ions cannot by itself provide a full explanation of the data, not even in

7626 DOI: 10.1021/la9042342

Sannomiya et al.

combination with electron depletion. However, formation of a lossy layer is partly supported by the electron depletion model, since a fully depleted layer can cause absorption. Although the observed spectral changes are most probably due to the combination of all the effects, the lossy layer seems to play the dominant role even at the capacitive charging regime. This work illustrates how LSPR-based sensors are handy and useful to monitor various phenomena, such as electrochemical reactions, in combined measurement systems. This is a great advantage compared to other optical sensors that require complicated setups. On the other hand, the quantification of the LSPR signal is not straightforward since the analytical solution is not available or complicated in most cases. More effective comparison with numerical simulation is needed for a perfect quantitative match to the observed signal. We have illustrated how taking advantage of several measurable parameters, such as resonance peak shift, height, and radius of curvature, becomes more important when the system to be described exhibits many different physical phenomena. We believe the results presented here can act as guidelines in the design of future applications of biointerface science, such as sensing schemes based on combined electrochemical and plasmonic readout. Acknowledgment. The authors thank Prof. F. H€oo€k for discussion and valuable advice and Mr. Stephen Wheeler for the flow cell production. This study was funded by ETH Z€urich and the Swedish Research Council. Supporting Information Available: Kinetics of radius of curvature data, SEM image, calculation of electron depletion, signal evaluation of diffusive layer, and discussion on inhomogeneous distribution of Stern layer. This material is available free of charge via the Internet at http://pubs.acs.org.

Langmuir 2010, 26(10), 7619–7626