Improved Biosensing Capability with Novel Suspended Nanodisks

ARTICLE pubs.acs.org/JPCC

Improved Biosensing Capability with Novel Suspended Nanodisks Marinus A. Otte, M.-Carmen Est
evez, Laura G. Carrascosa, Ana B. Gonz
alez-Guerrero, Laura M. Lechuga, and Borja Sep
ulveda* Nanobiosensors and Bioanalytical Applications Group, Research Center on Nanoscience and Nanotechnology (CSIC) & CIBER-BBN, 08193 Bellaterra, Barcelona, Spain

bS Supporting Information ABSTRACT: In this work, we provide a theoretical and experimental study of the adverse effects that inherently arise when plasmonic nanoparticles are attached to a solid support. Using the example of short-ordered arrays of gold nanodisks, we show that the right choice for the thin metal adhesion layer that is used to ensure stable binding of the nanodisks to the substrate is extremely important to achieve competitive limits of detection. Furthermore, we show that the presence of the high refractive index support undisputedly results in lower bulk and surface sensing sensitivities. To minimize this effect, we propose an isotropic chemical etch of the supporting substrate, resulting in dielectric nanopillars which distance the nanodisks from this high refractive index material. We demonstrate that this novel method not only preserves the mechanical stability but also strongly increases the sensing performance of these suspended nanodisk arrays. This is demonstrated via the label-free detection of DNA hybridization, reflecting the robustness and the surface-regenerative capabilities of these pillar-supported nanodisk arrays. Furthermore, through the successful implementation of this sensitivity enhancing method to more complex nanostructures, such as arrays of closely distanced nanodisk dimers, this method profiles itself as a simple strategy to improve the limits of detection of a wide variety of nanoplasmonic sensors.

’ INTRODUCTION The exceptional plasmonic properties of noble metal nanostructures have brought interesting insights and novel applications to a wide variety of research areas, including nanoplasmonic biosensing.1,2 Within this particular field, refractometric localized surface plasmon resonance (LSPR) sensing is one of the most employed techniques. This method exploits the strong dependence that the plasmonic resonance of the nanostructure has on the local dielectric environment,3,4 allowing the label-free detection of target analyte binding onto previously biofunctionalized nanostructures. These binding events can be detected via the tracking of corresponding spectral displacements of the LSPR peak. Motivated by the possibilities that these metal nanostructures offer to miniaturize and improve the sensitivity of propagating surface plasmon resonance sensors,5 different conceptual refractometric LSPR biosensing studies have been carried out using nanopores,6,7 -pyramids,8,9 -dimers,10 -rings,11 -tubes,12,13 or disks14-16 as the sensing platforms. All these different approaches generally aim at a reasonable trade-off between the shape-dependent sensitivity and the fabrication difficulty of the employed nanostructures. Logically, for the creation of high-throughput robust biosensors that are compatible with microfluidics for lab-on-a-chip applications, allowing kinetic analysis, and surface regeneration, attachment of the nanostructures to a solid substrate is a prerequisite. This can be achieved either through covalent linking of colloidal nanoparticles to chemically modified substrates or by using metal adhesion layers when lithographic nanopatterning r 2011 American Chemical Society

techniques are used. The latter is necessary since direct adhesion of noble metals on transparent substrates such as silica or organic polymers yields very poor results. However, previous work has shown that the presence of a supporting substrate negatively affects the sensing performance of these nanoplasmonic sensors in several ways.17,18 Next to the problems regarding unspecific binding of biomolecules to the substrate, the mere presence of the substrate and the linking layer can substantially reduce the sensitivity and the signal-to-noise ratio of the biosensing measurements. The aim of this work is, on the one hand, pointing out the problems and limitations that inherently arise from the attachment of the nanoparticles to a solid support and, on the other hand, to provide a very simple strategy to minimize such adverse effects, keeping the robustness of the sensing transducer. We will first demonstrate, theoretically and experimentally, that the thin metal layer used to stick metal nanostructures to the underlying substrate has a major impact on the sensing performance, illustrating that the right metal choice is of crucial importance to achieve low limits of detection. Furthermore, as a method to minimize the negative effects of the transparent substrate, we propose an isotropic wet chemical etch of the underlying substrate to create suspended nanostructures supported by dielectric nanopillars. As will be shown in the manuscript, this suspension technique profiles itself as a viable method Received: October 29, 2010 Revised: February 1, 2011 Published: March 04, 2011 5344

dx.doi.org/10.1021/jp110363a | J. Phys. Chem. C 2011, 115, 5344–5351

The Journal of Physical Chemistry C

ARTICLE

Figure 1. Influence of metal adhesion layers on the sensing performance of gold nanodisk arrays. (A) Schematic representation of a gold nanodisk on top of a glass substrate (n = 1.52) and surrounded by a dielectric medium (n = 1.33). (B) Calculated LSPR scattering spectra for a single gold nanodisk without a metal adhesion layer (black) and with a 1 nm thick Ti (red) and Cr (blue) sticking layer. (C) Experimental LSPR scattering spectra for shortordered arrays of gold nanodisks, attached to the substrate via a 1 nm thick Ti (black) and a 2 nm Cr (red) adhesion layer. (D) The corresponding realtime tracking of the relative LSPR peak displacement (ΔλLSPR) upon an induced bulk refractive index change of Δn = 0.02.

to noticeably increase the refractometric sensing performance, while keeping the mechanical stability of the nanostructures. For the experimental proof-of-concept, we have chosen short-range ordered arrays of gold nanodisks, a commonly employed nanostructured surface used for biosensing assays, due to its simple and large-scale lithographic fabrication technique.19 The large nanopatterned area (∼4 cm2) permits, in addition, the assessment of the homogeneity and robustness of the proposed strategies, which is vital for large scale and high-throughput assays. We present results showing that, with the creation of these isotropic dielectric nanopillars, sensitivities similar to those of nanodisks entirely surrounded by a homogeneous aqueous dielectric medium can be achieved. We corroborate these findings with the label-free detection of DNA strand hybridization, reflecting the high stability and exceptional surface-regenerative capabilities of these sensing substrates. Furthermore, to show that this sensitivity enhancing strategy can be used on a wide variety of nanoplasmonic structures, we demonstrate its successful implementation on more complex nanostructures, such as arrays of nanoparticle dimers.

’ EXPERIMENTAL METHODS FDTD Simulations. Scattering spectra and electromagnetic (EM) field profiles were numerically calculated using the finitedifference time-domain (FDTD) formalism, employing the commercially available Lumerical Software Package. Nanostructures were modeled as gold nanodisks (diameter = 100 nm, height = 20 nm) attached to the substrate by a Ti/Cr adhesion layer (thickness = 1-2 nm), resembling the nanostructures used in the experiments. The refractive index (RI) of the dielectric

environment surrounding the disks was set at 1.33, close to the RI of typical aqueous buffer solutions. Isotropic HF etches of the underlying glass substrate were simulated through the creation of dielectric pillars underneath the disks, with sizes corresponding to different etch times. EM field profiles were calculated with 0.5 nm grid accuracy, whereas for the generation of typical scattering spectra a grid size of 1 nm was used. Nanodisk Fabrication. Short ordered arrays of gold nanodisks or nanodisk dimers (diameter = 100 nm) were fabricated by colloidal hole-mask lithography.19 The nanodisks consisted of a 1 or 2 nm thick adhesion layer of Ti or Cr with 20 nm of gold on top. Before use, these nanodisk samples were cleaned by consecutive 1 min sonication cycles in acetone, ethanol, and deionized water (2
), after which the samples were blown dry with N2. The samples were then placed in an ozone generator for 20 min, after which they were thoroughly rinsed with deionized water and dried with N2. Reflectance Measurements. Reflectance measurements were carried out to test the surface and bulk sensitivities of the nanodisks. The LSPR of the nanostructures was excited by a collimated halogen light source (TE polarization), reaching the nanodisks through a hemicylindrical glass prism (n = 1.52), contacting the sample via refractive index matching oil. From the other side, the nanodisk samples were clamped by a custommade Delrin flow cell (volume ≈ 4 μL). LSPR excitation was achieved at a 15° angle. For the spectral analysis of the reflected light we used a spectrometer (Newton, Andor), connected to a cooled CCD camera (Shamrock SR-303i, T = -60 °C). LSPR spectra (2 ms exposure time) were accumulated every 3 s. Extraction of the LSPR peak position as a function of time was done after the measurements, using a high-degree polynomial fit.20 For the 5345

dx.doi.org/10.1021/jp110363a |J. Phys. Chem. C 2011, 115, 5344–5351

The Journal of Physical Chemistry C calculated standard deviations shown in Figure 1C, measurement data collected during the first 400 s of the experiment were used to calculate these values. Bulk Sensitivity Analysis. For the bulk sensitivity measurements, a 12.7% glycerol solution in deionized water was prepared with a RI of approximately 1.3522. Injection of this solution into the flow cell allowed us to compare the spectral peak displacement with respect to deionized water and extract the corresponding bulk sensitivity. Real-time monitoring of isotropic substrate etches was achieved through the injection of 0.05% HF into the flow cell. After each HF injection the bulk sensitivity was tested. All depicted ηB values are averaged over two measurements. DNA Hybridization Measurements. Nanodisks supported by dielectric pillars were obtained through an ex situ immersion of a nanodisk sample for 10 min in 0.05% HF, after which the sample was thoroughly rinsed with H2O and dried with N2. Immobilization of the 24-mer thiolated DNA probes on the surface of the suspended and nonsuspended nanodisks was done employing an overnight incubation protocol in which the nanodisk surfaces were covered with a droplet (volume ≈ 40 μL) of 2 μM thiolated DNA, diluted in a 750 mM NaCl, 50 mM phosphate buffer at pH 7. DNA hybridization measurements were carried out using a saline 750 mM NaCl, 75 mM sodium-citrate buffer at pH 7. The 32-mer single-strand DNA target contained a sequence of 24 bases complementary to the immobilized DNA receptor. DNA dehybridization was carried out using a 35% formamide solution.

’ RESULTS AND DISCUSSION Effect of Metal Adhesion Layers. Chemical modification of the substrate can be a feasible alternative to covalently bind colloidal metal nanoparticles to a substrate.21 In this case, the chemical link will not noticeably modify the optical properties of the nanostructure, and the main effect will be due to the high refractive index (RI) substrate. In contrast, most lithographic nanopatterning methods make use of thin metal layers, typically Cr and Ti, to provide the mechanical stability required for biosensing applications. Although these layers typically are only a few nanometers thick, their large optical absorption is expected to influence the plasmonic response of the gold nanostructures. To analyze this effect, numerical finite-difference time-domain (FDTD) calculations were carried out on single nanodisks. The nanodisks were modeled as 20 nm high gold disk-like geometries, with a diameter (D) of 100 nm, resembling the nanostructures used in the experiments, and we assumed a substrate refractive index of 1.52 and water (n = 1.33) as the external medium. A schematic representation is shown in Figure 1A. Figure 1B depicts the calculated plasmonic responses of a gold nanodisk contacted to the underlying glass in the absence of any adhering metal (black line) and via a 1 nm thick Ti (red line) and Cr (blue line) layer. Compared to the experimentally nonfeasible case in which the gold directly contacts the glass, these simulations show that a 1 nm thick layer of Ti is expected to result in a 54% reduction of the scattering intensity, while for the use of Cr this effect is even more dramatic (-73%). Furthermore, it can be observed that the full width half-maximum (fwhm) of the resonance peak increases when adhesion layers are introduced. In the absence of an adhesion layer, the fwhm equals 78 nm. However, introduction of a Ti or Cr layer results in fwhm values given by 118 and 158 nm, respectively. Both intensity attenuation and peak broadening increase even further for thicker adhesion

ARTICLE

layers, which can be ascribed to the absorptive properties of the employed adhesion metals and the subsequent damping of the LSPR. Even though the sensitivity to bulk changes of refractive index of these structures, defined as DλLSPR ð1Þ DnB is not significantly affected by the use of adhesion layers, both attenuation of the scattering intensity and peak broadening are expected to manifest themselves in a much lower signal-to-noise ratio of the measurements, thereby reducing the resolving power of these nanoplasmonic sensors. To confirm this effect experimentally, 20 nm high and 100 nm diameter Au nanodisk arrays were fabricated using two types of adhesion layers.To emphasize the adverse effects predicted by the simulations even more, 1 nm Ti and thicker 2 nm Cr adhesion layers were employed (samples without adhesion layer were not mechanically stable). Figure 1C compares the LSPR spectra in water for both latter cases, showing that, concurring with the theory, the plasmonic response is indeed strongly influenced by both the thickness and the composition of the adhesion layer. When comparing the scattering peaks, a 67% lower intensity and a much broader fwhm peak (fwhm ≈ 153 nm) is obtained for the nanodisks with the 2 nm Cr layer. To analyze how this affects the sensing performance of the nanodisks, the bulk sensitivity (ηB) was tested via low-angle reflectance measurements. For that purpose, a 12.7% glycerol solution (n = 1.3522) was employed to induce a RI change (Δn) of 0.02 refractive index units (RIU) with respect to water. A custom-made flow cell allowed us to carry out real-time, in-flow assays, thereby continuously monitoring the spectral position of the LSPR peak. Figure 1D shows the realtime tracking of the LSPR peak displacement (ΔλLSPR) for both nanodisk samples, showing a clear red-shift of the LSPR, while the higher RI glycerol solution flows through the flow cell. The normalized ηB-values extracted from these curves are approximately 95 nm RIU-1and 125 nm RIU-1 for the Cr and the Ti samples, respectively. Although the sensitivities, i.e., the absolute spectral shifts, are expected to be the same for both nanodisk substrates, the measurements indicate a different behavior. This observed discrepancy can be attributed to thickness uncertainties regarding the evaporated metal layer or different surface roughness. On top of the sensitivity, the signal-to-noise ratio of the measurements is a more critical parameter to evaluate the final limit of detection of the sensor. As shown in Figure 1D, the standard deviations of the signals for Ti and Cr linked nanodisks are given by σTi = 5.10-3 nm and σCr = 4.10-2 nm, respectively. These results clearly demonstrate that the lower intensity and larger fwhm obtained for the Cr sample deteriorate the limit of detection an order of magnitude when compared to the case in which Ti is used. Therefore, we can conclude that although absorption of light by metal adhesion layers inevitably affects the sensing performance the use of the less absorbing material is strongly advised to obtain a better refractometric sensing performance of the nanoplasmonic sensor. Therefore, we used nanodisks with a 1 nm Ti adhesion layer throughout all other experiments documented in this manuscript. Substrate Effects. As a next step, we studied the role that the underlying substrate refractive index has on the sensing performance of the nanoplasmonic nanodisks. A vast majority of transparent substrates on which plasmonic nanostructures can be fabricated have a larger RI than that of typical aqueous solutions used in biosensing assays. As a consequence, the symmetry of the ηB ¼

5346

dx.doi.org/10.1021/jp110363a |J. Phys. Chem. C 2011, 115, 5344–5351

The Journal of Physical Chemistry C

ARTICLE

Figure 2. (A) Near-field electric field profile of a single gold nanodisk located on a supporting substrate (n = 1.52) and surrounded by a homogeneous dielectric medium (n = 1.33). (B) Calculated bulk and surface sensitivities for a single gold nanodisk surrounded by a homogeneous dielectric surrounding (n = 1.33) and for nanodisks supported by substrates with different refractive indices. The inset figure shows the corresponding fwhm values of the calculated LSPR spectra.

EM near-field around the nanodisk is broken, shifting a large portion of such field toward the substrate, where it is almost entirely insensitive to RI changes of the external medium. Figure 2A, depicting the near-field EM intensity profile of a gold nanodisk, surrounded by water and located on a substrate (borosilicate glass, n = 1.52), clearly illustrates this effect. To analyze how this affects the sensing performance of these nanostructures, we calculated the sensitivity of the nanodisks for supporting substrates with different refractive indices. Typically, the sensitivity of nanoplasmonic sensors is quantified by the bulk sensitivity (ηB). However, when biosensing applications are considered, analyte binding takes place at the vicinity of the metal surface, and therefore, only a portion of the excited EM field is employed in the detection process. Consequently, the bulk sensitivity and the so-called surface sensitivity, defined as ηS ¼

DλLSPR Dd

ð2Þ

where d is the thickness of the adsorbed biolayer, can show different behavior.18 Therefore, distinction between both parameters is recommendable, making ηS a more interesting parameter to quantify and compare the biosensing performance of refractometric sensors. To check these differences, we calculated both ηB and ηS, assuming in the latter a homogeneous dielectric shell (n = 1.50) covering the exposed gold surface of the nanodisk. As shown by Figure 2B (left axis), the sensitivity of the nanodisk supported by borosilicate glass shows a drastic reduction (ηB = 225 nm RIU-1, ηS = 2.75 nm nm-1) compared to that of a disk in the absence of a substrate, that is, homogeneously surrounded by water (ηB = 356 nm RIU-1, ηS = 4.05 nm nm-1). For substrates

Figure 3. (A) Near-field electromagnetic field profile of a single gold nanodisk supported by a 22 nm isotropic pillar (n = 1.52) and surrounded by a dielectric medium (n = 1.33). (B) Calculated bulk and surface sensitivities for a single gold nanodisk supported by a dielectric pillar for different isotropic etch distances. The inset shows the corresponding LSPR spectral peak positions.

with an even higher RI, such as TiO2 (n = 2.4), the sensitivity decrease is even more pronounced, reaching values close to 45% decrease for the surface sensitivity and up to a 60% decrease for the bulk sensitivity (ηB = 145 nm RIU-1, ηS = 2.25 nm nm-1). Moreover, these simulations also show that a larger RI of the substrate is accompanied by a large broadening of the LSPR peaks (Figure 2B, right axis), thereby negatively affecting the ability to discriminate small spectral shifts. Suspended Nanodisks. The previous results clearly demonstrate that, to achieve higher sensitivities and an improved resolving power of the sensor, the influence of the high RI substrate needs to be minimized. To achieve this, different strategies have been reported, such as the use of low RI substrates22 or bottomup fabricated SiO2 nanoposts.23 Herein, we propose the isotropic wet chemical etch of the nanopatterned substrate to create suspended nanoparticles supported by dielectric pillars. The isotropic etching process not only frees up additional gold surface on the lower side of the nanodisk but also redistributes the EM-field in a more symmetrical way, thereby exposing the EM hot-spots to the dielectric environment surrounding the nanostructure, as Figure 3A displays. Both factors are expected to contribute to an enhancement of the sensitivity, which is confirmed in Figure 3B by the calculations of ηB and ηS as a function of the isotropic etching depth of the underlying substrate. It is worth noticing that ηB is steadily enhanced as the etching distance increases and slowly tends to the value of a nanodisk without substrate (356 nm RIU-1) when the etching depth is around 28 nm. In contrast, the surface sensitivity shows a faster increase, and already for an etching depth of 20 nm, the value is similar to that of the same nanodisk entirely surrounded by water. This can be explained by the liberation of the high EM intensity zone initially located at the interface between the lower side of the nanodisk and the substrate, 5347

dx.doi.org/10.1021/jp110363a |J. Phys. Chem. C 2011, 115, 5344–5351

The Journal of Physical Chemistry C

ARTICLE

Figure 4. (A) Real-time tracking of the LSPR peak position (λLSPR) for subsequent in-flow hydrofluoric acid etch cycles. (B) Experimental bulk sensitivities obtained after each of the etch cycles. The inset figure displays the real-time monitoring of the relative LSPR peak displacement (ΔλLSPR) before etching (red) and after the third etching cycle (blue).

thereby making it available for surface sensing purposes. As a consequence of the strong confinement and exponential decay of the excited EM field, the effect that the removal of high RI material has on ηS becomes smaller for additional etching steps, explaining the decreased enhancement slope for increasing pillar heights. Therefore, from an optical perspective, the nanodisk will behave as if it were “suspended” in the dielectric environment in the total absence of a supporting substrate, being the maximum achievable enhancement for the bulk and surface sensitivity at 70% and 60%, respectively. It is also relevant to observe that such sensitivity enhancement is induced despite the blue-shift of the LSPR caused by the substrate etching (associated with the decrease of the effective refractive index around the nanoparticle) since it has been previously shown that the sensitivity of plasmonic nanostructures improves as their LSPR is red-shifted.17 Such behavior highlights the importance of the redistribution of the EM generated by the LSPR and the improvement of its accessibility to the local changes of refractive index. Bulk Sensitivity Experiments. To corroborate these theoretical results, the bulk sensitivity of the nanodisks was experimentally tested through low-angle reflectance measurements. Since the employed substrate for the fabrication of the nanodisk arrays is borosilicate glass, we chose diluted hydrofluoric acid (HF, 0.05%) to accurately control the isotropic etching process to suspend the nanodisks. This process can be either performed via the in-flow injection of the diluted acid in the flow cell or externally, simply by immersion of the substrate in the acid solution, obtaining similar results. Figure 4A shows the real-time tracking of the spectral position of the resonance peak for three subsequent HF etches (numbered 0 to 3) inside the flow cell. As expected, every HF etch cycle causes a blue-shift of the LSPR, in agreement with the removal of high RI material, thereby diminishing the effective RI of the dielectric environment. Clearly, the first etching cycle provokes the largest blue-shift (ΔλLSPR ≈ 9.5 nm), in agreement with the theoretical predictions. As stated before, this effect is caused by “lifting” the EM hotspots that were originally located at the metal-substrate interface into the aqueous dielectric surroundings of the nanodisks. Further substrate etching shows smaller absolute blue-shifts since the effect of the underlying substrate becomes less dominant. A total of three consecutive etching cycles could be carried out (total ΔλLSPR ≈ -20 nm), before extreme LSPR peak-broadening was observed, accompanied with significant peak-intensity loss (data not shown). From this, two conclusions can be drawn.

First, from the constant peak widths during the first three etching cycles we can assume that the etching of the substrate is indeed isotropic and very homogeneous, affecting all disks in the same manner. Second, after any further substrate etching, the dielectric pillars become unstable, thereby tilting the nanodisks, or even cutting them loose from the substrate. Before and after each of the HF substrate etching steps, duplicated bulk sensitivity measurements were carried out by varying the RI of the dielectric surroundings of the nanodisks (Δn = 0.02 RIU). The corresponding results are shown in Figure 4B. Here, we can observe that ηB monotonically increases after each etching cycle, from 136 nm RIU-1 for nonsuspended to 222 nm RIU-1 for suspended nanodisks (three etching cycles), yielding an increment of approximately 67%. The inset of Figure 4B shows the real-time LSPR tracking of both of these measurements, exemplifying the enhanced bulk sensitivity obtained for the HF etched substrates. When comparing these results to the theoretically predicted values, we can conclude that although the same approximate linear trend is observed as a function of the etching distance the experimentally obtained values for ηB are substantially smaller. This difference can be attributed to geometrical differences between the ideal sharpedged disks used in the simulations and the fabricated nanodisks. Surface tension of the latter causes them to reshape, forming more oblate-shaped geometries. The accompanying rounding of the disk edges attenuates the local EM hotspots that are found at sharp particle features, thereby diminishing the sensitivity of the nanostructures. DNA Hybridization Experiments. To test whether the improved sensitivity observed in the bulk measurements also holds for the surface sensitivity, real biosensing measurements were carried out using the nanodisk samples as model systems for the detection of DNA hybridization. For that purpose, a short singlestranded DNA sequence (32-mer), part of the iap gene of bacteria Listeria innocua encoding a major part of the extracellular protein (p60), was used to monitor DNA hybridization. Thiolated, 24-mer receptor probes were previously immobilized on the surface of the nanostructures using an ex-situ immobilization protocol. Using a DNA target concentration of 200 nM, hybridization measurements were carried out in-flow, monitoring the spectral position of the LSPR as a function of time. To analyze the influence of the etching process on the surface-sensing performance of the nanodisks, these measurements were carried out on both nonsuspended and suspended nanodisk samples. Since an ex situ DNA immobilization protocol was used to 5348

dx.doi.org/10.1021/jp110363a |J. Phys. Chem. C 2011, 115, 5344–5351

The Journal of Physical Chemistry C

ARTICLE

Figure 5. (A) Real-time tracking of the relative LSPR peak displacement (ΔλLSPR) caused by DNA hybridization for nonsuspended (green) and suspended (red) nanodisks. The black line represents a control signal of a nonspecific DNA sequence. (B) Real-time monitoring of the relative LSPR peak displacement (ΔλLSPR) for several consecutive DNA hybridization and surface regeneration cycles carried out on suspended nanodisks.

functionalize the nanodisks, it was necessary to suspend the nanodisks ex situ. The latter was achieved via a 10 min HF immersion of the nanodisk substrate, resulting in an average isotropic etch distance of approximately 10-11 nm (see Supporting Information). Figure 5A depicts two typical hybridization signals (green and red curves) that were obtained in these experiments, belonging to nonsuspended and suspended nanodisks, respectively. From these graphs we can conclude that a considerable larger hybridization signal (ΔλLSPR ≈ 0.25 nm) is obtained for pillarsupported nanodisks, compared to those lying on a plain glass substrate (ΔλLSPR ≈ 0.16 nm). Injection of a noncomplementary control DNA sequence (black line) resulted in a small signal of ΔλLSPR ≈ 0.02 nm, thereby ruling out nonspecific adsorption of complementary DNA to the utmost extent. These hybridization measurements were carried out repeatedly on two different surfaces of both suspended and nonsuspended nanodisks, showing good levels of reproducibility. Whereas for nonsuspended nanodisks, DNA hybridization provoked an average ΔλLSPR of 0.176 ( 0.006 nm, for the suspended nanostructures ΔλLSPR equaled 0.248 ( 0.009 nm, yielding an average sensitivity enhancement of approximately 41%. Such enhancement could probably go further by improving the accessibility of the complementary DNA strands with the use of vertical and lateral spacers as we have proved previously for standard SPR.24 An important aspect for many label-free biosensing applications is the possibility of bioreceptor regeneration. For this, the binding of the analyte with its corresponding specific immobilized receptor must be broken, keeping the immobilized layer of bioreceptors active for further measurements. In our particular application, the regenerability of the functionalized nanostructured surfaces was tested by dehybridizing the duplex DNA with a formamide solution. The results of several consecutive hybridization and regeneration cycles, measured on an array of suspended nanodisks, are shown in Figure 5B. As can be seen in the graph, injection of formamide causes the λLSPR to blue-shift to its initial spectral position, that is, before DNA hybridization. As displayed by the graph, these hybridization and regeneration cycles could be carried out repeatedly. Subsequent injection of the target DNA always resulted in similar signal levels, throughout the “lifespan” of the receptor (more than 10 cycles), thereby emphasizing the stability and robustness of these suspended nanostructures as sensing platforms for real biosensing applications. Nanodisk Dimers. Once it has been established that suspended gold nanodisks exhibit a significantly improved sensing performance, as a next step, we tested this novel sensitivity enhancing

Figure 6. (A) Experimental scattering spectra of nonsuspended and suspended gold nanodisk dimer substrates, with the polarization of the incident light both parallel and perpendicular to the dimer axis. (B) Depicts the corresponding experimental bulk sensitivities. The inset figure shows the real-time tracking of the relative LSPR peak displacement (ΔλLSPR) for nonsuspended dimers for both longitudinal (blue) and transversal (red) polarization directions upon an induced bulk refractive index change of Δn = 0.02.

method on a more complex geometry to show that this technique can be easily extended to other nanoplasmonic structures. For this, short-ordered arrays of gold nanodisk (D = 100 nm, H = 20 nm) dimers were fabricated with interdisk distances of approximately 30 nm. In Figure 6A scattering spectra (continuous curves) of these closely distanced nanostructures are shown with the incident light polarization either parallel or perpendicular to the dimer axis. Note that the LSPR spectra corresponding to the longitudinal polarization, i.e., along the dimer axis, are red-shifted with respect to the transversal polarization, i.e., perpendicular to the dimer axis. As expected, suspension of these nanostructures, through the creation of isotropic dielectric pillars, results in 5349

dx.doi.org/10.1021/jp110363a |J. Phys. Chem. C 2011, 115, 5344–5351

The Journal of Physical Chemistry C blue-shifted LSPR spectra for both polarization directions (dashed curves). These results indicate that the plasmonic properties of these suspended dimer substrates are preserved. To compare the sensing characteristics of the dimers in both polarizations and to test whether this parameter is affected upon suspension of these structures, bulk sensitivity measurements were carried out. Figure 6B depicts the obtained ηB-values for both longitudinal and transversal polarizations, before and after suspension. When comparing both polarizations, it becomes clear that larger ηB-values are obtained for the direction parallel to the dimer axis, an effect attributed to the high EM intensity region created in the gap between the nanodisks. The inset of Figure 6B shows for both polarizations the real-time ΔλLSPR tracking (Δn = 0.02 RIU) for dimers located on the glass substrate. The extracted ηB-values for the transversal and longitudinal polarizations are given by 46 and 110 nm RIU-1, respectively, while suspension of these nanostructures approximately doubles these values to 90 and 245 nm RIU-1. From this, it can be concluded that, as expected, suspension of the dimers indeed improves their sensitivity. But more importantly, these results indicate that this sensitivity enhancing method can be used with more complex geometries with very short separation distances. As a consequence, this novel method presents itself as an easily applicable, general way to improve the sensitivity of a wide variety of refractometric nanoplasmonic sensing platforms.

’ CONCLUSIONS We have analyzed the main adverse effects for biosensing applications associated to the attachment of plasmonic nanostructures to a solid substrate. First, we have shown how the thin metal adhesion layer typically used to attach plasmonic nanostructures to the supporting substrate negatively influences the refractometric sensing performance, inducing a drastic reduction of the signal-to-noise ratio of the refractometric measurement. To diminish the consequences of this problem it is strongly advised to keep this layer as thin as possible and use a low absorptive metal. We have shown that 1 nm of Ti provides enough mechanical stability to allow in-flow biosensing measurements and surface regeneration. Moreover, the refractometric sensitivity is also severely affected by the presence of the high RI supporting substrate. Herein, an isotropic wet chemical etching process to distance the plasmonic nanostructures from the underlying substrate is introduced as a novel method to diminish the negative impact that the substrate has on the sensing sensitivity. Due to its facile implementation, this method can be easily extended to a wide variety of nanostructure geometries and substrates, exemplified in this manuscript with the use of short-ordered arrays of gold nanodisk dimers. To emphasize the potential use of these suspended nanostructures for real applications, the etched nanodisk platforms were used as model systems for DNA hybridization measurements. Duplex DNA formation was detected in both suspended and nonsuspended Au nanodisk arrays. In these experiments a 41% better sensitivity was observed for the suspended nanodisks, providing clear proof of the sensitivity enhancing properties of these dielectric nanopillars. The excellent surface-regenerative properties of the suspended nanostructures and the reproducibility of the measurements emphasize their robustness and stability, making them very suitable candidates for more exhaustive or even multiplexed biosensing assays.

ARTICLE

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional experimental details. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT A. Avi~n
o and R. Eritja from the Structural Biology Department of the Institute of Molecular Biology of Barcelona and CIBERBBN are greatly acknowledged for the synthesis of the DNA sequences. This research was carried out with the financial support of MULTIBIOPLAS of Spanish Ministry of Science and Innovation (TEC2009-08729); the M. Botín Foundation. M. A. Otte acknowledges the “Programa de Formaci
on de Profesorado Universitario (FPU)” of the “Ministerio de Educaci
on” of Spain; and B. Sepulveda acknowledges the “Ram
on y Cajal” program from “Ministerio de Ciencia e Innovaci
on” of Spain for financial support. ’ REFERENCES (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) Sep
ulveda, B.; Angelom
e, P. C.; Lechuga, L. M.; Liz-Marz
an, L. M. Nano Today 2009, 4, 244–251. (3) Liz-Marz
an, L. M. Langmuir 2006, 22, 32–41. (4) Murray, W. A.; Augui
e, B.; Barnes, W. L. J. Phys. Chem. C 2009, 113, 5120–5125. (5) Homola, J. Chem. Rev. 2008, 108, 462–493. (6) Feuz, L.; J€ onsson, P.; Jonsson, M. P.; H€o€ok, F. ACS Nano 2010, 4, 2167–2177. (7) Jonsson, M. P.; Dahlin, A. B.; Feuz, L.; Petronis, S.; H€o€ok, F. Anal. Chem. 2010, 82, 2087–2094. (8) Chung, P.-Y.; Lin, T.-H.; Schultz, G.; Batich, C.; Jiang, P. Appl. Phys. Lett. 2010, 96, 261108–261103. (9) Haes, A. J.; Van Duyne, R. P. J. Am. Chem. Soc. 2002, 124, 10596–10604. (10) Acimovic, S. S.; Kreuzer, M. P.; Gonz
alez, M. U.; Quidant, R. ACS Nano 2009, 3, 1231–1237. (11) Larsson, E. M.; Alegret, J.; Kall, M.; Sutherland, D. S. Nano Lett. 2007, 7, 1256–1263. (12) Kabashin, A. V.; Evans, P.; Pastkovsky, S.; Hendren, W.; Wurtz, G. A.; Atkinson, R.; Pollard, R.; Podolskiy, V. A.; Zayats, A. V. Nat. Mater. 2009, 8, 867–871. (13) McPhillips, J.; Murphy, A.; Jonsson, M. P.; Hendren, W. R.; Atkinson, R.; H€o€ok, F.; Zayats, A. V.; Pollard, R. J. ACS Nano 2010, 4, 2210–2216. (14) Chen, S.; Svedendahl, M.; Kall, M.; Gunnarsson, L.; Dmitriev, A. Nanotechnology 2009, 20, 434015. (15) Das, A.; Zhao, J.; Schatz, G. C.; Sligar, S. G.; Van Duyne, R. P. Anal. Chem. 2009, 81, 3754–3759. (16) Svedendahl, M.; Chen, S.; Dmitriev, A.; K€all, M. Nano Lett. 2009, 9, 4428–4433. (17) Miller, M. M.; Lazarides, A. A. J. Phys. Chem. B 2005, 109, 21556–21565. (18) Otte, M. A.; Sep
ulveda, B.; Ni, W.; P
erez-Juste, J.; Liz-Marz
an, L. M.; Lechuga, L. M. ACS Nano 2010, 4, 349–357. (19) Fredriksson, H.; Alaverdyan, Y.; Dmitriev, A.; Langhammer, C.; Sutherland, D. S.; Z€ach, M.; Kasemo, B. Adv. Mater. 2007, 19, 4297–4302. 5350

dx.doi.org/10.1021/jp110363a |J. Phys. Chem. C 2011, 115, 5344–5351

The Journal of Physical Chemistry C

ARTICLE

(20) Dahlin, A. B.; Tegenfeldt, J. O.; Hook, F. Anal. Chem. 2006, 78, 4416–4423. (21) Beeram, S. R.; Zamborini, F. P. ACS Nano 2010, 4, 3633–3646. (22) Brian, B.; Sep
ulveda, B.; Alaverdyan, Y.; Lechuga, L. M.; K€all, M. Opt. Express 2009, 17, 2015–2023. (23) Dmitriev, A.; H€agglund, C.; Chen, S.; Fredriksson, H.; Pakizeh, T.; K€all, M.; Sutherland, D. S. Nano Lett. 2008, 8, 3893–3898. (24) Carrascosa, L.; Calle, A.; Lechuga, L. Anal. Bioanal. Chem. 2009, 393, 1173–1182.

5351

dx.doi.org/10.1021/jp110363a |J. Phys. Chem. C 2011, 115, 5344–5351