Material-Selective Surface Chemistry for Nanoplasmonic Sensors

Jan 18, 2012 - ... to partly penetrate the biomolecular layers is taken into account. ...... Barbora ŠpačkováN. Scott Lynn, Jr.Jiří SlabýHana Š...
0 downloads 0 Views 3MB Size
Letter pubs.acs.org/NanoLett

Material-Selective Surface Chemistry for Nanoplasmonic Sensors: Optimizing Sensitivity and Controlling Binding to Local Hot Spots Laurent Feuz,† Magnus P. Jonsson,†,‡ and Fredrik Höök* Department of Applied Physics, Chalmers University of Technology, SE-41296 Gothenburg, Sweden S Supporting Information *

ABSTRACT: Optical sensors utilizing the principle of localized surface plasmon resonance (LSPR) offer the advantage of a simple label-free mode of operation, but the sensitivity is typically limited to a very thin region close to the surface. In bioanalytical sensing applications, this can be a significant drawback, in particular since the surface needs to be coated with a recognition layer in order to ensure specific detection of target molecules. We show that the signal upon protein binding decreases dramatically with increasing thickness of the recognition layer, highlighting the need for thin high quality recognition layers compatible with LSPR sensors. The effect is particularly strong for structures that provide local hot spots with highly confined fields, such as in the gap between pairs of gold disks. While our results show a significant improvement in sensor response for pairs over single gold disks upon binding directly to the gold surface, disk pairs did not provide larger signal upon binding of proteins to a recognition layer (already for around 3 nm thin layers) located on the gold. Local plasmonic hot spots are however shown advantageous in combination with directed binding to the hot spots. This was demonstrated using a structure consisting of three surface materials (gold, titanium dioxide, and silicon dioxide) and a new protocol for material-selective surface chemistry of these three materials, which allows for controlled binding only in the gap between pairs of disks. Such a design increased the signal obtained per bound molecule by a factor of around four compared to binding to single disks. KEYWORDS: Localized surface plasmon resonance (LSPR), nanoplasmonics, biosensing, hot spot, chemical surface modification, nanofabrication

B

ioanalytical sensors have become increasingly important tools in life science research as well as in medical diagnostics, drug discovery, environmental monitoring and food safety. In these different contexts, they are used to investigate biomolecular interactions with the prime aim to detect and determine the presence and concentration of specific biomolecular targets in complex mixtures. While many bioanalytical sensing concepts are based on fluorescence readout,1 methods that do not require chemical labeling have the advantage of making costly and time-consuming preparative steps redundant. In addition, the attachment of a chemical label to a target molecule may influence the interaction pattern between target and probe molecules. This has contributed to an increased interest in surface-based label-free bioanalytical sensors, which are typically based on biorecognition-induced signal generation caused by changes in optical, mechanical, or electrical properties of suitably designed substrates.2−4 One increasingly popular label-free bioanalytical sensing concept is based on the localized surface plasmon resonance (LSPR) associated with metal nanostructures.5−7 Localized surface plasmons are collective oscillations of the free electrons in the metal and can be excited by light at certain resonance conditions. As a result, distinct peaks appear in the optical extinction spectra of such structures. The use of LSPR for bioanalytical sensing is based on that the resonance condition is highly sensitive to changes in refractive index (RI) at the © 2012 American Chemical Society

interface of the metal nanostructure, as induced by, for example, biomolecular recognition events occurring close to the metal.8 The fact that the response is based on changes in interfacial refractive index makes LSPR sensors generic and highly versatile. At the same time, this generic feature makes it essentially impossible to discriminate between different types of biomolecules that bind to the surface. It is therefore of utmost importance to control the surface chemistry in such way that only the biomolecules of interest bind, while adsorption of other molecules is efficiently suppressed. This is particularly challenging in applications like biomarker identification9 and disease diagnostics10 in which case target proteins are typically present in concentrations many orders of magnitude lower than other biomolecules that need to be prevented from binding. For LSPR sensors, it is important to be aware that the sensitivity is inhomogeneously distributed on the sensor elements. This is particularly apparent for optically coupled nanoplasmonic structures, such as bow tie antennas,11,12 nanorod dimers13 and disk pairs,14,15 for which the plasmonic field, and hence the sensitivity to changes in RI, is highly localized to the gap between the nanostructures, while other regions provide significantly less sensitivity. As a result, if Received: November 7, 2011 Revised: January 14, 2012 Published: January 18, 2012 873

dx.doi.org/10.1021/nl203917e | Nano Lett. 2012, 12, 873−879

Nano Letters

Letter

biomolecular binding is allowed to occur everywhere on the sensor elements, only a very small fraction of the captured molecules will contribute to the observed signal. Note that this holds even in the case when the surrounding substrate is made inert for biomolecular binding, which in its own right has been demonstrated important for improved sensor performance.16 From a sensing perspective, especially when striving for detection of low abundance biomolecules, it would be preferable if the analyte can be controlled to bind only in the so-called hot spots possessing the highest sensitivity. We have in this work approached this challenge and present a method by which biomolecular binding can be directed to nanoscale gaps between pairs of disks, while binding to all other regions is efficiently suppressed by help of appropriate material-selective surface modifications. This was achieved using a new type of LSPR structure based on arrays of glass-supported gold disk pairs with a small region of titanium dioxide (TiO2) in each pair gap. Formation of a recognition layer for protein binding on TiO2 while making all other regions inert was accomplished using a new material-selective surface chemistry, compatible with selective modification of three different materials, gold (Au), silicon dioxide (SiO2), and titanium dioxide (TiO2) using disulfide-oligo(ethylene glycol) (OEG), poly(L-lysine)-graf tpoly(ethylene glycol) (PLL-g-PEG), and nitrodopamine-poly(ethylene glycol)-biotin (ND-PEG-biotin), respectively. The advantage of selective biomolecular binding to plasmonic hot spots is demonstrated and discussed in comparison with binding to well separated disks or pairs of disks, stressing the influence from the sensitivity being highest at and decaying rapidly away from the surface. The average exponential decay length of the plasmonic field is typically in the order of tens of nanometers or less.17,18 This is around 1 order of magnitude lower than for conventional surface plasmon resonance sensors19 and coincides with the dimension of many biomolecules. As a consequence, a recognition layer designed to promote binding of the analyte while suppressing nonspecific binding of all other biomolecules is likely to occupy a substantial fraction of the most sensitive part of the field. The signal induced by binding of target molecules to the modified surface may thus become significantly reduced compared with the situation when binding occurs directly on the metal surface. Using two recognition layers composed of either thiolpoly(ethylene glycol) (SH-PEG; 3.3 kDa) or disulfide-oligo(ethylene glycol) (OEG; 0.8 kDa), forming films with approximate thicknesses of 3 and 15 nm, respectively, the influence of the coating thickness on the response upon specific binding of a typically sized (65 kDa) target protein is addressed and analyzed in the context of the spatial sensitivity distribution of well separated LSPR-active disks and pairs of disks (named pairs below). These results were used to guide the design of the hot spot sensor. Characterization of the Bulk Sensitivity. Short-range ordered nanoplasmonic disks (155 nm in diameter and 30 nm thick) and pairs (consisting of two disks separated by around 15 nm, also of thickness 30 nm and with slightly different diameters of around 150 and 135 nm, respectively) were prepared by hole-mask colloidal lithography,20 as detailed in the Supporting Information. The extinction spectra in water of the resulting structures are shown in Figure 1. For the pairs, spectra were acquired both in longitudinal (L) and transverse (T) polarization with respect to the orientation of the pairs, as indicated in the figure. The bulk sensitivity of the structures, defined as the shift in their resonance wavelength per change in

Figure 1. Extinction spectra (in absorption units (abs)) in water are shown for disks (green) and pairs, the latter measured in longitudinal (L; blue) and transverse (T; red) polarization. The bulk sensitivity (S) is indicated with horizontal bars of proportional lengths for the different systems. Values for the figure of merit (FOM) are obtained by dividing the bulk sensitivity (S) by the full width at half-maximum (fwhm). The insets are scanning electron microscopy (SEM) micrographs of the disks and the pairs samples.

the refractive index of the surrounding environment, was determined by monitoring the changes in the centroid position21 of their extinction peak positions during exposure of the bare nanoplasmonic surface to aqueous glycerol solutions with increasing refractive indices. The disks displayed the highest bulk sensitivity (169 nm/RIU [refractive index units]) while the values are around 20% lower, but on the same order (between 125 and 141 nm/RIU, depending on the polarization settings) for the pairs. Although the bulk sensitivity is a common means to characterize LSPR sensors, this measure does not necessarily correlate with the accuracy by which changes in the peak position can be determined. In other words, the noise in the signal is not taken into consideration in the bulk sensitivity approach. Apart from details of the experimental setup, such as light source, detector, and the algorithms used to track the peak position,22 the noise depends strongly on the width and magnitude of the extinction peak. The first is commonly accounted for by considering the ratio between the bulk sensitivity and the full width at half-maximum (fwhm) of the extinction peak, which gives the so-called figure of merit (FOM) of the sensor.23 As shown in Figure 1, the disks display an almost 2-fold higher FOM than the pairs. Note, that while the longitudinal excitation displays higher bulk sensitivity than the transverse excitation, the situation with respect to FOM is the opposite. However, with the aim to investigate the role of hot spots and confined fields in the gaps of the pairs, we limit the results and discussion below to longitudinal excitation.24 While the actual noise in the signal also depends on other properties, such as the height of the extinction peak,22,25 the bulk sensitivity and the FOM here serve as rough evaluation of the bulk sensing capacity of the different structures. As analyzed in detail below, the signals and signal-to-noise ratios, which are the critical evaluation criteria in the case of biomolecular binding, depend on many other parameters, such as the quality of the recognition layer, probe density and the decay length and the spatial distribution of the evanescent field. 874

dx.doi.org/10.1021/nl203917e | Nano Lett. 2012, 12, 873−879

Nano Letters

Letter

Figure 2. (A) Binding curves shown by means of absolute peak centroid shifts of either dS-OEG-biotin (5%) or SH-PEG-biotin (5%) to disks (full lines) and pairs (dashed lines). (B) Final sensor response upon surface modification and protein binding. The recognition layer consists of either dSOEG-biotin (5%) or SH-PEG-biotin (5%) (striped red and orange bars, respectively). The areas around the gold structures were backfilled with PLL-g-PEG (gray bars) before exposing the surface to the protein NeutrAvidin (100 nM) (blue bars). The measured centroid shifts upon adsorption were normalized to a bulk sensitivity of 100 nm/RIU.

Surface Modification and Comparison of Decay Lengths. As stressed above, it is essential to use a recognition layer that promotes specific analyte binding while at the same time efficiently prevents unspecific binding of other biomolecules, which would otherwise generate false positives. The use of OEG and PEG as the base of the recognition layer was here chosen since they were previously shown to provide highly inert surfaces.26−29 Generally, the PEG systems are slightly superior to the OEG coatings in terms of protein repellence because of mainly two reasons. (i) Their larger thickness screens long-range attractive forces originating from the underlying substrate more efficiently. (ii) The functionality of OEG coatings relies on a highly ordered structure of the selfassembled monolayer (with all the hydrophilic end groups being on top of the layer and responsible for the protein repellence). Defects in the layer structure can thus lead to a loss in performance while the longer, flexible PEG chains are more forgiving in that respect. Figure 2A shows absolute resonance shifts induced by binding of the different recognition layers to the sensor surfaces. For both disks and pairs, adsorption of the longer PEG molecules resulted in markedly larger responses than those induced by the shorter OEG molecules. To facilitate a comparison between different samples, the absolute shifts at equilibrium were normalized to a common bulk sensitivity of 100 nm/RIU, so that the dependence on bulk sensitivity of the measured sensor response is cancelled. This is presented as the orange and red bars in Figure 2B, demonstrating that the trends persist after normalization. These results are attributed to the relatively long PEG molecules occupying a larger fraction of the sensitive plasmonic field than the shorter OEG molecules. Note that this is also in agreement with the significantly lower response observed upon subsequent NeutrAvidin binding (blue bars) to the PEG than to the OEG coatings, as discussed separately in detail below. This interpretation could also be independently verified by measuring the bulk sensitivity for the bare samples and for the

same samples after surface coating and subsequent protein binding (Figure 3). The reduction in bulk sensitivity (red

Figure 3. Bulk sensitivity measured before and after surface modification and protein binding for PEG and OEG on disks and pairs. The error bars correspond to measurements on two to four different samples. The decrease in bulk sensitivity after coating is presented in percentage in connection to the red arrows. Each value corresponds to the average of the decrease measured for two to four different samples.

arrows in Figure 3) gives a measure on the fraction of the sensitivity that is occupied by a specific surface coating on a given sample. In agreement with the results above, the use of a PEG coating resulted in around 2 times higher loss in sensitivity compared to when OEG was used, both for disks and for pairs. Adsorption of a biomolecular layer of thickness dtotal onto a sensor with a plasmonic field that decays with an average decay length L is expected to result in a fractional 875

dx.doi.org/10.1021/nl203917e | Nano Lett. 2012, 12, 873−879

Nano Letters

Letter

sensitivity loss of 1 − e−2dtotal/L.30 Using this relation and the measured bulk sensitivity prior to and after surface modification, the ratios between the thicknesses of PEG + protein and OEG + protein become 0.37 and 0.34 for the independent measurements on disks and pairs, respectively. These estimated values are based on the assumption that no glycerol solution penetrates the bound layers, yet they are in very good agreement with the thickness ratio of 0.36 obtained from literature values (using 3, 15, and 4 nm for the PEG, OEG, and protein layer, respectively).27,31 Notably, the peak shifts (normalized with bulk sensitivity) induced by OEG and PEG are more than a factor of 2 and 1.5 higher for the pairs than for disks (Figure 2). In analogy with the difference in shifts induced by OEG and PEG, this is also attributed to differences in the fraction of the sensitive field that the recognition layer occupies. The higher signals observed for the pairs thus indicate a more shallow extension of the evanescent field around the pairs compared with the disks, as expected for hot spots and highly confined plasmonic fields in the gap of the pairs.14 Again, this observation is in accordance with significantly higher reductions in bulk sensitivity for the pairs compared with the disks, both when PEG and when OEG is used (see Figure 3). With a fractional decrease in bulk sensitivity estimated as 1 − e−2dtotal/L (see above), the ratio between the decay lengths, Ldisks/Lpairs, can be estimated using either the results obtained with OEG or with PEG, which gives 2.1 and 2.2, respectively. This suggests an on average 2-fold higher confinement of the field in the case of pairs. Note however, that this ratio would increase further if the expected ability of glycerol to partly penetrate the biomolecular layers is taken into account. Since the degree of hydration/penetration is not known, we refrain from speculations on the exact values at this point, but state that the difference in decay length between the two structures is significant, in particular in the context of probing biomolecules. After modification of the gold structures with either PEG or OEG, but prior to protein binding, PLL-g-PEG was adsorbed on the SiO2 part of the sensor to minimize unspecific protein binding to the support. The importance of this step is evident from the small, yet clear, signals observed from the adsorption of PLL-g-PEG (gray bars in Figure 2B). Also note that the signals induced by PLL-g-PEG binding is considerably lower for the systems premodified with the thick PEG coating than for the short OEG coating, consistent with binding further away from the most sensitive regions of the sensor elements. Sensor Performance upon Protein Binding. With respect to the performance in real biosensing applications, different sensor platforms are preferably compared with respect to the signals and signal-to-noise ratios obtained during binding of the analyte to the recognition layer. We have investigated this aspect by using PEG and OEG molecules that carry a biotin group at their chain end. The protein NeutrAvidin was then used as it specifically binds to the biotin-activated part of the sensor surface. The response (change in centroid position, Δλpeak) upon binding of proteins forming a film with a thickness dprotein on a recognition layer of thickness drec can be estimated from

Δλ peak = ΔnSe−2drec / L(1 − e−2d protein / L)

protein binding to OEG-biotin compared to binding to PEGbiotin. Under the assumption that the number of bound molecules per area is the same for both recognition layers, the difference must be attributed to NeutrAvidin binding on locations of different sensitivity. Indeed, with a thicker recognition layer the proteins bind further away from the surface, where the sensitivity is lower (see first exponential factor in eq 1). This illustrates the importance of designing thin, yet high-quality, recognition layers for LSPR sensors, such that binding events can occur as close as possible to the metal surface. It should be stressed that despite the much larger sensor signals induced upon formation of the recognition layers (OEG and PEG) on the pairs compared with the disks, the resonance shifts upon protein binding is comparable for the disks and the pairs. This is again consistent with a shorter average decay length for the pairs. For binding on the thicker PEG-based recognition layer, the lower signal upon NeutrAvidin binding may, in addition to protein binding further away from the surface, also originate from reduced space in the gap between the pairs caused by the fact that the longer PEG chains occupy a larger volume than the OEG molecules. In any case, it is from these results clear that even for relatively thin recognition layers, the confined field in the hot spots of the pairs does not necessarily provide higher sensor signals upon protein binding. However, the situation may be different from the perspective of the signal induced per bound molecule, in particular if the molecules can be controlled to bind only in the hot spots, as addressed below. Directed Protein Binding to the Hot Spot of Disk Pairs. From the results presented above it is clear that (i) a thin surface coating is preferable for nanoplasmonic sensors and (ii) a system that provides large signals upon binding directly to the surface may not always be superior after surface coating, primarily due to differences in the average decay lengths in relation to the thickness of the surface coating. However, not considered so far is the fact that the plasmonic field, and hence the sensitivity, is highly inhomogeneous for LSPR sensors with both high-sensitivity regions and other regions with sensitivity close to zero. Hence, if binding is allowed everywhere on the sensor surface, the observed signal will correspond to only a small fraction of all captured biomolecules, while the majority will not contribute significantly to the signal response. One situation when this becomes particularly problematic is when the total number of biomolecules in the sample is low, such that the saturated surface coverage remains significantly lower than the equilibrium coverage governed by the nature of the interaction between the suspended protein and the surfaceimmobilized probe. In this case, it is highly preferable if all biomolecular binding reactions are directed to the regions with highest sensitivity. An additional limitation for measurements at low protein concentrations is the time required in order to reach a detectable surface coverage. By reducing the bioactive surface area of a nanoplasmonic sensor to the sensitive regions only, the binding kinetics can be improved by more than 1 order of magnitude, as we recently demonstrated for a nanoplasmonic sensor based on holes in a thin metal film.16 We have here explored the possibility of improving the sensor performance of the pair system described above by directing the biomolecular binding to the gap between the disks in each pair. This was achieved using a nanoplasmonic structure based on three different materials (Au, TiO2, and SiO2) combined with material-selective surface modification that

(1)

where Δn is the change in refractive index caused by the protein layer.32 As shown in Figure 2B (blue bars), for both disks and pairs the sensor response is significantly larger for 876

dx.doi.org/10.1021/nl203917e | Nano Lett. 2012, 12, 873−879

Nano Letters

Letter

Figure 4. (A) Schematic illustrating the fabrication process of hot spots structures, as described in more detail in the Supporting Information. The dashed ring in the top views of iii−vi illustrates the available area on the glass surface. The masking layers were omitted from the top views in iv−vi to better illustrate the current structure on the glass surface. Note that while only one nanostructure is shown in the schematic for simplicity, a large short-range ordered array is produced by the protocol on the whole glass sample. (B) SEM micrograph of an optimized nanoplasmonic structure composed of three materials. The two Au disks ensure nanoplasmonic activity. The TiO2 area exposed to the environment is kept small by partially covering the original TiO2 disk with two SiO2 disks. The surface is then chemically modified such that only the TiO2 area is bioactive.

Figure 5. (A) Comparison of Neutravidin (100 nM) binding to single Au disks (blue curve) and to a hot spot area between Au pairs (red curve). The signal-to-noise ratio has been evaluated by dividing the absolute binding signal by twice the standard deviation of the noise signal. (B) The data from panel A are here shown as signal per molecule, accounting for the fact that the available binding area (and thus the number of molecules contributing to the signal) of the hot spot structure is 30× smaller than for the disk structure.

illustrated in Figure 4A, vii and shown for a real sample in Figures 4B. In the first surface modification step, the surface was exposed to a disulfide-OEG solution. Disulfides bind to gold, but do not adsorb on either SiO2 or TiO2.34 In the second step, ND-PEG, where 5% of the end groups of the PEG carried a biotin moiety, were adsorbed on TiO2 exclusively.35 Finally, the remaining SiO2 surface was backfilled with PLL-g-PEG, to avoid unspecific binding also to SiO2 regions. The bioactive surface area created in this way matches the high-sensitivity region in the gap and corresponds to no more than 1% of the entire substrate. Figure 5A shows the temporal evolution of the peak shift upon addition of NeutrAvidin (100 nM) together with the corresponding response for the same binding reaction to OEG-biotin modified disks (see above), in which case the sensor area corresponds to around 20% of the entire substrate. The absolute signal and the noise level are higher and lower by a factor of 7 and 3, respectively, for the disks compared with the hot spot structure. The signal-to-noise ratio is about 4500 for the single disks, but only 190 for the hot-spot structure. However, it is in this context important to recall that the number of molecules that has triggered the total response (per nanostructure) is very different in the two cases. For a disk one can assume a bioactive area (the whole area of the disk) of approximately 30 × 103 nm2, while the bioactive area of the hot spot structure is only about 1 × 103 nm2. Since the bioactive area can be assumed to be proportional to the number of

allows for independent modification of these three materials. The structure is composed of a small area of TiO2 located between a pair of gold disks, while the rest of the surface is made of SiO2 (Figure 4B). The fabrication procedure is illustrated in Figure 4A and the details are provided in the Supporting Information. In brief, the structure was fabricated using hole-mask colloidal lithography.20 First, short-range ordered holes in a thin Au layer on a PMMA-coated glass slide were prepared using conventional colloidal lithography (Figure 4A, i,ii).33 These holes were transferred to the underlying PMMA layer by oxygen plasma etching (Figure 4A, iii). The etching was optimized to obtain an undercut in the PMMA layer (see Figure 4A, iii), such that each hole in the PMMA could be used to create multiple disks by evaporation at different angles. A thin (10 nm) Ti disk was produced on the glass surface by evaporation at normal incidence angle (0°, perpendicular to the surface, Figure 4A, iv). Two SiO2 disks with a common thickness of 10 nm were then fabricated by evaporation at angles of ±16°, only leaving a narrow strip of TiO2 between those disks (Figure 4A, v). By exposing the sample to ambient conditions, the Ti oxidized to TiO2. The substrate holder was then turned 90° and two gold disks of 30 nm thickness were fabricated by evaporation at angles of ±15° (Figure 4A, vi), thereby further decreasing the ambient-exposed TiO2 area and creating the nanoplasmonic disk pairs. The structure was finalized by PMMA lift off and consists of a pair of plasmonic gold disks with a thin region of TiO2 in the gap, as 877

dx.doi.org/10.1021/nl203917e | Nano Lett. 2012, 12, 873−879

Nano Letters



bound molecules (binding on TiO2 on the floor between the gold disk is not expected to be hampered by crowding effects in the same way as if binding occurs on the walls of the gold discs), the signal generated per molecule for the hot spot structure is around 4 times larger than in the case of single disks, as illustrated in Figure 5B. Such an increased signal per molecule is particularly beneficial when working with samples containing a low total number of molecules. Furthermore, with a total number of bound protein molecules per hot spot estimated to around 20 and assuming that a single hot spot can be probed with a signal-to-noise ratio of around 200, as here realized for multiple hot spots, each bound molecule would correspond to a signal that is 10 times larger than the noise level. This thus suggests a rational approach that will enable digital recording of single specific protein binding events. We also note that the actual noise level of the hot spot structure was a factor of 3 higher than that of the discs. This is primarily attributed to the broader plasmonic peak of our particular disk pair structure and could likely be improved using alternative dimer configurations, such as bow tie antennas11,12 or nanorod dimers,13 combined with TiO2 for material-selective surface modifications of the type presented here. By adjusting both the thickness of the recognition layer on the TiO2 and the thickness of the gold structure, one could further tune the vertical position of the hot spot to the exact height where the biomolecular interaction of interest is expected to occur. That way, an even better signal per molecule and thus overall performance of the nanoplasmonic sensor is conceivable, and the severe limitation imposed by the shallow evanescent field associated with discs and pairs may be possible to circumvent. In conclusion, it is clear from this study that the most suitable LSPR sensor system depends on the application of interest, for example, on whether low surface coverages need to be measured (low analyte concentrations) or if the aim instead is to detect a low number of total molecules (low concentrations and small sample volumes). Common for these applications is that as thin recognition layers as possible are required in order to obtain sufficiently high signals upon specific binding of analytes. While this is a direct result of the sensitivity being highly confined to the surface, it should be noted that this feature can also be beneficial in other contexts. For example, a short decay length can be utilized to monitor biomolecular structural changes.36,37 Furthermore, compared with conventional SPR sensors, the confined field of LSPR sensors leads to a lower relative effect from bulk RI changes compared with interfacial RI changes. This reduces the need for both reference channels and temperature stabilization, which is likely to be advantageous in, for example, the development of simple, cheap, and compact sensors, such as point-of-care instruments. Combined with the possibility to direct the biomolecular binding events to hot spot regions in between two adjacent gold nanostructures, this is expected to open new avenues for nanoplasmonic sensing applications.



Letter

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Present Address

‡ Kavli Institue of Nanoscience, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands.

Author Contributions †

These authors contributed equally to this work.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was financially supported by the SNSF (Swiss National Science Foundation, Grant PBEZA- 121238) and Swedish Foundation for Strategic Research. Erik Nilebäck (Chalmers University of Technology, Göteborg, Sweden) is acknowledged for providing the OEG solutions. M.P.J. also acknowledges support from the Wenner-Gren Foundations.



REFERENCES

(1) Ciruela, F. Curr. Opin. Biotechnol. 2008, 19 (4), 338−343. (2) Homola, J. Chem. Rev. 2008, 108 (2), 462−493. (3) Cooper, M. A.; Singleton, V. T. J. Mol. Recognit. 2007, 20 (3), 154−184. (4) Daniels, J. S.; Pourmand, N. Electroanalysis 2007, 19 (12), 1239− 1257. (5) Anker, J. N.; Hall, W. P.; Lyandres, O.; Shah, N. C.; Zhao, J.; Van Duyne, R. P. Nat. Mater. 2008, 7 (6), 442−453. (6) Stewart, M. E.; Anderton, C. R.; Thompson, L. B.; Maria, J.; Gray, S. K.; Rogers, J. A.; Nuzzo, R. G. Chem. Rev. 2008, 108 (2), 494−521. (7) Jonsson, M. P.; Dahlin, A. B.; Jönsson, P.; Höök, F. Biointerphases 2008, 3 (3), FD30−FD40. (8) Englebienne, P. Analyst 1998, 123 (7), 1599−1603. (9) Sawyers, C. L. Nature 2008, 452 (7187), 548−552. (10) Bell, J. Nature 2004, 429 (6990), 453−456. (11) Kinkhabwala, A.; Yu, Z.; Fan, S.; Avlasevich, Y.; Müllen, K.; Moerner, W. E. Nat. Photonics 2009, 3 (11), 654−657. (12) Jiao, X.; Goeckeritz, J.; Blair, S.; Oldham, M. Plasmonics 2009, 4 (1), 37−50. (13) Kottmann, J.; Martin, O. Opt. Express 2001, 8 (12), 655−663. (14) Acimovic, S. S.; Kreuzer, M. P.; González, M. U.; Quidant, R. ACS Nano 2009, 3 (5), 1231−1237. (15) Cao, W.; Huang, T.; Xu, X.-H. N.; Elsayed-Ali, H. E. J. Appl. Phys. 2011, 109 (3), 034310−6. (16) Feuz, L.; Jönsson, P.; Jonsson, M. P.; Höök, F. ACS Nano 2010, 4 (4), 2167−2177. (17) Jonsson, M. P.; Jönsson, P.; Höök, F. Anal. Chem. 2008, 80 (21), 7988−7995. (18) Rindzevicius, T.; Alaverdyan, Y.; Dahlin, A.; Hö ök, F.; Sutherland, D. S.; Käll, M. Nano Lett. 2005, 5 (11), 2335−2339. (19) Homola, J.; Yee, S. S.; Gauglitz, G. Sens. Actuators, B 1999, 54 (1−2), 3−15. (20) Fredriksson, H.; Alaverdyan, Y.; Dmitriev, A.; Langhammer, C.; Sutherland, D. S.; Zäch, M.; Kasemo, B. Adv. Mater. 2007, 19 (23), 4297−4302. (21) Dahlin, A. B.; Tegenfeldt, J. O.; Höök, F. Anal. Chem. 2006, 78 (13), 4416−4423. (22) Dahlin, A. B.; Chen, S.; Jonsson, M. P.; Gunnarsson, L.; Käll, M.; Höök, F. Anal. Chem. 2009, 81 (16), 6572−6580. (23) Sherry, L. J.; Chang, S. H.; Schatz, G. C.; Van Duyne, R. P.; Wiley, B. J.; Xia, Y. N. Nano Lett. 2005, 5 (10), 2034−2038. (24) Otte, M. A.; Estevez, M. C.; Carrascosa, L. G.; GonzalezGuerrero, A. B.; Lechuga, L. M.; Sepulveda, B. J. Phys. Chem. C 2011, 115 (13), 5344−5351.

ASSOCIATED CONTENT

S Supporting Information *

This information contains details about the fabrication process of the nanostructures, the surface modification procedures as well as about the LSPR measurement setup. A figure showing a bulk sensitivity measurement is also included. This material is available free of charge via the Internet at http://pubs.acs.org. 878

dx.doi.org/10.1021/nl203917e | Nano Lett. 2012, 12, 873−879

Nano Letters

Letter

(25) Mazzotta, F.; Wang, G.; Hägglund, C.; Höök, F.; Jonsson, M. P. Biosens. Bioelectron. 2010, 26 (4), 1131−1136. (26) Prime, K. L.; Whitesides, G. M. J. Am. Chem. Soc. 1993, 115 (23), 10714−10721. (27) Harder, P.; Grunze, M.; Dahint, R.; Whitesides, G. M.; Laibinis, P. E. J. Phys. Chem. B 1998, 102 (2), 426−436. (28) Pasche, S.; De Paul, S. M.; Vörös, J.; Spencer, N. D.; Textor, M. Langmuir 2003, 19 (22), 9216−9225. (29) Morra, M. J. Biomater. Sci., Polym. Ed. 2000, 11 (6), 547−569. (30) Jung, L. S.; Campbell, C. T.; Chinowsky, T. M.; Mar, M. N.; Yee, S. S. Langmuir 1998, 14 (19), 5636−5648. (31) Feuz, L.; Leermakers, F. A. M.; Textor, M.; Borisov, O. Langmuir 2008, 24 (14), 7232−7244. (32) Kedem, O.; Tesler, A. B.; Vaskevich, A.; Rubinstein, I. ACS Nano 2011, 5 (2), 748−760. (33) Hanarp, P.; Sutherland, D. S.; Gold, J.; Kasemo, B. Colloids Surf., A 2003, 214 (1−3), 23−36. (34) Love, J. C.; Estroff, L. A.; Kriebel, J. K.; Nuzzo, R. G.; Whitesides, G. M. Chem. Rev. 2005, 105 (4), 1103−1169. (35) Malisova, B.; Tosatti, S.; Textor, M.; Gademann, K.; Zürcher, S. Langmuir 2010, 26 (6), 4018−4026. (36) Jonsson, M. P.; Jönsson, P.; Dahlin, A. B.; Höök, F. Nano Lett. 2007, 7 (11), 3462−3468. (37) Hall, W. P.; Modica, J.; Anker, J.; Lin, Y.; Mrksich, M.; Van Duyne, R. P. Nano Lett. 2011, 11 (3), 1098−1105.

879

dx.doi.org/10.1021/nl203917e | Nano Lett. 2012, 12, 873−879