Article pubs.acs.org/ac
Curvature of the Localized Surface Plasmon Resonance Peak Peng Chen†,‡ and Bo Liedberg*,†,‡ †
Center for Biomimetic Sensor Science, 50 Nanyang Drive, Research Techno Plaza, Sixth Floor, Singapore 637553, Singapore School of Materials Science and Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
‡
S Supporting Information *
ABSTRACT: Localized surface plasmon resonance (LSPR) occurring in noble metal nanoparticles (e.g., Au) is a widely used phenomenon to report molecular interactions. Traditional LSPR sensors typically monitor shifts in the peak position or extinction in response to local refractive index changes in the close vicinity of the nanoparticle surface. The ability to resolve minute shifts/extinction changes is to a large extent limited by instrumental noise. A new strategy to evaluate LSPR responses utilizing changes in the shape of the extinction spectrum (the curvature) is proposed. The response of curvature to refractive index changes is investigated theoretically using Mie theory and an analytical expression relating the curvature to the refractive index is presented. The experimentally derived curvatures for 13 nm spherical gold nanoparticles (AuNPs) exposed to solvents with different bulk refractive indices confirm the theoretical predictions. Moreover, both the calculated and experimental findings suggest that the curvature is approximately a linear function of refractive index in regimes relevant to bio and chemical sensing. We demonstrate that curvature is superior over peak shift and extinction both in terms of signal-to-noise (S/N) ratio and reliability of LSPR sensors. With a curvature, one could readily monitor submonolayer adsorption of a low molecular weight thiol molecule (Mw = 458.6) onto 13 nm AuNPs. It is also worthwhile mentioning that curvature is virtually insensitive to instrumental instabilities and artifacts occurring during measurement. Instabilities such as baseline tilt and shift, shift in peak position as well as sharp spikes/steps in the extinction spectra do not induce artifacts in the sensorgrams of curvature.
T
(e.g., from the light source and/or detector). As a result, detecting molecules at ultralow concentrations using LSPR is very demanding especially for low molecular weight targets. Many efforts have been undertaken to increase the sensitivity of LSPR sensors by synthesizing novel nanomaterials with higher polarizability, such as nanorods,27 nanostars,25 nanoshells, 28 nanotriangles,29 and so forth. Physicists also have explored the potential of meta-materials by fabricating novel submicron plasmonic patterns using lithography techniques.30,31Although these efforts improve the refractive index sensitivity of LSPR sensors to some extent, they also bring many problems such as high cost, low yield, and tedious processing. In this contribution, we report a new strategy to monitor changes in the shape of LSPR peak of gold nanoparticles (AuNPs). Contrary to peak shifts or extinction changes that monitors the changes of the peak maximum at a certain point, the curvature of the LSPR peak takes into account the overall peak shape. Our theoretical and experimental findings demonstrate that the curvature (2nd order derivative at peak maximum) is approximately linearly related to refractive index in regimes relevant to sensing.
he collective oscillation of free electrons in the conduction bands of gold and other noble metals is known as surface plasmon resonance (SPR). Since the first report of SPR sensing in 1983,1 it has been widely explored and developed into a wellestablished commercial tool.2−4 Recently, noble metal nanoparticles have attracted tremendous interest in both fundamental5−11 and applied research including sensing,12−15 surface enhanced Raman scattering (SERS),16−18 and photovoltaics,19 among others. Contrary to propagating surface plasmon waves, the electron plasma in nanoparticles is localized to the particle, a phenomenon referred to as LSPR.20 The LSPR resonance frequency/wavelength of the nanosized noble metals depends on their size, shape, composition, interparticle distance, and the medium in which they are embedded/dissolved .6 Biomolecular sensing by LSPR sensors mostly relies on the peak shift or extinction changes induced by refractive index changes or interparticle aggregation phenomena occurring in response to molecular recognition and/or attachment/bridging.21−24 The refractive index sensitivity of LSPR sensors depends on the polarizability of the nanosized plasmonic structure/particles, which commonly varies from a few tens to a few hundred nmRIU−1.25 Despite the many advantages over propagating (planar) SPR sensors in terms of spatial resolution and cost, the refractive index sensitivity of LSPR sensors is still about 4 orders of magnitude smaller than of planar SPR.26 In addition, the ability to resolve minute changes in the LSPR peak position/extinction is limited by instrumental/chemical noise © 2014 American Chemical Society
Received: March 9, 2014 Accepted: July 1, 2014 Published: July 1, 2014 7399
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
■
Article
polynomial fitting of the absorption spectra using 25 data points, symmetric around the absorption maximum. The curvature is obtained similarly by polynomial fitting of the second derivative spectra using 25 data points, symmetric around the curvature maximum (more details can be found in Supporting Information, section 5). The peak shift obtained by the centroid method is calculated in a 60 nm wavelength window around the peak maximum. (A software for calculation of a curvature from experimental data is available upon request from the authors.)
EXPERIMENTAL SECTION Chemicals and Materials. Sodium citrate tribasic dehydrate (99%), gold(III) chloride trihydrate (99.9%), (3-aminopropyl)triethoxysilane (APTES) (97%), and glycerol (99%) were purchased from Sigma-Aldrich. Microscope slides from Sail Brand were used as glass substrates. Thiol PEG7 acid (O(2-carboxyethyl)-O′-(2-mercaptoethyl)-heptaethylene glycol) was purchased from Polypure AS, Norway. Spherical AuNPs with the size of 20 and 40 nm were purchased from BBI solutions. Nanoparticle Synthesis and Characterization. Gold nanoparticles were synthesized by reduction of HAuCl4 using trisodium citrate.32 In a typical synthesis, 90 mL of Milli-Q water and 10 mL of 10 mM HAuCl4 were mixed and heated to above 100 °C under reflux for 2 h. An aqueous solution of 10 mL of 77.6 mM trisodium citrate was added quickly to the reflux solution by piston and needle. The color of the solution changes from pale yellow to red in ca. 10 s. Heating was maintained for another 15 min before cooling to room temperature in air. The AuNPs were stored at 4 °C for future use. A Joel 2100 transmission electron microscope (TEM) was used to characterize the size of the nanoparticles and their distribution. Absorption spectra of AuNPs in solution were measured with a PerkinElmer Lambda 35 absorption spectrometer. Zeta-potential of the citrate capped AuNPs was measured using the spectrometer from Malvern Instruments. The size of AuNPs was derived from TEM images using the software ImageJ, and the distribution and standard deviation were calculated from over 200 randomly chosen individual particles. Surface Modification and Characterization. Pieces of glass substrate with 2 × 0.5 cm2 were cut from microscope slides and incubated in a mixture of water, 30% H2O2, and 25% ammonia solution (5:1:1 in volume) at 80 °C for 10 min for remove contaminants. Then they were rinsed copiously with Milli-Q water for a few times. After the cleaning, the glass substrates were dried by nitrogen gas. The cleaned glass substrates were incubated with 1% (volume) of APTES in 99.9% ethanol for 30 min for surface functionalization. After functionalization, the substrates were rinsed with Milli-Q water and dried in nitrogen gas. A drop of as-synthesized AuNPs (200 μL) was casted onto the APTES modified glass surface and incubated for 30 min to 4 h to generate samples possessing different surface densities. After incubation, the glass substrates with AuNPs were rinsed with Milli-Q water and dried with nitrogen gas. Secondary electron images on the AuNP coated glass were recorded using a Joel 7600F field emission scanning electron microscope (FESEM), and a thin layer of platinum coating was applied to avoid a charging effect. Absorption spectra were measured by immersing the substrate into a cuvette containing 1 mL of solution using a PerkinElmer spectrometer operating at a resolution of 1 nm. Solvents with different refractive indices were prepared by mixtures of water and glycerol. Volume fractions of glycerol in the five solvents are 0%, 25%, 50%, 75%, and 100%, respectively. Thiol PEG7 solution was diluted in cuvette from stock a solution of 1 mM. Calculations and Spectral Evaluation. Analytical derivation of derivatives and numerical calculations of absorption cross section are performed using the software Mathematica 9. Second-order derivatives are calculated from two sequential first order derivatives, and 128 data points are used in the calculation of curvature. The absorbance value is obtained by
■
RESULTS AND DISCUSSION For a spherical particle, the Maxwell’s equation is analytically solvable, which was first reported by Gustav Mie in 1908,33 and further developed by Bohren and Huffman.34 The extinction, absorption and scattering cross sections of an arbitrary spherical particle can be calculated from the Mie theory. The wavelength and refractive index dependent absorption A(λ, n) of a 13 nm AuNPs can be derived from Mie theory: A(λ , n) =
(a 2 + b 2 λ ) 18πVn2Nl 2 λ (a 2 + b2λ) (b1λ + a1 + 2n2) + 1 (1)
where, a1, a2, b1, b2, and π are constants, V is the volume of particle, N is the density of electrons, λ is the wavelength of light, l is the length of the optical path, and n is the refractive index of surrounding medium. The curvature (K) is related to the sharpness of the peak, and mathematically it can be defined as the absolute value of the second-order partial derivative of absorbance with wavelength at the peak position. K=
∂ 2A (A = A max ) ∂λ 2
(2)
The curvature as a function of refractive index n can be written as ⎧ 3 2⎡ b2 ⎤3 2 b n a (2 n a ) − + ⎪ ⎢ ⎥⎦ 1 2 1 ⎪ ⎣ b1 K (n) = 36πVNl ⎨− 2n2 + a1 ⎪ ⎪ ⎩ +
b12b2n2 (2n2 + a1)2
+
⎤⎫ + a1)⎥⎦ ⎪ ⎪ ⎬ 2 3 (2n + a1) ⎪ ⎪ ⎭
⎡ b13n2⎣⎢a 2 +
b2 (2n2 b1
(3)
The details of the calculation of curvature as a function of bulk refractive index are outlined in Supporting Information, section 1. Gold nanoparticles with a size of ∼13 nm were synthesized by reduction of chloroauric acid (HAuCl4) with sodium citrate.32 The size of the AuNPs can be tuned by the ratio of sodium citrate to chloroauric acid. Figure S2a shows a transmission electron microscope (TEM) image of the assynthesized AuNPs, and their size distribution is summarized in the histogram plot in Figure S2b. Most of the AuNPs have a size between 11 and 14 nm, with an average size of 12.9 nm and standard deviation of 1.1 nm, respectively. The absorption spectrum of the AuNPs in solution is shown in Figure S2c. A prominent peak is observed at 521 nm, which is due to LSPR. 7400
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
Article
Figure 1. (a) Scanning electron microscope (SEM) image of AuNPs assembled on the glass substrate. Scale bar 100 nm. (b) Absorption spectrum of AuNP assembly on glass in air.
Figure 2. (a) Absorption spectra of AuNPs on glass substrate immersed in solvents with different refractive indices: 1.333 (black), 1.364 (red), 1.399 (blue), 1.432 (green), and 1.474 (purple), respectively. Spectra are offset for better visualization. Inset: zoom-in absorption spectra of AuNPs in solvent with a refractive index of 1.333 (black) and 1.474 (purple). (b) Second-order derivatives of absorption spectra in (a). (c) Experimental (black circles/line) and calculated (red circles/line) curvature of AuNPs immersed in media with different refractive indices. Experimental (black) and calculated (red) data are fitted linear function to yield a slope of 5.19 × 10−6 nm−2 RIU−1 (R2 = 0.9804) and 5.49 × 10−6 nm−2 RIU−1 (R2 = 0.9876), respectively. (d) Experimentally measured peak shift (squares) and absorbance (triangles) at different refractive indices. The peak shift is fitted linearly, with a slope of 34 nmRIU−1 (R2 = 0.9970).
adsorb via electrostatic attraction to the positively charged APTES surface. Figure 1a shows the secondary electron image of a typical AuNP assembly on APTES-modified glass. The AuNPs are evenly distributed on the glass substrate, with a few exceptions where aggregation is observed. The absorption spectrum of the AuNP assemblies measured in air is shown in Figure 1b. The absorption spectrum of the AuNP assemblies resembles that of AuNPs in solution Figure S2c, except for a slight broadening and the appearance of a shoulder between
The peak is sharp, indicating narrow size and shape distribution of the AuNPs, which also agrees with the observations of TEM images, Figure S2a,b. The AuNPs are negatively charged due to the capping agent, sodium citrate, and the Zeta-potential of the AuNPs equals −37 mV. Nanoparticle assemblies were obtained by adsorbing the AuNPs on glass substrate.13,35 Briefly, the glass substrate was first functionalized with (3-aminopropyl)triethoxysilane (APTES), which renders the surface positively charged due the primary amine group. Citrate-capped AuNPs 7401
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
Article
density of AuNPs on the glass surface (see Supporting Information section 4, Figure S5). This enhancement of refractive index sensitivity has been reported theoretically and experimentally elsewhere.37 In addition, Figure 3b summarizes the average value and standard deviation of the peak shift, absorbance, and curvature for the samples 1−3 with similar density of AuNPs (178 ± 21/μm2). Besides noise (normal variation), other artifacts (abnormal variations) may also be present in the absorption spectra. In typical absorption measurements recorded over long periods of time, artifacts in absorption and peak position occur from time to time because of instrumental instabilities (e.g., caused by temperature variation, mechanical vibrations, and/or unpredictable spikes in the power line). Another source of artifact is of the chemical nature and might arise from nonspecific adsorption of contaminants on the AuNPs. Both types of artifacts may shift the absorption spectrum in wavelength (horizontal direction) or absorption (vertical direction), tilt the overall spectrum, produce sharp spikes/steps, and so forth, all of which may make measurements based on single-point recording inaccurate. Fortunately, they do not affect the curvature as long as the overall peak shape remains the same. To illustrate the performance of the curvature in relation to the peak shift and absorption algorithms, a series of absorption spectra displaying typical instrumental artifacts (tilt, baseline shift, and step) are depicted in Figure 4a, and the results of peak shift, absorbance, and curvature are shown in Figure 4b− d. The absorption spectra overlap during the first 19 min of measurement. However, the spectra start to change after 20 min. Three kinds of artifacts associated with the measurements are recognized. Tilt of spectra occurs first (20−24 min), where the absorbance values at lower wavelengths increase faster than those at longer wavelengths. After the initial tilting, the spectra uniformly shift upward at all wavelengths (25−31 min). In addition, a step-like feature is observed near the peak maximum (23−31 min). The evolutions of the corresponding peak shift, absorption, and curvature are plotted in Figure 4b, c, and d, respectively. From Figure 4b and c, it is obvious that these artifacts move the response signal away from a stable baseline. Interestingly, these artifacts have very limited effects on curvature, Figure 4d. The curvature is very stable, and no significant trend of deviation from the baseline can be observed. The significance of variation is illustrated using 3 times of the standard deviation (3σ) of baseline (dotted red lines). When the spectra are stable (1−19 min), the value of curvature oscillate inside the channel formed by 3σ. When the artifacts start to become visible in the spectra in Figure 4a, three cases of breaking of the channel formed by 3σ are observed at 23, 25, and 28 min, respectively. Most importantly, the curvature baseline values immediately return back after breaking the 3σ lines and therefore cannot be considered as a significant new trend (as seen in the large baseline drifts in absorption and peak shift, Figure 4b,c). Instead, the effect of these artifacts on curvature can be better described as a slight increase in the magnitude of oscillation of the baseline, Figure 4d. This ability to suppress artifiacts can be explained by eq 2. Mathematically, the overall shift of peak position σλ or absorption σA can be described as a zero order change and the tilt as a first order change (kλ + σA). They are both expected to be completely eliminated in the second-order derivative, eq 2. This claim is confirmed using artificially generated zero order shifts (λ, A, and combinations thereof) and first-order tilt on measured
600 and 700 nm. This shoulder is attributed to the small fraction of aggregated AuNPs in the sample. Its presence has a marginal effect in the analysis of peak shift, absorption, and curvature of the peak at 521 nm, because the two LSPR peaks are well separated from each other. The effect of refractive index on the LSPR response was investigated experimentally by measuring the absorption spectra of AuNPs in solvents with different refractive indices varying from 1.333 to 1.474 (prepared by varying the concentration of glycerol in water), Figure 2a. As the refractive index increases, the LSPR peak shifts toward the red, the absorbance values and the sharpness of the LSPR peak increase (sharpness = Amax/fwhm), Figure 2a inset. Second-order differentiation with wavelength was performed on the spectra in Figure 2a, and the corresponding curves are plotted in Figure 2b. Generally, the second-order derivatives are related to the sharpness of the curve. Local maxima are observed in the spectra of the second-order derivatives at the LSPR peak wavelength, the absolute value of which is defined as the curvature, see eq 2 and Figure 2b. More details on the calculations of second-order derivatives are shown in the Supporting Information, section 2. The curvature of AuNPs on APTES-modified glass in solvents of different refractive indices (black circle) is plotted in Figure 2c. The experimental measured curvature (black circles) matches well with the calculated values (red circles). In the small range between 1.333 and 1.474, both the experimental curvature and calculated curvature, eq 3 can be approximated by a linear function (Figure 2c). The slope of measured curvature versus refractive index is 5.19 × 10−6 nm−2 RIU−1, which is very similar to the calculated one 5.49 × 10−6 nm−2 RIU−1. A linear relationship is also observed for 20 and 40 nm sized AuNPs (Supporting Information in section 3, Figure S4). The corresponding peak shift (squares) and absorption (triangles) are plotted in Figure 2d. The peak shift increases linearly with refractive index, with a sensitivity of 34 nmRIU−1, which is in line with observations reported elsewhere.25 The absorption, on the other hand increases rapidly for small refractive index changes but appears to start leveling off when the refractive index increases further, above n = 1.36. In addition to sensitivity, the noise level is an equally important factor in sensing. To compare the performance of peak shift, absorption, and curvature in LSPR sensing, the signal-to-noise ratio (S/N) is calculated. The signal-to-noise ratio S/N is defined as the refractive index sensitivity S divided by the standard deviation of baseline noise N (see Supporting Information section 4, Table S1). The S/N ratio of peak shift, absorbance, and curvature of four distinct samples with different surface density, (Supporting Information section 4, Figure S5) are plotted in Figure 3a, respectively. For all the four samples, the S/N of peak shift is the lowest among the three, which is due to the high noise of peak position. The S/N of absorption is significantly improved as compared to the peak shift. This improvement has been observed by other investigators and used for sensing.36 The S/N ratio of curvature is better than that of both the peak shift and absorption for all the samples. It is worthwhile noticing though that the S/N ratio varies from sample to sample. Sample 4 shows significantly better S/N ratio in peak shift, absorption, and curvature than the other three samples. We propose that the higher S/N ratio in sample 4 is due to interparticle coupling originating from the higher surface 7402
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
Article
Figure 3. (a) S/N ratio of peak shift calculated using centroid algorithm, absorbance calculated with polynomial fitting, and curvature calculated with polynomial fitting of four different AuNP assemblies. (b) S/N ratio and standard deviation of peak shift calculated using centroid algorithm, absorbance calculated with polynomial fitting, and curvature calculated with polynomial fitting of samples 1−3 (with similar surface density of 178 ± 21/μm2).
Figure 4. (a) Summary of 31 absorption spectra of AuNPs on glass immersed in water measured over time. The time interval of measurement is 1 min. Arrow: stable baselines are observed from time 1 to 19 min. Artifacts start to induce spectral shifts at time 20 min. Left rectangle (20−24 min): spectra start to tilt from spectrum 20. Right rectangle (25−31 min): spectra shift upward. Middle rectangle (23−31 min): a step-like feature appears near the peak maximum. (b) Evolution of peak shift (the maximum) over time. (c) Evolution of absorbance at the maximum over time. (d) Evolution of curvature over time. Red dashed lines: 3 times the standard deviation (3σ) of baseline from 1 to 19 min.
AuNPs through the −SH moiety was used. In a typical experiment, the AuNP assemblies on glass were incubated in 10 μM SH-PEG7-COOH, and the adsorption process (molecular binding) was monitored over time. The evolution of the LSPR peak shape and absorption is shown in Figure 5a, and it is obvious that the absorption increases and the LSPR peak become narrower with time upon adsorption of thiol molecules. The peak shift, on the other hand, seems to be very small for this particular model system, < 1 nm. Figure 5b summarizes the response of the peak shift. The time of injection of thiol molecules is indicated by the star.
absorption spectra (see Supporting Information section 6, Figure S7). Sharp steps/spikes change absorbance values very abruptly, which theoretically also could affect the overall shape of the spectrum. However, practically the effect of sharp steps/ spikes could be significantly reduced/almost eliminated, because they only occur at single-point/few points and contribute therefore only marginally to the overall shape and curvature. To further illustrate the advantages of curvature in monitoring local refractive index changes, a thiol molecule (SH-PEG7-COOH, Mw = 458.6) which readily adsorbs to 7403
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
Article
Figure 5. (a) Evolution of absorption spectra of AuNP assemblies on glass upon incubation in 10 μM SH-PEG7-COOH. The SH-PEG7-COOH was introduced after a 10 min stabilization of the baseline (the red line is the sum of nine overlaid spectra obtained prior to injection). The point of injection of thiol is indicated by the red * at 9 min. The spectrum in green is discussed in (c) below. (b) Peak shift of the absorption spectra in (a). (c) Absorbance of the absorption spectra at the maximum in (a). An overall downward shift of the spectrum occurs at time 24 min, see negative spike (arrow) and inset (23, 24, and 25 min). (d) Curvature of the absorption spectra in (a).
Finally, the S/N ratio of this particular monolayer forming molecule obtained for peak shift, absorption, and curvature equals 7.4, 87.6, and 179.2, respectively. The ability to monitor small molecules using nanoparticles is advantageous compared to traditional planar SPR sensors as a result of the localization of the evanescent field. The improvement of S/N ratio of curvature over absorption and peak shift originates from two factors. First, the noise level of curvature is lower than that of peak shift and absorption, because it eliminates some zeroth and first order variations in peak shift and absorption (eq 2). Second, the sensitivity to the refractive index is higher than the absorbance, because it captures both changes in absorbance and width of peak. As shown in Figure 5a, with molecular binding of the thiol molecule, both the increase in absorption and decrease in width contribute the response of curvature. In addition, the sharp feature in absorbance due to instrumental imperfections at time at 24 min (see arrow in Figure 5c) did not show up in the curvature, which is consistent with the observation in Figure 4. Taken together, the low noise and superior S/N ratio of curvature in combination with the ability to suppress experimental artifacts open the avenue to push the limit of detection for LSPR sensors and to make reliable analyses of binding phenomena, including, for example, kinetic determination of rate constants and affinity.
A red shift of the LSPR peak starts to appear above the noise level at 18 min, and it levels off after 24 min. Figure 5c shows the corresponding response of absorption due to the adsorption of the thiol molecule on the AuNP surface. Note that the absorbance values changes instantly upon injection of the thiol moleculemuch earlier than the shift in peak position. Thus, it seems that the peak shift is not able to capture the initial adsorption event in this particular experiment. This marginal response in peak position in favor for a substantial change in absorbance has been observed before by Nath et al.38 and explains the distinct differences seen in the kinetics (Figure 5b,c). The binding of the thiol molecule induces a significant increase in the absorbance over the time, and it seems that the adsorption occurs via a two-step process. An initially fast process is followed by a slower one that levels off after ∼20 min. Moreover, at time 24 min, a sharp negative spike due to the overall downward shift of the absorption spectrum appears in the sensorgram (see arrow and inset in Figure 5c and the green curve in Figure 5a). Figure 5d shows the response of curvature. Again it is obvious that the adsorption occurs via a two-step process. It is also evident that the noise is further reduced in comparison to both the peak shift in absorbance. The binding of the thiol molecules appears to saturate after 20 min adsorption, and an additional injection of 10 μM SHPEG7-COOH did not introduce any further changes. This suggests that with curvature it is able to accurately monitor the adsorption process of this low molecular weight thiol molecule. 7404
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
■
Article
(7) Aili, D.; Gryko, P.; Sepulveda, B.; Dick, J. A. G.; Kirby, N.; Heenan, R.; Baltzer, L.; Liedberg, B.; Ryan, M. P.; Stevens, M. M. Nano Lett. 2011, 11, 1401−1405. (8) Zhao, Y.; Zhou, F.; Zhou, H.; Su, H. Phys. Chem. Chem. Phys. 2013, 15, 1690−1698. (9) Hentschel, M.; Saliba, M.; Vogelgesang, R.; Giessen, H.; Alivisatos, A. P.; Liu, N. Nano Lett. 2010, 10, 2721−2726. (10) Fu, Y. H.; Zhang, J. B.; Yu, Y. F.; Luk’yanchuk, B. ACS Nano 2012, 6, 5130−5137. (11) Lovera, A.; Gallinet, B.; Nordlander, P.; Martin, O. J. F. ACS Nano 2013, 7, 4527−4536. (12) Chen, P.; Selegård, R.; Aili, D.; Liedberg, B. Nanoscale 2013, 5, 8973−8976. (13) Nath, N.; Chilkoti, A. Anal. Chem. 2001, 74, 504−509. (14) Wang, Z.; Zhang, J.; Yin, Z.; Wu, S.; Mandler, D.; Zhang, H. Nanoscale 2012, 4, 2728−2733. (15) Feuz, L.; Jonsson, M. P.; Höök, F. Nano Lett. 2012, 12, 873− 879. (16) Cao, Y. C.; Jin, R.; Mirkin, C. A. Science 2002, 297, 1536−1540. (17) Chen, P.; Yin, Z.; Huang, X.; Wu, S.; Liedberg, B.; Zhang, H. J. Phys. Chem. C 2011, 115, 24080−24084. (18) Lu, G.; Li, H.; Wu, S.; Chen, P.; Zhang, H. Nanoscale 2012, 4, 860−863. (19) Wu, K.; Rodríguez-Córdoba, W. E.; Yang, Y.; Lian, T. Nano Lett. 2013, 13, 5255−5263. (20) Odom, T. W.; Schatz, G. C. Chem. Rev. 2011, 111, 3667−3668. (21) Rodríguez-Lorenzo, L.; de La Rica, R.; Á lvarez-Puebla, R. A.; Liz-Marzán, L. M.; Stevens, M. M. Nat. Mater. 2012, 11, 604−607. (22) Su, X.; Kanjanawarut, R. ACS Nano 2009, 3, 2751−2759. (23) Liu, X. H.; Wang, Y.; Chen, P.; Wang, Y.; Zhang, J.; Aili, D.; Liedberg, B. Anal. Chem. 2014, 86, 2345−2352. (24) Rosman, C.; Prasad, J.; Neiser, A.; Henkel, A.; Edgar, J.; Sönnichsen, C. Nano Lett. 2013, 13, 3243−3247. (25) Chen, H.; Kou, X.; Yang, Z.; Ni, W.; Wang, J. Langmuir 2008, 24, 5233−5237. (26) Haes, A.; Van Duyne, R. Anal. Bioanal. Chem. 2004, 379, 920− 930. (27) Holzner, F.; Kuemin, C.; Paul, P.; Hedrick, J. L.; Wolf, H.; Spencer, N. D.; Duerig, U.; Knoll, A. W. Nano Lett. 2011, 11, 3957− 3962. (28) Sun, Y.; Xia, Y. Anal. Chem. 2002, 74, 5297−5305. (29) Hall, W. P.; Modica, J.; Anker, J.; Lin, Y.; Mrksich, M.; Van Duyne, R. P. Nano Lett. 2011, 11, 1098−1105. (30) Xu, X.; Peng, B.; Li, D.; Zhang, J.; Wong, L. M.; Zhang, Q.; Wang, S.; Xiong, Q. Nano Lett. 2011, 11, 3232−3238. (31) Verellen, N.; Van Dorpe, P.; Huang, C.; Lodewijks, K.; Vandenbosch, G. A. E.; Lagae, L.; Moshchalkov, V. V. Nano Lett. 2011, 11, 391−397. (32) Carpay, F.; Cense, W. Nature Phys. Sci. 1973, 241. (33) Mie, G. Annalen der Physik 1908, 330, 377−445. (34) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light by Small Particles; Wiley: New York, 2008. (35) Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter, D. G.; Natan, M. J. Science 1995, 267, 1629−1632. (36) Nusz, G. J.; Marinakos, S. M.; Curry, A. C.; Dahlin, A.; Hook, F.; Wax, A.; Chilkoti, A. Anal. Chem. 2008, 80, 984−989. (37) Martinsson, E.; Sepulveda, B.; Chen, P.; Elfwing, A.; Liedberg, B.; Aili, D. Plasmonics 2014, DOI: 10.1007/s11468-11013-19659-y. (38) Nath, N.; Chilkoti, A. Anal. Chem. 2004, 76, 5370−5378.
CONCLUSIONS A convenient strategy to monitor LPSR responses utilizing the overall shape of LSPR peak (curvature) is reported. The relationship between the curvature of the LSPR peak and the refractive index of the surrounding medium is investigated theoretically and experimentally. In a narrow regime relevant to biomolecular sensing, the curvature of the LSPR peak can be approximated with a linear function of the refractive index. It is demonstrated that monitoring of the curvature offers improved S/N ratio and reliability over monitoring of absorbance and peak shift. The proposed strategy to follow adsorption phenomena in real time offers a convenient way to push the limit of detection and improve the reliability of LSPR sensing. We further demonstrate that with curvature the LSPR sensor could monitor submonolayer adsorption of small molecules, which for example could be useful in metabolic sensing and drug screening. In addition, the curvature appears not to be sensitive to instrumental artifacts causing tilt, shift, and sharp steps/spikes in the absorption spectra, which could be advantageous when data collection is performed over extended periods of time and when monitoring interactions in complicated matrices as serum, juice etc. We also anticipate that the use of curvature can improve the read out of other transducers technologies based on monitoring the movement of resonance or absorption-like features over time.
■
ASSOCIATED CONTENT
S Supporting Information *
Theoretical calculation of the relationship of curvature and bulk refractive index, method of calculation of curvature value from experimental measured spectra, effect of gold nanoparticle size, comparison of peak shift, absorbance and curvature using four different samples, illustration of peak fitting methods, illustration on the effect of different artifacts of LSPR spectrum on curvature. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: [email protected]. Tel.: (+65) 6316 2957. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work was supported by the School of Materials Science and Engineering and the Provost office, Nanyang Technological University, Singapore. We also thank Tran Thi Nhung for assisting us in the TEM characterization of gold nanoparticles.
■
REFERENCES
(1) Liedberg, B.; Nylander, C.; Lundström, I. Sens. Actuators 1983, 4, 299−304. (2) Liedberg, B.; Nylander, C.; Lundström, I. Biosens. Bioelectron. 1995, 10, i−ix. (3) Löfas, S.; Malmqvist, M.; Ronnberg, I.; Stenberg, E.; Liedberg, B.; Lundström, I. Sens. Actuators, B 1991, 5, 79−84. (4) Wang, Y.; Huang, C.-J.; Jonas, U.; Wei, T.; Dostalek, J.; Knoll, W. Biosens. Bioelectron. 2010, 25, 1663−1668. (5) Yin, X.; Schäferling, M.; Metzger, B.; Giessen, H. Nano Lett. 2013, 13, 6238−6243. (6) Link, S.; El-Sayed, M. A. J. Phys. Chem. B 1999, 103, 4212−4217. 7405
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Curvature of the Localized Surface Plasmon Resonance Peak Peng Chen†,‡ and Bo Liedberg*,†,‡ †
Center for Biomimetic Sensor Science, 50 Nanyang Drive, Research Techno Plaza, Sixth Floor, Singapore 637553, Singapore School of Materials Science and Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
‡
S Supporting Information *
ABSTRACT: Localized surface plasmon resonance (LSPR) occurring in noble metal nanoparticles (e.g., Au) is a widely used phenomenon to report molecular interactions. Traditional LSPR sensors typically monitor shifts in the peak position or extinction in response to local refractive index changes in the close vicinity of the nanoparticle surface. The ability to resolve minute shifts/extinction changes is to a large extent limited by instrumental noise. A new strategy to evaluate LSPR responses utilizing changes in the shape of the extinction spectrum (the curvature) is proposed. The response of curvature to refractive index changes is investigated theoretically using Mie theory and an analytical expression relating the curvature to the refractive index is presented. The experimentally derived curvatures for 13 nm spherical gold nanoparticles (AuNPs) exposed to solvents with different bulk refractive indices confirm the theoretical predictions. Moreover, both the calculated and experimental findings suggest that the curvature is approximately a linear function of refractive index in regimes relevant to bio and chemical sensing. We demonstrate that curvature is superior over peak shift and extinction both in terms of signal-to-noise (S/N) ratio and reliability of LSPR sensors. With a curvature, one could readily monitor submonolayer adsorption of a low molecular weight thiol molecule (Mw = 458.6) onto 13 nm AuNPs. It is also worthwhile mentioning that curvature is virtually insensitive to instrumental instabilities and artifacts occurring during measurement. Instabilities such as baseline tilt and shift, shift in peak position as well as sharp spikes/steps in the extinction spectra do not induce artifacts in the sensorgrams of curvature.
T
(e.g., from the light source and/or detector). As a result, detecting molecules at ultralow concentrations using LSPR is very demanding especially for low molecular weight targets. Many efforts have been undertaken to increase the sensitivity of LSPR sensors by synthesizing novel nanomaterials with higher polarizability, such as nanorods,27 nanostars,25 nanoshells, 28 nanotriangles,29 and so forth. Physicists also have explored the potential of meta-materials by fabricating novel submicron plasmonic patterns using lithography techniques.30,31Although these efforts improve the refractive index sensitivity of LSPR sensors to some extent, they also bring many problems such as high cost, low yield, and tedious processing. In this contribution, we report a new strategy to monitor changes in the shape of LSPR peak of gold nanoparticles (AuNPs). Contrary to peak shifts or extinction changes that monitors the changes of the peak maximum at a certain point, the curvature of the LSPR peak takes into account the overall peak shape. Our theoretical and experimental findings demonstrate that the curvature (2nd order derivative at peak maximum) is approximately linearly related to refractive index in regimes relevant to sensing.
he collective oscillation of free electrons in the conduction bands of gold and other noble metals is known as surface plasmon resonance (SPR). Since the first report of SPR sensing in 1983,1 it has been widely explored and developed into a wellestablished commercial tool.2−4 Recently, noble metal nanoparticles have attracted tremendous interest in both fundamental5−11 and applied research including sensing,12−15 surface enhanced Raman scattering (SERS),16−18 and photovoltaics,19 among others. Contrary to propagating surface plasmon waves, the electron plasma in nanoparticles is localized to the particle, a phenomenon referred to as LSPR.20 The LSPR resonance frequency/wavelength of the nanosized noble metals depends on their size, shape, composition, interparticle distance, and the medium in which they are embedded/dissolved .6 Biomolecular sensing by LSPR sensors mostly relies on the peak shift or extinction changes induced by refractive index changes or interparticle aggregation phenomena occurring in response to molecular recognition and/or attachment/bridging.21−24 The refractive index sensitivity of LSPR sensors depends on the polarizability of the nanosized plasmonic structure/particles, which commonly varies from a few tens to a few hundred nmRIU−1.25 Despite the many advantages over propagating (planar) SPR sensors in terms of spatial resolution and cost, the refractive index sensitivity of LSPR sensors is still about 4 orders of magnitude smaller than of planar SPR.26 In addition, the ability to resolve minute changes in the LSPR peak position/extinction is limited by instrumental/chemical noise © 2014 American Chemical Society
Received: March 9, 2014 Accepted: July 1, 2014 Published: July 1, 2014 7399
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
■
Article
polynomial fitting of the absorption spectra using 25 data points, symmetric around the absorption maximum. The curvature is obtained similarly by polynomial fitting of the second derivative spectra using 25 data points, symmetric around the curvature maximum (more details can be found in Supporting Information, section 5). The peak shift obtained by the centroid method is calculated in a 60 nm wavelength window around the peak maximum. (A software for calculation of a curvature from experimental data is available upon request from the authors.)
EXPERIMENTAL SECTION Chemicals and Materials. Sodium citrate tribasic dehydrate (99%), gold(III) chloride trihydrate (99.9%), (3-aminopropyl)triethoxysilane (APTES) (97%), and glycerol (99%) were purchased from Sigma-Aldrich. Microscope slides from Sail Brand were used as glass substrates. Thiol PEG7 acid (O(2-carboxyethyl)-O′-(2-mercaptoethyl)-heptaethylene glycol) was purchased from Polypure AS, Norway. Spherical AuNPs with the size of 20 and 40 nm were purchased from BBI solutions. Nanoparticle Synthesis and Characterization. Gold nanoparticles were synthesized by reduction of HAuCl4 using trisodium citrate.32 In a typical synthesis, 90 mL of Milli-Q water and 10 mL of 10 mM HAuCl4 were mixed and heated to above 100 °C under reflux for 2 h. An aqueous solution of 10 mL of 77.6 mM trisodium citrate was added quickly to the reflux solution by piston and needle. The color of the solution changes from pale yellow to red in ca. 10 s. Heating was maintained for another 15 min before cooling to room temperature in air. The AuNPs were stored at 4 °C for future use. A Joel 2100 transmission electron microscope (TEM) was used to characterize the size of the nanoparticles and their distribution. Absorption spectra of AuNPs in solution were measured with a PerkinElmer Lambda 35 absorption spectrometer. Zeta-potential of the citrate capped AuNPs was measured using the spectrometer from Malvern Instruments. The size of AuNPs was derived from TEM images using the software ImageJ, and the distribution and standard deviation were calculated from over 200 randomly chosen individual particles. Surface Modification and Characterization. Pieces of glass substrate with 2 × 0.5 cm2 were cut from microscope slides and incubated in a mixture of water, 30% H2O2, and 25% ammonia solution (5:1:1 in volume) at 80 °C for 10 min for remove contaminants. Then they were rinsed copiously with Milli-Q water for a few times. After the cleaning, the glass substrates were dried by nitrogen gas. The cleaned glass substrates were incubated with 1% (volume) of APTES in 99.9% ethanol for 30 min for surface functionalization. After functionalization, the substrates were rinsed with Milli-Q water and dried in nitrogen gas. A drop of as-synthesized AuNPs (200 μL) was casted onto the APTES modified glass surface and incubated for 30 min to 4 h to generate samples possessing different surface densities. After incubation, the glass substrates with AuNPs were rinsed with Milli-Q water and dried with nitrogen gas. Secondary electron images on the AuNP coated glass were recorded using a Joel 7600F field emission scanning electron microscope (FESEM), and a thin layer of platinum coating was applied to avoid a charging effect. Absorption spectra were measured by immersing the substrate into a cuvette containing 1 mL of solution using a PerkinElmer spectrometer operating at a resolution of 1 nm. Solvents with different refractive indices were prepared by mixtures of water and glycerol. Volume fractions of glycerol in the five solvents are 0%, 25%, 50%, 75%, and 100%, respectively. Thiol PEG7 solution was diluted in cuvette from stock a solution of 1 mM. Calculations and Spectral Evaluation. Analytical derivation of derivatives and numerical calculations of absorption cross section are performed using the software Mathematica 9. Second-order derivatives are calculated from two sequential first order derivatives, and 128 data points are used in the calculation of curvature. The absorbance value is obtained by
■
RESULTS AND DISCUSSION For a spherical particle, the Maxwell’s equation is analytically solvable, which was first reported by Gustav Mie in 1908,33 and further developed by Bohren and Huffman.34 The extinction, absorption and scattering cross sections of an arbitrary spherical particle can be calculated from the Mie theory. The wavelength and refractive index dependent absorption A(λ, n) of a 13 nm AuNPs can be derived from Mie theory: A(λ , n) =
(a 2 + b 2 λ ) 18πVn2Nl 2 λ (a 2 + b2λ) (b1λ + a1 + 2n2) + 1 (1)
where, a1, a2, b1, b2, and π are constants, V is the volume of particle, N is the density of electrons, λ is the wavelength of light, l is the length of the optical path, and n is the refractive index of surrounding medium. The curvature (K) is related to the sharpness of the peak, and mathematically it can be defined as the absolute value of the second-order partial derivative of absorbance with wavelength at the peak position. K=
∂ 2A (A = A max ) ∂λ 2
(2)
The curvature as a function of refractive index n can be written as ⎧ 3 2⎡ b2 ⎤3 2 b n a (2 n a ) − + ⎪ ⎢ ⎥⎦ 1 2 1 ⎪ ⎣ b1 K (n) = 36πVNl ⎨− 2n2 + a1 ⎪ ⎪ ⎩ +
b12b2n2 (2n2 + a1)2
+
⎤⎫ + a1)⎥⎦ ⎪ ⎪ ⎬ 2 3 (2n + a1) ⎪ ⎪ ⎭
⎡ b13n2⎣⎢a 2 +
b2 (2n2 b1
(3)
The details of the calculation of curvature as a function of bulk refractive index are outlined in Supporting Information, section 1. Gold nanoparticles with a size of ∼13 nm were synthesized by reduction of chloroauric acid (HAuCl4) with sodium citrate.32 The size of the AuNPs can be tuned by the ratio of sodium citrate to chloroauric acid. Figure S2a shows a transmission electron microscope (TEM) image of the assynthesized AuNPs, and their size distribution is summarized in the histogram plot in Figure S2b. Most of the AuNPs have a size between 11 and 14 nm, with an average size of 12.9 nm and standard deviation of 1.1 nm, respectively. The absorption spectrum of the AuNPs in solution is shown in Figure S2c. A prominent peak is observed at 521 nm, which is due to LSPR. 7400
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
Article
Figure 1. (a) Scanning electron microscope (SEM) image of AuNPs assembled on the glass substrate. Scale bar 100 nm. (b) Absorption spectrum of AuNP assembly on glass in air.
Figure 2. (a) Absorption spectra of AuNPs on glass substrate immersed in solvents with different refractive indices: 1.333 (black), 1.364 (red), 1.399 (blue), 1.432 (green), and 1.474 (purple), respectively. Spectra are offset for better visualization. Inset: zoom-in absorption spectra of AuNPs in solvent with a refractive index of 1.333 (black) and 1.474 (purple). (b) Second-order derivatives of absorption spectra in (a). (c) Experimental (black circles/line) and calculated (red circles/line) curvature of AuNPs immersed in media with different refractive indices. Experimental (black) and calculated (red) data are fitted linear function to yield a slope of 5.19 × 10−6 nm−2 RIU−1 (R2 = 0.9804) and 5.49 × 10−6 nm−2 RIU−1 (R2 = 0.9876), respectively. (d) Experimentally measured peak shift (squares) and absorbance (triangles) at different refractive indices. The peak shift is fitted linearly, with a slope of 34 nmRIU−1 (R2 = 0.9970).
adsorb via electrostatic attraction to the positively charged APTES surface. Figure 1a shows the secondary electron image of a typical AuNP assembly on APTES-modified glass. The AuNPs are evenly distributed on the glass substrate, with a few exceptions where aggregation is observed. The absorption spectrum of the AuNP assemblies measured in air is shown in Figure 1b. The absorption spectrum of the AuNP assemblies resembles that of AuNPs in solution Figure S2c, except for a slight broadening and the appearance of a shoulder between
The peak is sharp, indicating narrow size and shape distribution of the AuNPs, which also agrees with the observations of TEM images, Figure S2a,b. The AuNPs are negatively charged due to the capping agent, sodium citrate, and the Zeta-potential of the AuNPs equals −37 mV. Nanoparticle assemblies were obtained by adsorbing the AuNPs on glass substrate.13,35 Briefly, the glass substrate was first functionalized with (3-aminopropyl)triethoxysilane (APTES), which renders the surface positively charged due the primary amine group. Citrate-capped AuNPs 7401
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
Article
density of AuNPs on the glass surface (see Supporting Information section 4, Figure S5). This enhancement of refractive index sensitivity has been reported theoretically and experimentally elsewhere.37 In addition, Figure 3b summarizes the average value and standard deviation of the peak shift, absorbance, and curvature for the samples 1−3 with similar density of AuNPs (178 ± 21/μm2). Besides noise (normal variation), other artifacts (abnormal variations) may also be present in the absorption spectra. In typical absorption measurements recorded over long periods of time, artifacts in absorption and peak position occur from time to time because of instrumental instabilities (e.g., caused by temperature variation, mechanical vibrations, and/or unpredictable spikes in the power line). Another source of artifact is of the chemical nature and might arise from nonspecific adsorption of contaminants on the AuNPs. Both types of artifacts may shift the absorption spectrum in wavelength (horizontal direction) or absorption (vertical direction), tilt the overall spectrum, produce sharp spikes/steps, and so forth, all of which may make measurements based on single-point recording inaccurate. Fortunately, they do not affect the curvature as long as the overall peak shape remains the same. To illustrate the performance of the curvature in relation to the peak shift and absorption algorithms, a series of absorption spectra displaying typical instrumental artifacts (tilt, baseline shift, and step) are depicted in Figure 4a, and the results of peak shift, absorbance, and curvature are shown in Figure 4b− d. The absorption spectra overlap during the first 19 min of measurement. However, the spectra start to change after 20 min. Three kinds of artifacts associated with the measurements are recognized. Tilt of spectra occurs first (20−24 min), where the absorbance values at lower wavelengths increase faster than those at longer wavelengths. After the initial tilting, the spectra uniformly shift upward at all wavelengths (25−31 min). In addition, a step-like feature is observed near the peak maximum (23−31 min). The evolutions of the corresponding peak shift, absorption, and curvature are plotted in Figure 4b, c, and d, respectively. From Figure 4b and c, it is obvious that these artifacts move the response signal away from a stable baseline. Interestingly, these artifacts have very limited effects on curvature, Figure 4d. The curvature is very stable, and no significant trend of deviation from the baseline can be observed. The significance of variation is illustrated using 3 times of the standard deviation (3σ) of baseline (dotted red lines). When the spectra are stable (1−19 min), the value of curvature oscillate inside the channel formed by 3σ. When the artifacts start to become visible in the spectra in Figure 4a, three cases of breaking of the channel formed by 3σ are observed at 23, 25, and 28 min, respectively. Most importantly, the curvature baseline values immediately return back after breaking the 3σ lines and therefore cannot be considered as a significant new trend (as seen in the large baseline drifts in absorption and peak shift, Figure 4b,c). Instead, the effect of these artifacts on curvature can be better described as a slight increase in the magnitude of oscillation of the baseline, Figure 4d. This ability to suppress artifiacts can be explained by eq 2. Mathematically, the overall shift of peak position σλ or absorption σA can be described as a zero order change and the tilt as a first order change (kλ + σA). They are both expected to be completely eliminated in the second-order derivative, eq 2. This claim is confirmed using artificially generated zero order shifts (λ, A, and combinations thereof) and first-order tilt on measured
600 and 700 nm. This shoulder is attributed to the small fraction of aggregated AuNPs in the sample. Its presence has a marginal effect in the analysis of peak shift, absorption, and curvature of the peak at 521 nm, because the two LSPR peaks are well separated from each other. The effect of refractive index on the LSPR response was investigated experimentally by measuring the absorption spectra of AuNPs in solvents with different refractive indices varying from 1.333 to 1.474 (prepared by varying the concentration of glycerol in water), Figure 2a. As the refractive index increases, the LSPR peak shifts toward the red, the absorbance values and the sharpness of the LSPR peak increase (sharpness = Amax/fwhm), Figure 2a inset. Second-order differentiation with wavelength was performed on the spectra in Figure 2a, and the corresponding curves are plotted in Figure 2b. Generally, the second-order derivatives are related to the sharpness of the curve. Local maxima are observed in the spectra of the second-order derivatives at the LSPR peak wavelength, the absolute value of which is defined as the curvature, see eq 2 and Figure 2b. More details on the calculations of second-order derivatives are shown in the Supporting Information, section 2. The curvature of AuNPs on APTES-modified glass in solvents of different refractive indices (black circle) is plotted in Figure 2c. The experimental measured curvature (black circles) matches well with the calculated values (red circles). In the small range between 1.333 and 1.474, both the experimental curvature and calculated curvature, eq 3 can be approximated by a linear function (Figure 2c). The slope of measured curvature versus refractive index is 5.19 × 10−6 nm−2 RIU−1, which is very similar to the calculated one 5.49 × 10−6 nm−2 RIU−1. A linear relationship is also observed for 20 and 40 nm sized AuNPs (Supporting Information in section 3, Figure S4). The corresponding peak shift (squares) and absorption (triangles) are plotted in Figure 2d. The peak shift increases linearly with refractive index, with a sensitivity of 34 nmRIU−1, which is in line with observations reported elsewhere.25 The absorption, on the other hand increases rapidly for small refractive index changes but appears to start leveling off when the refractive index increases further, above n = 1.36. In addition to sensitivity, the noise level is an equally important factor in sensing. To compare the performance of peak shift, absorption, and curvature in LSPR sensing, the signal-to-noise ratio (S/N) is calculated. The signal-to-noise ratio S/N is defined as the refractive index sensitivity S divided by the standard deviation of baseline noise N (see Supporting Information section 4, Table S1). The S/N ratio of peak shift, absorbance, and curvature of four distinct samples with different surface density, (Supporting Information section 4, Figure S5) are plotted in Figure 3a, respectively. For all the four samples, the S/N of peak shift is the lowest among the three, which is due to the high noise of peak position. The S/N of absorption is significantly improved as compared to the peak shift. This improvement has been observed by other investigators and used for sensing.36 The S/N ratio of curvature is better than that of both the peak shift and absorption for all the samples. It is worthwhile noticing though that the S/N ratio varies from sample to sample. Sample 4 shows significantly better S/N ratio in peak shift, absorption, and curvature than the other three samples. We propose that the higher S/N ratio in sample 4 is due to interparticle coupling originating from the higher surface 7402
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
Article
Figure 3. (a) S/N ratio of peak shift calculated using centroid algorithm, absorbance calculated with polynomial fitting, and curvature calculated with polynomial fitting of four different AuNP assemblies. (b) S/N ratio and standard deviation of peak shift calculated using centroid algorithm, absorbance calculated with polynomial fitting, and curvature calculated with polynomial fitting of samples 1−3 (with similar surface density of 178 ± 21/μm2).
Figure 4. (a) Summary of 31 absorption spectra of AuNPs on glass immersed in water measured over time. The time interval of measurement is 1 min. Arrow: stable baselines are observed from time 1 to 19 min. Artifacts start to induce spectral shifts at time 20 min. Left rectangle (20−24 min): spectra start to tilt from spectrum 20. Right rectangle (25−31 min): spectra shift upward. Middle rectangle (23−31 min): a step-like feature appears near the peak maximum. (b) Evolution of peak shift (the maximum) over time. (c) Evolution of absorbance at the maximum over time. (d) Evolution of curvature over time. Red dashed lines: 3 times the standard deviation (3σ) of baseline from 1 to 19 min.
AuNPs through the −SH moiety was used. In a typical experiment, the AuNP assemblies on glass were incubated in 10 μM SH-PEG7-COOH, and the adsorption process (molecular binding) was monitored over time. The evolution of the LSPR peak shape and absorption is shown in Figure 5a, and it is obvious that the absorption increases and the LSPR peak become narrower with time upon adsorption of thiol molecules. The peak shift, on the other hand, seems to be very small for this particular model system, < 1 nm. Figure 5b summarizes the response of the peak shift. The time of injection of thiol molecules is indicated by the star.
absorption spectra (see Supporting Information section 6, Figure S7). Sharp steps/spikes change absorbance values very abruptly, which theoretically also could affect the overall shape of the spectrum. However, practically the effect of sharp steps/ spikes could be significantly reduced/almost eliminated, because they only occur at single-point/few points and contribute therefore only marginally to the overall shape and curvature. To further illustrate the advantages of curvature in monitoring local refractive index changes, a thiol molecule (SH-PEG7-COOH, Mw = 458.6) which readily adsorbs to 7403
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
Article
Figure 5. (a) Evolution of absorption spectra of AuNP assemblies on glass upon incubation in 10 μM SH-PEG7-COOH. The SH-PEG7-COOH was introduced after a 10 min stabilization of the baseline (the red line is the sum of nine overlaid spectra obtained prior to injection). The point of injection of thiol is indicated by the red * at 9 min. The spectrum in green is discussed in (c) below. (b) Peak shift of the absorption spectra in (a). (c) Absorbance of the absorption spectra at the maximum in (a). An overall downward shift of the spectrum occurs at time 24 min, see negative spike (arrow) and inset (23, 24, and 25 min). (d) Curvature of the absorption spectra in (a).
Finally, the S/N ratio of this particular monolayer forming molecule obtained for peak shift, absorption, and curvature equals 7.4, 87.6, and 179.2, respectively. The ability to monitor small molecules using nanoparticles is advantageous compared to traditional planar SPR sensors as a result of the localization of the evanescent field. The improvement of S/N ratio of curvature over absorption and peak shift originates from two factors. First, the noise level of curvature is lower than that of peak shift and absorption, because it eliminates some zeroth and first order variations in peak shift and absorption (eq 2). Second, the sensitivity to the refractive index is higher than the absorbance, because it captures both changes in absorbance and width of peak. As shown in Figure 5a, with molecular binding of the thiol molecule, both the increase in absorption and decrease in width contribute the response of curvature. In addition, the sharp feature in absorbance due to instrumental imperfections at time at 24 min (see arrow in Figure 5c) did not show up in the curvature, which is consistent with the observation in Figure 4. Taken together, the low noise and superior S/N ratio of curvature in combination with the ability to suppress experimental artifacts open the avenue to push the limit of detection for LSPR sensors and to make reliable analyses of binding phenomena, including, for example, kinetic determination of rate constants and affinity.
A red shift of the LSPR peak starts to appear above the noise level at 18 min, and it levels off after 24 min. Figure 5c shows the corresponding response of absorption due to the adsorption of the thiol molecule on the AuNP surface. Note that the absorbance values changes instantly upon injection of the thiol moleculemuch earlier than the shift in peak position. Thus, it seems that the peak shift is not able to capture the initial adsorption event in this particular experiment. This marginal response in peak position in favor for a substantial change in absorbance has been observed before by Nath et al.38 and explains the distinct differences seen in the kinetics (Figure 5b,c). The binding of the thiol molecule induces a significant increase in the absorbance over the time, and it seems that the adsorption occurs via a two-step process. An initially fast process is followed by a slower one that levels off after ∼20 min. Moreover, at time 24 min, a sharp negative spike due to the overall downward shift of the absorption spectrum appears in the sensorgram (see arrow and inset in Figure 5c and the green curve in Figure 5a). Figure 5d shows the response of curvature. Again it is obvious that the adsorption occurs via a two-step process. It is also evident that the noise is further reduced in comparison to both the peak shift in absorbance. The binding of the thiol molecules appears to saturate after 20 min adsorption, and an additional injection of 10 μM SHPEG7-COOH did not introduce any further changes. This suggests that with curvature it is able to accurately monitor the adsorption process of this low molecular weight thiol molecule. 7404
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
Analytical Chemistry
■
Article
(7) Aili, D.; Gryko, P.; Sepulveda, B.; Dick, J. A. G.; Kirby, N.; Heenan, R.; Baltzer, L.; Liedberg, B.; Ryan, M. P.; Stevens, M. M. Nano Lett. 2011, 11, 1401−1405. (8) Zhao, Y.; Zhou, F.; Zhou, H.; Su, H. Phys. Chem. Chem. Phys. 2013, 15, 1690−1698. (9) Hentschel, M.; Saliba, M.; Vogelgesang, R.; Giessen, H.; Alivisatos, A. P.; Liu, N. Nano Lett. 2010, 10, 2721−2726. (10) Fu, Y. H.; Zhang, J. B.; Yu, Y. F.; Luk’yanchuk, B. ACS Nano 2012, 6, 5130−5137. (11) Lovera, A.; Gallinet, B.; Nordlander, P.; Martin, O. J. F. ACS Nano 2013, 7, 4527−4536. (12) Chen, P.; Selegård, R.; Aili, D.; Liedberg, B. Nanoscale 2013, 5, 8973−8976. (13) Nath, N.; Chilkoti, A. Anal. Chem. 2001, 74, 504−509. (14) Wang, Z.; Zhang, J.; Yin, Z.; Wu, S.; Mandler, D.; Zhang, H. Nanoscale 2012, 4, 2728−2733. (15) Feuz, L.; Jonsson, M. P.; Höök, F. Nano Lett. 2012, 12, 873− 879. (16) Cao, Y. C.; Jin, R.; Mirkin, C. A. Science 2002, 297, 1536−1540. (17) Chen, P.; Yin, Z.; Huang, X.; Wu, S.; Liedberg, B.; Zhang, H. J. Phys. Chem. C 2011, 115, 24080−24084. (18) Lu, G.; Li, H.; Wu, S.; Chen, P.; Zhang, H. Nanoscale 2012, 4, 860−863. (19) Wu, K.; Rodríguez-Córdoba, W. E.; Yang, Y.; Lian, T. Nano Lett. 2013, 13, 5255−5263. (20) Odom, T. W.; Schatz, G. C. Chem. Rev. 2011, 111, 3667−3668. (21) Rodríguez-Lorenzo, L.; de La Rica, R.; Á lvarez-Puebla, R. A.; Liz-Marzán, L. M.; Stevens, M. M. Nat. Mater. 2012, 11, 604−607. (22) Su, X.; Kanjanawarut, R. ACS Nano 2009, 3, 2751−2759. (23) Liu, X. H.; Wang, Y.; Chen, P.; Wang, Y.; Zhang, J.; Aili, D.; Liedberg, B. Anal. Chem. 2014, 86, 2345−2352. (24) Rosman, C.; Prasad, J.; Neiser, A.; Henkel, A.; Edgar, J.; Sönnichsen, C. Nano Lett. 2013, 13, 3243−3247. (25) Chen, H.; Kou, X.; Yang, Z.; Ni, W.; Wang, J. Langmuir 2008, 24, 5233−5237. (26) Haes, A.; Van Duyne, R. Anal. Bioanal. Chem. 2004, 379, 920− 930. (27) Holzner, F.; Kuemin, C.; Paul, P.; Hedrick, J. L.; Wolf, H.; Spencer, N. D.; Duerig, U.; Knoll, A. W. Nano Lett. 2011, 11, 3957− 3962. (28) Sun, Y.; Xia, Y. Anal. Chem. 2002, 74, 5297−5305. (29) Hall, W. P.; Modica, J.; Anker, J.; Lin, Y.; Mrksich, M.; Van Duyne, R. P. Nano Lett. 2011, 11, 1098−1105. (30) Xu, X.; Peng, B.; Li, D.; Zhang, J.; Wong, L. M.; Zhang, Q.; Wang, S.; Xiong, Q. Nano Lett. 2011, 11, 3232−3238. (31) Verellen, N.; Van Dorpe, P.; Huang, C.; Lodewijks, K.; Vandenbosch, G. A. E.; Lagae, L.; Moshchalkov, V. V. Nano Lett. 2011, 11, 391−397. (32) Carpay, F.; Cense, W. Nature Phys. Sci. 1973, 241. (33) Mie, G. Annalen der Physik 1908, 330, 377−445. (34) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light by Small Particles; Wiley: New York, 2008. (35) Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter, D. G.; Natan, M. J. Science 1995, 267, 1629−1632. (36) Nusz, G. J.; Marinakos, S. M.; Curry, A. C.; Dahlin, A.; Hook, F.; Wax, A.; Chilkoti, A. Anal. Chem. 2008, 80, 984−989. (37) Martinsson, E.; Sepulveda, B.; Chen, P.; Elfwing, A.; Liedberg, B.; Aili, D. Plasmonics 2014, DOI: 10.1007/s11468-11013-19659-y. (38) Nath, N.; Chilkoti, A. Anal. Chem. 2004, 76, 5370−5378.
CONCLUSIONS A convenient strategy to monitor LPSR responses utilizing the overall shape of LSPR peak (curvature) is reported. The relationship between the curvature of the LSPR peak and the refractive index of the surrounding medium is investigated theoretically and experimentally. In a narrow regime relevant to biomolecular sensing, the curvature of the LSPR peak can be approximated with a linear function of the refractive index. It is demonstrated that monitoring of the curvature offers improved S/N ratio and reliability over monitoring of absorbance and peak shift. The proposed strategy to follow adsorption phenomena in real time offers a convenient way to push the limit of detection and improve the reliability of LSPR sensing. We further demonstrate that with curvature the LSPR sensor could monitor submonolayer adsorption of small molecules, which for example could be useful in metabolic sensing and drug screening. In addition, the curvature appears not to be sensitive to instrumental artifacts causing tilt, shift, and sharp steps/spikes in the absorption spectra, which could be advantageous when data collection is performed over extended periods of time and when monitoring interactions in complicated matrices as serum, juice etc. We also anticipate that the use of curvature can improve the read out of other transducers technologies based on monitoring the movement of resonance or absorption-like features over time.
■
ASSOCIATED CONTENT
S Supporting Information *
Theoretical calculation of the relationship of curvature and bulk refractive index, method of calculation of curvature value from experimental measured spectra, effect of gold nanoparticle size, comparison of peak shift, absorbance and curvature using four different samples, illustration of peak fitting methods, illustration on the effect of different artifacts of LSPR spectrum on curvature. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: [email protected]. Tel.: (+65) 6316 2957. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work was supported by the School of Materials Science and Engineering and the Provost office, Nanyang Technological University, Singapore. We also thank Tran Thi Nhung for assisting us in the TEM characterization of gold nanoparticles.
■
REFERENCES
(1) Liedberg, B.; Nylander, C.; Lundström, I. Sens. Actuators 1983, 4, 299−304. (2) Liedberg, B.; Nylander, C.; Lundström, I. Biosens. Bioelectron. 1995, 10, i−ix. (3) Löfas, S.; Malmqvist, M.; Ronnberg, I.; Stenberg, E.; Liedberg, B.; Lundström, I. Sens. Actuators, B 1991, 5, 79−84. (4) Wang, Y.; Huang, C.-J.; Jonas, U.; Wei, T.; Dostalek, J.; Knoll, W. Biosens. Bioelectron. 2010, 25, 1663−1668. (5) Yin, X.; Schäferling, M.; Metzger, B.; Giessen, H. Nano Lett. 2013, 13, 6238−6243. (6) Link, S.; El-Sayed, M. A. J. Phys. Chem. B 1999, 103, 4212−4217. 7405
dx.doi.org/10.1021/ac500883x | Anal. Chem. 2014, 86, 7399−7405
