Nanoplasmonic sensing is an emerging technique that utilizes the interaction of light with surface plasmons — collective oscillations of free electrons in the conduction band —in metallic nanostructures to transduce chemical bindings to signal for remote optical readout. With receptive biological macromolecules immobilized on nanostructured metal surfaces, whose length scale is close to the wavelength of light, the binding of target biomolecules leads to an increase of the local refractive index, which is then transduced into the change of spectral features observed at the far-field. The greatest attraction of nanoplasmonic sensing is it does not require extrinsic fluorescent or radioactive labeling (i.e. “label-free”), which would potentially change the binding properties of biomolecules and cause other significant problems like background binding and autofluorescence.1
The small sensing volume resulting from short electromagnetic field decay length leads to high sensitivity for detection biomolecular binding events in local environments.
Significant improvement in nanofabrication and nanomaterials synthesis in the past decade has allowed a variety of nanostructures to be utilized as nanoplasmonic sensors. A wide range of nanostructure shapes such as triangles, cubes, rods, and rings 2, 3
have been reported to exhibit high sensitivity based on localized surface plasmon resonance (LSPR) or coupled plasmonic resonance. The discovery of extraordinary light transmission through metallic subwavelength-hole arrays,4
with an efficiency that is orders of magnitude greater than that predicted by classic optical theory, has created another exciting opportunity for nanoplasmonic sensing. Surface plasmon polaritons (SPPs) are generally proposed to explain this phenomenon observed on periodically arrayed nanoholes, although some works have claimed that other mechanisms like diffraction and LSPR also play important roles.5-7
Several studies exploiting periodically arrayed nanoholes fabricated by focused ion beam (FIB),8-12
and soft lithography14, 15
have demonstrated their excellent sensitivity when used as label-free chemical sensing devices. Label-free plasmonic sensing is traditionally performed on prisms (Kretschmann geometry) or grating couplers, on which biomolecules are immobilized on a continuous metal film and the angular distribution or the reflected intensity is measured.16
Those real-time techniques have been successful in performing binding affinity analysis of proteins or small molecules, and commercial instruments are available.1
Compared with conventional coupling techniques, the nanohole geometry has a much smaller probing area and can be fabricated in high spatial density. The optical transmission configuration enables multiplexed detection and offers better opportunities to develop low-cost devices with integrated microfluidics for rapid bioanalytical measurements.
Nanohole plasmonic sensing devices that have been reported to have excellent detection limits are mostly based on wavelength interrogation with relatively low sampling rates on infinite or quasi-infinite nanohole arrays. High spatial density multiplexed real-time measurements are difficult to achieve in those designs. For screening binding affinity of thousands of protein in parallel, small-dimension array nanoholes are of more interest because they offer high packing density of sensing elements, and can be integrated with compact microfluidic devices that require less analyte volume. Small-dimension nanohole arrays have been reported to exhibit strikingly different transmission features from infinite arrays15
because the contribution from the edge becomes significant.17
Their refractive-index sensing properties have not been extensively studied.
Another key point in developing a nanohole sensing system is the interface between the nanostructure surface and the chemical or biological systems. For lipid-membrane-mediated reactions, especially those associated with membrane proteins or membrane fusion processes, a smooth surface is greatly desired to help building a high-quality single lipid-bilayer that incorporates protein receptors and improving experimental reproducibility.18, 19
Visualization of chemisorbed proteins and molecules can also be achieved by using scanning probe techniques (AFM / STM) on atomically-smooth surfaces.19, 20
On the other hand, from the view of nanofabrication, smoother surface will be beneficial on improving the quality of fabricated features, especially when they are in nanometer sizes.
Here, we present a nanoplasmonic sensing platform which exploits small-dimension ordered arrays of nanoholes (≤ 4.5 μm × 4.5 μm) for multiplexed (30 arrays in a 48 μm × 58 μm area) and real-time (time resolution ≤ 1s) analysis on biorecognition events. The measuring platform was built on a conventional inverted microscope with regular bright-field optics. Intensity interrogation was performed instead of commonly used wavelength interrogation to achieve multiplex and real-time sensing capability. Nanohole arrays were perforated on super smooth gold surfaces fabricated by replica techniques. We chose fluorinated ethylene propylene copolymer (FEP) as the replica substrate because it is chemically inert, thermoplastic, transparent in the visible region, and has a refractive index (1.341 at λ = 590 nm) close to biological solutions. It has been reported that when the refractive indices on both sides of the gold film match, the intensity of extraordinary transmission through nanoholes would be strongly enhanced because of the overlap of transmission maxima from two interfaces and the higher lifetime of a coupled state from strong coupling of SP modes on each side of the nanohole arrays.21-23
Meanwhile, although the change in local dielectric environments normally causes more shift of the spectral features in the near-infrared and infrared range than in the visible range 10, 14
, we used visible light as the sensing wavelength to prevent potential near-infrared adsorption of water and many biological species. Moreover, shorter excitation wavelengths generate surface plasmon wave with shorter propagation lengths in the X-Y plane, leading to higher possible packing density and hence higher throughput on multiplexed biosensing.
The substrate for nanoholes was prepared by the template-stripped-gold (TSG) method18, 19
to obtain a super smooth surface for biomolecule immobilization. FEP was used to replace epoxy resins used in the regular TSG method (see supporting information
). The resulting TSG substrate has a 220 nm-thick gold film on FEP with the smooth surface facing up (). The surface roughness is ~ 1.2 nm (peak to peak, RMS = 2.7 Å, see supporting info
). 220 nm-thickness was used to effectively block the direct transmitted light. For comparison, a regular gold substrate was prepared by thermally depositing 8 nm chromium and 220 nm gold on a cleaned glass slide. FEI DB235 focused ion beam system with 30 kV acceleration voltage and 10 pA ion current was applied to mill nanoholes arrays on the gold film. Ion image magnification was calibrated before each milling. After immobilizing biomolecules, a PDMS block with a microfluidic channel (volume ~ 3 μL) was mounted on the nanohole film. A syringe pump was used to deliver solutions through this channel at a 30 μL/min flow rate.
Figure 1 (a) A fluoropolymer-based nanohole array device with the super smooth gold surface facing up, fabricated by the template-stripped method (b) Schematic of the experimental setup (c) CCD image of 30 sets of nanohole arrays at λ = 540 nm (d) SEM (more ...)
Transmission spectra and real-time sensing responses were collected using an inverted microscope with bright-field accessories (Nikon eclipse TE300, ). For multiplexed spectral acquisition, a multispectral illumination source and an intensity imaging CCD were used instead of a polychromatic spectrometer.24
Non-polarized white light from a 75W Xenon lamp was passed through a double monochromator (Jobin-Yvon SPEX 1680B) and a condenser lens. Divergence of light (5° ~ 39°) can be adjusted using the diaphragm on the condenser lens. The light coming out from the nanohole arrays was collected with a 20x objective (NA = 0.5) and then focused on an uncooled CCD (QImage, Intensified Retiga). The spectra were corrected by the lamp transmission profile through the glass substrate. The bandpass of the monochromator was set to 1 nm for acquiring transmission spectra. An image processing program was created to extract intensity value of individual nanohole arrays from CCD images (see supporting info
A serial of nanohole patterns with different nanohole diameters (d), periodicity (P, hole-to-hole distance), and numbers was fabricated on TSG substrates. is the CCD image of one of the nanohole array patterns we studied. This pattern contains thirty 9×9 nanohole arrays spaced by 11 μm. Since calculated SPP propagation length on gold along the X-Y direction is less than 3 μm for excitation wavelengths shorter than 630 nm,16
11 μm spacing ensures sufficient separation between individual arrays to avoid interference, while allowing a maximum number of nanohole arrays in the CCD field of view. The hole diameter was changed from 130 nm to 180 nm, and the D value (hole periodicity/diameter) was varied from 1.67 to 3.00. show SEM micrographs of the top view and cross section of a nanohole array. A protruded ring of gold ca. 160 nm in height was found underneath the hole. It is mainly due to Ga ion bombardment on soft FEP since this phenomenon was not observed on normal gold substrates. We also found that it was relatively easier to fabricate high-quality nanoholes ( insert) on TGS substrate by FIB than on fresh thermally deposited gold surface (surface roughness ~10 nm).
The optical setup and design of nanohole arrays allow collecting transmission spectra of all thirty small nanohole arrays simultaneously. This feature enables fast screening of small nanohole array patterns in term of sensing sensitivity in different wavelengths. The black curve in represents a typical transmission spectrum of one of the nanohole arrays (d = 180 nm, D = 2.33, TSG substrates) in when flowing DI water and the red curve represents the spectrum upon switching flowing media from DI water (RI = 1.3331) to 21% NaCl (RI = 1.3702). Bulk plasmon resonance peak around λ = 510 nm was found in almost all spectra and did not shift upon the change of flowing media. The wavelength shift sensitivity is 323 nm/RIU for λ ~ 600 nm. The blue curve in represents the differential intensity (ΔT/T in percentage) between two spectra in . Two maxima (562 nm and 618 nm) and one minima (588 nm) are found in the visible range. These wavelengths (denoted as S-wavelengths in the following context) will be useful for single-wavelength sensing by intensity interrogation. We compared the spectra collected with different collimation conditions. Larger light divergence (39°) normally excites a wider distribution of wavevectors and leads to lower sensitivity than smaller light divergence (5°). However, the signal to noise ratio (S/N) with 5°divergence is much worse than 39°divergence because of lower photon flux. We hence used 39°divergence (NA = 0.34) for the rest of experiment.
Figure 2 (a) Typical transmission spectra of a nanohole array (d = 180 nm, P = 419 nm) on the TSG substrate. Black plot: in DI water; red plot: in 21% NaCl solution. (b) Intensity difference (ΔT/T) between two plots in Fig. 2a. (c) The relation between (more ...)
From analyzing the transmission spectral features from a series of nanohole arrays, we first discover that the hole diameters only slightly affect the transmission maxima and the periodicity is the dominant factor. Larger diameters lead to higher transmission. For the same periodicity, the transmission intensity from a d = 180 nm array (9×9, P = 360) is around 7 times that of a d = 130 nm array at the resonance peak. This observation is consistent with previous reports using large nanohole arrays and collimated / polarized light sources.4, 25-27
shows the relation between S-wavelengths (plotted if higher than 5% intensity change) and hole periodicity from thirty different small nanohole arrays in from TSG and regular gold substrates. Maxima and minima of differential intensity form four different linear relations (Lines 1~4) in the visible region. Two different nanohole substrates (glass and FEP) are of the same linear relations of maxima and minima of differential intensity, although different peak positions in transmission spectra due to the different interfaces between gold and substrates. This observation confirms that the refractive-index sensing occurs exclusively on the gold-solution interface. For periodicity < 450 nm, differential intensity maxima mostly appear at shorter wavelengths than minima; for periodicity > 450 nm, the few minima and maxima at the right-bottom corner of indicate another set of trends extending to the large-periodicity region. No maxima or minima of differential intensity were found on the left-hand side of Line (1) in the visible region. Each line in represents a red-shift edge of transmission peaks. For a small refractive index change (e.g. 0.0371 in this figure), in most spectra we collected, only the shift of a single edge of transmission peaks was observed. A full theoretical analysis of the nature of transmission features is beyond the scope of this article. Nevertheless, is helpful in determining nanohole periodicity when using a monochromatic excitation source that has better intensity and stability than white light sources.
The magnitude of intensity change of Lines (1) and (2) was plotted in in the bubble plot format. A strong trend can be observed that shorter periodicity mostly leads to weaker intensity change for both lines, that is, poorer sensing performance. For the same diameter, a long-periodicity array always has more intensity change than the short-periodicity one. In average, Line (2) has slightly better sensing performance than Line (1), especially for P < 400 nm. Lines (3) and (4) have the same trend as Lines (1) and (2) (data not shown for clarity). The extraordinary transmission from nanohole arrays on TSG substrates is typically more intense than on normal gold substrates by 10 ~ 20 % and the intensity change (sensing performance) is in average also about 20 % better.
The fact that longer periodicity leads to higher sensitivity implies that a larger nanohole array (for the same nanohole number) is a better sensor. This suggests the interaction area between gold and flowing solutions substantially affects the sensor sensitivity. Meanwhile, we also investigated the effect of the number of nanoholes on sensitivity. In , the number of nanohole arrays was changed from 9 (3×3) to 196 (14×14) on TSG substrates. The transmission intensity was proportional to the number of nanoholes (). The FWHM (Full Width at Half Maximum) was found to be smaller when increasing the number of nanoholes, which is consistent with a previous report.5
For intensity interrogation used in this work, small FWHM (sharper resonance peaks) will be preferable for obtaining higher sensing sensitivity. According to this result, we suggest that when comparing the sensitivity of different nanohole sensing systems based on intensity interrogation or wavelength interrogation, the size of nanohole arrays should be taken into consideration.
Figure 3 Transmission intensity at λ = 568 nm (max. intensity) of nanohole arrays from 3×3 to 14×14 (d = 150 nm, P = 450 nm) on TSG substrates. (b) Full-Width at Half-Maximum (FWHM) of nanohole arrays from 3×3 to 14×14. (more ...)
For real-time sensing measurement, the monochromator was set to the wavelength around the S-wavelengths of target nanohole arrays in . The bandpass was increased to 4 nm to improve the S/N ratio. More than 30 real-time sensing signals could be collected simultaneously. show the real-time response from a 9×9 and a 3×3 array (with different d and P). When the refractive index varied from 1.3331 and 1.3702, both arrays exhibited linear relation between intensity changes and the refractive index (Inserts in ). For the 9×9 array, the highest intensity change was observed as a 37.5% increase for Δn = 0.0371 (DI water to 21% NaCl) at λ = 590 nm; for the best 3×3 array, it was only 7.5% at best. Less intense transmission from the 3×3 array also leads to a low S/N ratio. Hence, we observe an inevitable trade-off between the sensitivity and size of the nanohole arrays. Nevertheless, 1010 %/RIU sensitivity for the 9×9 array is fairly close to the regular grating coupler-based SPR sensors operated in the intensity interrogation mode,16
with the sensor size of nanohole arrays is only 1/2000 of that used in commercial instruments (typical spot size ~ 200 μm in diameter).
Although the TSG surface offers better opportunity for more uniform immobilization of biomolecules, it also has a smaller surface area than the regular gold for immobilization (3% less, estimated by AFM). In order to demonstrate the capability of detecting biorecognition events, Glutathione S-Transferase (GST) and anti-GST antibody binding, a high affinity interaction that has been extensively used in Western blots and dot blots to detect GST-fusion proteins, was selected in this work (). We immobilized GST and used anti-GST as the target molecule since anti-GST is a bigger protein (MW ~ 150kD) that will induce more change in the local dielectric environment. The TSG surface was first functionalized with a thiol terminated with a primary amine. The GST proteins were then immobilized onto the gold surface with a cross-linker that covalently attached to the amino terminus of GST and thiol (see supporting info.
for details). Different concentrations of anti-GST antibodies prepared in PBS 7.4 buffer were introduced to the flow cell through a sample injector. After the binding event, 10 mM glycine (pH 2.2) was used to dissociate GST and anti-GST complex to regenerate the surface. In these experiments we functionalized all the nanohole arrays on the same substrate to demonstrate the sensing capability. For a practical multiplexed sensing, commercial AFM-based nano-spotters28
may be used to functionalize individual arrays with different biomolecules.
Figure 4 (a) Sensing response upon injecting 615 nM anti-GST and subsequently the regenerator. (b) Schematic illustration of surface chemistry on nanohole arrays (drawing not to scale) (c) Sensing response (black plot) of a 9×9 nanohole array (d = 200 (more ...)
shows that the transmission intensity from a nanohole array immobilized with GST changed significantly after 615 nM anti-GST was introduced. Subsequent stripping with 10 mM glycin (pH 2.2) successfully removed anti-GST from the gold surface and the signal baseline came back to the previous level. The reaction was fully reversible for at least ten times. In , three different concentrations of anti-GST were injected in a random order. Red curves represent the best fit of the binding responses to a simple bimolecular reaction model (A+B ↔ AB)29
using BIAevaluation (GE Health). The dissociation affinity constant (KD
) of this specific assay was determined by commercially available BIACORE 3000 (GE Health) as 2.97 nM, which is typical for a high affinity binding. Although immobilized GST could bind with more than one anti-GST and the bimolecular model may not be the best model to describe this reaction pathway, the calculated association constant (ka
) from nanohole arrays was 2.1×104
, fairly close to 3.7×104
determined by BIACORE 3000 with dissociation and association constants fit separately. The dissociation constant (1.1×10-4
, determined by BIACORE 3000) was too small to be accurately verified in this nanohole system because of the high affinity of this assay and the baseline drift. For smaller nanohole arrays (), although the S/N ratio was not as good as the 9×9 array, similar association constants (2.1×104
for 5×5 and 2.6×104
for 3×3) still can be determined by the fitting software. The detection limit (2σ) is around 10 nM for the 9×9 array. It should be mentioned that the measuring setup was designed to acquire multiple transmission spectra so highly intense and stable laser sources were not used. Low detection limits, though also highly desirable, are not the highest priority in this work.
Although TSG substrates show better sensing performance than the regular thermally evaporated gold substrates, the transmission enhancement and performance improvement are far less than we expected from literature.21, 23
The possible reasons are that the refractive index of FEP still not exactly matches the biological solutions and there might be a small amount of air trapped between gold and FEP during the template-stripping process. The improvement of the TSG process with FEP is in progress. The protruded gold ring in the FEP layer caused by ion bombardment might have also played an important role. The sharp edge could contribute to some localized surface plasmon characters,9
which have been proven to markedly affect the transmission intensity.
In summary, we present a plasmonic sensing platform that exploits small-dimension ordered arrays of nanoholes perforated on a super smooth gold surface that adhered on fluoropolymer substrate. It is capable of performing multiplexed and real-time binding affinity analysis on biorecognition events. The spectral study on 1 μm2 ~ 20 μm2 nanohole arrays indicates that higher gold-solution contact area results in higher plasmonic sensing sensitivity. There is an inevitable trade-off between the sensitivity and the size of the periodically arrayed nanoholes. Nevertheless, our result demonstrated that nanohole arrays with dimension as small as 1 μm2 could be used to detect biomolecular binding events effectively and analyze the binding kinetics. TSG on FEP substrates provides a smooth environment for higher-quality nanofabrication and more uniform immobilization. The refractive index matching between FEP and water solutions contributes to ~20% improvement on the sensing performance. This nanohole approach presents a sophisticate platform to studies ranging from highly multiplexed protein binding screening, to viral fusion and transmembrane protein analysis that require a high-quality supported lipid membrane.