Optical excitation of surface plasmon polariton (SPP) on a thin metallic surface is widely applied for label-free and real-time detection of surface biomolecular interactions [1
S. A. Maier, Plasmonics: Fundamentals and Applications (Springer 2007).
]. The common method employs attenuated total reflection (ATR) on a glass prism to excite SPP on a thin gold film. The ATR biosensors are sensitive to surface environmental changes. However, the setup is typically bulky, expensive and requires a large amount of sample volume. Due to its optical configuration, it is hard to be used for high-throughput and chip-based detections, such as DNA microarrays and protein microarrays [2
M. Schena, D. Shalon, R. W. Davis, and P. O. Brown, “Quantitative monitoring of gene expression patterns with a complementary DNA microarray,” Science
270(5235), 467–470 (
1995). [CrossRef] [PubMed]
G. MacBeath and S. L. Schreiber, “Printing proteins as microarrays for high-throughput function determination,” Science
289(5485), 1760–1763 (
]. Recently, much attention has been placed on gold nanostructures for making high-throughput biosensing devices [4
A. G. Brolo, R. Gordon, B. Leathem, and K. L. Kavanagh, “Surface plasmon sensor based on the enhanced light transmission through arrays of nanoholes in gold films,” Langmuir
20(12), 4813–4815 (
2004). [CrossRef] [PubMed]
J. Ji, J. G. O’Connell, D. J. D. Carter, and D. N. Larson, “High-throughput nanohole array based system to monitor multiple binding events in real time,” Anal. Chem.
80(7), 2491–2498 (
2008). [CrossRef] [PubMed]
]. The SPP is excited directly by subwavelength metallic apertures. In periodic nanostructures, an extraordinary optical transmission (EOT) occurs when the SPP wavelength meets the Bragg condition [11
T. W. Ebbesen, H. J. Lezec, H. F. Ghaemi, T. Thio, and P. A. Wolff, “Extraordinary optical transmission through sub-wavelength hole arrays,” Nature
391(6668), 667–669 (
Where i, j
are the orders in the x-y directions, P
is the period of the nanostructure, εm
is the dielectric constant of the metal and n
is refractive index of the environment. When the surface molecular density increases, the resonant wavelength is red-shifted due to the increase of n
. Using the peak EOT wavelength as signals, high density sensing arrays have been demonstrated [7
A. De Leebeeck, L. K. S. Kumar, V. de Lange, D. Sinton, R. Gordon, and A. G. Brolo, “On-chip surface-based detection with nanohole arrays,” Anal. Chem.
79(11), 4094–4100 (
2007). [CrossRef] [PubMed]
L. Pang, G. M. Hwang, B. Slutsky, and Y. Fainman, “Spectral sensitivity of two-dimensional nanohole array surface plasmon polariton resonance sensor,” Appl. Phys. Lett.
91(12), 123112 (
]. The theoretical and experimental studies indicate that the wavelength sensitivity (Sλ
) is close to the period of the nanostructures, i.e. Sλ
, where RIU
is the refractive index unit. Such wavelength sensitivity is similar to the sensitivity of conventional grating-based biosensors, but is much smaller than that of ATR-based biosensors [12
J. Homola, S. S. Yee, and G. Gauglitz, “Surface plasmon resonance sensors: review,” Sens. Actuators B Chem.
54(1–2), 3–15 (
]. To increase the wavelength sensitivity, the period has to be increased to several micrometer long and the resonant wavelength will be shifted to the infrared region.
On the other hand, the redshift of the resonant wavelength also induces an intensity change at a fixed wavelength near the resonant condition. Such intensity changes are more suitable for real-time and multiple detections [9
K. L. Lee, C. W. Lee, W. S. Wang, and P. K. Wei, “Sensitive biosensor array using surface plasmon resonance on metallic nanoslits,” J. Biomed. Opt.
12(4), 044023 (
2007). [CrossRef] [PubMed]
]. A low-noise and large-area CCD simultaneously records the multiple intensity changes in the sensing array during the biomolecular interactions. Dynamic studies of multiple bio-interactions are thus feasible. For a small wavelength shift, the intensity sensitivity (SI
) is proportional to the wavelength sensitivity and differential of the spectrum.
If we assume a Gaussian profile for the resonant spectrum,
, the differentiation is
is the resonant wavelength and d
is the half of the 1/e
width. The maximum I’
and the maximum value,
. It indicates the maximum intensity sensitivity is proportional to the inverse of the bandwidth. For the same period, a small bandwidth will have a large sensitivity in the intensity measurement. The resonant bandwidth is dependent on the size and shape of the metallic structures. In this paper, we first compared the resonant bandwidths for two different metallic nanostructures: nanohole arrays and nanoslit arrays. These two nanostructures have been proposed for high-density label-free microarrays. The finite-difference time-domain (FDTD) calculations and experimental results show that gold nanoslit array has a much narrower bandwidth than that of the nanohole array. In addition, decreasing the slit width can narrow the bandwidth and greatly increase the intensity sensitivity. A microfluidic device combined with 10 × 10 nanoslit arrays were made to test the dynamic biomolecular interactions. The measurement results verify the real-time and high-throughput detections of the gold nanoslit arrays. Using the intensity measurement, the detection sensitivity for Anti-BSA molecules is below 100 pM when the intensity stability is 0.2%.
2. Device fabrication and measurement setup
The microarray chips used glass slides as the substrate. Gold nanostructures were fabricated by using electron beam lithography and reactive ion etching method. Gold has poor adhesion to glass surface, hence a 5-nm-thick Ti film was deposited before a 150-nm-thick gold film by using an electron gun evaporator. After the deposition, a 350nm-thick PMMA resist (MicroChem) was spun-coated on the sample. A field-emission scanning electron microscope modified with a nano-pattern generating system (NPGS) was used to write the nanostructures on the PMMA resist. The patterns were then transferred into Au/Ti film by using argon sputtering in a reactive ion etching machine (Oxford Instrument). The power of the radio frequency wave in the reaction chamber was 200 W. The flow rate of argon gas was 40 sccm. The PMMA resist was removed by rinsing the samples in acetone for a few hours. To make sure that no PMMA resist residue remained on the gold nanostructures, we further cleaned the samples by using ozone sputtering in the RIE chamber. The samples were then put in ultra-pure water, placed in an ultrasound bath for 20 minutes, and purged dry by nitrogen.
In our experiments, we fabricated gold nanohole arrays and nanoslit arrays with different hole sizes and slit widths. The periods for nanoholes and nanoslits were the same 600 nm. At this period, the resonant wavelength was about 825 nm in water environment as estimated from the Eq. (1)
, where εm
= −29 + 2.0i and n = 1.332, respectively. Figure 1(a)
shows the optical image of a biochip combined with 10 × 10 nanoslit arrays. Each array size was 150 μm in square and the spacing between adjunct arrays was 100 μm. Both dimensions are comparable with modern DNA microarrays. Figure 1(b)
shows the SEM images for gold nanoholes. The hole diameter was about 150 nm. Figure 1(c)
shows the periodic gold nanoslits on the glass substrate. The slit width was about 80 nm. It is noted that the slit width can be further reduced without losing transmission intensity due to none cut-off condition for the transverse magnetic (TM) polarized wave. However, transmission through the gold nanoholes is greatly reduced when the hole size is smaller that 100 nm.
Fig. 1 (a) The optical image of a 10 × 10 microarray on a glass slide. Each green area was consisted of a 600nm-period gold nanoslit array. The area was 150 μm in square. (b) The SEM image of a 600nm-period gold nanohole array. (c) The SEM image of a 600nm-period gold nanoslit array.
Water mixtures with different refractive indices were injected into the microfluidic devices to test the bulk sensitivities. To measure the responses for the wavelength and intensity, the devices were tested by two different optical setups: the transmission spectrum measurement and the intensity measurement at a fixed wavelength. Figure 2(a)
shows the setup for the spectroscopic measurement. A 12W halogen light was spatially filtered by using an iris diaphragm and a collimation lens. Its incident polarization was controlled by a linear polarizer. The polarized light was then focused on a single array by using a 10 × objective lens. The transmission light was then collected by another 10 × objective lens and focused on a fiber cable. The transmission spectrum was taken by using a fiber-coupled linear CCD array spectrometer. Figure 2(b)
shows the intensity measurement, we took the array images by using a 4 × objective and a large-area TE-cooled CCD (SBIG 2000ME). The incident light wavelength was selected near the resonant wavelength.
Fig. 2 (a) The optical setup for measuring the transmission spectrum of single periodic gold nanostructure in aqueous condition. (b) The optical setup for measuring the transmission intensities of multiple periodic gold nanostructures at a fixed wavelength.
3. Calculations and measurement results
We performed the EOT calculations by using a commercial FDTD software (FullWAVE 4.0, RSoft). Both gold nanoslit and nanohole arrays had the same period of 600 nm. The hole size was 100 nm, 150 nm and 200 nm, respectively. The slit gap was 50 nm, 100 nm and 150 nm, respectively. The thickness of gold film was 150 nm. The incident light was a plane wave with its wavelength varied from 800 nm to 880 nm. It is noted that the FDTD program used coherent single-wavelength light for the incident wave. The transmission spectrum is obtained by scanning the wavelength. However, in the experiments, the incident light was a white light with low coherent length. Since the SPP resonance comes from the coherent interference effects. The iteration times in the FDTD calculations thus are important for fitting the experimental results. In our calculations, the time step (cΔt) was 10 nm and the iteration time was 6000. Figure 3(a)
shows the calculation spectra for the nanohole arrays. Those arrays have large bandwidths. The bandwidths are about 20 nm and independent with the hole size. It is noted that the resonant wavelength peak is red-shifted with the hole size. Such spectrum changes can be attributed to the cut-off behavior of nanoholes. As the hole size decreased, the decreased transmission will be less for shorter wavelengths than for larger wavelengths. It thus results in a decrease of peak wavelength with a decrease of hole diameter. Figure 3(b)
shows the transmission spectra for the nanoslit arrays. Contrary to the nanohole arrays, the transmission peak is blue-shifted with the slit width. The blueshift can be explained by the Fabry–Perot resonant effect. In the nanoslits, there is no cut-off for the TM polarized wave. The TM wave can propagate along the slit gaps. Due to the reflection at upper and lower interfaces, a Fabry-Perot cavity is formed [13
R. Gordon, “Light in a subwavelength slit in a metal: Propagation and reflection,” Phys. Rev. B
73(15), 153405 (
A. Moreau, C. Lafarge, N. Laurent, K. Edee, and G. Granet, “Enhanced transmission of slit arrays in an extremely thin metallic film,” J. Opt. A, Pure Appl. Opt.
9(2), 165–169 (
]. This vertical cavity mode is coupled to the horizontal surface plasmonic modes at the outside surface. Therefore, the resonant peaks are correspondent not only to periodicity of the metallic structures (surface plasmon mode) but also the slit gaps (cavity mode). The effective index of the TM wave in the nanoslit is increased with the decrease of the slit width [13
R. Gordon, “Light in a subwavelength slit in a metal: Propagation and reflection,” Phys. Rev. B
73(15), 153405 (
]. As a result, the resonant wavelength in the slit is increased when the slit width is reduced. It is noted that the bandwidth of the nanoslit array was about 10 nm which is much smaller than the nanohole array. The bandwidth becomes smaller when the slit gap is reduced. It indicates the nanoslit array is better for the intensity measurement and its intensity sensitivity can be further increased by using very narrow slits.
Fig. 3 (a) The calculated transmission spectra of gold nanohole arrays with different hole sizes. (b) The calculated transmission spectra of gold nanoslit arrays with different slit widths. All the structures have the same 600 nm period. The hole sizes and slit widths are indicated in the insets.
shows the measured spectra for the nanohole arrays. The hole sizes were 130 nm, 140 nm, 150 nm and 180 nm, respectively. No obvious EOT peak is found when the hole size is smaller than 100 nm. Figure 4(b)
shows the normalized transmission spectra. The large bandwidth and the redshift of the spectra with the hole size are consistent with the FDTD calculations. The measured bandwidth was larger than 25 nm. Figures 5(a)
show the measured results for the nanoslit arrays. The slit widths as observed from the SEM images were 35 nm, 67 nm, 92 nm, 113 nm and 166 nm, respectively. Figure 5(a)
is the transmission spectrum. Obvious EOT peak is found even when the slit width is smaller than 35 nm. Figure 5(b)
shows the normalized transmission spectrum. The relationship between the bandwidth and the slit width was indicated in the inset. As the slit width increases, the resonant spectrum is blue-shifted and the bandwidth is increased. The measurement results are quite consistent with the FDTD calculations. The measured bandwidth was 23 nm for the 166 nm nanoslits and was reduced to 13.6 nm for the 35 nm nanoslits. Therefore, the nanoslit array with a smaller slit width is expected to have a better response in the intensity measurement.
(a) The measured transmission spectra of gold nanohole arrays with different hole sizes. (b) The normalized transmission spectra of Fig. 4(a)
. All the structures have the same 600 nm period. The hole sizes are indicated in the insets.
(a) The measured transmission spectra of gold nanoslit arrays with different slit widths. (b) The normalized transmission spectra of Fig. 5(a)
. All the structures have the same 600 nm period. The bandwidths for different slit widths are shown in the inset.
To verify that the intensity is dependent on the bandwidth of the spectrum and find the optimal wavelength for the measurement, we compared the transmission spectra of the nanohole and nanoslit arrays surrounded with pure water and 20% (w/w) NaCl solution, respectively. The refractive indexes of water and NaCl solution are 1.3325 and 1.339 as measured by a precise refractometer. Figure 6(a)
shows the transmission spectra for the nanoslit arrays. The increase of refractive index results in a redshift of the resonant spectrum. The amount of the wavelength shift was about 3.74 nm and independent with the slit width. Such wavelength shift yields a wavelength sensitivity of 575.38 nm/RIU. This sensitivity is close to the theoretical prediction, 600 nm/RIU. Nevertheless, the intensity changes are varied with the wavelength and slit width. The intensity sensitivities at different wavelengths can be estimated from following equation,
Fig. 6 (a) The measured transmission spectra of gold nanoslit arrays in different surrounding media. (b) The spectrum of intensity sensitivity for nanoslits. The period was 600 nm and slit widths were 35 nm, 92 nm and 166 nm, respectively. (c) The spectrum of intensity sensitivity for nanoholes. The period was 600 nm and hole sizes were 105 nm, 158 nm and 191 nm, respectively. (d) The maximum sensitivities for different nanoholes and nanoslits.
shows the intensity sensitivity as a function of wavelength for different slit widths. The spectrum of the intensity sensitivity is similar to the differentiation of the EOT spectrum. The optimal wavelength for the intensity sensitivity is near the peak EOT wavelength. For the 35 nm nanoslits, the optimal wavelengths are 803.5 nm and 818 nm. The corresponding intensity sensitivities are −3800%/RIU and 3700%/RIU, respectively. The peak sensitivities for 92 nm and 166 nm nanoslits are 2800%/RIU and 1700%/RIU. The 35 nm slit array has an intensity sensitivity twice larger than the 166 nm slits. As compared with the bandwidths (13.6 nm for 35 nm slit and 23 nm for 166 nm slit), the intensity sensitivity is roughly proportional to the inverse of the bandwidth. Figure 6(c)
shows the spectrum of intensity sensitivity for nanoholes with different hole sizes. The spectrum is also close to the differentiation of the EOT spectrum. The maximum intensity sensitivities are ~1800%/RIU for 105 nm holes, ~1500%/RIU for 158 nm holes and ~1000%/RIU for 191 nm holes. The nanohole sensitivity is much smaller than the sensitivity of nanoslits. Figure 6(d)
shows the comparison of maximum sensitivities between nanoholes and nanoslits. The sensitivity of nanoslits is greatly increased as the slit width decreases. The sensitivity for nanoholes is also increased as the hole size decreases. Nevertheless, the intensity improvement is not as significant as the nanoslits. Decreasing the slit width can reduce the bandwidth and thus increase of sensitivity. However, bandwidth of the nanoslits cannot approach to zero. The skin depth (~20 nm) and absorption of optical wave within the gold film will limit the bandwidth.
In the intensity measurement, the linearity of the response is important for interpreting the surface interactions. To verify the linearity of the intensity change, we injected the water/glycerin mixtures with different concentrations into the microfluidic device. The ratios of mixtures were 0%, 10%, 20%, 30% and 40% (v/v). The corresponding refractive indexes are 1.3325, 1.3355, 1.338, 1.34 and 1.3418, respectively. We used a CCD to take the intensity changes as a function of time. The inset of the Fig. 7(a)
shows three nanoslit arrays (150 μm × 150 μm) with various slit widths, 80 nm, 110 nm and 150 nm, respectively. The increase of the refractive index decreases the intensity. The changes are stable and reproducible with the refractive index change. The 80 nm nanoslit array has a larger intensity change than the 150 nm one. The noise control is important for the intensity measurement. In this experiment, we used a TE-cooled CCD and maintained the CCD temperature at −20°C to reduce the detection noise. The environment temperature was kept at 25°C. All the setup was placed on an optical table. The major noise level came from the intensity fluctuation of the light source, which was about 0.2% as estimated from the fluctuations in Fig. 7(a)
. Figure 7(b)
shows the normalized intensity change against the refractive index. The change is linear. The intensity sensitivities as calculated from the slopes are 2900%/RIU, 2400%/RIU and 1250%/RIU for the 80 nm, 110 nm and 150 nm nanoslits. This result again identifies that reducing the slit width increasing the intensity sensitivity.
Fig. 7 (a) The normalized intensity changes for different surrounding water/glycerin mixtures covering on the nanoslit arrays. The inset shows three nanoslit arrays (150 μm × 150 μm) with various slit widths, 80 nm, 110 nm and 150 nm, respectively. (b) The normalized intensity change against the refractive index. The intensity sensitivities are calculated from the slopes.
4. High-throughput bio-detections
Using the sub-100nm nanoslit arrays, we performed the high-throughput and label-free detections of antigen-antibody and DNA-DNA interactions. In the antigen-antibody studies, the interactions between BSA (bovine serum albumin) (Sigma-Aldrich) and anti-BSA (Sigma-Aldrich) were measured. Figure 8(a)
shows the setup for the intensity detection of the microarray. In the experiment, the buffer solution, 10 mM phosphate-buffered saline (PBS) (UniRegion Bio-Tech), was first injected to the microarray chip. Then 50 μΜ BSA was injected on the nanoslit surface. Due to the physical absorption of BSA on gold surface, the BSA will be coated on the slit arrays. The BSA solution flew for one hour in order to make sufficient BSA fixed on the gold surface. The PBS buffer then washed the chip to remove the unbound proteins. Finally, 5 nM anti-BSA was injected into the microarray surface. After three hours protein-protein interactions, the unbound anti-BSA was washed away by the PBS buffer. We take the intensity images of the microarray at different interaction time. Figure 8(b)
shows the movie of intensity images of the 10 × 10 nanoslit arrays captured at different interaction time. The images were obtained by subtracting the measured time-lapsed images I(x,y;t) with the initial CCD image I(x,y;t = 0). The intensities indicate the amounts of the surface refractive index changes. It can be found that the slit arrays decreased the transmission intensity due to surface binding of BSA and Anti-BSA molecules. There were two defects (1,1) and (4,1) in the microarray, which had no nanoslit structures inside. These two areas did not show any intensity changes during the interactions. This is an evidence that the measured intensity changes are related to the nanoslit structures.
(a) The setup for measuring the antibody-antigen interactions on the microarray. The interactions were monitored by the intensity changes as detected by the CCD camera. (b) The movie of intensity images of the 10 × 10 nanoslit arrays at different interaction times. The images were obtained by subtracting the measured time-lapsed images I(x,y;t) with the initial CCD image I(x,y;t = 0)(Media 1
shows the normalized transmission intensity as a function of the interaction time for one of the nanoslit array (5,5). The intensity signal is stable with time when the PBS buffer is injected into the microfluidic device. The BSA coated on the gold surface resulted in an intensity change of 5%. For the Anti-BSA, the 5 nM concentration caused an intensity change up to 12%. The Anti-BAS has a larger intensity change because the molecular weight of anti-BSA (150 kDa) is much larger than that of BSA (66 kDa). Our measurement results verify the multiple dynamic detections of the gold nanoslit arrays. Using the intensity measurement, the detection sensitivity for Anti-BSA molecules is up to 83.3 pM when the intensity stability is 0.2%. Figure 9(b)
shows another biomolecular interactions using DNA hybridizations. The 20 × SSC buffer (87.65 g sodium chloride, 44.1 g sodium citrate, adjusted to pH 7.0 with NaOH) (HopeGen Biotech) was used to wash the sample. The DNA sequence (MDBio, Inc) for the probe DNA was 5′-HS-CAGAAAAAAAAGGTAG-3′ and the target DNA was 5′-CTACCTTTTTTTTCTG-3. The 16 mer oligonucelotides correspond to a length of about 5 nm. The concentration was 5 μM for probe DNA and 50 nM for target DNA. The probe DNA can be immobilized on the nanoslit surface through the covalent bonding between thiol group and gold surface. A clear intensity decrease can be found during the immobilization of the probe DNA. After the wash of SSC buffer, the target DNA was injected to the device. A small decrease of intensity (~0.5% intensity change) was measured when the target DNA hybridized with the probe DNA. It is noted that the second DNA strand did not change the length and only affect the refractive index at the monolayer. The intensity change for probe DNA is much larger than the target DNA because of the large concentration difference. In the experiments, we also tested the DNA interactions by using non-complementary target DNAs. There was no obvious intensity response for the nonspecific binding.
(a) The normalized intensity as a function of the interaction time for one of the nanoslit array (5,5) shown in Fig. 8(b)
. The BSA results in an intensity change of 5%. The 5 nM Anti-BSA caused an intensity change of 12%. (b) The normalized intensity as a function of the interaction time for DNA-DNA interactions. The probe DNAs were immobilized to the gold surface. The target DNA of 16 mer oligonucelotides was detected when hybridized with the probe DNAs.