We present the dynamic study of optical trapping of fluorescent molecules using high-density gold nanodisk arrays. The gold nanodisks were fabricated by electron beam lithography with a diameter of 500 nm and a period of 1 μm. Dark-field illumination showed ∼15 times enhancement of fluorescence near edges of nanodisks. Such enhanced near-field generated an optical trapping force of ∼10 fN under 3.58 × 103 W/m2 illumination intensity as calculated from the Brownian motions of 590 nm polystyrene beads. Kinetic observation of thiolated DNA modified with Cy5 dye showed different binding rates of DNA under different illumination intensity. The binding rate increased from 2.14 × 103 s−1 (I = 0.7 × 103 W/m2) to 1.15 × 105 s−1 (I = 3.58 × 103 W/m2). Both enhanced fluorescence and binding rate indicate that gold nanodisks efficiently improve both detection limit and interaction time for microarrays.

Microarrays have been attracted significant attention for the detection of genetic expression and protein-protein interactions.1 It takes advantages of high throughput and small volume of probe molecules. One drawback of the microarray for pathogen detection is that it exhibits long interaction time, up to several hours. It makes the microarray unattractive for fast detection applications. The interaction time has been investigated and assessed to be limited by the diffusion of biomolecules. For example, a typical diffusion coefficient for DNA strands is about D = 1 × 10−7 cm2/s. The time required for DNA strands to migrate via diffusion from the edge of a hemisphere to the center can be estimated by the equation, |$N/6.20 \times 10^{23} = 2/3\pi (\sqrt {Dt})^3 \times M$|N/6.20×1023=2/3π(Dt)3×M, where M is the molar concentration and N is the detectable oligo number. Assuming N = 106, the diffusion time (t) will be 23.8 hrs for 1 pM DNA concentration. As the concentration of strands in solution decreases, a greater volume of liquid is required and the reaction time is increased to provide required strands for the detection. To improve the binding rate, other active components for providing convention of flow have to be used. For example, a microfluidic device that pushes the sample fluid back and forth past the chip can be used to reduce the DNA hybridization time.2 

In recent years, metal structure enhanced fluorescence (MEF) method have been developed to improve signal to noise ratio of fluorescent detection.3 The localized surface plasmon resonance (LSPR) on gold or silver nanoparticles/nanostructures has been used for enhancing fluorescent detection of biological samples.4,5 On the other hand, the strong optical near-field near plasmonic nanostructure can generate optical trapping force in liquid environment. This localized near-field helps attracting biomolecules to the nanostructures with a trapping force in femto-newton range.6 In this letter, we present an improved microarray substrate to enhance both the detection sensitivity and binding rate by using high-density gold nanodisks (GNDs). Using the dark-field illumination, strong LSPR occurs at the edge of nanodisks. It produces a MEF factor up to 15 and a trapping force of 15 fN. With this optical trapping force, the binding rate is increased about two orders of magnitude under 3.58 × 103 W/m2 illumination.

GNDs were fabricated using electron-beam lithography. First, PMMA photoresist was spun casted on a glass substrate and soft baked on a hot plate at 170 °C for 90 seconds. The electron-beam writer generated 500-nm diameter nanoholes with 1-μm period on PMMA. Then 5-nm-thick Ti and 100-nm-thick Au film were deposited on the patterned substrate by an electron gun evaporator. The sample was rinsed and sonicated in acetone and DI water for 10 minutes to remove PMMA photoresist. Fig. 1(b) shows the SEM image of GNDs on glass substrate. In order to observe MEF effect, we used a super-resolution dark-field optical microscopy (Cyto-Viva) with an oil condenser to illuminate the sample as showed in Fig. 1(a). The dark-field illumination takes advantages of high spatial resolution, low background light and large off-angle incidence. It produced LSPR along the depth direction of the GNDs. The resonance can by tuned by the gold film thickness. For the 500-nm-diameter GNDs with 100-nm-thick film, there were two LSPR peaks at 550 nm and 660 nm as seen in Fig. 1(c). The small 550-nm-peak was the in-plane LSPR and the 660-nm-peak was the out-of-plane LSPR. The 660-nm LSPR is much stronger due to the dark-field illumination which has the major electric polarization along the depth direction. Because the excitation was based on the dark-field illumination, the background light was very low. It is unnecessary to use color filters to block the incident light. In the optical setup, we illuminated the sample by a white light source and took the images using a color CCD camera (Pixelfly qe, PCO.). Fig. 1(d) shows the optical image of GNDs with Cy5-ssDNA on GNDs. The single strand DNA sequence was (5- CTACC TTTTT TTCT G-3). The first sequence was modified with thiol linker at the 5 end and labeled with fluorescence marker (Cy5) at 3 end. The Cy5 maker had excitation/emission peak wavelengths at 650 nm/670 nm, respectively. In the experiments, a buffer solution of SSC buffer (5x) was injected to the chamber for the reference. Then 10μM ssDNA solution with thiolated Cy5-ssDNA was injected into the sample for 30 minutes. The sample was washed three times for removing unbounded ssDNA by SSC buffer. The ssDNAs were immobilized on the gold surface through the covalent bonding of S-Au bonds. As can be seen in Fig. 1(d), obvious red emission was observed at the edges of GNDs. In order to measure the value of MEF factor, we coated an uniform layer (15 nm) of red fluorescent dye on a glass substrate with the GNDs by using a thermal evaporator. Fig. 1(e) shows the red fluorescent intensity on the glass and near the edge of GNDs. It shows about 15 times fluorescence enhancement near the GND.

FIG. 1.

(a) A schematic diagram of experimental setup of dark-field optical microscopy for DNA biosensor. The inset figure was ssDNA which has a fluorescence marker (Cy5) and immobilized with GNDs. (b) The SEM image of GND arrays on glass substrate. The image was tilted at 45° to view the thickness. (c) Optical scattering spectrum of GNDs in liquid environment. The 550-nm-peak is the in-plane LSPR and the 660-nm-peak is the out-of-plane LSPR. (d) Dark-field optical image of GNDs. The sample was treated with 10 μM Cy5 labeled ssDNA for 30 minutes. (e) The intensity distribution near the GND. A 15-nm fluorescent red dye was coated on the surface as indicated in the inset.

FIG. 1.

(a) A schematic diagram of experimental setup of dark-field optical microscopy for DNA biosensor. The inset figure was ssDNA which has a fluorescence marker (Cy5) and immobilized with GNDs. (b) The SEM image of GND arrays on glass substrate. The image was tilted at 45° to view the thickness. (c) Optical scattering spectrum of GNDs in liquid environment. The 550-nm-peak is the in-plane LSPR and the 660-nm-peak is the out-of-plane LSPR. (d) Dark-field optical image of GNDs. The sample was treated with 10 μM Cy5 labeled ssDNA for 30 minutes. (e) The intensity distribution near the GND. A 15-nm fluorescent red dye was coated on the surface as indicated in the inset.

Close modal

The plasmonic GNDs produced strong localized near fields. It not only enhanced the fluorescence but also created a potential well on metal surface to trap dielectric and biological specimen.6–8 To estimate the trapping force, we used polystyrene (PS) beads (diameter = 590 nm) with a concentration of 0.05% W/V. The incident light was a 300 W Xeon lamp. Because of the dark-field illumination, the scattering images of PS beads can be directly observed without the use of fluorescent labeling. It thus enables the long-term and dynamic observation of Brownian motions of beads. Fig. 2(a) shows the multimedia file of the recorded motion of PS beads near GNDs surface. Strong scattering light from the LSPR was observed when the bead was near the GND. We tracked the PS bead and analyzed its velocity and optical scattering intensity. From the multimedia, it can be found that the velocity of the PS bead was decreased as it approached to the GND. Fig. 2(b) shows the velocity and optical scattering intensity of PS beads as a function of time. When the bead's velocity was decreased, the scattering intensity was significantly increased. It confirmed that trapping of PS bead was in the near-field of the GNDs. Fig. 2(c) shows the statistic histogram of the motion of PS beads (number = 296) on the GND array. These were two distinct velocity in histogram as fitted by Gaussian distributions. The average velocity of PS bead was 1.59 μm/s and 3.57 μm/s when they were near and away from the GNDs. According to Stokes’ Law, we could estimate the trapping force by using F = 6πηrv, where F is the drag force, η is the viscosity of liquid solution, r is the radius of the PS bead, and v is the velocity. For the velocity reduction from 3.57 μm/s to 1.59 μm/s, the drag force of GNDs was about 8 fN. This result was similar with other reports of plasmonic optical tweezers in femtonewton range.6 The intensity of 300 W Xe-lamp illumination was not strong enough to completely trap the PS beads, but it can produce a small trapping force to decrease the bead velocity.

FIG. 2.

(a) (multimedia view) The dark-field images of a polystyrene bead moving near GNDs. (b) The average velocity and optical scattering intensity when polystyrene beads close to GNDs. (c) Histogram of average velocities of polystyrene beads. There was two peaks at 1.59 μm/s and 3.57 μm/s, respectively. (sample number = 296). [URL: http://dx.doi.org/10.1063/1.4869640.1]

FIG. 2.

(a) (multimedia view) The dark-field images of a polystyrene bead moving near GNDs. (b) The average velocity and optical scattering intensity when polystyrene beads close to GNDs. (c) Histogram of average velocities of polystyrene beads. There was two peaks at 1.59 μm/s and 3.57 μm/s, respectively. (sample number = 296). [URL: http://dx.doi.org/10.1063/1.4869640.1]

Close modal

To demonstrate that such trapping force can increase the binding rate of the DNA on the sample surface, we studied the dynamic immobilization of thiolated Cy5-ssDNA with 5μM concentration. We recorded the Cy5 fluorescence in real time by using a color CCD and dark-field illumination. The CCD camera exposure time was 30 ms and the readout time was 87 ms. The multimedia file in upper left of Fig. 3(a) shows the dynamic binding of DNA on a GND array when incident optical intensity was increased from 20% to 100% (I = ∼3.58 × 103 W/m2) of the 300 W Xe-lamp. Obvious red color changes on GNDs were observed. The changes were due to the immobilization of Cy5-ssDNA on the GNDs. In order to describe the binding rate changes, we defined T0 was a initial state when input light intensity was I = ∼0.7 × 103 W/m2. The illumination intensity was increased to I = ∼3.58 × 103 W/m2 at T1 state. T2 was the time required for the trapped fluorescent molecules reaching to the steady state. The representative images of 9 × 9 GNDs at different interaction time, T0, T1 and T2, were shown in Fig. 3(a). Due to the immobilization of Cy5-ssDNA, numbers of red GNDs increased with interaction time. Fig. 3(b) shows the RGB intensities of the GND as a function of time. In the beginning, R, G and B pixel intensities have similar values. When the input light intensity was increased to I = ∼3.58 × 103 W/m2, scattering intensities at three pixels increased simultaneously. The intensities dropped due to the absorption of Cy5-ssDNA. The red light was much brighter than G and B colors due to the excitation of Cy5 fluorescence. At the steady state, the GNDs became red color.

FIG. 3.

(a) (upper left) (multimedia view) The dynamic binding of DNA on a GND array when incident optical intensity was increased from 20% to 100% (I = ∼3.58 × 103 W/m2) of the light power. Dark-field optical images of GNDs recorded at different interaction time: T0 state (lower left), T1 state (upper right) and T2 state (lower right). The scale bar was 3 μm. (b) Dynamic binding process of Cy5-labled ssDNA on single GND. The intensities changes in R, G and B pixels as a function of time. (c, d) Cy5-labled ssDNA immobilization on 50 × 50 gold nanodisks. The experimental data (blue dots) show the interaction time and the corresponding numbers of GNDs at steady state. The red line shows the Langmuir fitting curve. The illumination intensity was (c) 0.7 × 103 W/m2 and (d) 3.58 × 103 W/m2. [URL: http://dx.doi.org/10.1063/1.4869640.2]

FIG. 3.

(a) (upper left) (multimedia view) The dynamic binding of DNA on a GND array when incident optical intensity was increased from 20% to 100% (I = ∼3.58 × 103 W/m2) of the light power. Dark-field optical images of GNDs recorded at different interaction time: T0 state (lower left), T1 state (upper right) and T2 state (lower right). The scale bar was 3 μm. (b) Dynamic binding process of Cy5-labled ssDNA on single GND. The intensities changes in R, G and B pixels as a function of time. (c, d) Cy5-labled ssDNA immobilization on 50 × 50 gold nanodisks. The experimental data (blue dots) show the interaction time and the corresponding numbers of GNDs at steady state. The red line shows the Langmuir fitting curve. The illumination intensity was (c) 0.7 × 103 W/m2 and (d) 3.58 × 103 W/m2. [URL: http://dx.doi.org/10.1063/1.4869640.2]

Close modal

To study the enhanced binding ability due to the optical trapping force, we measured the numbers of steady-state GNDs as a function of interaction time. The binding between GNDs and DNA was described by modifying a kinetic binding model.9 In our experiment, we assumed the GND as a “probe” and the thiolated DNA a “Target”. Then we fitted it with a Langmuir binding model.

\begin{eqnarray}N &=& N_{eq} \left( {1 - e^{ - \left( {k_a C + k_d } \right)\left( {t - t_0 } \right)} } \right) \nonumber\\N_{eq} &=& \displaystyle\frac{{k_a C}}{{k_a C + k_d }} \cdot N_{\max }\end{eqnarray}
N=Neq1ekaC+kdtt0Neq=kaCkaC+kd·Nmax
(1)

Where C is the concentration of DNA analyte, t0 is start time, Nmax is the maximum GND binding number, Neq is the GND number at the steady state. An association rate constant (ka) was calculated from the fitting of Eq. (1) to the experimental data. In this fitting, we assumed the dissociation rate constant, kd ∼ 0 because the S-Au was a strong covalence binding. Fig. 3(c) and Fig. 3(d) show the experimental data and curve fitting of 50 × 50 GNDs for I = ∼0.7 × 103 W/m2 and I = ∼3.58 × 103 W/m2, respectively. The x-axis is the interaction time. The y-axis is the corresponding number of GNDs where Cy5-labeled ssDNA reached at steady state. It can be seen that interaction efficiency between GND and DNA was significantly increased by increasing optical intensity. The ka was 2.14 × 103 M−1s−1 for low illumination and increased to 1.15 × 105 M−1s−1 at I = ∼3.58 × 103 W/m2. The low-light result was similar with Nirmal's group10 who used DNA immobilization on gold thin film structure. The result was also at the same order of association constants with other groups.11,12 When strong light intensity was incident onto the high-density GNDs, two orders of magnitude enhancement of binding constant was measured.

In conclusion, we used GNDs with a diameter of 500 nm and a period of 1 μm as a substrate to simultaneously enhance the fluorescent intensity and binding rate of biomolecules. The fluorescence was increased up to 15 times and the binding constant was enhanced two orders of magnitude for a 300 W lamp under dark-field illumination. We proved the optical trapping force and LSPR effect of GNDs by the velocity and intensity changes of submicron polystyrene spheres. Using a color CCD, the dynamic binding process of Cy5-ssDNA and GNDs was studied in real time. The statistics show the binding rate was significantly increased with the illumination intensity. The GND array provides a high-throughput detection with simultaneous fluorescence and binding rate enhancement. These kinds of gold nanostructure substrate can be applied for studying DNA/RNA hybridization, protein-protein interaction and extracellular signals of cell secretion.

This work was support by the grand of National Science Council, Taipei, Taiwan, under contract No. of NSC-99-2120-M-007-009 and NSC-97-3112-B-001-022. Technical support from NanoCore, the core facilities for nanoscience and nanotechnology at Academia Sinica in Taiwan, is acknowledged.

1.
Gavin
MacBeath
and
L.
Schreiber
,
Science
289
,
1760
(
2000
).
2.
C. W.
Wei
,
J. Y.
Cheng
,
C. T.
Huang
,
M. H.
Yen
, and
T. H.
Young
,
Nucleic Acids Research Methods
33
,
e78
(
2005
).
3.
P. P.
Pompa
,
L.
Martiradonna
,
Della
Torre
,
F.
Della Sala
,
L.
Manna
,
M.
De Vittorio
,
F.
Calabi
,
R.
Cingolani
, and
R.
Rinaldi
,
Nature nanotechnology
1
,
126
(
2006
).
4.
Mark P.
Kreuzer
,
Romain
Quidant
,
J.-Pablo
Salvador
,
M.-Pilar
Marco
, and
Gonçal
Badenes
,
Anal Bioanal Chem
391
,
1813
(
2008
).
5.
Greg J.
Nusz
,
Adam C.
Curry
,
Stella M.
Marinakos
,
Adam
Wax
, and
Ashutosh
Chilkoti
,
ACS Nano
3
(
4
),
795
(
2009
).
6.
Maurizio
Righini
,
Giovanni
Volpe
,
Christian
Girard
,
Dmitri
Petrov
, and
Romain
Quidant
,
Phys. Rev. Lett.
100
,
186804
(
2008
).
7.
Lina
Huang
,
Sebastian J.
Maerk
, and
Olivier J. F.
Martin
,
Opt. Express
17
(
8
),
6018
(
2009
).
8.
Mathieu L.
Juan
,
Maurizio
Righini
, and
Romain
Quidant
,
Nature photonics
5
,
349
(
2011
).
9.
Edward C.
Hulme
and
Mike A.
Trevethick
,
British Journal of Pharmacology
161
,
1219
(
2010
).
10.
Nirmal
Prabhakar
,
Kavita
Arora
,
Sunil K.
Arya
,
Pratima R.
Solanki
,
M.
Iwamoto
,
Harpal
Singh
, and
B. D.
Malhotra
,
Analyst
133
,
1587
(
2008
).
11.
Jeroen
Pollet
,
Filip
Delport
,
Kris P. F.
Janssena
,
Karolien
Jans
,
Guido
Maes
, and
Helge
Pfeiffer
,
Biosens. and Bioelectron.
25
,
864
(
2009
).
12.
Shu-ichi
Nakano
,
Takayuki
Kanzaki
,
Mariko
Nakano
,
Daisuke
Miyoshi
, and
Naoki
Sugimoto
,
Anal. Chem.
83
,
6368
(
2011
).