2.1.

Assembly of the Microfluidic Platform

The microfluidic structure incorporates three or more parallel microfluidic channels to activate the microfluidic flow between two liquid reservoirs [Fig. 3(a)]. The width of the channel was chosen to be 500 μm by considering two factors: first, the width of the light beam which was $\sim 2\mathrm{mm}$ and second, the sealing of the channel. Due to the high flexibility of PDMS, sealing of the channel having the width of more than 500 μm is a difficult task as the oxygen plasma treatment is used for the bonding. When the plasma-treated substrates of glass and PDMS were kept together, it was observed that the channel having width higher than 500 μm collapsed. The plasma-treated surfaces exhibit a strong surface attraction force that could be the reason for the collapse of the channel. For this reason, the microfluidic channel was designed with three parallel microchannels of 500 μm, separated by 200 μm. The depth of the microchannel was 100 μm. The microfluidic structure was fabricated on PDMS by soft lithography. The mold for casting the PDMS was fabricated on SU8 photo-resist. Patterning of SU8 was done on a silicon wafer by using UV photolithography. The SU8 mold was silanized to promote the removal of the PDMS structure from the mold. For the fabrication of the microfluidic structure, the PDMS base and curing agent (SYLGARD 184 from Dow Corning, MIDLAND MI, USA) are mixed in a 10∶1 wt ratio, degassed in a vacuum desiccator to remove the gas bubbles, and casted to the mold and baked at 80°C for 5 h.

Fig. 3

Assembly process of the microfluidic device.

The substrate holding the nanoisland structures is shown in Fig. 3(b). The device [Fig. 3(c)] was fabricated by bonding the glass substrate to a PDMS substrate with the help of oxygen plasma treatment. The glass substrate containing gold nanoisland structures and the PDMS slab containing the microfluidic structure were treated with oxygen plasma for 35 s in a plasma cleaner and bonded together immediately. The inlet and outlet tubes were fixed with the liquid reservoirs using silicone glue. Figure 4 shows photograph of the nanoparticle integrated miocrofluidic bio-sensor.

Fig. 4

Photograph of the nanoparticle-integrated microfluidic bio-sensor.

2.2.

Experimental Procedure for the Bio-Sensing Experiment

Gold nanoparticles were first functionalized with a solution of mercaptoundecanoic acid in ethanol by covering the active surface of the platform (around 150 to 200 μL). Then the cross-linker solution (N, N′-diisopropylcarbodiimide and N-hydroxysuccinimide) was dropped on the sample to covalently attach the antibody to the gold nanostructures. After 10 min, the spectrum of the functionalized gold is measured. Afterwards, an antibody solution is introduced into the sensing platform and kept in contact with the gold nanostructure for at least 1 h. The change in the position of LSPR corresponding to the binding of antibody was measured. The concentration of antibody on the functionalized gold structure was kept constant and the concentration of antigen was varied between 5 and $\mathrm{10,000}\mathrm{ng}/\mathrm{mL}$. The excess of antibody is washed away with a phosphate buffered saline (PBS) solution and a blocker (nonfat milk powder solution) is passed, then the corresponding antigen is added to the sample and the LSPR spectrum is recorded to monitor the change of LSPR corresponding to the antigen-antibody interaction. The procedure is repeated in the microfluidic device by pumping the solution into the device.

3.1.

Tuning the Morphology of the 3-D Gold Nanostructure from Nanoaggregates to Nanoislands

The nonannealed samples showed a dark blue coloration [Fig. 5(a)] while the sample annealed at 550°C [Fig. 5(b)] had a red tint. As it can be seen in the Fig. 5(c), the films showed the striping behavior characteristic to samples deposited slowly by convective self-assembly.^6,38 It is found that the stripes are uniformly distributed and the spacing between them is constant as long as the deposition temperature is the same. For samples annealed at lower temperatures, the color of the sample is not uniform. The scanning electron microscope (SEM) images corresponding to the two samples are shown in Fig. 6. It can be seen that, before annealing, the nanoparticles are aggregated, forming long chain-shaped structures with several layers of nanoparticles [Fig. 6(a)]. Aggregation begins in the evaporating meniscus because of the strong attractive van der Waals forces between gold nanoparticles.³⁸ The annealed samples show nanoisland morphology [Fig. 6(b)], with large well-separated nanoislands and smaller particles around them. The nanoisland films formed by nucleation and coalescence of gold nanoparticles at 550°C are polydisperse with an average diameter of grains around 70 nm. The density of nanoislands is estimated to be around $64\text{islands}/\mu {\mathrm{m}}^2$ and it strongly depends on the concentration of gold in the colloidal solution.

Fig. 5

Images of the nonannealed sample (a), sample annealed at 550°C for 1 h (b) and enlarged images of the samples showing the stripping behavior (c).

Fig. 6

SEM images of 3-D gold nanostructures (a) as deposited and (b) after annealing at 550°C for 1 h (the magnification is $\times \mathrm{100,000}$ in both images and $\times \mathrm{200,000}$ in inset).

At temperatures higher than 350°C, for a sample with around $25\mathrm{\mu g}/\mathrm{ml}$ of Au in the colloidal solution, clearly separated nanoislands are formed. The islands look like droplets, are elongated (ellipsoid shaped) and their size varies between 10 and 100 nm. The formation of nanoislands upon annealing (at temperatures above 350°C) was widely reported for gold nanostructures prepared by LbL (layer-by-layer) deposition,^28,29,40 as well as by thermal³³ and electron-beam⁴¹ evaporation. The morphology of gold nanoislands generated from LbL assembly depends on the number of layers and the nature of linkers used for building up the multilayer. It has been reported⁴² that the morphology of ultrathin thermally evaporated films (less than 10 nm) can be tuned to nanoisland-type structures at temperatures as low as 200°C using long annealing times (20 h). The morphology change is thought to be induced by sintering and agglomeration of neighboring nanoparticles and diffusion coalescence, which results in the formation of a single layer of laterally-spaced islands. In the case of the multilayers deposited by angled convective assembly [Fig. 6(a)], having the intertwined chain structure, the gold nanoparticles would melt prior mobility and coalescence. Indeed, it is known that the melting temperature of gold nanoparticles is strongly depressed (tens to hundreds of degrees) due to the particle size. Due to high surface tension of liquid gold, nanoislands with convex surfaces are formed during the annealing process and percolation occurs at a high density of nanoislands.

Compared to the aggregated multilayer structure, the nanoisland structure is likely to provide multiple binding sites and enough room for the diffusion of bio-molecules.²⁸ The change in morphology, induced by annealing, is accompanied by a dramatic change in the optical properties as shown in Fig. 7(b).

Fig. 7

UV-visible spectra of gold nanostructures deposited on glass, (a) before annealing and (b) after annealing at 550°C for 1 h.

The spectra of the nonannealed and the annealed samples were measured in the UV-visible range (Fig. 7). The gold multilayer was deposited by using a solution of $45\mathrm{\mu g}/\mathrm{mL}$ gold. Aggregated gold multilayers show a broad band in the 550 to 750 nm regions [Fig. 7(a)]. When heated at 550°C for an hour, the band is blue-shifted to 545 nm and becomes considerably narrower [Fig. 7(b)]. This band belongs to the nanoisland structures shown in Fig. 6(b). The position and the shape of this band clearly show that the island structures are particulate, not aggregated. When the concentration of the gold in solution was reduced to $25\mathrm{\mu g}/\mathrm{mL}$, the band appears to be changed, as shown in Fig. 8. When the gold nanoparticle solution was diluted ($25\mathrm{\mu g}/\mathrm{mL}$ of Au), the spectrum [Fig. 8(a)] shows two bands. The band corresponding to aggregated gold nanostructure is situated around 620 nm showing a lower degree of aggregation and a band at 526 nm. After annealing, the spectrum shows only one narrow band at 529 nm with a noticeable higher absorbance [Fig. 8(b)].

Fig. 8

UV-vis spectra of gold multilayers, (a) nonannealed and (b) annealed at 500°C for 1 h (multilayers deposited from a colloidal solution with a low concentration of Au, $25\mathrm{\mu g}/\mathrm{mL}$).

3.2.

Sensing Experiments: Study of the Reproducibility and Sensitivity of Nanoisland Sensors

Several sensing platforms have been prepared and annealed, and both the position and the absorbance of the LSPR band were measured immediately after annealing and again, after several days. The results show a good reproducibility: for a number of 50 platforms, the position of the peaks after annealing was at $540+/-10\mathrm{nm}$. The position of the Au LSPR band belonging to the annealed sample may vary because of the slight differences in the preparation and annealing conditions. However, these variations would not affect the precision of the detection as the shift of the band would be the same.

The sensitivity has to be determined because it gives a preliminary assessment of the ability of the platform to be used in the actual biosensing experiment. Hence, it is important to investigate the sensitivity of the LSPR peak to the refractive index, as during the biosensing process, the analytes bind to the gold nanostructure and the refractive index of the environment of the gold nanostructure changes which results in a shift of the LSPR peak. The shift of Au-LSPR ($\mathrm{\Delta }\lambda$) upon the change of refractive index of the local environment of the nanoparticles ($\mathrm{\Delta }n$) is plotted (Fig. 9) and the slope of the curve is measured as the sensitivity of the platform. The shift of Au LSPR band in different solvents, with refractive indices in the range of 1.330 to 1.479 for both the nonannealed and annealed structures is shown in Fig. 9. In order to assess the sensitivity of the sensing platform, the change of UV-visible absorbance of gold nanostructures immersed in various media was investigated before and after annealing. The sensing platforms made of nonannealed and annealed samples are introduced subsequently in the solvents given in Table 1.

Fig. 9

Sensitivity of the platforms to the change in the refractive index of the environment for the annealed and for the nonannealed sample. (The error bar represents the standard deviation of 10 measurements.)

Table 1

Solvents used for measuring the refractive index sensitivity of annealed and nonannealed sensing platforms.

The sample with gold nanostructures is immersed in a cuvette filled with the solvent with a known refractive index. In order to assess the refractive index sensitivity, the following formulas are used:

The sensitivity is assessed by plotting the graph of $\mathrm{\Delta }n$ vs. $\mathrm{\Delta }\lambda$ as illustrated in Fig. 9. The sensitivity is measured by the slope of the graph (Fig. 9). The sensitivity of the nanoisland (annealed) sensor platform is much higher than that of the nonannealed one, more specifically, the sensitivity as calculated from the slope of the lines is found to be $63\mathrm{nm}/\mathrm{r}\mathrm{efractive}\mathrm{i}\mathrm{ndex}u\mathrm{nit\; (RIU)}$, compared to $33\mathrm{nm}/\mathrm{RIU}$ for the nonannealed platform. The higher sensitivity of the annealed samples, in spite of the broad size distribution of nanoislands shown in the SEM image [Fig. 6(b)], can be explained by their sensibly larger surface area than that of particles and aggregates. The large surface area allows the binding of a larger number of molecules (linkers and biomolecules) to gold resulting in a high sensitivity.

The binding kinetics of antigen–antibody interaction was investigated by using the annealed gold nanostructures. The antibody and antigen were immobilized on the sample as explained in Sec. 2.2. The binding time as function of shift of LSPR peak is investigated in this study. The experimental results (Fig. 10) carried out on four samples for various time intervals show that around 30 to 40 min are required to complete the antigen—antibody interaction. After 40 min, the shift of LSPR peak remained constant.

Fig. 10

The shift of LSPR band over the binding time of the antibody ($100\mathrm{ng}/\mathrm{ml}$) and antigen ($40\mathrm{ng}/\mathrm{ml}$). (Error bar represents the standard deviation of 4 measurements.)

The change of LSPR spectra corresponding to each step in the bio-sensing procedure are recorded (Fig. 11) as explained in the Sec. 2.2. A shift of 10 nm was obtained when the concentration of antigen was $40\mathrm{ng}/\mathrm{ml}$. The same experiment was repeated for various concentrations of the antigen. Figure 12 shows the shift in LSPR band versus the concentration of the antigen. It can be seen that there is quite a large red shift of the band due to the antibody-antigen recognition event. For the sensing experiments, only the most concentrated ($45\mathrm{\mu g}/\mathrm{ml}$) gold solution was used, which resulted in high density nanoisland samples. Sensing experiments have been performed with concentrations of antigen in the range of 5 to $1000\mathrm{ng}/\mathrm{mL}$ and a calibration curve was established. The calibration curve (Fig. 12) shows that the shift of AuLSPR band ($\mathrm{\Delta }\lambda$) is proportional to the concentration in the range of 1 to $150\mathrm{ng}/\mathrm{mL}$ (with $0.2\mathrm{nm}/\mathrm{ng}/\mathrm{mL}$ sensitivity). At higher concentrations, the shift is still linear but the sensitivity is considerably lower. However, for practical purposes, the low concentrations range (high sensitivity) is important. The detection limit for bST, by using a nanoisland film was found to be $5\mathrm{ng}/\mathrm{mL}$.

Fig. 11

Spectra corresponding to a concentration of $40\mathrm{ng}/\mathrm{ml}$ bST (antigen). Annealed gold nanostructure (550°C) (a), antibody was adsorbed on the sample (b) and the antigen was adsorbed on the gold holding the antibody (c). The shift due to antibody-antigen interaction is 10 nm. For clarity, the spectra corresponding to different intermediate states (functionalized gold nanoparticles, activated carboxyl group, etc.) are omitted.

Fig. 12

Sensing curve: Shift in the wavelength ($\mathrm{\Delta }\lambda$) versus antigen concentration. Inset: the sensitivity for low concentration of polypeptides between 1 and $100\mathrm{ng}/\mathrm{ml}$. Error bars represent the standard deviation of 10 measurements.

3.3.

Sensing Experiments in a Microfluidic Device

The microfluidic device was used instead of the cuvette in the spectrophotometer for the sensing experiments carried out in the device. A similar device, without nanoparticles was used as reference cell for the measurement. In this setup (Fig. 13), the light beam from a UV-visible source is splitted into two parallel beams and the sensing and the reference devices were placed vertically in the light paths. The sensing procedure as explained in Sec. 2.1 was repeated with the device, and all the sensing steps were carried out by pumping the solutions into the microfluidic channel by using the syringe pump.

Fig. 13

Experimental setup of PDMS/glass nano-integrated micro-bio-sensor.

The absorbance spectra were recorded with the microfluidic device (Fig. 14). The curve in Fig. 14(a) shows the Au LSPR band before starting the sensing experiment. The Au-LSPR band was recorded [Fig. 14(b)] after absorbing the antibody of bST on the functionalized nanoparticles through pumping the solution in the microfluidic channel. The curve in Fig. 14(c) shows the LSPR band upon introducing the corresponding antigen. A shift of LSPR band of around 10 nm is recorded upon the interaction of the antigen ($40\mathrm{ng}/\mathrm{ml}$) and antibody ($100\mathrm{ng}/\mathrm{ml}$). The shift of LSPR band is the same as the sensing results obtained on the glass sample (Fig. 11). Several experiments (3 to 4 measurements) were carried out in the device and they confirmed the reproducibility of the method (Fig. 15). The sensing curve of gold nanoisland structure on glass samples (Fig. 12) is also included in the Fig. 15 for comparison.

Fig. 14

Sensing results in the device.

Fig. 15

Sensing curve for the microfluidic device.