2.1.

Materials

A SU-8 50 photoresist, developer, and SU-8 thinner were purchased from MicroChem Inc. and used as received. Crystal Violet (CV) was purchased from Sigma Aldrich. Polystyrene nanosphere (PNS) was obtained from Corpuscular Inc. Sodium dodecyl sulphate (SDS), sodium citrate dihydrate, and hydrogen tetrachloroaurate (III) hydrate $\left(\mathrm{H}\mathrm{Au}{\mathrm{Cl}}_4\right)$ were purchased from Sigma Aldrich Inc.

2.2.

Experiment Setup

The schematic for the experiment setup is shown in Fig. 1 . Briefly, a $17\text{-}\mathrm{mW}$ He–Ne $632.8\mathrm{nm}$ laser (Thorlab) was attenuated to $\sim 5\mathrm{mW}$ using a neutral density filter (Edmund Inc.). A set of lenses, which acts as both a beam expander and spatial filter, was used to produce a $7\text{-}\mathrm{mm}$ (Ø) beam of uniform profile. The so-obtained beam was then focused onto the microchannel in the microfluidics via a dichroic mirror and through an Olympus $40\times$ , $0.90\mathrm{NA}$ microscope objective. The laser power at the sample was measured to be $3.5\mathrm{mW}$ . Raman signals generated from the microfluidics were collected by the same objective and focused into a $400\text{-}\mu \mathrm{m}$ optical fiber (Ocean Optics, Inc.), which delivered the signals to a single-stage monochromator (DoongWo, Inc.). Grating used in this study was $600\mathrm{g}∕\mathrm{mm}$ , and the CCD detector (ANDOR Inc.) operating temperature was set at $-60°\mathrm{C}$ . A syringe pump (BARUN Inc.) was used to allow for delivering the fluidic sample into the microfluidics at a constant flow rate of $0.1\mathrm{ml}∕\min$ . Waste materials from the microfluidics were collected in a beaker. An ANDOR software was used to acquire the Raman spectra as well as to control the spectrometer. Microsoft Excel was used to process the acquired spectra.

Fig. 1

Schematic diagram of the experiment setup.

2.3.

Fabrication of SU-8 Microfluidics

The procedure employed in the current study for fabricating SU-8 microfluidics was similar to those published previously but with slight modifications.^{32, 33} Briefly, the SU-8 50 photoresist was first mixed well with the SU-8 thinner (MicroChem) at a ratio of 45:1. The mixture was then centrifuged at $500\mathrm{rpm}$ for $1\min$ to remove air bubbles. The addition of SU-8 thinner is necessary in order to reduce the viscosity and to aid in uniform spreading of the SU-8 50 photoresist during spinning. To produce a thin SU-8 layer, the prepared mixture was pipetted onto a cleaned $22\times 22\mathrm{mm}$ coverslip and made to spread out over the glass surface with a pipette tip. The sample was then subjected to spinning at $2000\mathrm{rpm}$ for $20\mathrm{s}$ to produce a SU-8 layer of uniform thickness. Prior to the UV-exposure, the sample was baked for $5\min$ at $65°\mathrm{C}$ . Microchannel pattern to be fabricated on the SU-8 layer was printed in black using a laser printer (HP Colourjet) on a $22\times 30\mathrm{mm}$ transparency, and transferred to the photoresist layer by placing the transparency between a UV-light source and the baked SU-8 layer. Upon UV exposure for $10\mathrm{s}$ , regions outside the shadow of the printed pattern polymerized, while areas defined by the image of the pattern did not and could be dissolved out with the SU-8 developer (MicroChem). The resultant microfluidics is shown in Fig. 2a . Microchannels derived in this manner were $45\mu \mathrm{m}$ in depth and $216\mu \mathrm{m}$ wide as indicated by the profile plot shown in Fig. 2b. The same pattern was used throughout the experiment.

Fig. 2

(a) SU-8 50 microfluidics, (b) line profile cutting across the microchannel [along the blue line shown in Fig. 2a], showing a $216\times 40\mu \mathrm{m}$ channel, and (c) FE-SEM image of the in-channel Au-coated periodic nanostructures. (Color online only.)

2.4.

Deposition of Polystyrene Nanospheres

A periodic monolayer of carboxylated polystryene nanospheres of $\sim 397\mathrm{nm}$ in diameter was deposited in the microchannel via the technique of spin-coating. The choice of the particle size was such that the dimension of the final nanostructures would be suitable for SERS at $633\mathrm{nm}$ , which was the wavelength of the laser currently used in the experiment. Prior to spin-coating, the stock PNS solution was diluted to 2.5%, after which 0.16% of sodium dodecyl sulfate was added to minimize surface tension. SDS was used to assist uniform spreading of the nanospheres in the channel during spinning. About $10\mu \mathrm{l}$ of the PNS solution was applied onto the microchannel each time. Different spin conditions have been tried, and the resultant distributions of the particles in the channel were examined under a field-emission scanning electron microscope (FE-SEM). It was found that the formation of PNS monolayers was significantly impeded by the 3-D structure of the microchannel, with no close-packed monolayer formed in most cases. Anyhow, an optimal spin condition under which a monolayer of hexagonally packed particles formed was eventually obtained: Spin $\text{speed}=2000\mathrm{rpm}$ , spin $\text{time}=20\mathrm{s}$ .

2.5.

Development of Metal Nanostructures

The monolayer of polystyrene spheres was overcoated with $100\text{-}\mathrm{nm}$ Au layer by sputtering under a pressure of ${10}^{-6}\text{Torr}$ . The finished product is checked under FE-SEM [see Fig. 2c] and then stored in a dry environment until needed.

2.6.

Encapsulation of SU-8 Microfluidics

Prior to a SERS experiment, the SU-8 microfluidics bearing the SERS-active metal nanostructures was encapsulated, as depicted in Fig. 3a , using a piece of parafilm as the sealer. Two holes matching, in position, the input and output port of the microfluidic pattern were made in the top plastic slab to allow access of liquid into the channel. The encapsulated SERS microfluidics was then placed under the microscope objective as shown in Fig. 3b for SERS measurements. Silicon tubing of $\sim 2\mathrm{mm}$ diam was used to deliver the sample into and out of the microfluidics.

Fig. 3

(a) Encapsulation of the SU-8 50 microfluidics and (b) Raman measurements on encapsulated microfluidics. Microscope Objective: $40\times$ , 0.90 NA. Laser power at the sample: $3.5\mathrm{mW}$ . Two silicon tubings are used to deliver the sample into and out of the microfluidics.

2.7.

Preparation of Au Colloid

A $15\text{-}\mathrm{nm}$ Au colloid was synthesized by following procedures described by Grabar .³⁴ Briefly, $25\mathrm{mg}$ of hydrogen tetrachloroaurate (III) hydrate powder was added to $200\mathrm{ml}$ of distilled water and the mixture was brought to rolling boil on a hot plate with vigorous stirring. After which, $\sim 34.2\mathrm{mg}$ of sodium citrate dihydrate powder dissolved in $3\mathrm{ml}$ of distilled water was rapidly added to the vortex of the stirring tetrachloroaurate solution. Boiling continued for $10\min$ , during which the solution exhibited several color changes starting with yellowish then purplish and, finally, ruby-red. The final solution was stored in $4°\mathrm{C}$ until needed.

3.1.

Stability Test

As mentioned above, metal (Au or Ag) colloids were normally used as the Raman-enhancing substrates in the majority of previous SERS microfluidics.^{15, 20} The colloid can be prepared outside the system or alternatively by pumping metal salt into the system and reducing it in situ.^{15, 35} Analyte would then be introduced via a different input port and allowed to mix with the colloid inside the system. For a maximum Raman enhancement, aggregating agents were often added to induce agglomeration.³⁶ This unfortunately could bring about such issue as difficulties in maintaining a good control over the aggregation process. As will be shown later, indeed there exists an optimal aggregate size at which a maximum Raman enhancement can be obtained. This is mainly due to the fact that the plasmon peak of an aggregate correlates strongly with its overall size, and only those that can resonate at the excitation laser energy would give rise to strong Raman signals.²² Thus, it is crucial that the state of aggregation of the colloid at the point of signal accumulation be reproducibly controllable.²⁰ This can be achieved by acquiring the Raman signals in a continuous-flow condition with a flow rate precisely adjusted in such a way the agglomerating colloid forms an optimal cluster as it passes through the laser-probe volume.²⁰ Though it has been demonstrated that such an approach increases the analytical precision and improves quantitation, the volume of the required materials substantiates if a long signal accumulation time is needed [e.g., to improve the signal to noise ratio (S/N)]. Additionally, any variation in the salt content of the samples would affect the aggregation process, and the flow rate must be readjusted to maintain a maximum enhancement. This would surely complicate the analysis as the ionic strength of each sample would have to be known a priori—an inconvenient practice especially in a large-scale clinical study.

To demonstrate the adverse effect the ionic strength of a sample has on a colloidal-based SERS, a model sample consisting of $1\mu \mathrm{M}$ CV dissolved in Au colloid was used and injected into a microfludic system bearing no SERS nanostructures. The reason for choosing CV as the test analyte is its well-established SERS characteristic.²² About $1\mathrm{ml}$ of the solution was used each time and pumped into the system at a constant flow rate of $0.1\mathrm{ml}∕\min$ using the BARUN syringe pump. To simulate the effect of ionic strength, sodium chloride (NaCl) was added to the sample. There were five measurements in total, each time with a solution containing a different amount of NaCl. The NaCl concentrations used in the current study were 0, 6.5, 10.3, 34, and $65\mathrm{mM}$ . Every sample was prepared immediately before use and introduced into the microchannel within $\sim 1.5\min$ . A new microfluidic device was used each time to avoid cross-contamination. Three spectra were taken under continuous flow with a $40\text{-}\mathrm{s}$ integration time for each experiment condition. Typical CV SERS spectra are shown in Fig. 4a showing the effect of the NaCl concentrations on the SERS intensities. All spectra were agreeable with those published previously for CV.³⁵ A curve of the average peak intensity at $1200{\mathrm{cm}}^{-1}$ against NaCl concentrations is plotted in Fig. 4b. It can be clearly seen that signal strengths vary by as much as six times over the concentration range considered here. Furthermore, there also exists an optimal concentration $\left(10.3\mathrm{mM}\right)$ for which a maximum enhancement occurs. This corresponds to a specific aggregate size (see insets for transmission electron microscopy (TEM) images of the aggregates) that exhibits a plasmon peak matching the laser wavelength, thereby giving a maximum Raman-enhancing electromagnetic field between aggregated particles.

Fig. 4

(a) CV SERS spectra taken with different amount of NaCl added to the colloid solution: $\mathrm{A}=0\mathrm{mM}$ , $\mathrm{B}=6.5\mathrm{mM}$ , $\mathrm{C}=10.3\mathrm{mM}$ , $\mathrm{D}=34\mathrm{mM}$ , and $\mathrm{E}=65\mathrm{mM}$ . All spectra are background subtracted. Integration $\text{time}=40\mathrm{s}$ . (b) Plot of the peak intensity at $1200{\mathrm{cm}}^{-1}$ vs different NaCl concentrations. Inset: TEM images of $15\mathrm{nm}$ Au aggregates at different NaCl concentrations.

Next, the stability of the in-channel SERS nanostructures toward variations in the sample’s ionic strength was examined. The experiment procedure described above was again used, except that no colloid was mixed in the 1 μM CV solution. Typical SERS spectra obtained with this setup is shown in Fig. 5a and the plot of the average $1200\text{-}{\mathrm{cm}}^{-1}$ peak intensity in Fig. 5b. As expected, the immobilized nanostructures provide a much better reproducibility with a signal variation of $\sim 20\%$ over the entire concentration range compared to 137% in the colloid-based system. Furthermore, the SERS signal derived from the nanostructure bearing microfluidics remains virtually unchanged over a very long period of time $\left(10\min \right)$ . By virtue of triplicate measurements, it was estimated that signals obtained from the nanostructures can be reproduced to within 11% at a given NaCl concentration, in agreement with those previously published for a periodic SERS-active substrate.^{37, 38} Last, mention should be made about the huge errors $(\sim 30\%)$ seen in a stop-flow microfluidic system as reported previously,²⁰ which is a result of inhomogeneous hot-spot distribution within the solution. The smaller error seen in the current system thus implies a uniform hot-spot distribution within the periodic substrate.

Fig. 5

(a) CV SERS spectra taken with different amount of NaCl added to the CV solution using microfluidics bearing the periodic nanostructures. $\mathrm{A}=0\mathrm{mM}$ , $\mathrm{B}=6.5\mathrm{mM}$ , $\mathrm{C}=10.3\mathrm{mM}$ , $\mathrm{D}=34\mathrm{mM}$ , and $\mathrm{E}=65\mathrm{mM}$ . All spectra are background subtracted. Integration $\text{time}=40\mathrm{s}$ . (b) Plot of the peak intensity at $1200{\mathrm{cm}}^{-1}$ vs different NaCl concentrations.

3.2.

SERS Study of Urine

For a real-world test, one author donates his urine sample for SERS-microfluidic analysis. The choice of a urinal fluid is twofold. First, a Raman study of urine offers a convenient tool for disease diagnosis.^{23, 39, 40} For example, detection of urinal creatinine using Raman can provide information about muscular dystrophy, hyperthyroidism, and poliomyelitis.⁶ Second, the high salt content as well as the large fluctuations in the ionic strength of urine offers a perfect challenge for biofluid analysis using a SERS microfluidics.⁴¹

First, a colloid-based system was used. Initially, mixing the urine sample with the Au colloid in a 1:5 ratio results in rapid aggregation and no signal can be obtained. A second trial employing a lower urine-to-Au ratio (1:10) produces a lower aggregation rate, allowing the temporal changes of the urine SERS spectrum to be monitored. Shown in Fig. 6a are urine SERS spectra acquired at different time points over a period of $14\min$ . The S/N of the spectra is somewhat poor due to the dilution of the urine sample. Nonetheless, Raman peaks corresponding to urea $\left(1000{\mathrm{cm}}^{-1}\right)$ , uric acid $\left(442{\mathrm{cm}}^{-1}\right)$ , Creatinine (690, 730, and $1082{\mathrm{cm}}^{-1}$ ) based on previous study,⁴² can be identified in the 560- and $680\text{-}\mathrm{s}$ spectra. At $880\mathrm{s}$ , however, the spectrum diminishes due to the formation of very large aggregates, which have a plasmon peak longer than the laser line. The fact that the signal undergoes a continuous change—peaks at 560s followed by a decay—suggests the instability of the colloid-based system toward the high ionic strength in the urine.

Fig. 6

(a) Temporal changes in the urine SERS spectra obtained using a colloid-based microfluidics system. Integration $\text{time}=40\mathrm{s}$ for each spectrum. (b) Urine SERS spectrum (upper curve) obtained using microfluidics bearing periodic nanostructures, and unenhanced Raman (lower curve) of the urine sample. Integration time for the urine SERS spectrum is $40\mathrm{s}$ , and for the unenhanced Raman $200\mathrm{s}$ . Note the intensity of the unenhanced Raman spectrum has been reduced by five times from the original data in order to reduce the large background arising from the long integration time used to obtain the spectrum.

Next, microfluidics bearing the periodic nanostructures was used to study the urine sample. About $2\mathrm{ml}$ of urine was loaded into the syringe and injected into the microchannel using the syringe pump at $0.1\mathrm{ml}∕\min$ . The SERS spectrum derived from this particular system is shown in Fig. 6b (upper curve). No appreciable signal change was observed over the $14\text{-}\min$ interval. This suggests the stability and robustness of the nanostructures in urine analysis. Note also the better S/N in comparison to that attainable with the colloid-based system. The Raman peak corresponding to urea is more profound in this case. A prominent peak at $1607{\mathrm{cm}}^{-1}$ is not known previously in urine; thus, band assignment for this particular peak is not possible at this moment. Although one may be tempted to attribute this particular band to the SU-8 polymer, such an assumption might not be true for the following reasons: (i) no such band was observed in the Raman spectrum of the SU-8 photoresist; (ii) the SERS signals were originated mainly from the Au-coated polystyrene spheres layer; and (iii) the laser focal spot was positioned on the Au-coated spheres layer near the center of the microchannel at some distance from the SU-8 channel wall. Further investigation is thus needed with regard to the source of the $1607{\mathrm{cm}}^{-1}$ band.

For comparison, an unenhanced Raman spectrum of urine is also shown in Fig. 6b (lower curve). The urine Raman spectrum was obtained from an undiluted urine sample using an integration time of $\sim 200\mathrm{s}$ . Note that in comparison to the SERS spectrum, very few peaks are present in the unenhanced spectrum, suggesting the sensitivity of our in-channel SERS nanostructures. Also under development is an in-channel Ag-overcoated PNS nanostructure, which should, in principle, give a better enhancement than the Au counterpart, albeit some work would need to be done to improve the shelflife of Ag-coated microfluidics.