The silica microtoroids were fabricated using 2 μm thermal oxide on Si wafers (Montco Silicon, P-doped), following a well-established, three-step process: oxide patterning, silicon dry-etching, and laser reflowing [48]. First, the silica surface was lithographically patterned with 100 μm diameter circular pads using S1813 photoresist (Shipley). The oxide was then etched using Buffered Oxide Etch (Transene). Second, the silicon was isotropically dry-etched using a custom-built XeF₂ etching system, forming a microdisk. Finally, the silica microdisks were reflowed to form toroids using a 50 W CO₂ laser (Synrad).

All anhydrous solvents used during device functionalization were stored under UHP Ar. All chemicals and other solvents for the following reactions were used as received from their respective suppliers. The silica microtoroids fabricated following the above procedures were functionalized via three primary reaction steps: surface hydroxylation, amination, and finally, biotinylation. First, the silica surface was terminated with hydroxyl groups via typical literature procedures to promote covalent attachment of the silane coupling agent, using either O₂ plasma treatment (120 W, 200 mTorr, 2 minutes) or a 70:30 H₂SO₄ (fuming, Aldrich): H₂O₂ (30 wt %, Aldrich) piranha etch for 45 minutes [68–70]. Following the piranha etch, the samples were cleaned with DDI H₂O, and air-dried between 25–60°C for 30 minutes–24 hours to completely remove the water from the sample surface. Second, aminated silica surfaces were prepared by grafting silane coupling agents to the surface using two different methods: organic solvent deposition or chemical vapor deposition [41,67,71,72]. In the first method, a 2 mM solution of 3-aminopropyltrimethoxysilane (APTMS, Aldrich) was prepared in anhydrous toluene in a glove bag under a UHP Ar environment. The sample was then placed in the solution, and reacted at room temperature on an incubating rocker (VWR) for 10 minutes–4 hours to form a monolayer of silane on the surface. The sample was then removed from the reaction solution, and washed with toluene, ethanol, and then water for 5 minutes each on the incubating rocker. The sample was again air-dried between 25–60 °C for 30 minutes–24 hours to completely remove the water from the sample surface. In the second method (chemical vapor deposition), the hydroxylated sample was exposed to APTMS vapor in a vacuum desiccator under aspirator vacuum for 15 minutes–4 hours. Lastly, the aminated samples were then biotinylated via reaction in a 10 mM solution of NHS-Biotin (Pierce) in dimethylsulfoxide (DMSO, anhydrous, Aldrich) at room temperature for 30 minutes, on an incubating rocker. The surface was then washed with water and then acetone, as above, to remove any physically absorbed biotin from the surface, and air-dried at room temperature for 30 minutes.

Amine-terminated samples were fluorescently labeled using fluoroscein-5-isothiocyanate (FITC, Pierce), which forms a stable thiourea bond with the primary amines present on the surface in a 1:1 ratio. The FITC reaction solution was prepared in a darkened room by dissolving 1 mg FITC in 1 mL DMSO (anhydrous), and diluting it via addition of 50 μL of the solution to 1 mL of 0.1 M sodium carbonate buffer at pH 9.0 (VWR). The samples were gently inserted into this reaction solution in covered vials, placed in an ice bath, and reacted for at least 8 hours in an incubating rocker. Physically adsorbed FITC was removed via rinses with DMSO. The samples were then air-dried at 25–60 °C for 30 minutes–24 hours. Biotin-terminated samples were fluorescently labeled with Texas Red fluorescent dye conjugated to avidin protein (Invitrogen). Biotin-avidin binding was accomplished by reacting the biotinylated samples in a covered, 10 μg/mL solution of Texas Red-avidin protein conjugate in phosphate buffered saline (PBS, VWR) for 30 minutes at room temperature in an incubating rocker. The sample was removed from the solution and washed with PBS to remove physically-adsorbed reactant.

The as-fabricated and surface-modified microtoroid devices were characterized qualitatively using a Nikon H550S optical microscope equipped with a Nikon digital camera, and NIS-Elements imaging software. The surface chemistry of these devices was explored via X-ray photoelectron spectroscopy (XPS, M-Probe ESCA) using a 1,487 eV Al Kα source, and a survey scan from 0 to 1,000 binding eV to identify all chemical species on the surface. Reaction efficacy and quality was monitored using fluorescent imaging with a Nikon ECLIPSE LV100D-U fluorescence microscope equipped with a Nikon digital camera at each reaction step. Intensity measurements were obtained using NIS-Elements imaging software and Texas Red and FITC excitation and emission filters, where appropriate. Lastly, ellipsometric measurements were taken using a Gaertner L166C ellipsometer to determine the film thickness of the initial silane layer. For the XPS and ellipsometry techniques, control samples consisting of surface-modified, non-patterned, silica-on-silicon wafers were used in place of the patterned samples.

The devices were characterized quantitatively by microcavity analysis at each reaction step to determine the impact of the surface functionalization methods on the device sensitivity and to evaluate the bioconjugation conditions best suited to ensuring the devices’ performance. The resonator quality factor was measured using a tapered optical fiber waveguide to couple power into the devices from a narrow linewidth, CW tunable diode laser centered at 635 nm (New Focus) (Figure 3b) [73–75]. The F-SV optical fiber (Newport) was tapered to ∼500 nm waist diameter by heating with an oxyhydric torch while stretching the fiber with a two-axis stage controller (Sigma Koi). During testing, the device was placed on a 3-axis nanopositioning stage (Optosigma), and the device was monitored using side- and top-view cameras simultaneously. Coupling into the resonator results in the excitation of the whispering gallery modes of the microcavity (Figure 3c). The resonance linewidth data was recorded using a digitizer/oscilloscope card which was directly integrated into the computer for automated data recording (NI, PCI-5114). The laser scan speed and scan direction was optimized to ensure that neither distorted the resonance lineshape. The quality factor of the microtoroid resonator was determined by calculating the resonance linewidth (full width at half-maximum) from the recorded spectra taken in the undercoupled regime [48,60,63,75–77].

Our exploration of these specific combinations grew out of the requirements of grafting organosilanes to inorganic surfaces, such as silica, alumina, titania, etc. Typically, silane coupling agents (R(CH₂)_nSiX₃, where R is an organic functional group and X is a hydrolyzable leaving group) undergo hydrolysis to labilize the leaving groups and form a reactive silanol intermediate species. This silanol species can couple to surface hydroxyl groups via condensation reactions and form stable siloxane bonds. However, the condensation reaction that couples the silane to the surface requires a high density of hydroxyl groups to form a uniform surface coverage [35,67,71,78]. Typically, the –OH surface density is increased by an oxidative piranha treatment, which leaves the surface extremely hydrophilic. This treatment, however, may be too harsh in terms of temperature and acid strength with regard to microstructured optical devices, and may lead to difficulties in preserving the structural and surface integrity of the devices. An alternative is the use of O₂ plasma, which is typically used in fabrication processes to clean photoresist from structured surfaces. O₂ plasma etching is simple, environmentally benign, minimizes device handling, and does not require post-treatment to remove excess water from the surface. On the other hand, O₂ plasma treatment may not yield as high of a surface hydroxyl density, leading to low surface coverage of the organosilane. Therefore, it is important to identify which route is most compatible with our devices.

After hydroxylation, the silane coupling agent can be grafted to the surface. Here, we used APTMS as our silane coupling agent due to its short organic tether between the silica head and the terminal amine, which should ensure that the final biotin probe is well within the evanescent tail of the functioning device. Additionally, the amine portion should not cause additional material losses due to absorption at the wavelength of interest. Surface grafting of APTMS may be accomplished via reaction in aqueous or organic solvent, or through vapor phase deposition. Typically, alkoxysilanes (where X is an alkoxy group) are used to functionalize the surface of microstructured devices, since they minimize the formation of a silica polymer network on the surface and help achieve monolayer coverage, due to the lower hydrolyzability of the alkoxy leaving groups [67]. This lower hydrolyzability can lead to the necessity of hydrolyzing the alkoxysilane in water first, as they are unreactive to surface hydroxyls at ambient conditions in their native form. Unfortunately, aqueous solvent deposition causes the same multi-layer deposition problem as working with non-alkoxysilanes. Fortunately, chloro- and methoxysilanes, such as APTMS, do not need pre-hydrolysis, and can be coupled to the substrate in a uniform monolayer through siloxane bonding via anhydrous organic solvent deposition or vapor deposition [67]. In both deposition cases, since water is not present in the system, the three-dimensional silica polymer network does not form. Either route may be appropriate for microstructured optical devices, but vapor deposition typically requires less chemical use, post-treatment, and handling than organic solvent deposition. Here, we investigate both routes, along with the hydroxylation alternatives, to determine their impact on the devices.

Figure 4, below, shows optical micrographs of typical silica microtoroids before and after hydroxylation using either piranha or O₂ plasma treatment. The primary issue with piranha treatment is that it is carried out in an aqueous environment, and upon reaction completion and device removal from the solution, water must be removed from the now hydrophilic surface. Typically, complete removal is not possible via air-drying at room temperature. Additionally, drying with a heat gun or air nozzle results in fractured toroids. To overcome this, the device-laden substrate may be placed on a hot plate, or in an oven, and the temperature raised to 35 °C. Temperatures above this result in structural damage to the toroid itself (as indicated by the white circles in Figure 4c). Typically, this damage presents itself either as thin cracks radiating outward from the pillar (thin white lines in the image), or small divots (black spots in the image) in the surface of the toroid. Either type of damage will reduce the performance of the device. In comparison, O₂ plasma treatment requires minimal device handling and no post-treatment, and does not result in surface- or structurally-damaged devices. Additionally, the time requirements for O₂ plasma treatment are significantly less than those for piranha treatment.

Figure 5 shows the ellipsometry measurement of the film thickness of the silane film resulting from amination of the surface of control wafers that had been hydroxylated via O₂ plasma treatment over various reaction times. It is interesting to note that while organic solvent deposition resulted in an approximately constant film thickness over time, increasing the time of vapor deposition resulted in non-uniform film thicknesses, increasing in an approximately exponential trend. This behavior is typical for chemical vapor deposition processes [79]. The difference in these deposition rates is due to the difference in film deposition mechanisms between chemical vapor deposition and organic solvent deposition of alkoxysilanes [80]. We refer the interested reader to Duchet et al. for a discussion of these mechanisms [80]. From this data, it is apparent that 10 minutes of chemical vapor deposition is sufficient to create a uniform monolayer, approximately 8–10 Å thick.

Figure 6 shows fluorescent micrographs of FITC- and Texas Red-labeled, aminated and biotinylated (respectively) toroids obtained using the four combinations of hydroxylation/amination procedures. The combination of O₂ plasma treatment and vapor deposition results in the most uniform surface coverage of the toroids. Interestingly, in each case, amination occurred only on the silica toroid itself, and not on the silicon wafer. This was interesting to us, as the process of hydroxylating the surface of the device and the substrate should lead to the ability to aminate both surfaces (the silica of the device as well as the silicon of the substrate). As an exploration of why this preferential amination appeared only on the silica, we hydroxylated (using O₂ plasma) and aminated three different sets of planar substrates: a bare silicon wafer, a bare silicon wafer subjected to XeF₂ etching using the same parameters as used to make the devices, and a 2 μm SiO₂ on silica wafer. The aminated samples were subjected to FITC labeling, and the average intensities of the resulting surfaces were measured, normalized by the background intensity of each sample. Interestingly, the average intensity (a.u.) of the FITC labeled surfaces showed that XeF₂ etching seems to reduce the overall amination of the silicon surface (intensity of the bare wafer: 1161, intensity of the bare wafer subjected to XeF₂ etching prior to functionalization: 364, and intensity of the thermal oxide wafer: 924). Clearly, the exposure to the XeF₂ is inducing the improved functionalization of the device surface. However, at this point, the precise mechanism is unclear.

The combination of piranha treatment and organic solvent deposition typically led to the poorest results, with severe surface clumping, indicating either a non-uniform coverage by the silane coupling agent, or the presence of water drops remaining on the surface after hydroxylation. Clumping of the silane layer during organic solvent deposition could be minimized by carrying out the reaction on a rocker; however, it was never entirely prevented, as shown by Figure 6c, which shows the O₂ plasma treatment/organic solvent deposition combination. Note that the image was deliberately overexposed to help show the detail of surface clumping and surface damage from organic solvent deposition.

The results from the series of experiments are summarized in Table 1. Overall, the difficulties involved with proper drying of the piranha-treated samples and with clumping of the organic solvent deposition samples indicated that these methods were inappropriate for further use. Note that, due to its poor performance, the piranha treatment/organic solvent deposition combination was not pursued to the point of biotinylation. Additionally, several biotinylated samples were labeled with FITC to determine the extent of the NHS ester reaction with surface amine groups; fluorescent microscopy did not yield samples with intensities above the substrate background (no fluorescent labeling occurred), indicating that the density of amines remaining after biotinylation was too low to be detected by fluorescent imaging techniques. This implies that the reaction biotinylated all the amine groups present or that steric hindrance prevented access of the FITC to the remaining amines on the surface. Regardless, the inaccessibility of the amine groups for reaction with FITC, a relatively small molecule compared to antibodies and proteins, indicates that the presence of the amine groups should lead to minimal activity towards non-targeted species in the environment.

Although fluorescent labeling provides confirmation of the presence and uniformity of the surface species, it is an indirect method to do so. An additional confirmation may be obtained using X-ray photoelectron spectroscopy (XPS), which directly probes the surface composition through irradiation of the sample with X-rays, resulting in the ejection of electrons from the top ∼10 nm of the sample. Simultaneously measuring the kinetic energy and number of electrons ejected allows us to quantify the elemental source of these electrons. Figure 7 show the resulting surface composition at each reaction step obtained through XPS data collection. The C (1s) signal that appears in the spectra is from primarily adventitious carbon in the ultra-high vacuum chamber, so direct comparison in terms of intensity of these peaks between runs is inappropriate, as the signal from the sample is masked by varying amounts of carbon in the chamber. Additionally, it is difficult to directly compare the intensities of the peaks between runs, since slight differences in spectrometer environment can lead to significant differences in peak intensities between runs (for instance, the Si (2p) and Si (2s) signals in each spectra); hence the designation of arbitrary units for the intensity data. Overall, the XPS data shown provide an additional confirmation of the success of each reaction step via the introduction of nitrogen (N) and sulfur (S) peaks into the spectra after amination and biotinylation, respectively. The combination of fluorescence imaging and XPS verify both the surface chemistry and surface quality of the samples.

To determine the stability of the biotinylated devices for long-term storage in ambient conditions, seven chips, each containing 1–5 toroids, were biotinylated at the same time (day 1), then stored for various time-frames, bound with an avidin-Texas Red complex, and subjected to fluorescent microscopy. The intensity due to the binding reaction of biotin and avidin, and how it changes over storage time, can be used as a measure of the degradation or instability of the biotin probe with respect to avidin over time. Figure 8 shows the average device intensity, normalized by toroid surface area, over approximately 1.5 months of storage. It is important to note that long-term storage does not have a significant negative impact on the stability of the probe molecules attached to the device in terms of its ability to bind with avidin, as the normalized intensity holds steady around a normalized intensity of 0.35. Therefore, these devices could be fabricated and bioconjugated in advance, and stored until they were needed for use.

The as-fabricated and functionalized microtoroids were tested at each reaction step to determine the effects of functionalization on the sensitivity of the device at 633 nm. This wavelength was chosen based on the typical sensing environment (buffer or serum) in which these devices would be used [81]. Absorption of water in the near-IR to IR is very high (for example, at 1540 nm the loss is 11.8 cm⁻¹) [82], so device sensitivity would be limited in this regime. However, at shorter wavelengths, absorption is reduced (for example, at 600 and 625 nm, the absorption of water is 0.002 and 0.003 cm⁻¹, respectively) [82]. Therefore, the 635 nm wavelength was selected for this study. Figure 9 shows the results of a quality factor study of the bioconjugated devices after each reaction step. It is interesting to note that hydroxylation leads to a slight overall decrease in Q-value; however this is recovered after amination, indicating the formation of siloxane bonds and reduction of –OH density, which has a higher absorbance at 635 nm than the siloxane groups [78]. In this case, we require the maintenance of the Q above 10⁶ after biotinylation for these devices to be used as sensors in aqueous environments, for such applications as medical diagnostics or environmental monitoring. As Figure 9 shows, 8 of the 11 biotinylated microtoroids showed resonances corresponding to Q-values above 10⁶. A typical transmission spectra of the final, biotinylated toroids is shown in Figure 10. This Q-study demonstrates the effectiveness of our bioconjugation method, and shows that this method does not adversely impact the sensitivity (Q-factor) of the devices.