Improving surface enhanced Raman signal reproducibility using gold-coated silver nanospheres encapsulated in silica membranes

Gold(III) chloride trihydrate, sodium citrate dihydrate, Amberlite MB-150 mixed bed exchange resin, (3-aminopropyl) trimethoxysilane (APTMS), sodium chloride (NaCl), sodium trisilicate (27%), tetraethyl orthosilicate (TEOS), silver perchlorate, sodium borohydride, and hydroxylamine hydrochloride were purchased from Sigma. Ethanol, ammonium hydroxide (NH₄OH), hydrochloric acid (HCl), and nitric acid (HNO₃) were purchased from Fisher Scientific (Pittsburgh, PA). Ultrapure water (18.2 MΩ cm⁻¹) was obtained from a Barnstead Nanopure System and used for all experiments. All glassware items were cleaned with aqua regia (3:1 HCl:HNO₃) and rinsed thoroughly with water and oven (glass) or air (plastic) dried overnight before use.

Ag@Au nanoparticles were synthesized using a seeded growth method previously described in the literature [30, 31]. Briefly, 100 mL of a 0.3 mM sodium citrate solution prepared in nitrogen-purged water was stirred on an ice bath in the dark. Freshly prepared sodium borohydride (final concentration = 1 mM) was added to the citrate solution. Next, 1 mL of 10 mM silver perchlorate was added to the solution within 2 min, and the resulting silver nanoparticle solution was stirred for 3 min. Stirring was stopped, and silver nanoparticles with average diameters of 11.5 ± 3.2 nm formed within 3 h. Next, 20 mL of water was added to 25 mL of as-synthesized Ag seeds and stirred for ∼2 min (4 °C). Fifteen mL of both 6.25 mM hydroxylamine hydrochloride and 0.465 mM gold salt were added slowly (3 mL min⁻¹) using a syringe pump. This Ag@Au nanoparticle solution was stirred for 1 h to ensure nanoparticle formation and stored at 2–4 °C until use. The concentration of these materials was estimated using a standard estimation model for the silver seeds [32], and an average diameter of 18.5 ± 2.3 nm was determined using TEM.

Silica shells on Ag@Au (Ag@Au@SiO₂) nanoparticles were synthesized via a modified Stöber method [24, 33–35]. Briefly, the pH and conductivity of 25 mL of the as synthesized Ag@Au nanoparticles were adjusted to 5 and ∼110 μS cm⁻¹ using NH₄OH and Amberlite resin, respectively. After resin removal via filtration, 129.2 μL of 1 mM APTMS was added drop-wise to the nanoparticle solution (with stirring). After 30 min, 201 μL of 2.7% sodium silicate was added slowly to the solution and stirred for 24 h. The silica shell thickness was further increased by adding ethanol (final ratio of 1 part water to 4.4 parts ethanol). After 6 h, 20 μL of 1 mM APTMS and 20 μL TEOS were added. The pH of the solution was increased to ∼11 using concentrated ammonium hydroxide. After 16 h, the Ag@Au@SiO₂ nanoparticles were centrifuged (45 min, 9383xg) three times with ethanol then three times with water. The Ag@Au@SiO₂ nanoparticles were then passed through Sephadex-50 column to remove Ag@Au nanoparticles that did not contain complete silica shells [2] and stored in ethanol until use. These samples exhibited average diameters of 48.7 ± 6.4 nm with an average silica shell thickness of 15.1 ± 6.8 nm.

Silica shells were converted into silica membranes via an internal silica etching process induced at basic pH values. Because the Ag@Au@SiO₂ nanoparticles were stored in ethanol, the samples were triply centrifuged and redispersed in water to a concentration of 3 nM. Concentrated NH₄OH was added to the solution to induce internal etching [1, 24]. The reaction was quenched by adding 100 mM HNO₃ until the solution pH was ∼6. Finally, the nanoparticles were washed 3 times in water and passed through a Sephadex-G50 column to remove defect particles. These samples exhibited average diameters of 46.1 ± 6.0 nm with an average silica shell thickness of 13.8 ± 6.5 nm.

TEM was performed using a JEOL JEM-1230 microscope equipped with a Gatan CCD camera. Samples were prepared on 400 mesh copper grids that were coated with a thin film of Formvar and carbon (Ted Pella). The nanoparticle solution was diluted in a 50% water−ethanol mixture, and ∼10 μL of the solutions were pipetted onto grids and dried. At least 200 nanoparticles were analyzed (Image Pro Analyzer) to estimate average nanoparticle diameters, and average silica shell thicknesses were determined by calculating the differences between Ag@Au and Ag@Au@SiO₂ nanoparticle diameters (error is from propagated error in these measurements).

LSPR spectra were collected using disposable methacrylate cuvettes (pathlength = 1 cm) and an ultraviolet–visible (UV–vis) spectrometer (Ocean Optics USB4000). Either deuterium or halogen lamps were used for UV and visible excitation, respectively. LSPR spectra were collected in transmission geometry every minute for 2 h (integration time = 60 msec, average = 25 scans, and boxcar = 10), and extinction maximum wavelengths (λ_max) were determined from the zero-point crossing of the first derivative. SERS spectra were collected simultaneous to the LSPR measurements at a 90° angle from the UV–vis light sources. SERS measurements were performed using 6 nM Ag@Au@SiO₂ nanoparticle solutions prepared in 10 mM phosphate buffer (pH 7.4) and 4-aminothiophenol. Prior to SERS measurements, samples were mixed and incubated for at least 1 h. SERS data were collected using a BW Tek iRaman spectrometer with an excitation wavelength (λ_ex) of 785 nm. Blank spectra were collected using cuvettes containing everything but analyte. All SERS spectra shown represent sample data minus these blank spectra.

Improvements in the direct SERS detection of target molecules using various silver or gold nanoparticles is well-established in the literature [36–45], and plasmonic solution-phase nanomaterials are widely used as SERS substrates because of their tunable LSPR and inherent structural properties (i.e., radius of curvature, shape and size tunability, etc) [25, 46, 47]. Ag@Au nanoparticles are selected as SERS substrates given their increased SERS activity relative to Au nanoparticles, greater dielectric sensitivity versus Au nanoparticles (vide infra), and superior chemical stability versus Ag nanoparticles [48]. Example TEM images of some Ag@Au solution-phase SERS substrates is shown in figure 1(a). For instance, Ag nanoparticles (diameter, d = 11.5 ± 3.2 nm) are coated in a thin Au shell (thickness = 3.5 ± 3.9 nm) to promote larger SERS intensities versus yet promote improved chemical stability in basic conditions from the Ag cores. These materials can be stabilized in silica shells to provide controlled electromagnetic coupling between nanostructures. The sample shown in figure 1(a2) represents Ag@Au@SiO₂ nanoparticles (total diameter = 48.7 ± 6.8 nm, silica shell thickness = 15.1 ± 6.8 nm). These samples can be converted into internally etched Ag@Au@SiO₂ nanoparticles (figure 1(a3), total diameter = 46.1 ± 6.0 nm with an average silica thickness of 13.8 ± 6.5 nm).

Zoom In Zoom Out Reset image size

Importantly, satisfying both the chemical and electromagnetic enhancement mechanism and their distance dependencies are critical for achieving large SERS signals [49]. For this reason, SERS is used to improve detection; however, molecules must diffuse toward the plasmonic nanostructured surface and interact at short separation distances. Surface functionalization can be used to promote selective interactions between molecules and the nanostructured surface. For instance, molecular adsorption can be promoted using specific surface chemistry, which facilitates binding and detection [23]. Silica, for instance, is a structurally versatile, biocompatible, and optically transparent protective material which can be subsequently modified with a variety of different chemical functionalities. If these layers are porous, target molecules can interact with the metal through the porous regions of this layer and be detected using SERS [2].

An example of SERS detection of 4-aminothiophenol on Ag@Au nanoparticles stabilized by (1) ions, (2) microporous silica shells, and (3) internally etched silica membranes is shown in figures 1(b) and (c). Ion stabilized nanoparticles reveal the largest SERS signals (figure 1(b1)); however, these signals vary rapidly with time. This occurs as molecules bind to the metal surface and destabilize the materials and impacts the LSPR and SERS spectral properties (figure 1(c1)). In contrast, silica shell-stabilized Ag@Au nanoparticles (figures 1(a)/(b)/(c2)) are stable yet exhibit no SERS activity because the analytes do not interact with the large electric fields near the metal surface [2, 49]. Finally, silica membrane stabilized Ag@Au nanoparticles (figures 1(a)/(b)/(c3)) exhibit time dependent SERS signals, which are smaller than those observed with ion stabilized nanoparticles because the silica membranes prevent the formation of hot-spots between nanostructures, but the signals stabilize with time once the nanoparticle surfaces are saturated by molecules [1].

While membrane-stabilized nanoparticles result in predictable SERS signals as a function of time once surface saturation occurs, silica membrane formation depends on silica cross-linking and density [4]. Conditions such as pH, temperature, storage conditions, and storage time impact the formation of the silica membranes, which are important requirements for SERS [1, 4]. As previously reported for silica membrane stabilized Au nanoparticles, considerations of synthesis and storage parameters are important to achieve silica coated nanoparticles with similar silica cross-linking. Although attempts are made to achieve consistent silica with uniform refractive indices, some variations in silica cross-linking can still exist because of the amorphous nature of the silica shell.

As a result, the optical properties of the Ag@Au nanoparticle cores can be used to characterize the dielectric properties of the silica shells and membranes. To do this, well-established dielectric models can be utilized [4, 50, 51] using the LSPR properties of these materials. First, LSPR spectra are collected for Ag@Au and Ag@Au@SiO₂ nanoparticle samples are collected in water and varying bulk refractive index. Example LSPR spectra for these two samples are shown in figure 2(a). As expected, the presence of the silica shell induces a 14 nm red shift in the extinction maximum wavelength (λ_max) upon shell formation.

Zoom In Zoom Out Reset image size

These samples can be rinsed and redispersed in solvents with known bulk refractive indices. This is shown in figure 2(b). To achieve varying bulk refractive indices, sucrose solutions varying from 0 to 85% (w/v) and 2 nM nanoparticle solutions are used. After a 30 min incubation period, an increase in bulk refractive index causes the extinction maximum wavelengths for both samples to red shift. Equation (1), which describes the LSPR wavelength shift response (Δλ_max), can be used to describe these data as follows:

$$\begin{eqnarray}&&\Delta {\lambda }_{{\rm{max}}}=\;{m}_{1}\left({n}_{{\rm{eff}}}-{n}_{{\rm{water}}}\right)-{m}_{2}{\left({n}_{{\rm{eff}}}-{n}_{{\rm{water}}}\right)}^{2},\end{eqnarray} \tag{ 1 }$$

where m₁ and m₂ are the linear and nonlinear refractive index sensitivity terms that account for linear and nonlinear changes in the local electric fields near the Ag@Au nanoparticle surfaces, n_eff is the effective local refractive index, and n_water is the refractive index of the water 1.33₃.

Both qualitative and quantitative differences are observed between LSPR spectra collected at various sucrose concentrations. For instance, figure 2(b) shows that increasing bulk refractive index from 1.33 to 1.47₅ of ion stabilized Ag@Au and Ag@Au@SiO₂ nanoparticles (silica shell thickness = 15.1 ± 6.8 nm) red-shifts the maximum LSPR wavelengths nonlinearly relative to initial λ_max of Ag@Au nanoparticles in water. From these data, the refractive index sensitivity is easily estimated using equation (1). Second, the linear and nonlinear refractive index sensitivities for Ag@Au nanoparticles are estimated as 170 and 360 nm/RIU, respectively. Third, the effective refractive index of the local condensed silica (including voids) around Ag@Au nanoparticles can be estimated using the intersection of the refractive index sensitivity responses as shown in figure 2(b). As such, the silica refractive index is found to be 1.45₈, a value that is consistent with silica condensed on Au nanoparticles synthesized using a similar synthetic protocol [4]. Finally, the characteristic LSPR decay length (l_d) [51] for Ag@Au nanoparticles can be estimated as follows [4, 50]:

$$\begin{eqnarray}&&{\rm{\Delta }}{\lambda }_{{\rm{max}}}=\;{m}_{1}\left[\left({n}_{{\rm{silica}}}-{n}_{{\rm{water}}}\right)\left(1-{e}^{-\displaystyle \frac{2a}{{1}_{{\rm{d}}}}}\right)\right]\\ &&\quad \quad \quad -{m}_{2}{\left[\left({n}_{{\rm{silica}}}-{n}_{{\rm{water}}}\right)\left(1-{e}^{-\displaystyle \frac{2a}{{l}_{{\rm{d}}}}}\right)\right]}^{2},\end{eqnarray} \tag{ 2 }$$

where a is the silica shell thickness, n_silica is the refractive index of silica, and n_water is the refractive index of water. As such, the LSPR decay length is estimated at ∼11 nm and can be subsequently used to quantify the dielectric characteristics of the silica membrane on Ag@Au@SiO₂ nanoparticles.

In addition, the effective refractive index (n_eff) of the silica-containing layer surrounding Ag@Au@SiO₂ nanoparticles can be estimated so that the LSPR properties of the metal cores can be correlated to effective silica density. First, 3 nM silica shell stabilized Ag@Au nanoparticles are immersed in 1.5 M NH₄OH for varying periods of time to induce membrane formation. The reaction is quenched by the addition of acid, and the samples are rinsed and redispersed in water. Example TEM images and LSPR spectra for three internally etched Ag@Au@SiO₂ nanoparticle samples are shown in figures 3(a) and (b), respectively. As expected, membrane formation is visible from varying silica contrast near the metal surfaces in the TEM images and results in systematic blue-shifts in the LSPR spectra.

Zoom In Zoom Out Reset image size

Next, equation (3) can be used to relate nanoparticle morphology to the optical properties of the internally etched Ag@Au@SiO₂ nanoparticles. The relationship between n_eff and the void near the surface of the metal nanoparticles is related as follows:

$$\begin{eqnarray}&&\;{n}_{\text{eff}{\rm{\ }}}={n}_{{\rm{void}}}\left(1-{e}^{-\displaystyle \frac{2V}{{l}_{{\rm{d}}}}}\right)+{n}_{{\rm{silica}}}\left({e}^{-\displaystyle \frac{2V}{{l}_{{\rm{d}}}}}-{e}^{-\displaystyle \frac{2S}{{l}_{{\rm{d}}}}}\right)\\ &&\quad \quad \;+{n}_{{\rm{water}}}\left({e}^{-\displaystyle \frac{2S}{{l}_{{\rm{d}}}}}\right)\;\;,\end{eqnarray} \tag{ 3 }$$

where n_void is the refractive index of the void volumes in the silica shell, V is the void thickness, and S is the silica membrane thickness (i.e., total silica shell thickness—void thickness). The observed LSPR wavelength shift data in figure 3(b) are then related to effective refractive index around the Ag@Au cores. This result is shown in figure 3(c). As expected, shifts in the λ_max increase in magnitude as etching progresses and voids around the nanoparticle cores form. As the effective refractive index decreases from 1.45₈ (silica) to 1.33₃ (water), an 18 nm blue-shift is clearly observed for internally etched Ag@Au@SiO₂ nanoparticles. By combining these straight-forward LSPR measurements, TEM images, and dielectric model, the effective refractive index for the porous silica in figures 3(a1)–(a3) are estimated at 1.44₈, 1.40₈, and 1.36₂, respectively. The impact of the effective refractive index and as a result, effective silica density are expected to impact the diffusion of molecules near the metal nanoparticle core [1], which should influence the size of the SERS-active volume near the metal surface and as a consequence, the magnitude of the SERS signal of small molecules (vide infra).

SERS signals arise from molecules occupying the volume nearest to the metal substrate [25] and is often approximated in terms of distance dependence [3]. For internally etched Ag@Au@SiO₂ nanoparticles, the silica membrane and interior void is the matter that occupies the SERS-active volume and facilitates molecular diffusion through the silica shells for SERS detection. As such, SERS intensities are expected to increase as the effective refractive index decreases. To evaluate this hypothesis, nine internally etched Ag@Au@SiO₂ nanoparticle samples are prepared so that the effective refractive index varies from 1.36 to 1.46. The nanoparticle samples are diluted to final concentrations of 6 nM and incubated with 30 μM 4-aminothiophenol for at least 1 h. SERS signals were collected, and the CS/CC stretching frequency centered at 1079 cm⁻¹ is plotted as a function of effective refractive index. These data are summarized in figure 4 and reveal that SERS intensity increases with decreasing effective refractive index (i.e., increasing void volume). As the effective refractive index decreases from 1.45₈ to 1.33₃ nm, Si–O–Si bonds in the polymer matrix are broken thereby generating silica voids near the metal surface, which can then be occupied by molecules or solvent. Of note, decreasing the effective refractive index from 1.44 to 1.38 shows significant changes in SERS intensities. As more molecules occupy the internal silica volume, SERS signals increase. Above an effective refractive index of 1.48, insufficient internal volume is available thereby hindering molecular diffusion and as a result, reduces SERS intensities. When the effective local refractive index is less than 1.38, SERS intensities saturate and are limited by monolayer formation of 4-aminothiophenol on the Ag@Au nanoparticle surfaces. Because small changes in effective refractive index near the metal core impact SERS intensities, these data suggest that SERS signals are reproducible if effective refractive index of the local environment of the local SERS-active volume surrounding the SERS substrate is considered.

Zoom In Zoom Out Reset image size

If SERS signals from analytes depend on the silica-free surface area and volume near the Ag@Au nanoparticles, SERS signals should vary as a function of analyte to molecule concentration. To understand how molecular concentration impacts SERS signals, 6 nM internally etched Ag@Au@SiO₂ nanoparticles were incubated with 4-aminothiophenol concentrations ranging from 1 to 30 μM for at least 1 h in 10 mM phosphate buffer (pH 7.4). The effective refractive index is maintained at 1.38 to ensure consistent internal void volumes. Representative SERS spectra are shown in figure 5(a) and reveal that the unique molecular vibrational modes for 4-aminothiophenol centered at 1173 cm⁻¹ (CH bend, CC stretch), 1079 cm⁻¹ (CC and CS stretches), and 393 cm⁻¹ (CS bend) [52–55] are consistent with the a₁ vibrational modes for 4-aminothiophenol and the formation of a monolayer of 4-aminothiophenol on the metal surface.

Zoom In Zoom Out Reset image size

The largest SERS intensity in each spectrum is observed for the CC/CS stretching transitions, which is consistent with molecules attached to the metal surface through sulfur atom and enhanced by the strong electromagnetic fields near the metal surface [52–54]. Of note, only a₁ symmetric vibrational modes are observed for 4-aminothiophenol, which suggests that the molecules are oriented on the metal surface as a loosely packed monolayer with an average tilt angle of ∼30° [56, 57].

Previously, silica membranes were shown to prevent plasmon coupling between gold nanoparticle cores [1, 2, 4], and similar observations are observed here for internally etched Ag@Au@SiO₂ nanoparticles. Thus, SERS signal changes are a consequence of local 4-aminothiophenol concentration inside the silica membrane and are expected to provide quantitative molecular detection. Because SERS signals are directly proportional to the number of molecules within the localized SERS-active volume near the metal surface, saturated SERS signals should be limited by 4-aminothiophenol monolayer formation. To evaluate these trends, SERS intensities for CC stretch/CH bend, CS/CC stretch, and CS bend modes are plotted as a function of 4-aminothiophenol concentration and 6 nM internally etched Ag@Au@SiO₂ nanoparticles with an effective refractive index of 1.38 (1 + hour incubation).

As shown in figure 5(b), the SERS intensities for each vibrational frequency systematically increase then saturate after the addition of ∼20 μM 4-aminothiophenol. First, the CC/CS stretching and CC stretching/CH bending modes, which are orientation independent, represent SERS signals from all molecules that occupy SERS-active volume inside each silica membrane. Second, the CS bending mode centered at 393 cm⁻¹ is consistent with molecules bound to the metal surface [58–60]. Finally, because of the absence of b₂ vibrational modes, 4-aminothiophenol signals are limited to molecules inside the silica shells in the absence of strong electric fields induced by hot-spots [55, 61], and the effective molecular footprint of molecules are estimated at 0.535 nm²/molecule [57]. As such, a monolayer of 4-aminothiophenol would be achieved if ~2000 molecules/nanoparticle were added to solution (assuming an average Ag@Au nanoparticle diameter = 18.5 nm). For a 6 nM nanoparticle solution, this equates to a 12 μM 4-aminothiophenol concentration. As a result, any molecular concentration below 12 μM should provide a concentration dependent response while concentrations above this value should be concentration independent. This is observed in figure 5(b). These results demonstrate that quantitative and reproducible SERS detection is possible once the effective refractive index surrounding internally etched Ag@Au@SiO₂ nanoparticles as well as metal surface area are considered.

Finally, SERS signals are evaluated as a function of internally etched Ag@Au nanoparticle concentration. The concentration of internally etched silica coated Ag@Au nanoparticles with an effective refractive index of 1.38 were varied from 0 to 10 nM and incubated with 30 μM 4-aminothiophenol for at least 1 h to ensure that a molecular excess was added to solution. Representative SERS spectra are shown in figure 6(a) and show a systematic increase in SERS intensity with increasing nanoparticle concentration. The SERS intensity for the CS\CC stretch at 1079 cm⁻¹ is plotted as a function of internally etched Ag@Au@SiO₂ nanoparticle concentration and increases linearly over the concentration range studied (figure 6(b)). Above this concentration, signals and background vary significantly from measurement to measurement and are attributed to detector limitations and/or light throughput. This suggests that nanoparticle concentration directly impacts SERS signals, and can be used to lower detection limits as well as to expand the dynamic range for quantitative SERS detection of small molecules.

Zoom In Zoom Out Reset image size