Nanoclustered Gold Honeycombs for Surface-Enhanced Raman

Dec 4, 2012 - Zhenli Qiu , Jian Shu , Dianping Tang. Chemical Communications 2018 54 (52), 7199-7202. Applications. Pratima Bajpai. 2017,105-212 ...
0 downloads 0 Views 441KB Size
Article pubs.acs.org/ac

Nanoclustered Gold Honeycombs for Surface-Enhanced Raman Scattering Weinan Leng†,‡,§ and Peter J. Vikesland*,†,‡,§ †

Department of Civil and Environmental Engineering, Virginia Tech, 415 Durham Hall, Blacksburg, Virginia 24061, United States Institute for Critical Technology and Applied Science (ICTAS), Virginia Tech, Blacksburg, Virginia 24061, United States § NSF-EPA Center for the Environmental Implications of NanoTechnology (CEINT), Duke University, Durham, North Carolina 27708, United States ‡

S Supporting Information *

ABSTRACT: A honeycomb-shaped gold substrate was developed for surface-enhanced Raman imaging (SERI). The honeycombs are composed of clusters of 50−70 nm gold nanoparticles and exhibit high Raman enhancement efficiency. An average surface enhancement factor (ASEF) of 1.7 × 106 was estimated for a monolayer of L-cysteine molecules adsorbed to gold via a thiol linkage. The presence of a linear relationship in the low concentration region was observed in SERI detection of malachite green isothiocyanate (MGITC). These results together with the high reproducibility and simple and cost-effective fabrication of this substrate suggest that it has utility for applications of surface-enhanced Raman scattering (SERS) in quantitative diagnoses and analyte detection. urface-enhanced Raman scattering (SERS) is a fieldenhanced vibrational spectroscopic technique that shows great promise for the sensitive and rapid identification of molecules associated with roughened metal surfaces or metal nanoparticles.1−3 The surface enhanced Raman effect is generally explained as the combination of a dominant electromagnetic (EM) enhancement that results from localized surface plasmon resonance (LSPR) and a chemical enhancement related to charge transfer (CT) between the substrate and surface associated molecules.4−7 A current focus of SERS research is on the design of robust metal nanostructures that exhibit predictable LSPR. Nanostructure design involves control of the composition, shape, size, and interparticle spacing between nanoparticles, each of which tunes the LSPR to obtain optimized SERS substrates at a desired wavelength.8 Numerous SERS-active substrates with appropriate nanometer architectures have been developed over the years.9 They include electrochemically roughened electrodes,10,11 metallic nanoparticle arrays,12−14 and periodic ordered nanostructured substrates,8,15−19 each of which have specific applications, advantages, and disadvantages. Although Raman enhancements of up to 1014−1015 have been achieved on “hot spots” under conditions of electronic resonance,5,7 these existing approaches generally rely on LSPR developed by nanoparticle aggregates with random morphologies and thus suffer from a lack of reproducibility and the absence of simple engineering design rules. Even for highly ordered SERS substrates, molecular adsorption sites on the macroscopic substrates exhibit high variability with respect to their Raman enhancement factors.20,21 The random spatial localization of

S

© 2012 American Chemical Society

hot spots and variations in adsorption sites pose a problem for reliable quantitative measurements of trace molecules on a given substrate. Our intent with this paper is to contribute to the continued development of SERS. As such, we present a method for quantitative detection of trace analytes on an ordered SERS substrate produced by the combination of Langmuir−Blodgett (LB) methods and low-cost sputter coating. The Raman method involves XY-surface-enhanced Raman imaging (SERI) of sample spots dried on the substrate. The use of XY-imaging and the acquisition of thousands of analysis points reduces the signal variance often observed when recording single SERS spectra. Raman imaging enables systematic scans of the surface that facilitate evaluation of the spatial variability of the surface enhancement. Furthermore, imaging minimizes thermal and photo degradation of target analytes because of the continual scanning of the laser across the substrate. Low power and short exposure times are always preferable when measuring analytes on substrates, especially when a high magnitude microscope objective is employed in the measurement, as the focused beam may introduce heat into the sample, which in turn may yield conformational changes. The SERS substrates described herein were produced using the LB method followed by sputter-coating with gold and the subsequent removal of the silica spheres. The LB technique provides a flexible methodology for the fabrication of integrated Received: April 18, 2012 Accepted: December 3, 2012 Published: December 4, 2012 1342

dx.doi.org/10.1021/ac301028w | Anal. Chem. 2013, 85, 1342−1349

Analytical Chemistry

Article

LB Lithographic Preparation of SERS Substrates. Ordered silica particle films23,25,26 were obtained by the vertical lifting method using a KSV2000 LB trough at room temperature. These films were then used as a mold to create honeycomb-shaped gold films. Prior to LB deposition, glass coverslips (2.5 × 5 cm2) were cleaned with “piranha” solution (Boiling 3:1 H2SO4/30% H2O2 for an hour. Caution: Piranha solution is both strongly acidic and strongly oxidizing. It must be handled with extreme caution!), and then sonicated in a 5:1:1 H2O:30%H2O2:NH3·H2O solution for an hour to produce a hydrophilic surface. A diluted suspension of functionalized silica particles in an 80%/20% (v/v) mixture of chloroform and ethanol was prepared. Following dispersion of the solution into ultrapure water (>18 MΩ·cm), a layer of randomly distributed silica spheres remained at the interface after a 30 min solvent evaporation period. Under the compression process, an ordered layer of particles was formed at a surface pressure of 6 mN·m−1. The particle film was then deposited on the hydrophilic glass slip (initially immersed in the water at a speed of 10 cm·min−1 and then slowly pulled out at a speed of 0.1 cm·min−1). When the films were dry, a 40 nm thick layer of gold was deposited onto the silica film using a BAL-TEC SCD 005 Cool Sputter Coater with a current of 30 mA, a sputtering time of 300 s, and a working distance of 50 mm between gold target and substrate. Following gold application, the substrate was sonicated for 1−3 min in chloroform to remove the silica particles, thus creating a gold “honeycomb” on the glass surface. Finally, the substrates were rinsed with chloroform, dried by N2 and annealed using a vacuum heating system (OV-12 Vacuum oven, JEIO TECH, Inc. Korea) at 200 °C for 2 h. Instrumentation. Substrate morphology was evaluated with a Veeco MultiMode Atomic Force Microscope (Veeco Metrology, Santa Barbara, CA) using Tapping mode and a FEI Quanta 600 FEG Scanning Electron Microscope in low vacuum mode. Analysis of the AFM images was done using Digital Instruments, NanoScope IIIa software (version 5.30r3sr3). Image analysis was conducted using a 100 × 100 μm2 scan area and three randomly chosen scan areas were selected to evaluate the difference, Dif f, between the three-dimensional surface area and the two-dimensional, footprint of a given scan. The average Dif f value provides an estimate of gold surface coverage in three dimensions. The absorptivity of the gold substrate was recorded using a Cary 5000 UV−visible-NIR Spectrometer (Varian Analytical Instruments, Walnut Creek, CA). Raman measurements were collected with a WITec alpha500R instrument using either a 785 nm or a 633 nm excitation source. Sample solutions were carefully transferred onto a SERS substrate or alternatively the substrate was immersed in sample solution, dried in ambient air, and then mounted on the sample stage of the Raman spectrometer. Backscattered light was collected using 10× and 100× objectives at integration times ranging from 50 ms to 1 s. An edge filter placed in the path of the signal effectively cuts off the excitation radiation. The signal was dispersed using a 300 groove/mm or 1200 groove/mm grating and the dispersed light was collected by a Peltier cooled charge coupled device (CCD). All SERS spectra obtained from the substrates were average spectra collected over the scan area.

nanoparticle substrates that are advantageous both for their large surface area and a high degree of uniformity that facilitates high sensitivity SERS applications.16,22 In this effort, the silica particles used for LB deposition were functionalized with capping agents to make them hydrophobic and thus amenable for the LB technique. Relative to electron beam deposition, sputter-coating is relatively low-cost and can be done using instrumentation that is more widely available. SERS substrates produced by chemical synthesis or chemical assembly methods are generally covered by chemical agents that can affect the interaction between the target molecule and the nanoparticle surface.16 By using an LB lithographic method, the SERS substrates used in this work are preorganized as groups of “hot spots” that occur uniformly across the surface.23 The resulting substrates are comprised of round honeycombs with an average periodicity of 1100 nm. The combs consist of clusters of 50−70 nm gold nanoparticles formed by sputter coating and sonication. This substrate offers spatially uniform SERS activity over the visible and near-infrared spectral regions. This novel substrate should facilitate the rapid, accurate, and cost-effective development of clean SERS-based biosensors for the detection of trace levels of contaminants. In the present study, we couple the optimal gold honeycomb platform generated by LB lithography with SERI to quantitatively detect trace organic molecules in diminutive sample volumes, and we evaluate the SERS substrate for its applications in analytical sensing and imaging.



EXPERIMENTAL SECTION Chemicals and Materials. Tetraethoxysilane (TEOS, Fluka), ammonium hydroxide (28−30% in water, SigmaAldrich, ACS), ethanol (Fisher Scientific, HPLC), allyltrimethoxysilane (95%, Aldrich), chloroform (Alfa Aesar, HPLC), and L-cysteine (≥98%, Sigma) were reagent grade and used as received. Malachite green isothiocyanate (MGITC) was acquired from Invitrogen Corp. (Grand Island, NY). All other reagents were obtained from Sigma-Aldrich at the highest purity available. Preparation of Amphiphilic Silica Particles. Amphiphilic silica particles of average diameter (±standard deviation of replicate measurements of particle diameter obtained via ImageJ analysis) 1100 ± 50, 810 ± 20, and 360 ± 12 nm were prepared according to the literature.24−26 Ethanol solutions of TEOS were continuously added to an ammonia−ethanol solution at a rate of 8 mL/h by a single-syringe pump, and the reaction mixture was stirred for 22 h at room temperature. The particles were then functionalized by adding excess allyltrimethoxysilane (ATMOS). After 5 h, the system was boiled for 2 h to ensure covalent attachment of the coupling agent to the silica particle surface. Excess reactants and small particles in the suspensions were removed by three repeated cycles of washing and centrifugation. Table 1 lists the experimental conditions for silica nanoparticle preparation. Table 1. Experimental Conditions for Synthesis of Silica Particles TEOS (mL)/ ethanol (mL)

ammonia (mL)/ ethanol (mL)

centrifuge speed (rpm)/time (min)

final particle size (nm)

25/25 10/40 10/40

20/200 20/100 20/200

1500/10 2000/10 5000/10

1100 ± 50 810 ± 20 360 ± 12



RESULTS AND DISCUSSION Substrate Preparation and Characterization. Langmuir−Blodgett assembly is a powerful technique that can be used to produce large-area monolayers of amphiphilic silica spheres.25,26 The silica spheres used herein for LB film 1343

dx.doi.org/10.1021/ac301028w | Anal. Chem. 2013, 85, 1342−1349

Analytical Chemistry

Article

e-beam vacuum evaporation. In the e-beam process the deposited gold that penetrates the 3-fold holes of the silica LB film forms the well-known Fischer triangle footprint pattern. The gold sputtering process facilitates the enhanced diffusive movement of gold atoms ejected from the target and results in the observed honeycomb pattern. In comparison to nanoparticle films produced by direct deposition from suspension, the honeycomb substrates are less dense and have controllable thickness. Following sonication-mediated liftoff of the silica particles, a 15−33 nm-high honeycomb-shaped structure composed of clusters of gold nanoparticles is produced. Figure 1C−E shows that the distribution density of the clustered gold nanoparticles on the substrates could be varied by changing the particle size of the silica nanoparticles employed. The size of the nanoparticles within these clusters depends on the sonication time employed for silica sphere liftoff (Figure 2).

production were prepared by hydrolysis of tetraethoxysilane in an alcoholic medium in the presence of water and ammonia. Following surface functionalization by the coupling agent allyltrimethoxysilane, the particles are highly uniform with diameters of 1100, 810, and 360 nm. By dispersing the modified silica nanoparticles on the surface of a water phase, well organized particle arrays form at the air−water interface following solvent evaporation. The transfer of the resulting nanoparticle film onto a microscope coverslip was achieved at a surface pressure of 6 mN·m−1. Constant surface pressure during particle transference is critical to the fabrication of a homogeneous monolayer film. Surface pressure was controlled by setting the forward and backward barrier speeds at low velocities. The final area coated by the self-assembled spheres was limited only by the initial amount of particles used for the compression process. Monolayers can be prepared on any hydrophobic substrate over an arbitrarily large area. The Langmuir−Blodgett lithographic process for the preparation of gold honeycomb substrates for SERS is summarized briefly in Figure 1. The SEM and AFM images

Figure 1. Langmuir−Blodgett lithographic process for gold honeycomb production. (A) Langmuir−Blodgett deposition of functionalized silica nanoparticles, (B) Gold deposition by a sputter coater and a cross-sectional view illustrating the penetration of gold atoms into a 3-fold hole in the silica film, (C) A SEM image of a honeycomb substrate produced using 1100 nm silica nanoparticles (scale bar = 10 μm), and (D and E) AFM images of honeycomb substrates produced using 810 and 360 nm silica particles, respectively.

Figure 2. AFM images of honeycomb-shaped gold substrate using 1100 nm silica spheres with different sonication times of one minute (A) and three minutes (B).

in Figure 1C−E illustrate the honeycomb-shaped substrate produced by the LB lithographic process with 1100, 810, and 360 nm functionalized silica particles. In this process, a 20−60 nm-thick gold film (whose depth is controlled by the sputter time, applied current, and working distance) is sputter coated onto the functionalized silica template made by the LB technique. The applied sputtered gold is transported through gaps between the silica particles and printed onto the glass substrate. This process is different from substrates produced by

For a one minute sonication period, the substrate was composed of relatively uniform gold nanoparticles with a size of 50−70 nm. Longer sonication periods, however, result in the break-up of the nanoparticles into smaller sizes. Using AFM image roughness analysis, this difference in sonication time decreases the difference between the three-dimensional surface area and the two-dimensional footprint area (100 × 100 μm2) from 1.15% to 0.776% with 3 min. 1344

dx.doi.org/10.1021/ac301028w | Anal. Chem. 2013, 85, 1342−1349

Analytical Chemistry

Article

The UV−vis−NIR absorption characteristics of the films were evaluated in the visible−NIR range. The spectrum shown in Figure 3 illustrates that this substrate (produced by 1100 nm

This value tends to a constant upon an increase in gold deposition (Supporting Information Figure S1B). Because of their optimal properties, only honeycomb substrates produced using ∼40 nm gold deposition were subjected to further study. This parameter will be used subsequently to evaluate the average surface enhancement factor. Surface-Enhanced Raman Imaging. The SERS performance of the gold honeycomb substrate produced using 1100 nm silica nanoparticles was compared with those of a simple gold film, gold film over silica nanoparticles before the process of sonication, and with gold triangular footprints fabricated by thermal vacuum gold deposition instead of sputter coating (Supporting Information Figure S2). Using a drop casting method, MGITC molecules were attached to the substrates. Figure 4 shows the measured Raman images of MGITC on

Figure 3. Absorption spectrum of gold honeycomb in the visible-NIR range. Inset is a tapping mode AFM image that shows the nanoclustered gold film.

silica nanoparticles and a sonication time of 1 min) exhibits a LSPR band at 580 nm characteristic of individual gold nanoparticles and a broad plasmon band in the NIR that we attribute to the coupling of electromagnetic waves between neighboring gold nanoparticles (i.e., gold aggregates).27 Additional optical spectra in Supporting Information Figure S1A were generated using honeycomb arrays prepared with the same diameter silica spheres, but with gold deposition thicknesses that ranged from 20 to 60 nm. The LSPR band for the individual gold nanoparticles only varies by 13 nm. This result suggests the LSPR spectra are sparingly sensitive to the deposited gold thickness. The aggregated gold nanoparticles within the honeycombs serve as extremely versatile SERS substrates because they facilitate laser excitation over the wide frequency range of the plasmon band produced by the aggregates. As noted previously, the honeycomb is composed of gold aggregates with an average nanoparticle size of 50−70 nm. Gold nanoparticles with a core size of 63 nm are most efficient for SERS under red (630−650 nm) and near-infrared (785 nm) excitation.28 For practical SERS applications in aqueous environments, the substrate was further stabilized by annealing under vacuum at 200 °C for 2 h. This annealing process leads to self-organization of the clustered gold particles and enhances the adherence of the nanoparticles to the glass substrate.29−31 The percentage of the substrate covered by gold (μ) is estimated in Supporting Information Figure S1B based upon graphical analysis of collected AFM images. After accounting for surface roughness, the three-dimensional surface area was estimated from the two-dimensional area and the difference (Dif f) between them was determined to be around 1% for 40 nm gold deposition following a one-minute sonication period and heat annealing. The average size of the holes (d) that are not covered by gold after the removal of the 1100 nm (D) silica spheres was ∼720 ± 50 nm. On the basis of the following equation the percentage of the glass substrate covered by gold was determined to be 62% ⎛ πd 2 ⎞ μ = ⎜1 + Diff − ⎟ × 100% ⎝ 2 3 D2 ⎠

Figure 4. Raman images obtained by mapping the Raman intensity integrated over 1568−1675 cm−1 from malachite green isothiocyanate (MGITC, 2 μL, 1.33 × 10−5 M) adsorbed on a gold honeycomb film (A), gold triangular footprints (B) and gold film over silica nanoparticles (FON) (C). Averaged Raman spectra (D) from the MGITC locations and also the reference spectrum obtained at the sputter coated gold film outside the honeycomb array. The shaded column indicates the spectral ranges used for Raman imaging, and the color scale bar from 0−100 CCD cts for three images is located in C. Image area 4 × 4 mm2, 100 × 100 pixels, integration time 50 ms/ spectrum, 633 nm excitation wavelength with an Olympus 10× objective.

each the latter three substrates (the gold film did not exhibit significant enhancement and thus a Raman image could not be acquired) and the corresponding Raman spectra. As a result of differences in their relative surface polarities, the dried MGITC spots on the substrates exhibit diameters of 1.7 ± 0.1 mm for Au film over silica, 2.0 ± 0.1 mm for Au triangular footprints, and 2.5 ± 0.1 mm on the Au honeycomb film, respectively. These different spot sizes result in average surface coverages (defined as the applied MGITC concentration divided by the spot area) of 7, 5, and 3 molecules per nm2, respectively. Impressively, the measured Raman intensities on the honeycomb substrates are approximately 2.5-fold higher than the film over silica nanoparticles and triangular footprints. Given the

(1) 1345

dx.doi.org/10.1021/ac301028w | Anal. Chem. 2013, 85, 1342−1349

Analytical Chemistry

Article

lower surface coverage of the honeycomb substrates we estimate they exhibit a 5-fold greater enhancement relative to the other substrates. This finding suggests that the ordered structure of the honeycomb matrix is partially responsible for the SERS enhancement observed for this system since fewer target molecules per unit area are required to achieve a particular enhancement level. Comparison of the results obtained for the randomly arrayed sputtered gold film alone, and those obtained with the honeycomb matrix provides additional evidence that the physical organization of the gold nanoparticles in the honeycomb shape is responsible for the observed SERS enhancement. To investigate the adhesive strength of the interaction between the particles and the glass, the substrate was immersed in an aqueous 10−5 M solution of MGITC for approximately 1 h and subsequently rinsed with Milli-Q water to remove nonsurface associated MGITC. The substrate with an adhered monolayer of MGITC was then immersed in a deionized water solution and mixed on a Vari-Mix Platform Rocker (Thermolyne) for one day. Periodically, micro-Raman imaging of the substrate was conducted using 785 nm excitation through a 100× microscope objective with an integration time of 0.1 s. For each time period, three to five random 25 × 25 μm2 areas of substrate were scanned and the average Raman spectra was calculated from the 2500 spectra obtained at each scan position (Figure 5A). Relative to Figure 4, we note that the Raman peak at 1611 cm−1 loses much of its intensity when the excitation wavelength was tuned to 785 nm from 633 nm due to the decreased resonance enhancement effect. Figure 5B shows the average SERS intensity at 1170 cm−1 varied by less than 6% over a 25 h period. The