Ultraflat Transparent Gold Nanoplates—Ideal Substrates for Tip‐Enhanced Raman Scattering Experiments

Abstract

Thin crystalline gold plates provide a perfect surface to immobilize biomolecules for further label-free investigations on the nanometer scale. These first tip-enhanced Raman experiments provide evidence for local variations of cystine adsorption on such ultrasmooth gold surfaces (see image). Ultraflat gold nanoplates were found to be very attractive substrates for tip-enhanced Raman scattering (TERS) measurements. The transparent flat triangular or hexagonal nanoplates were synthesized by citrate reduction of HAuCl4 in aqueous solution. The obtained nanoparticles with a height of 15–20 nm had a smooth homogeneous surface with a roughness of about 100–200 pm. After spreading the gold plates on glass slides, cystine was successfully immobilized on them and the first TERS spectra of an amino acid were recorded. The spectra revealed a local variation of the attachment of cystine on the gold surface in two conformers. The production of well-defined gold particles is of general interest due to their geometry-dependent physical properties. Controlling and tuning the properties of these particles down to the nanometer regime extends their present applications, such as fluorescence quenching of organic molecules,1 surface-enhanced Raman scattering (SERS),2-4 and, as will be presented here, TERS. In SERS, molecules are adsorbed on rough metal substrates like colloids, electrodes, or evaporated films, resulting in a Raman signal-enhancement of several orders of magnitude. The irregular surface structure of those metallic surfaces, however, restricts the application of the SERS technique. With SERS, chemical information on low sample concentrations down to single molecules is accessible but the spatial information is lost. This argument is also valid for defined and highly reproducible SERS structures like electron-beam lithographic masks (see for instance Reference 5) and for such structures the field enhancement invariably varies across the surfaces. Due to this lack of spatial resolution, a quantitative molecular analysis is hardly possible. If such information is required, TERS is the method of choice.6 In TERS a scanning tunneling microscope (STM)1 or an atomic force microscope (AFM) is coupled with a Raman spectrometer. Here the field-enhancing feature is confined to the very end of the probe. Usually a sharp metal tip or a single small metal particle acts as the sole source of enhancement. For many applications it is essential to immobilize single molecules on a surface. In order to localize the molecules by scanning probe techniques it would be desirable to fix the sample molecule on an atomically flat substrate. This limits the selection to only a few substrates like mica, gold, or HOPG (highly ordered pyrolytic graphite). Of these materials gold is often preferred as its immobilization chemistry is well-established. Atomically flat gold surfaces have already been employed in STM–TERS experiments. Under such conditions, no band intensity changes caused by different shape- or size-induced field-enhancing properties are expected since all adsorption sides on the substrate surface are equal.7 The lateral resolution capabilities of TERS are therefore not affected by such substrates. The main problem when using gold is its opaqueness and therefore incompatibility with back-reflection TERS instrumental designs. This design, on the other hand, is very powerful due to its high collection efficiency and compatibility with standard microscopy experiments. Consequently, the challenge is to produce atomically flat or nearly atomically flat gold substrates with thicknesses that still allow signal collection through the gold film itself. In addition to the established immobilization chemistry, another benefit was expected for TERS experiments on a flat gold surface: an additional Raman signal enhancement caused by the close proximity of the gold plate and the silver tip should be observed. This so-called “gap-mode” should potentially overcome or even exceed the adsorption losses due to the gold plate. The gap between the gold surface and the silver tip should provide an additional enhancement as theoretical considerations predict.8, 9 The advantage of a smooth gold plate is the consistent surface leading to a constant enhancement factor of the Raman intensity as stated above. Earlier results on gap-mode TERS employing AFM on evaporated gold films showed that a further enhancement can be only observed when the tip–sample distance is smaller than 30 nm.10 A uniform transparent flat gold surface fulfills this requirement and allows a back-reflection setup. So far, all published gap-mode experiments used a side reflection setup AFM10, 11 or STM,12 and therefore, transparency of the substrate was not crucial. In these experiments, smooth gold or platinum surfaces were utilized but no single gold nanoplates. To date many gold nanoparticle shapes have been identified. Spheres,13 triangular and hexagonal plates,4, 14-20 wires,19 and branched species,2 to name the most important, have been discussed in the literature. The physical properties, in particular the optical properties, of such nanoparticles strongly depend on their shape, size, and morphology. Typical classification methods are scanning electron microscopy (SEM), UV–Vis spectroscopy,14 or X-ray diffraction.2 While the optical characterization of an ensemble of nanoparticles is straight forward (UV–Vis, etc.), the investigation of a single gold nanoparticle requires elaborate experimental conditions.21 In this letter we present the first TERS experiments on immobilized biomolecules on a single transparent gold nanoplate. As a model substance, cystine, which plays an important role in building and stabilizing the tertiary structure in proteins, was chosen. The formation of disulfide (SS) bonds within and between most protein molecules leads to the 3D structure. Cystine is the oxidized dimer of cysteine, which reacts with thiol side-chain groups to form the SS moiety. Functional groups containing sulfur atoms are known to have a strong tendency to bind on gold and silver surfaces. From theoretical calculations and SERS measurements on cystine it is known that several conformers coexist and the molecules interact via the disulfide bridge with the metallic surface.22-24 The specific properties of the gold substrates used in this report, in particular their transparency and flatness, allow the use of the very efficient back-reflection setup to investigate the properties of the gold–cystine interaction on a nanometer scale. The experimental setup is schematically outlined in Figure 1. The advantages of this illumination geometry in contrast to the side reflection geometries are the better collection efficiency that can be obtained by using high numerical aperture microscope objectives or even oil immersion objectives, and the flexibility of the system to perform standard Raman or fluorescence control experiments. The crucial difference between the gold substrates discussed here compared to previously used atomically flat substrates is the transparency, which provides much more flexibility with respect to the optical setup. Figure 2 shows the penetration depth of gold for various wavelengths. From these values a practical thickness of <20 nm can be deduced as suitable for a transmission setup as shown in Figure 1. Schematic outline for a back-reflection TERS setup for measurements through a single gold nanoplate. Penetration depth of gold calculated from optical constants. A major effort therefore was the synthesis of kink-free gold plates with a maximum thickness below 20 nm and hopefully a size larger than 1 µm to provide visual control underneath an optical microscope. The flatness is crucial to avoid enhancement variations inherent to all SERS substrates. To keep the synthetic effort to a minimum, the preparation of gold plates was based on the reduction of HAuCl4 in aqueous solution. This reduction can involve different reducing reagents, such as trisodium citrate,14 (polyamidoamine) dendrimer,25 ammonium formate,2 ascorbic acid,18, 19 azacryptand,13 lemongrass,4 and poly(diallyldimethylammonium) chloride.16 These nanoplates show a size range between 20 nm and 10 µm. The formation of crystalline plates was explained by the transformation of an unstable intermediate to an energetically favorable state.25 Additionally, stabilizers like polyvinylpyrrolidone,2, 17 azacryptand,13 and poly(diallyldimethylammonium) chloride,16 and cetyltrimethylammonium bromide (CTAB)14, 19 as a capping agent are reported to benefit the growth of a certain particle shape. CTAB was reported to favor the formation of the plates due to specific interactions with gold seed intermediates. Hence, temperature, time, and reagent ratios influence the formation of the nanoplates and can be used to tune the size and thickness of the plates.14, 19 For our purposes, a synthesis reported by Chu et al. was adopted.14 This procedure is reproducible in contrast to the synthesis treating HAuCl4 with citric acid in absence of a surfactant.18 Size characterization of the nanoplates was accomplished by SEM and AFM (Figure 3a and b). The height of a typical nanoplate was determined to be 20 nm with a standard roughness on the gold of 120 pm (root mean square, rms) and 400 pm maximum peak-to-peak variation, which is essentially the noise level of our instrument (Figure 3b and c). a) SEM image of gold nanoplates. b) AFM topography (baseline corrected) of a single gold nanoplate. c) Height profile on the gold plate corresponding to the black line in (b). For the TERS measurements, a monolayer of cystine was adsorbed onto a cover slip covered with gold plates. The quality of the topography images (Figure 4) measured prior to the actual TERS is slightly compromised compared to the one shown in Figure 3b due to the evaporated silver on the AFM probe required for the TERS experiments. However, the quality is by far sufficient for the determination of location necessary in these experiments. The actual TERS spectra were recorded at different positions as indicated in Figure 4b and are presented in Figure 5. To exclude tip contamination, reference measurements on the glass next to the gold plate were always carried out (Figure 4b, position 0). a) AFM topography of a gold nanoplate with a cystine monolayer. b) Zoomed-in region of interest. The numbers refer to the TERS spectra in Figure 5. TERS spectra of cystine on distinct positions on the gold plate (see Figure 4b), * refers to the Si signal of the AFM tip, λ = 568 nm, Pon sample = 130 µW, acquisition time = 5 s The enhancement due to the TERS probe was only roughly estimated to be at least on the order of 106–107 based on signal-to-noise arguments and area differences between the TERS probe and the laser spot. The loss in laser power when irradiating the sample through a 20-nm gold layer can be calculated as ≈25%. If the signal is collected again through the gold, the signal intensity is ≈6% compared to a directly irradiated sample. Hence, the tip in combination with the gold surface must make up for this disadvantage. The results clearly show the high quality of the TERS spectra when the signal was collected on top of the gold plate, strongly indicating an additional enhancement component such as the gap-mode. A quantitative assessment of such an effect is presently difficult, as the cystine could only be reliably adsorbed on the gold surface. TERS of a cystine monolayer on the glass could not be observed in the experiments, nor could a cystine monolayer on the glass be confirmed by other experimental means, indicating that the amino acid was only adsorbed on the gold. A direct comparison between TERS spectra of the compound on glass and on gold to reveal the enhancing mechanisms was therefore not possible. Qualitatively, the TERS experiments show a high signal-to-noise ratio even though the effective power on the sample (Pon sample) was only 130 µW considering the attenuation of the laser due to the gold layer. Comparing TERS spectra at different points of the sample in Figure 4b, one can see that the spectra resemble each other and can be divided in two main groups. This leads to the assumption that cystine was attached in at least two states. The spectra at positions 1 and 2 show bands for a neutral carboxyl (1685 cm−1) and amino group (1197 cm−1) in contrast to a zwitterionic form of cystine (positions 3–6) with a band at 1631 cm−1 for ν(COO−) and at 1614 and 1119 cm−1 for ν(NH3+). Considering all measured spectra it can be concluded that the latter form is more abundant than the neutral compound. The most prominent band at 676 cm−1 (662 cm−1 for free cystine) can be assigned to the CS vibration. The signal at 746 cm−1 also originates from the CS vibration and implies the presence of a second conformer (Table 1). Due to the SS moiety in cystine different conformations of the molecule are possible, depending on the dihedral angle of the CSSC bonds.26, 27 In SERS experiments on cystine using colloidal silver or gold solutions the presence of these bands was explained by a weak interaction of the disulfide bridge with the gold surface, exposing several conformers.23, 24 The assignment of the SS bond in the TERS experiments turned out to be ambiguous. The band is expected around 500 cm−1.27 Actually, a weak signal at 489 cm−1, which may arise from that moiety, can be detected. As can be seen from the spectra in Figure 5, the Si vibration from the tip is located at 520 cm−1, which may suppress the signal and render an assignment in this particular region difficult. SERS Au colloid24 SERS Ag colloid23 In the SERS experiments in colloidal Au or Ag solutions, Podstawka et al. assigned the observed bands at 491 and 518 cm−1 to the disulfide bridge, but also argue that CO and CO can be expected in this range. The authors excluded a dissociation of cystine. From our point of view this assignment of the signals to ν(SS) has to be regarded critically since the signal-to-noise ratio is rather poor. Furthermore, these studies stand in contrast to other observations where the SS bridge was cleaved on absorption on silver sol surfaces.28 In addition, density functional theory (DFT) calculations of cystine adsorbed on gold in the gas phase showed that one sulfur atom is closer to the gold surface than the other. For energetic reasons the SS bond was predicted to be cleaved, resulting in an AuS bond.22 We could not show this experimentally since it was not possible to measure beyond 400 cm−1 and the AuS vibration was expected at around 267 cm−1 with respect to cysteine.29 The dominant signals of ν(CS) can be recognized as an indicator that the molecules were chemically bound via the sulfur to the gold surface. The enhanced signals of the COO−/COOH and NH3+/NH2 groups can also be taken as evidence for additional interactions with the metal. The slightly varying Raman signal intensities and intensity ratios at different positions can be explained by the changing number of molecules contributing to the signal due to slight discontinuities in the cystine monolayer and to rotation of the molecules on the surface. This is a typical effect seen in TERS experiments and can be related to the small number of scatterers involved that do not provide a statistical average usually detected in Raman experiments. In summary, it was demonstrated that flat gold nanoplates with a height of around 20 nm are ideal substrates for the immobilization of cystine in order to measure TERS spectra in back-reflection geometry. Triangular and hexagonal gold plates were obtained in a straightforward one-pot synthesis by reduction of HAuCl4 with sodium citrate in the presence of a surfactant and turned out to be almost atomically flat with a roughness of 100–200 pm. The experiments reveal that cystine is attached via the sulfur atoms to the gold surface in two conformers. A decisive argument for the bond breaking of cystine to directly bound cysteine was not found. The expected signal losses due to the absorption of the gold plate were most likely compensated by an additional Raman signal enhancement caused by a cavity effect between the TERS probe and the substrate. The results are encouraging for extending the research to peptides in general and, in the long term, even to protein chains. Gold plates were synthesized mainly following the procedure given in the literature:14 15 mg (0.05 mM) sodium citrate are dissolved in 30 mL water and heated to 50 °C. A solution of 9.0 mg (0.03 mM) HAuCl4 and 55 mg (0.15 mM) CTAB in 20 mL water are heated to 50 °C and added to the citrate solution under stirring. The orange reaction mixture decolorizes immediately and turns light blue after heating to 82 °C within 15 min. While stirring for a further 10 min the nanoplates precipitate. After cooling to room temperature, 3 mL of the reaction mixture are diluted with 20 mL water and centrifuged at 3000 rpm for 10 min on clean glass cover slides. The slides are washed in 70 °C water to remove any remaining CTAB. Spherical nanoparticles that had also formed during the reduction are removed with ultrasonication and rarely found. In general, these spherical species have no influence on the TERS experiments as they can be easily avoided during measurements. Finally, the glass slides are dried in vacuum. For the TERS measurements a cover slide with gold plates was immersed for several hours in an aqueous cystine solution and washed three times with water. After drying in vacuum, TERS spectra and AFM topography are recorded. The detailed description of the workflow and the instrument have been described previously.30 The excitation wavelength was 568 nm with a power after the last lens of 530 µW and an acquisition time of 5 s.