Introduction

Since it was discovered in 2004, graphene, a single atomic layer of graphite, has attracted vast interest due to its unique properties1. Recently, the two vibrant and rich fields of investigation of graphene and plasmonics have strongly overlapped2. A combination of graphene with plasmonic nanostructures has substantially improved the photodetection capabilities of graphene3,4. The combination of graphene with conventional plasmonics based on noble metals could be beneficial for both fields of investigation: plasmonic nanostructures can enhance the properties of graphene (stronger Raman signature and more effective graphene-plasmonic photocells), and graphene can influence the optical response of plasmonic nanoarrays2,5,6. Because the interaction between light and graphene is determined by the local electromagnetic fields (induced on the graphene sheet), this interaction can be increased by placing metal nanostructures close to the graphene, such as well-defined dimensions gold nanodisks arrays on the top of graphene7, nanodisks arrays8, gold nanopyramids9, a photonic crystal nanocavity10, and a pair of gold nanodisks11 covered with graphene.

From a surface-enhanced Raman scattering (SERS) point of view, graphene provides the ideal prototype test material for investigation of SERS12. In a graphene-mediated SERS (G-SERS) substrate13,14,15,16,17,18; the monolayer graphene (1G) provides an atomically flat surface for Raman enhancement. Because the graphene surface is chemically inert, signals from G-SERS substrates have great advantages over normal SERS by providing cleaner vibrational information free from various metal–molecule interactions (molecules cannot interact with the chemically active metal by graphene-mediated interactions) and being more stable against photo-induced damage13. Furthermore, the ultrasensitive graphene-plasmonic hybrid platform for label-free SERS detection is approaching single-molecule detection18.

However, previous studies ignored the fact that molecules can interact with (can even be photocatalyzed by) the chemically active metal through the high electron mobility of graphene during charge transfer in SERS measurements. Furthermore, the mechanism of SERS employing graphene has not been fully explored. Recently, increasing attention has turned to graphene–molecule12,19,20, molecule–metal21,22, and metal–grapheme7,8,9,10,11 interactions; however, a great challenge remains and clarification is needed. Currently, little is known about the complex interaction among molecules, graphene, and metal processes. This is the motivation for this work.

In this paper, we start by devising graphene-coated on Ag bowtie nanoantenna arrays (ABNA), as shown in Figure 1, where the para-aminothiophenol (PATP) on the substrate is an increasingly important probe molecule in SERS because of the intense SERS signal23,24,25,26. PATP-to-p, p′-dimercaptoazobenzene (DMAB) conversion is used to study charge transfer between the metal and molecules27,28,29,30,31, during which the catalytic reaction is monitored by SERS. Large-area, well-ordered, uniform-sized ABNA replaces previously disordered gold/silver nanoislands as the SERS substrate. Well-ordered ABNAs can insure the same large-area graphene–metal interaction. The PATP-to-DMAB conversion is achieved on the chemically inert and high electron mobility 1G-coated on ABNA. The 1G further enhances the plasmon-driven, surface-catalyzed reaction of the conversion of PATP-to-DMAB. In contrast, the conversion probability is lower on bilayer grapheme (2G)-coated on ABNA. The chemical reaction can be controlled by the distribution of localized surface plasmon. This discovery represents an important advancement toward the understanding of the complexity of the interaction among molecules, graphene, and metal processes and is also important for both plasmon-driven photocatalysis and applications of graphene. Furthermore, the DMAB molecule is an important organic optical electronic and molecular wire material32. There are potentially new photoelectric characteristics and potential applications of DMAB-graphene-plasmonic composites in transparent conductors and photodetection.

Figure 1
figure 1

(a) Schematic view of graphene-assisted, plasmon-driven reaction of the transformation of PATP-to-DMAB. (b) Scanning electron microscopic (SEM) image of the large-area, well-ordered, uniform-sized, graphene-coated Ag bowtie nanoantenna arrays (below the yellow dashed line); the yellow dashed line shows the graphene border. (c) SEM image of bare Ag bowtie nanoantenna arrays. (d) SEM images of the bare Ag bowtie nanoantenna arrays and (e) chemically inert, graphene-covered Ag nanoantenna arrays after a month. The Ag bowtie nanoantenna arrays are protected by the graphene monolayer.

Materials and methods

Figure 1a is a schematic view of the graphene-assisted, plasmon-driven reaction of the transformation of PATP-to-DMAB. Layer-controlled graphene by chemical vapor deposition (CVD) was transferred to the ABNA. Large-area, well-ordered, uniform-sized ABNAs are presented in Figure 1b (above the yellow dashed line). The yellow dashed line shows the graphene border. Graphene-coated ABNA is shown in Figure 1b (below the yellow dashed line). A typical Raman spectrum of graphene-coated ABNA is shown in Supplementary Fig. S1. Only the G and 2D bands are observed, which indicate that the graphene is still of high quality on ABNA. After a month, bare Ag nanoparticles show large morphological changes without the 1G, as show in Figure 1d. However, the chemically inert, 1G-covered Ag nanoantennas did not show signs of changes; the surfaces of the particles kept their triangle shape, as shown in Figure 1e. In a previous study, by passivating the surface of silver nanostructures with 1G, atmospheric sulfur-containing compounds (hydrogen sulfide (H2S) and carbonyl sulfide) were unable to degrade the surface of the silver33. The ABNA is protected by the graphene monolayer. The H2S molecule is smaller and more active than the PATP molecule. Therefore, ABNA cannot come into contact with the PATP molecules covered with the graphene without laser exposure.

Results and discussion

PATP-to-DMAB conversion on chemically inert graphene on ABNA

The time-dependent PATP-to-DMAB conversions were monitored in situ by SERS on the ABNA, 1G and 2G-coated ABNA, respectively. PATP was deposited on the surface of the samples by thermal evaporation. The SERS spectrum of PATP on ABNA is seen in Figure 2a (SERS spectrum was collected with an exposure time of 10 s and a laser power of 3 mW). The characteristic Raman peaks of DMAB are at 1387 and 1432 cm−1, which correspond to the ag16 and ag17 vibrational modes of N = N of DMAB28,29,34,35. Therefore, the ratio of Inn/Icc was observed to monitor the catalytic reaction, where Inn and Icc are the Raman intensities of the 1432 cm−1 band (from the generated DMAB) and the 1589 cm−1 band (from the benzene ring of the generated DMAB and unreacted PATP), respectively. The PATP-to-DMAB conversion is almost completed on ABNA with prolonged irradiation time of 11 min, as shown in Figure 2a.

Figure 2
figure 2

In situ SERS of PATP on (a) ABNA, (b) 1G-coated ABNA, (c) 2G-coated ABNA. (d) The relationship between the reaction rate and the irradiation time for ABNA (black line), 1G-coated ABNA (red line), and 2G-coated ABNA (blue line).

The characteristic Raman peaks of DMAB were also found on PATP on the 1G-coated ABNA after a shorter time of 1 min (11 min for ABNA without graphene), indicating that the PATP-to-DMAB conversion occurred on the chemical inertness and high electron mobility 1G on the ABNA. For comparison, the time-dependent normal Raman spectra of PATP on Si and graphene are also shown in Supplementary Fig. S2. The Raman spectrum of PATP on graphene shows that the vibrational modes at 1084 and 1589 cm−1 are vcc + vcs and vcc29,36, respectively (the Raman spectra were collected with an exposure time of 60 s and a laser power of 10 mW). Only the G (superposed with the vcc of PATP) and 2D bands are observed, indicating that PATP deposition does not affect the quality of our synthesized graphene. A comparison between these two spectra of PATP on graphene and 1G-coated ABNA shows that PATP cannot covert to DMAB on chemically inert graphene without metal, even though a higher laser power is employed. Furthermore, Zhang’s group reported the Raman spectra of a self-assembled monolayer of PATP on a flat Au and Ag surface by graphene-mediated SERS tapes13. Although PATP was anchored on the flat gold surface and had been isolated from the hot spot (Ag and Au nanoparticle gap) by graphene that was one atom thick, the weak and ignored characteristic Raman peaks of DMAB still emerged at approximately 1390 and 1430 cm−1. Therefore, the plasmon-driven transformation of PATP-to-DMAB occurs on graphene-covered Ag nanoparticles.

The number of graphene layers controlled the plasmon-driven reaction on graphene-covered ABNA

To investigate the mechanisms of the graphene plasmon-driven, surface-catalyzed reaction of the conversion of PATP-to-DMAB, the charge difference densities of PATP on graphene-coated Ag nanostructures were calculated. The orientation of the charge transfer for electronic transitions has been extensively studied by charge difference density36,37,38. As a single sheet of sp2 carbon atoms, the delocalized π-bonds of graphene act as a natural ‘magnet’ for collecting aromatic molecules, where the aromatic molecules lie parallel to the graphene. Electrons can efficiently transfer from PATP to graphene and Ag clusters with excited states near 488 nm (see Figure 3a).

Figure 3
figure 3

Charge difference densities for the electronic transitions of PATA/1G/ABNA (a) and PATA/2G/ABNA (b), where green and red stands for holes and electrons, respectively.

Furthermore, by comparing the Inn/Icc of the two Raman spectra of PATP on ABNA and 1G-coated ABNA within 1 min, we can see that the reaction probability of the conversion of PATP-to-DMAB is larger on the 1G-coated ABNA, where Inn and Icc are the Raman intensity of the 1432 cm−1 band (from the generated DMAB) and the Raman intensity of the 1589 cm−1 band (from the benzene ring of the generated DMAB, unreacted PATP and graphene(very weak)), respectively. For a more quantitative comparison, the spatial distributions of the electromagnetic field intensity for ABNA and 1G-coated ABNA (as measured by scanning electron microscope (SEM)) at the XY plane are calculated with the finite-difference time-domain (FDTD) method, using periodic boundary conditions (see Figure 4a and 4b). The 1G-coated ABNA has similar resonance positions to those of ABNA, but the electromagnetic enhancement |E|2 on the tip of the nanoparticle at an incident wavelength of 488 nm for 1G-coated ABNA is stronger than that for ABNA (1.5 × 103 versus 1.1 × 103), where |E| = |Elocal/Ein| and Elocal and Ein are the local and incident electric fields, respectively. Similarly, Xu et al. reported the maximum |E|2 value in the graphene-containing gold hemisphere dimer is twice as large as that of the pristine gold hemisphere dimer13. In previous studies, a dielectric SiO2 layer introduced additional enhancement through the multi-scattering process39. A similar case may also exist in graphene-containing metal nanostructures. Thus, stronger resonance plasmon can be obtained, and more energetic, ‘hot’ electron-hole pairs40,41 are produced on 1G-coated ABNA. Conversion of PATP-to-DMAB takes place with higher probability on 1G-coated ABNA. Therefore, 1G can enhance the plasmon-driven surface-catalyzed reaction of the conversion of PATP-to-DMAB.

Figure 4
figure 4

The FDTD calculated local electric field enhancement (log |E|2) of (a) ABNA, (b) 1G-coated ABNA, (c) 2G-coated ABNA at the XY plane at an incident wavelength of 488 nm.

For the 2G-coated ABNA, the ratio of Inn to Icc is still low after the longest irradiation time of 11 min, as shown in Figure 2c. The relationships between the reaction rate and the irradiation time for ABNA (black line), 1G-coated ABNA (red line), and 2G-coated ABNA (blue line) are shown in Figure 2d. This result indicated that the reaction probability of the conversion of PATP-to-DMAB is small on 2G-coated ABNA. The characteristic Raman peaks of DMAB are found on the 1G-coated ABNA with a more efficient reaction in 1 min. The spatial distributions of the electromagnetic field intensity for 2G-coated ABNA at the XY plane were also calculated by the FDTD method using periodic boundary conditions (see Figure 4c). The electromagnetic enhancement |E|2 on the tip of the nanoparticle at an incident wavelength of 488 nm for 2G-coated ABNA is slightly weaker than that for ABNA (0.9 × 103 versus 1.1 × 103). Similarly, the increase of the SiO2 layer thickness compromises the SERS activity39. Part of the reason for the smallest reaction probability is the least hot electron generation on the 2G-coated ABNA.

More importantly, the charge difference densities of PATP on the bilayer graphene on the Ag clusters were calculated. For the two-layer case, electrons can efficiently transfer only to the top of the bilayer graphene where the molecule is adsorbed, and the electrons cannot efficiently transfer to the Ag clusters and the bottom of the bilayer graphene, which is directly adsorbed on the Ag clusters (see Figure 3b). Thus, the bilayer graphene effectively decreases the probability of the PATP-to-DMAB conversion.

Process of the graphene enhanced plasmon-driven, surface-catalyzed reaction

To further investigate the PATP-graphene composite interaction with ABNA, three types of changed ABNA were fabricated. SiO2 films approximately 2 nm and 5 nm in thickness were deposited on as-prepared ABNA prepared by radio frequency magnetron sputtering to obtain samples 1 and 2. For sample 3, the previous thickness of silver film of approximately 60 nm was decreased to 20 nm. The graphene was transferred onto the three samples before PATP deposition. The SEM of graphene-covered sample 1, sample 2, and sample 3 are shown in Figure 5. The SERS spectra of PATP on the three graphene-coated samples are shown in Figure 5a. The characteristic Raman peaks of DMAB (1387 cm−1 and 1432 cm−1) rise to different extents for the three samples. For samples 1 and 3, the conversion of PATP-to-DMAB occurred within certain limits. However, for sample 2, the conversion of PATP-to-DMAB minimally occurred. Therefore, the thicker SiO2 film can prevent electrons from the PATP-graphene composite from recombining with strong hot-holes arising from the localized surface plasmon resonance (LSPR) of the ABNA as in samples 1 and 2. The lesser produced hot-holes limit PATP-to-DMAB conversion in sample 3.

Figure 5
figure 5

(a) The SERS spectra of PATP on graphene-covered sample 1, sample 2, and sample 3. And sample 1 (b), sample 2 (c), and sample 3 (d) corresponding SEM.

In previous studies28,29,36, the conventional plasmon-driven reaction of the conversion of PATP-to-DMAB consisted of two processes, as confirmed by ultrahigh-vacuum, tip-enhanced Raman spectroscopy28 shown in Figure 6a. The first process is the chemical adsorption of the PATP molecules onto the silver nanoparticles through a thiol group. The second process is that a couple of PATP molecules lose electrons to hot-holes arising from LSPR42,43; then the couple of PATP molecules convert to a DMAB molecule. For the graphene-controlled, plasmon-driven, surface-catalyzed reaction, the following mechanism is demonstrated: electron transfer first from the PATP molecule to graphene; next, the electrons recombine with hot-holes arising from LSPR on the ABNA surface42,43; the 1G introduces a strong dipole allowed transition between PATP and graphene, which is enhanced by the EM fields; then, a couple of the PATP molecules lose electrons to become DMAB molecules on the graphene surface, as shown in Figure 6b. For the bilayer graphene, the electronic transition is weak and hot electron transfer from Ag is difficult.

Figure 6
figure 6

(a) Schematic view of the structure of PATP and the formation process of DMAB by the conventional plasmon-driven reaction. (b) Schematic view of the formation process of DMAB by the graphene-assisted, plasmon-driven reaction.

The distribution of the ABNA effected plasmon-driven reaction

To demonstrate the large-area, ordered, graphene-controlled, plasmon-driven, surface-catalyzed reaction, which is dependent on the localized surface plasmon distribution, point-by-point Raman mapping was recorded on a randomly selected 3.5 μm × 3.75 μm area with a step size of 250 nm on the 1G-coated ABNA. The SEM of the selected area is shown in Figure 7a marked by red dotted lines. By marking the silicon substrates with a UV lithography-defined coordinate system, we are able to observe, in the optical microscopy of the Raman system, the same field of view identified in the SEM.

Figure 7
figure 7

(a) The SEM of the randomly selected Raman mapping area of 3.5 × 3.75 μm2 marked by red dotted lines. The semi-circle is a marker of the UV lithography-defined coordinate system. (b) SERS intensity maps at 1432 cm−1 (νnn) corresponding to the area shown in (a). (c) overlay image of (a) and (b). (d) A list of randomly selected Raman spectra; their corresponding SEM and Raman mappings are marked by blue dotted lines, as shown in (c).

First, the selected area is 1G-coated ABNA confirmed by the 2D to G peak ratio Raman mapping. After PATP deposition, the Raman intensity mapping at 1432 cm−1 (the characteristic Raman peak of DMAB) is shown in Figure 7b. As illustrated in Figure 7b, the graphene-assisted, plasmon-driven, surface-catalyzed reaction occurred periodically on well-ordered ABNA covered by large-area 1G. The regions of chemical reaction distribute at the croci strips in the same way as the hot spots are distributed in Figure 4. The overlay (Figure 7a and 7b) of the Raman mapping with the corresponding SEM shows more clearly that PATP-to-DMAB conversion mostly occurs on the 1G-covered bowtie nanoantenna tip. A list of randomly selected Raman spectra, A0–A10, is shown in Figure 7d. The corresponding SEM and Raman maps of the spectra A0–A10 are marked by blue dotted lines, as shown in Figure 7c. The emergence of the characteristic Raman peaks of DMAB at 1387 and 1432 cm−1 change periodically with the 1G-coated ABNA orderly distribution. The chemical reaction could be controlled by localized surface plasmon distribution.

Conclusions

We demonstrate that PATP can be converted to DMAB on chemically inert, graphene-coated, ABNA during optical excitation, which was monitored by in situ SERS spectra. The PATP-to-DMAB conversion occurred on monolayer and bilayer graphene-covered ABNA to different degrees. 1G can further enhance the plasmon-driven, surface-catalyzed reaction. Bilayer graphene decreased the probability of the PATP-to-DMAB conversion. The chemical reaction was controlled by the localized surface plasmon distribution, as confirmed by the Raman mapping. To reveal the mechanisms, the charge difference density shows that electrons efficiently transferred from the PATP molecule to the 1G-coated Ag nanostructures. Subsequently, electrons recombined with hot-holes arising from LSPR (LSPR is excited by incident light). Some of the PATP molecules lose electrons to become DMAB molecules on the graphene surface. The 1G enhanced and the bilayer graphene suppressed the plasmon-driven, surface-catalyzed reaction, which represents an important advancement toward the understanding of the complex interaction among molecules, graphene, and metal processes and is also important for both plasmon-driven photocatalysis and applications of graphene.