Plasmonic Core–Shell Silicon Carbide–Graphene Nanoparticles

Abstract

We demonstrate the synthesis of silicon carbide nanoparticles exhibiting monolayer to few-layer graphene coatings and characterize their optical response to confirm their plasmonic behavior. A multistep, low-temperature plasma process is used to nucleate silicon particles, carbonize them in-flight to give small silicon carbide nanocrystals, and coat them in-flight with a graphene shell. These particles show surface plasmon resonance in the infrared region. Tuning of the plasma parameters allows control over the nanoparticle size and consequently over the absorption peak position. A simplified equivalent dielectric permittivity model shows excellent agreement with the experimental data. In addition, optical characterization at high temperatures confirms the stability of their optical properties, making this material attractive for a broad range of applications.

Introduction

Ever since its initial discovery, graphene has attracted tremendous attention because of its unique electrical, thermal, and optoelectronic properties.¹ Graphene nanostructures have been demonstrated in computational and experimental works to exhibit surface plasmon resonance (SPR).² This behavior has important consequences from an application point of view since graphene structures with tunable and controllable optical response could find use in sensing,^2c for the development of photonic circuits,³ for thermophotovoltaic devices,⁴ etc. While most of the research in this area is based on lithographically patterned graphene placed onto a flat surface, various attempts have been made to realize three-dimensional graphene nanostructures. Graphene sheets have been wrapped around metal oxide nanowires and nanoparticles to form hybrid structures through a liquid-phase chemical process.⁵ Graphene has been grown on hexagonal-phase silicon carbide by high-temperature annealing of silicon carbide micropowders.⁶ However, due to size effects on SPR of nanoparticles, significantly smaller particles must be realized to harness SPR in the infrared regime.⁷

In this work, we demonstrate tunable infrared SPR in ceramic core, graphene shell nanoparticles. To achieve that, we leverage the capabilities of low-temperature (as referred to the gas temperature that remains close to room temperature⁸), nonthermal plasmas to produce particles with a relatively narrow size distribution,⁹ i.e., to avoid the formation of large particles that would compromise the optical performance of the sample. Nanoparticles are heated to temperatures considerably higher than the gas temperature in these reactors because of exothermic reactions with plasma-produced species (ions, radicals, metastables, etc.). This property has been extensively investigated by many groups,¹⁰ and it is confirmed by recent reports on the nonthermal plasma synthesis of graphitic particles¹¹ and ceramic-like carbides and nitrides.¹² For the process used for this study, silicon nanoparticles produced in a first nonthermal plasma stage are rapidly carbonized by a second nonthermal plasma into β-phase silicon carbide nanoparticles.¹³ By carefully controlling the carbon precursor concentration, it is possible to tune between bare silicon carbide, single layer, and few layers of graphene coating. Fourier transform infrared (FTIR) measurements show a broad absorption feature in the infrared region that we attribute to plasmon-induced resonance. This conclusion is supported by the results of an “equivalent dielectric permittivity” model that has been demonstrated to be in good agreement (<10% error) to rigorous Mie Theory calculations for similar systems.^2a FTIR measurements at temperatures as high as 650 °C show little dependence of the absorption cross section on temperature, confirming that the material is highly thermally stable.

Results and Discussion

The material discussed here was produced by a nanoparticle synthesis reactor very similar to the one previously reported by our group.¹³ It is comprised of two continuous flow capacitively coupled nonthermal plasma reactors placed in series. A photograph of the reactor is shown in Figure S1 of the Supporting Information. Each of the plasma discharges is sustained using a copper electrode wrapped around the quartz tube, which is biased by a 13.56 MHz radio frequency power supply. One hundred standard cubic centimeters per minute (sccm) of argon containing 1.37% silane is flown through the primary plasma reactor to continuously nucleate and grow silicon nanoparticles, which are then continuously carried by the flow into the second plasma reactor. Following the primary reactor, methane is added to the flow downstream of a 2 mm orifice with flow rates ranging from 0.6 to 2 sccm. The silicon nanoparticles are carbonized in the secondary plasma to form β-phase silicon carbide nanoparticles and grow the graphitic shell. The particles are then carried downstream of the reactor where they are collected onto a fine stainless steel mesh acting as a filter. Figure 1a shows a transmission electron microscopy (TEM) micrograph of the resulting core–shell β-SiC–graphene nanoparticle. TEM was performed on a Tecnai T12 microscope. Fast Fourier transform analysis of the image confirms that the particle core is composed of silicon carbide, while a graphene monolayer surrounds the particle core. Additional TEM images of samples produced under various conditions are provided in Figure S2 of the Supporting Information. While a detailed discussion of the mechanism leading to the formation of the graphitic layer is not the primary focus of this contribution, it is likely the result of a chemical vapor deposition (CVD) process onto the nanoparticle surface. As already mentioned, nanoparticles are heated to temperatures that significantly exceed those of the background gas in these reactors. In addition, the plasma-enhanced CVD growth of graphene has been experimentally confirmed in several studies.¹⁴ The appearance of a graphenelike shell that grows in thickness with increasing methane concentration is consistent with a CVD mechanism. The pressure is kept at either 3 or 6 Torr for the samples discussed in this contribution. We have found that varying pressure is the simplest way of controlling particle size due to changes in the residence time. The diameters of 100 particles were measured from TEM micrographs to collect particle size statistics. The results of this analysis are shown in Figure 1b,c for the particles produced at 3 and 6 Torr, respectively. Average particle diameters were found to be 11.5 and 22 nm. There is no discernible change in particle size as a function of methane flow.

Figure 1.

(a) TEM micrograph of a silicon carbide nanocrystal with a single layer graphene shell. (b) and (c) Particle size distributions for samples produced at 3 and 6 Torr, respectively.

Figure 2 shows the results of structural characterization. X-ray diffraction (XRD, performed on a Panalytical Empyrean with a Cu Kα source) patterns are shown in Figure 2a. Some silicon is still present at the lower methane flow conditions, although both Raman (performed on a Horiba LabRam with a 532 nm laser excitation) in Figure 2b and TEM confirm the presence of graphitic carbon. This suggests that while the majority of the silicon particles has been carbonized, a flow of 0.6 sccm of methane is not sufficient to ensure that the sample is structurally uniform. At higher methane flows, there is practically no residual silicon at the reactor outlet. Raman indicates the presence of two clear peaks present at around 1350 and 1580 cm^–1, corresponding to the D and G bands of carbon. The D/G peak intensity ratio ranges from 1.9:1 to 3:1, indicating a mixture of disordered and graphitic material.¹⁵ However, there is no significant peak present at the two-dimensional position at around 2700 cm^–1, which correlates to the presence of graphene.¹⁶ These Raman spectra are clearly different from those observed for large graphene flakes. This difference is justified by the presence of defects in the graphene coating. For instance, the coating does not appear to be uniform and continuous around the SiC core in Figure 1a. Figure S2 also clearly shows the presence of exposed edges and blisters. While further work is necessary to understand how to improve the quality of the graphene coating, the material described in this report has sufficiently high quality to show plasmonic behavior.

Figure 2.

(a) XRD patterns for samples produced at 3 Torr, with increasing amounts of methane flown through the system. The diffraction patterns for samples produced at 6 Torr show a similar trend and are not shown for brevity. (b) Raman spectra for samples produced at 3 and 6 Torr, with increasing amounts of methane flown through the system.

FTIR spectra have been measured with a Nicolet iS50 by drop-casting a dispersion of particles, ultrasonicated in chloroform, onto a ZnSe ATR window. Results of FTIR are summarized in Figure 3. Figure 3a shows the spectra for the particles produced at 3 Torr, whereas Figure 3b shows those for the particles produced at 6 Torr. All of the spectra are normalized to the peak at 860 cm^–1, corresponding to the vibrational mode of Si–C,¹⁷ to account for differences in the amount of drop-casted material. The samples exhibit distinct broad absorption features. At the higher pressure condition (i.e., larger particles), this peak position redshifts considerably from ∼5000 cm^–1 (2 μm, 0.62 eV) to ∼3000 cm^–1 (3.3 μm, 0.375 eV). With the increasing methane flow, the peak initially increases in magnitude and then drops considerably and redshifts to lower energies.

Figure 3.

(a) FTIR spectra for the samples produced at 3 Torr, with a varying amount of methane. The calculated absorption spectra shown good agreement with respect to the peak position. Accounting for the dispersion in particle sizes (labeled as “Hi mob—PSD”) and for an increase in plasmon dephasing rate due to defects (“low mobility”) is not sufficient to reproduce the measured absorption spectra. Both these effects are accounted for to obtain a good fit between the experimental data and the calculated results. (b) Same as (a) but for samples produced at 6 Torr.

A modified version of the equivalent dielectric permittivity model developed by Shi et al.^2a is used to theoretically reproduce the broad features shown in the spectra in Figure 3. The nanoparticle core has a radius r_SiC and graphene shell thickness t_g. Although a monolayer of graphene measures 0.34 nm in thickness, the value here is taken to be 1 nm in following with preceding computational works on optical conductivity of graphene. This accounts for inhomogeneity in the growth of graphene.^2a,18 The dielectric constant of graphene is calculated from the optical conductivity relation

1

where ε_∥ is the in-plane dielectric constant of graphene, ε_⊥ is the out-of-plane dielectric constant of graphene (set to the value of graphite, 2.5), ω is the excitation light frequency, and ε_o is the vacuum dielectric permittivity constant. t_g is the graphene shell thickness, as stated earlier. The optical conductivity of graphene is calculated from the random phase approximation with the local limit condition based on the Kubo formalism¹⁹

2

3

4

where σ_intra and σ_inter are the intraband and interband conductivities, respectively, e is the elementary electron charge, k_B is the Boltzmann constant, ℏ is the reduced Planck constant, T is the room temperature (300 K), E_f is the Fermi level of graphene (taken to be 1 eV, ba sed on the discu ssion in ref (2a)), and τ_g is the carrier relaxation time. Since the nanoparticles are much smaller than the wavelengths under consideration (1–10 μm), the core–shell particles can be treated as an isotropic homogeneous medium with an effective dielectric constant, ε_np, obtained as

5

where f is the volume fraction of graphene, ε_SiC is the dielectric constant of the silicon carbide core, taken to be 6.52,²⁰ and ε_eq-G is the equivalent dielectric constant of the graphene shell. Given the spherical symmetry of the shell, the perpendicular ε_⊥ (out-of-plane) and the two parallel ε_∥ (in-plane) components of graphene’s dielectric constant equally contribute to the equivalent dielectric constant of the shell

6

The absorption, scattering, and extinction cross sections can then be calculated by^2b

7

8

9

where k is the wave vector of the incident light and ε_c is the permittivity of the environment, taken as unity for air. R is the particle radius. The absorption spectrum for the case of particles with sizes equivalent to the ensemble average size, and with a mobility of 10 000 cm²/(V s) corresponding to the bulk value for graphene, is shown in Figure 3a, labeled as “high mobility”. While the peak position matches quite well with the measured one, both the width of the particle size distribution and an increased carrier relaxation frequency need to be included to reproduce the measured peak shape. The synthesized SiC particles do not have a monodispersed particle size distribution, as shown in Figure 1b,c. This is easily accounted for by convoluting the single-size theoretical spectrum over the Gaussian fit of the measured particle size distribution. The resulting absorption spectra are labeled as “High mob – PSD” in Figure 3a,b. While this increases the width of the theoretical extinction spectrum, it is not sufficient to reproduce the measured data. An additional reduction in electron mobility from the bulk value of 10 000 cm²/(V s) to 1000 cm²/(V s) gives an excellent agreement between the theoretical and the measured absorption peak shapes. The use of a reduced mobility is justified by both TEM and Raman characterizations (Figures 1a and 2b), which confirm the less-than-perfect structure of the coating. The model predicts a shift in peak position that is also in excellent agreement with the measurement, as shown in Figure 3b.

The dependence of the absorption spectra on the methane flow deserves further comment. At each pressure, the peak intensity initially increases with methane flow, with no change in peak position. Then, as methane is further increased, the peak intensity drops drastically and the position redshifts significantly. This behavior is consistent with the observation of residual silicon in the samples carbonized with the smallest amount of methane. This explains the increase in the intensity of the plasmon peak when methane is increased to 1 sccm, i.e., the fraction of fully carbonized particles with a graphene shell is higher in this sample. Additional increase of the methane flow induces growth in the layer thickness and appearance of defects, as shown in Figure S2. This is accompanied by a red shift in plasmon energy and by a broadening in the absorption feature. This is consistent with a significant increase in the scattering and plasmon dephasing rates in these samples.

These results suggest that improvements with respect to both the particle size distribution and quality of the graphene layer will help achieving a sharper plasmon resonance feature in the absorption spectrum. We believe that this is attainable by further engineering of the plasma reactor. These efforts go beyond the scope of this manuscript, which provides the first experimental evidence of tunable plasmon resonance in graphene by its direct growth onto sufficiently small nanoparticles. The nanopowder format is advantageous for various applications since it is inherently compatible with well-established coating techniques and with top-down manufacturing schemes. Its thermal stability is also among its advantages. The area of refractory plasmonics aims at overcoming the limitations of small metallic particles with respect to compatibility with high-temperature environments and has recently attracted significant interest of the community.²¹ In Figure 4a, we show the temperature dependence of the absorption for particles produced at 3 Torr with 1 sccm of methane flow, corresponding to the “Mid CH4 (1 sccm)” spectrum in Figure 3a. The absorption has been measured at temperatures as high as 650 °C. These measurements have been performed in diffuse reflectance mode under an argon atmosphere using the Praying Mantis tool from Harrick Scientific Products Inc. We do not observe any major change in the spectra, confirming the stability of the material and its potential for the many applications that require operation at high temperatures.

Figure 4.

(a) Schematic of the diffuse reflectance measurement scheme. The samples are placed in an environmental chamber with temperature control. (b) Stacked FTIR absorption spectra acquired at increasing temperatures.

Conclusions

We have demonstrated the synthesis of silicon carbide nanoparticles exhibiting graphitic coatings with reproducible and tunable SPR absorption. A low-temperature plasma process is used to continuously and rapidly convert silane and methane into a core–shell plasmonic structure. The tunability of the absorption peak intensity was investigated and found to agree with calculations done by the equivalent dielectric permittivity model. The broadening of the absorption features compared to previous computational works arises from the lowered electron mobility and the width of the particle size distribution. The nanoparticle format makes this material particularly promising from the point of view of handling and ease of integration into functional structures. In particular, their thermal stability makes them, in our opinion, particularly interesting for application in high-temperature environments.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsomega.9b00933.

• Photograph of the continuous flow, two-stage plasma reactor (Figure S1) used to synthesize the core–shell SiC–graphene nanoparticles and TEM micrographs of the nanoparticles under consideration, for all of the conditions discussed in the paper (PDF)

The authors declare no competing financial interest.