1. Introduction

Graphene has attracted much attention for its unique properties including constant optical absorption [1], high carrier mobility [2], bipolar tunability [3], and metal-liked property in the terahertz region [4]. Therefore, graphene has brilliant prospects for applications such as field effect transistors [5], integrated circuits [6], detectors [7], optical modulators [8], photonic devices [9]. In particular, plasmons in graphene nanostructures can confine electromagnetic energy on a super-diffraction scale, which has huge advantages for the development of quantum optics devices [10], photodetectors [7], optical antennas [11], modulator [12] and biosensors [13]. Besides, based on the coupling between local surface plasmon polaritons on adjacent graphene structures, wave plates [9], perfect absorption devices [14] and filter [8,15] were demonstrated to intensely modulate the propagation, absorption, and polarization of the incident light. Thus, the fabrication ability of graphene nanostructures can be regarded as the cornerstone of the development of graphene-based photonic, electronic and plasmonic devices.

The commonly used fabrication methods of structured graphene can be classified into bottom-up [16] and top-down methods [17–19]. Among them, laser direct writing (LDW), focus ion beam (FIB) etching and electron beam lithography (EBL) are most commonly used top-down methods. However, FIB etching and EBL methods suffer several knotty problems [20], including the limited patterning speed by the mutual repulsion effects of the electron/ion flux and the complex and expensive systems with vacuum environment and radiation shielding. Alternatively, LDW is an advanced green fabrication technology and is considered as one of the most promising micro-/nano- fabrication approaches [21–24], for its advantages of none-mask, no second deposition characteristic, high speed, and low cost. More importantly, super-diffraction fabrication had been realized in a variety of materials [25]. However, for all we know, super-diffraction LDW fabrication of 2D materials, such as graphene structures, which has a huge meaning for graphene optical-electronic applications, has not been realized yet.

LDW techniques include laser ablation [26], photopolymerization [27], laser-induced oxidation [28] and near-field optical lithography through near-field structures and superlenses [29]. Being attracted by the outstanding properties of LDW methods, researchers have made great efforts in structuring graphene by pulsed laser ablation LDW (PLA-LDW) [30]. R. Sahin et al. utilized the third harmonic of an infrared femtosecond laser and transformed a Gaussian beam to a Bessel beam to realize high resolution PLA-LDW fabrication of structured graphene [23]. The best fabrication resolution of the ablation lines could be reduced to 480 nm, but with poor precision (minimum width 400 nm and maximum width 545 nm), which did not break the diffraction limit yet (343 nm, third harmonics). Since both the resolution and the precision determine the final fabrication quality, a fabrication approach with poor precision is not acceptable. Therefore, it is urgent and significant to further strive for better super-diffraction LDW fabrication methods that combine high resolution with high precision.

Here, we report on a novel LDW fabrication method, which is based on laser-induced graphene oxidation (LIGO) in silica-graphene-SiC sandwich-type samples, named as LIGO-LDW. The fabrication resolution can easily break the diffraction limit without sacrificing the fabrication precision. The precision is well kept within ± 6 nm when the resolution (width of destructed lines) is 186 nm, only ± 3.2% of the resolution. Moreover, neat edge graphene structures can be achieved with this novel method without using photoresist masks thus avoiding organic pollution of the graphene structures. Such a superior method enables us to easily fabricate large scale graphene structures. Scattering scanning near-field optical microscope (s-SNOM) measurements were performed to characterize the near-field properties of the fabricated samples. They show good agreement with the simulation analyses.

2. Experiments

2.1 Super-diffraction fabrication

A processing flow chart of graphene structures is shown in Fig. 1. A polished 6H-SiC wafer is selected as substrate. Then a graphene flake grown by chemical vapor deposition (CVD) is transferred onto the SiC substrate. Finally, after depositing a silica nanofilm on the graphene by e-beam evaporation, a silica-graphene-SiC sandwich sample is prepared. Figure 1(d) shows the fabrication scheme of samples using the LIGO-LDW method. The laser used in our experiments is a Ti: sapphire femtosecond laser (λ = 800 nm, 120 fs pulse duration, 1 kHz repetition rate). A Gaussian laser beam is focused on the graphene layer of a sample through a 100 × immersion objective. The scanning speed is kept at 40 μm/s. When a femtosecond laser beam irradiates the samples, laser pulses cross over the thin silica layer and focus on the graphene layer. Redox reaction of graphene and silica are induced by the pulses in the high intensity area. The reaction consumes carbon atoms and finally forms CO₂ (or CO when carbon is abundant) and C-Si bonds or Si-Si bonds on the SiC substrate. Finally, after the silica layer is etched by a hydrofluoric acid, the graphene structure is obtained.

 

Fig. 1 The processing flow chart. (a) The substrate: 6H-SiC. (b) Graphene transfer. (c) Silica-graphene-SiC sandwich structure. (d) Fabrication scheme of the LIGO-LDW method. (e) Structured graphene.

Download Full Size | PDF

Using the LIGO-LDW method, destructed lines with different width can be fabricated in a wide range of laser power. As illustrated in Fig. 2(a), the width of destructed lines is narrowing with the decrease of the laser power. When the power is reduced to 4.9 μW, the resolution is as high as 100 nm which is only 1/8 of the working wavelength in vacuum, far beyond the diffraction limit, as shown in the rightmost picture of Fig. 2(a). Remarkably, high precision is thereby always kept around ± 7 nm for different resolutions ranging from 100 nm to 250 nm.

 

Fig. 2 SEM images of graphene nanostructures fabricated by the LIGO-LDW method and the dependence of the fabrication resolution on laser power. (a) SEM images of destructed lines with different width (dark lines in the images). (b) Experimental values (black squares) and the fitting curve (red) of the destructed linewidth as a function of laser power. (c) Large scale graphene ribbon array. The width of destructed lines is about 186 nm. (d) Large scale graphene square array. The side length of each graphene square is about 220 nm.

Download Full Size | PDF

The relationship between the width of destructed lines and laser power is shown in Fig. 2(b). The fitting curve is obtained from the formula (5) in [31]. From the fitting result, we can obtain a defocusing of about 300 nm, and an average power threshold of about 1.0 μW. This indicates the resolution could be further improved by reducing the defocusing of the sample and the average power of the laser.

Utilizing this method, one can fabricate various kinds of graphene structures. As shown in Figs. 2(c) and 2(d), we fabricated a large scale graphene ribbon array and a graphene square array by the LIGO-LDW method. In Fig. 2(c), the destructed linewidth is about 186 nm. The SEM image shows that the edges of the fabrication lines are very smooth, i.e. a high precision. The standard deviation of the fabrication lines is only ± 6 nm, just ± 3.2% of the resolution (186 nm). Moreover, the remaining graphene ribbons had no curling, which could be attributed to the silica layer above the graphene that prevented curling to happen.

2.2 Clarification of the fabrication mechanism

Raman spectroscopy, Auger electron spectroscopy (AES) and atomic force microscopy (AFM) were used to characterize the fabricated graphene structures. Figure 3(a) is the Raman spectra of the processed and unprocessed regions of the samples. It should be noted that the Raman spectroscopy has a focal spot diameter of 1.00 μm with an excitation wavelength of 532 nm, which was much larger than the width of a single destructed line. To exclude the influence of Raman excitation by a focal spot larger than the processed region in the super-diffraction fabrication area, we continuously fabricated several lines without any spacing, which combined to a wide destructed region with a width of 4.00 μm. Thus we could measure the Raman spectra of the pure destructed region. The black curve corresponding to the unfabricated region shows an obvious D peak and a 2D peak of graphene. In contrast, both peaks disappear in the red curve, which means that the graphene in the processed region has been destructed. As the graphene flake was isolated during fabrication from the atmosphere by a thin silica layer, a reaction of graphene and oxygen of the air can be excluded. Therefore, we presume that graphene reacted with silica. In addition, there might be some residue of nanostructured or amorphous carbon in the fabricated region resulting from a transformation of graphene under the action of the femtosecond laser pulses. Since the Raman excitation laser has a penetration depth of several hundred nanometers, the Raman spectra not only contains the characteristic peaks of the graphene, but also of the SiC substrate. They give rise to two characteristic SiC peaks at 1516 cm⁻¹ and 1712 cm⁻¹. Compared to these strong SiC peaks, the G peak of CVD graphene is too weak to be distinguished from the background.

 

Fig. 3 Characterization of samples fabricated by the LIGO-LDW method. (a) Raman spectra of the sample. The black and red curves correspond to the unprocessed and processed regions, respectively. Scale bar, 1 μm. (b) AFM image indicating a 1.2 nm embossment at the destructed region. (c) AES spectra of the fabricated sample.

Download Full Size | PDF

In order to further verify the mechanism of the LIGO-LDW method, AFM images of samples were measured, as shown in Fig. 3(b). One can see that the processed region had an embossment compared with the unprocessed region. The height of the embossment is about 1.2 nm, which means that the silica layer took part in the reaction and the resultant of the reaction had a good bonding with the SiC substrate. Silica can be reduced via carbothermal reduction, which is commonly used in industrial silicon preparation [32]. Since the thermal effect of our femtosecond laser pulses is negligible, the fabrication mechanism might be based on a laser-induced photochemical reaction. In other words, we can presume that the redox reaction happens between graphene and silica and the resultant is SiC and/or Si, which is determined by the ambient parameters including the laser intensity and the quantity of carbon atoms. Both the graphene and the silica are excited to excitation states via multi-photon absorption due to the high peak intensity of the femtosecond laser, and the high energy electrons in both graphene and silica break the Si-O bonds. Then, oxygen atoms of silica combine with the carbon atoms as C-O bonds and silicon atoms of silica form C-Si bonds or Si-Si bonds with the SiC substrate.

Figure 3(c) illustrates the AES spectra of the samples, which present the percentage changes of the carbon and silicon elements before and after processing. The red and black curves correspond to the processed and unprocessed regions, respectively. The statistic results show that the carbon percentage decreases, while the silicon percentage slightly increases. Thus, we conclude that graphene reacted with silica. Besides, the spectral lineshape of silicon element has an obvious change, which corresponds to the specific types of chemical bonds. The Si element only exists in the SiC substrate in the unprocessed region and the main type of bonds in the unprocessed region are C-Si bonds. The processed region has two types of bonds, the Si-Si bond and the C-Si bond, caused by the LIGO. The newly formed Si-Si bonds in the fabricated region change the line shape and peak width of the AES spectrum. The realization of super-diffraction width of the destructed lines is attributed to the reaction threshold of the LIGO: only graphene in the high intensity region of the focal spot can react with silica.

2.3 Plasmonic characteristics

Simple graphene structures play a significant role in near-field research. Different plasmon modes and dispersion relationship can be investigated through near-field detection of a simple graphene structure [4,33]. In order to study graphene plasmons more intuitively, we fabricated two crossed lines and thus formed an “X” structure, then four graphene wedges could be achieved. The width of the destructed lines is about 200 nm and the cross angle is 30°. The near-field distributions of this structure were performed by an s-SNOM [4,18]. Figure 4(a) is the near-field image under an excitation wavelength of λ₀ = 10.195 μm, and Fig. 4(b) is the corresponding SEM image. The near-field profiles taken along the corresponding dash lines in Fig. 4(a) are plotted in Fig. 4(b) to clearly show the characteristics at different locations. The white and red lines represent two different local modes of surface plasmons. And in the upper region, we can observe more peaks, which stem from the interference of plasmons.

 

Fig. 4 Near-fields of graphene wedge structures. (a) The near-field distributions of an “X” structure at an excitation wavelength of 10.195 μm. (b) SEM image. Scale bar, 500 nm. The upper near-field profiles were taken along the corresponding dash lines in (a). (c) and (d) The near-field distributions at an excitation wavelength of 9.588 μm.

Download Full Size | PDF

Figures 4(c) and 4(d) are the near-field distributions images with an excitation wavelength of 9.588 μm. Compared with Fig. 4(a), the peaks in Fig. 4(c) appear at a lower position and the spacing between two neighboring peaks decreases, which indicates out-shifts of plasmon peaks caused by the dispersion of graphene plasmons [18,34]. The excellent near field responses of this graphene wedge show that the graphene structure processed by the LIGO-LDW method can be used for near field regulation of the light field. It is a very exciting result that such plasmon modes can be observed on CVD-grown graphene samples. The near-field images suggest that the LIGO-LDW fabrication forms a perfect reflection edge. There is no curling and obvious sawteeth of the graphene edges and thus the scattering loss at the reflection edge is very low. This is attributed to the special sandwich structure of the samples and the high precision of the fabrication.

Figure 5 shows the plasmonic characteristics of graphene ribbons with different widths at an excitation wavelength of 9.588 μm. The SEM images and corresponding near-field distributions of graphene ribbons are shown in Figs. 5(a) and 5(b). All near field signals here are normalized to the near field signal of graphene far away from the boundary. Figures 5(c)-5(f) shows the near-field profiles extracted along the white dashed lines from left to right in Figs. 5(a) and 5(b). The thickness of graphene is set as d = 0.34 nm in the simulation. The Fermi energy level is E_F = 0.38 eV and the carrier mobility is 1000 cm²/Vs. We set the processed area as the silicon dioxide (1.2 nm), for the surface Si or SiC is easily oxidized into SiO₂ in the air. A dipole is used to act as the exciting source and it is placed 150 nm above the sample. And then we solved the electric field intensity of E_y 15 nm above the sample. The simulated peaks are in good agreement with the experimental results inside the graphene ribbons, while the changing trend of simulated plasmon intensity is different from that of the measured intensity outside the graphene ribbons. This difference is mainly due to the increased absorption by adsorbed impurities in the processed area. In the simulation, we take the edge of the graphene as a truly isolated boundary and do not take into account the surrounding amorphous carbon or other impurities. From the fitting results, we can obtain a plasmon wavelength of ${\lambda }_{\text{p}}=2\pi /\mathrm{Re}\left(q_p\right)$ that is approximately equal to 230 nm ($q_p$ represents the wave vector of plasmon). The fitted damping ratio ${\gamma }_p=\mathrm{Im}\left(q_p\right)/\mathrm{Re}\left(q_p\right)$ is 0.135, which is comparable to previous results [33,35].

 

Fig. 5 Plasmonic characteristics of graphene ribbons. The linewidth of the destructed graphene (dark regions) is 230 nm. The incident wavelengths λ₀ is 9.588 μm. (a) SEM and near-field images of graphene ribbons with widths of 480, 430, 370, and 330 nm. (b) SEM and near-field images of graphene ribbons with widths of 150, 140, 80, and 70 nm. (c)-(f) Experimental (black curves) and simulated (red curves) near-filed profiles perpendicular to graphene ribbons along the white dashed lines from left to right in (a) and (b).

Download Full Size | PDF

With the decreasing of graphene ribbon width, the number of plasmon peaks is reduced from four to one. The principal plasmon peaks are marked with blue arrows, and the secondary peaks are marked with green arrows, which are weaker than the principal ones due to propagation loss [19,36]. Distinctly, a strong resonance enhancement of the plasmon intensity is observed in Fig. 5(f). The peak intensity of Fig. 5(f) appears to be about the sum of the two principal peaks of Fig. 5(c). When the graphene ribbon width drops to 80 or 70 nm, the plasmon intensity has an obvious decease. This means that one can modulate the intensity of the plasmons by changing the width of the graphene strip.

3. Conclusions

In conclusion, we report a novel LIGO-LDW method to fabricate structured graphene beyond the diffraction limit. The mechanism is based on laser-induced redox reaction of graphene and silica in samples with a specific sandwich structure. Super-diffraction fabrication of graphene structures can be reached with high resolution as well as high precision. The width of the destructed lines could be reduced to as narrow as 100 nm with a relative precision of around ± 7%. Benefiting from the advantages of the LIGO-LDW method, the fabricated graphene structures have very smooth edges. Excellent near-field distributions of graphene plasmons were observed experimentally in these structures, which verified that our method should be useful for the development of graphene integrated plasmon circuits and devices. Moreover, we show that a chemical reaction can be confined to a nano-scale region by a laser beam. Modified methods can be applied to fabricate nanostructures of other 2D materials via proper chemical reactions.