1. Introduction

With the development of optical device miniaturization, micro high-performance emitters are required. Due to its excellent electronic and mechanical properties, graphene can sustain a colossal current in an atomically thin monolayer, which makes it one of the prominent candidates for micro emitting sources through Joule heating [1–4]. However, the emissivity of the suspended monolayer graphene is very low according to Kirchhoff’s law, because its light absorptivity is only 2.3% in the visible to near-infrared range [5,6]. In addition, in order to obtain high emission intensity in the visible wavelength, graphene need to be heated up to a high temperature (1500–3000 K), leading it in the tremendous risk of oxidation [7–9]. Hence, improving the emissivity of graphene and preventing its oxidation are critical for visible graphene-based thermal emitters.

Coupling graphene with various resonant structures are effective ways to enhance the graphene’s absorptivity (i.e. emissivity, according to Kichhoff’s law). Experimental absorption improvement demonstrations include Fabry-Pérot cavity [10,11], photonic crystal cavity [12], guided mode structure [13,14], metal plasmonic grating [15], metal-photonic crystal hetero-structure [16], and some have been successfully developed to tailor the thermal emission spectra of graphene [17–25]. For example, planar micro-cavity consisting of metallic mirrors could tune the graphene thermal light emission peak at near-infrared spectra [23]. And a silicon photonic crystal nano-cavity could also modify the graphene emission spectrum from flat line to the spectra with several narrow peaks at resonant modes [24]. Except for planar cavity and photonic crystal cavity, Geoffrey R. Nash et al. demonstrated a dual-band emitter based on multilayer graphene encapsulated by planar metamaterial in the mid-infrared and far-infrared [25].

To prevent graphene from oxidation at high temperature, it is effective to use vacuum environment or a dielectric protecting layer. Early in 2015, Young et al. realized bright visible light emission from suspended graphene in vacuum [26]. The radiation spectra showed interference between thermally emitted light from graphene and light reflection from the separate substrate surface. The local thermal luminescence at the channel restriction was also observed in hexagonal born nitride or alumina encapsulated bowknot graphene [27,28]. Benefiting from the Al₂O₃ protection, Yusuke Miyoshi et al. evenly reported ∼10 GHz graphene-on-silicon emitters in the near-infrared region in air [29].

In this work, we propose and fabricate a visible monolayer-graphene-based metal FP cavity emitter with large polarization-insensitive angle tolerance. The metal Fabry-Pérot cavity not only effectively isolates oxygen, ensuring device work at ambient, but also improves emissivity of graphene at 637 nm to 0.28, which is much higher than that of suspended graphene [26], and graphene on silica/silicon substrate [30,31]. In addition, the full width at half height (FWHW) of the emission peak of our graphene emitter is about 57 nm, which is much smaller than that of the graphene encapsulated by the h-BN or Al₂O₃ [27–29,32,33]. Through analyzing the emission spectra and Raman spectra of graphene in the emitter, we find the electron temperature of graphene surface could beyond 1200 K, about 300 K higher than the phonon temperature. Moreover, several emitters with different dielectric thickness have also been fabricated and their emission peaks red-shifts because of the increase of dielectric thickness, demonstrating emissivity wavelength tunability characteristic. The proposed emitters are entirely compatible with the bulk MOSFET process and have excellent optoelectronic application prospects for future graphene-based systems.

2. Device design and absorption measurement

The schematic and a front view of the proposed graphene-based FP cavity thermal emitter is shown in Fig. 1(a). A rectangular graphene stripe is connected with two metal electrodes, and the graphene stripe with two metal electrodes are sandwiched by two layers of Al₂O₃. A 200 nm thick gold plate is at the back side of the bottom Al₂O₃ layer to inhibit the emission transmission, and a 30 nm thick gold plate is on the top of the top Al₂O₃ layer.

 

Fig. 1. (a) The schematic and a front view of the proposed graphene-based FP cavity thermal emitter. (b) Measured absorption spectrum and simulated absorption spectra. The black line denotes the measured total absorption of the structure. The dashed blue line, simulated total absorption of the structure. The dashed red line, simulated absorption of graphene. (c) Normalized electric field intensity distribution of the emitter under normal incidence of a plane wave with on-resonant wavelength. Simulated absorption of the graphene as a function of the wavelength and incident angle for TE (d) and TM (e) polarization.

Download Full Size | PDF

The absorption and electro-thermal properties of the proposed graphene-based thermal emitter are analyzed using finite-element simulation software (Comsol Multiphysics), and the specific material parameters in the simulation are listed in the Supplement 1. Figure 1(b) shows the measured and calculated absorption spectra of the whole structure in the designed thermal emitter under normal incidence. The absorption spectrum of the fabricated emitter was measured by our microscope setup. At first, a tungsten-halogen source and a 50X objective (NA = 0.4) were used in the microscope setup to measure the reflection spectra of the graphene thermal emitter and a gold mirror. Then, the reflectivity (R) of the emitter as a function of wavelength was obtained by normalizing the reflection spectrum of the emitter with that of the gold mirror (the absorption of the gold mirror was considered in the normalizing process). At last, the absorptivity (A) of the emitter was derived based on the formula A = 1 – R, since the transmission of the emitter was blocked by the thick bottom gold layer of the emitter. Compared with simulated results, the absorption peak in measured spectrum is shifted by 2 nm and broadened by 5 nm. The differences of the measured and simulated absorption spectra could be attributed to the fabrication errors, such as the thickness deviations of the Al₂O₃ layers or the top gold layer, and the roughness of the top gold layer. Meanwhile, the reflection light collected by the 50X objective in the experiments not only includes the normal incident light, but also some oblique incident light, which leads to the broadening of the resonant absorption peak. Through measured absorption spectrum, the cavity quality factor of the fabricated cavity is 13, determined by the formula, Q = λ_cavity / Δλ_cavity, where Δλ_cavity (48 nm) is the FWHM of the peak at λ_cavity (637 nm). Figure 1(c) shows the structure’s normalized electric field intensity distribution under a normal incidence of a planer wave at the resonance condition. As we can see, the maximum electric field energy (the antinode of the optical field) locates around position of graphene, resulting in the absorption improvement of graphene. The absorption of graphene towards incident light at 637 nm is increased to 0.28, about 12 times larger than that of suspended graphene (2.3%), quantitatively consisted with the enhancement of electric field intensity.

The absorption of graphene dependent on the incident angle and wavelength for TE polarization and TM polarization are shown in Fig. 1(d) and 1(e), respectively. In the angle range from 0 to 25 degrees, the absorption rate and resonant wavelength of maximum absorption are nearly unchanged, indicating the graphene emitter is insensitive to the incident angle and polarization of light. From another point of view, heated graphene emission light is non-angle-sensitive and non-polarization-sensitive in a large range.

3. Graphene thermal emission

The proposed graphene emitters were fabricated on a 300 nm / 500 µm silica/silicon substrate. And the fabrication processes are listed in the section 2 of Supplement 1. The size of the rectangular graphene thermal emitter was 10 µm * 3 µm. The optical image of devices with the top layer was shown in the left picture of Fig. 2(a), which appears blue to the eye due to the cavity mode of 637 nm. The control group without the top gold layer shows no particular color because of featureless absorption in the visible range.

 

Fig. 2. (a)The optical photo of thermal graphene emitter with the top gold plate (left panel) and without the top gold plate (right panel). Scale bar, 5 µm. (b) Current (blue) and injected power (red) of the graphene emitter. (c) Emission spectra of the graphene emitters with (solid) and without (dot) the top gold layer at different voltages. (d) The optical picture of illuminating graphene device at V = 20 V. (e) Temperature distribution calculated by COMSOL at 20 V. (f) Normalized emission intensity of theoretical emission spectra (dash) and measured emission spectrum (solid) at 20 V.

Download Full Size | PDF

Figure 2(b) plots the current and injected power of the graphene device as a function of the applied voltage during the emission collecting process. The emission spectra of metal FP cavity under different voltage were collected by the same microscope setup for the absorption measurement and plotted in Fig. 2(c) in solid line. At low biased voltage, the emission signal from graphene is very weak and can hardly be detected. With the injected power rising, the signal intensity becomes stronger since the current Joule heating increases the temperature of graphene. As shown in Fig. 2(c), the peak emission intensity increases sharply while the peak position around 637 nm keeps unmoved when biased voltage increased from 15 V to 20 V. In addition, a bright light spot on the graphene stripe was captured by a visible CCD at 20 V as shown in Fig. 2(d). It is interesting that the light spot is not located in the middle of graphene stripe, which is related to non-uniform doping of graphene [30,34]. When large biased voltage is applied to graphene, there would be a significant potential drop along the device and thus the Femi level becomes position-dependent. The minimized carrier density point is the place where the Fermi level crossed the Dirac point, and produces the most heat due to Joule heating. Considering the doping level of graphene (4 V) and gate voltage (0 V) in experiments, Fig. 2(e) shows the calculated temperature distribution of graphene surface at 20 V by COMSOL, where the position with highest temperature is consistent with that of light spot in Fig. 2(d).

As for the control group without the top gold plate, the measured emission spectra under the same voltage are plotted in dotted line in Fig. 2(c). Due to lack of cavity-induced optical confinement, the emission intensity around 637 nm is far weak than that from the emitter with the top gold layer at the same voltage.

In order to extract the electron temperature from the emission spectra, we calculated the theoretical emission spectra from graphene at certain temperatures. The emission intensity from the graphene thermal device can be expressed as $I(\lambda ,T) = \varepsilon {I_B}(\lambda ,T)$, where emissivity ε equals the graphene absorptivity and I_B is blackbody emission intensity defined as ${I_B}(\lambda ,T) = \frac{{8\pi hc}}{{{\lambda ^5}}}\frac{1}{{\exp (\frac{{hc}}{{\lambda {k_B}{T_e}}}) - 1}}$ (Planck’s law), where h is Planck constant, c is the light speed, λ is the wavelength, k_B is Boltzmann’s constants and T_e is the electron gas temperature. We can get the ideal emission spectra of the FP emitter at different temperatures by using the simulated absorptivity of graphene in Fig. 1(b) as the emissivity. In order to clearly understand and compare the relationship between intensity and temperature in the visible range, all emission spectra have been normalized by the maximum value of their own intensity spectra from 400 nm to 900 nm. The theoretical emission spectra are plotted in dashed line and the measured emission intensity under 20 V is plotted in solid line in Fig. 2(f). As we can see, the higher is the temperature, the higher is the ratio of emission intensity at 637 nm to that at 900 nm. When the temperature below 900 K, the emission intensity of 637 nm is lower than that of 900 nm, which is contrary to the measured results under 20 V. Although measured emission peak is wider than the theoretical peak, which is probably due to combined influence of the uneven temperature distribution in the heated graphene and non-ideal optical collected system, we could draw a conclusion that the electric temperature of heated graphene under 20 V is supposed to be around 1200 K.

Next, the lattice temperature of graphene at different biased voltage was investigated by Raman spectroscopy [35–38]. We collected the Raman spectra of the light emitting point (position 1) and the position (position 2) away from the light emitting point under the same voltage. Figure 3(a) shows the derived temperature of these two points according to the 2D peak thermal coefficient [36]. As shown in Fig. 3(a), the temperature distribution along graphene is inhomogeneous and position 1 is always higher than position 2, approaching nearly 900 K at 20 V. Moreover, the temperature derived from 2D peak shift of position 1 is almost the same as that derived from G peak shift at large biased voltage, shown in Fig. 3(b), indicating the acoustic phonon is in equilibrium with photonic phonon. Generally, the 2D peak Raman shift is mainly affected by the temperature and less affected by other factors, while the G peak Raman shift is influenced by several factors, such as temperature, thermal expansion, stress and etc [35]. In the metal FP cavity, graphene is sandwiched by Al₂O₃ and the expansion or stress caused by Al₂O₃ would cancel each other out, so that the temperature is the main reason accounting for the Raman shifts of G peak. The measured temperature change trend of graphene in Fig. 3(b) is in agreement with the trend calculated by using COMSOL in Fig. 3(c), and the maximum temperature in the graphene is higher than that in the gold plates. The big temperature difference between the gold layer and graphene is caused by the following reasons. Firstly, the thermal interface resistance between the graphene and the Al₂O₃ layer is large, inhibiting the heat transfer from graphene to the Al₂O₃ layer and leading to the heat accumulation in the graphene. Secondly, the thermal interface resistance exists between the Al₂O₃ and the gold layer, which further inhibits the heat transfer in the vertical direction. In addition, the gold has a high thermal conductivity, 317 W/(m•K), inhibiting the heat accumulation in the gold layer. Therefore, the temperature of the gold layer is far below the temperature of graphene. It is worth noting that when graphene emitted visible light at 20 V as mentioned before, the maximum of electron temperature T_e derived from Fig. 2(c) could reach about 1200 K, far higher than the optical phonon temperature T_2D (derived from the 2D peak shifts) and T_G (derived from the G peak shifts) in Fig. 3(a), T_e = ∼1.4T_2D, indicating electrons and phonons are not in equilibrium [28].

 

Fig. 3. (a) Derived T_2D of two points along the graphene. (b) Derived T_G and T_2D of position 1 at different voltage. (c) The relationship between the bias voltage and maximum temperature of graphene, the top gold plate and the bottom gold plate calculated by COMSOL.

Download Full Size | PDF

The resonant emission wavelength in the metal FP cavity is sensitive to the thickness of dielectric layer. The total absorption and graphene absorption is calculated in Fig. 4(a) with Al₂O₃ thickness varying from 120 nm to 180 nm. Besides the resonant wavelength red shifting, the graphene absorption is gradually increasing from 0.2 to 0.5, about 25 times of that for suspended graphene. The absorption improvement means the emissivity improvement. Figure 4(b) shows the measured emission spectra from the graphene emitters with Al₂O₃ thickness of 130 nm, 150 nm and 170 nm. As shown in Fig. 4(a), there is an absorption enhancement trend in the structures with Al₂O₃ thickness over 150 nm when the wavelength is shorter than 600 nm. However, this absorption enhancement is mainly caused by the gold plates of the structure, rather than the graphene, so, no emission enhancement is observed when the wavelength is shorter than 600 nm as shown in Fig. 4(b). Theoretically, the emissivity enhancement of any desired wavelength could be designed. For example, with 400 nm thickness Al₂O₃ and 20 nm thickness top gold layer, the absorption wavelength could extended to 1550 nm and the emissivity of graphene could be up to 0.6.

 

Fig. 4. (a) Theoretical absorption spectra of the whole structure and graphene with different dielectric layer. (b) Measured emission spectra of graphene emitter with three different thickness dielectric layers.

Download Full Size | PDF

4. Conclusions

In summary, we proposed and demonstrated a narrow band graphene emitter based on metal FP cavity in the visible range. The demonstrated structure can significantly enhance the emission of graphene in a selective wavelength range, and can protect graphene from oxidation when graphene is heated to a high temperature. Furthermore, the temperature of electrons and phonons of graphene under high biased voltage has been extracted from the emission spectra and Raman spectra. Meanwhile, the emission wavelength of the emitter can be easily tuned by changing the dielectric layer thickness, and three graphene emitters with different Al₂O₃ thickness have been experimentally presented, whose peak emission wavelengths are 634 nm, 693 nm and 751 nm, respectively. The demonstrated emitters have broader potential applications in thermophotovoltaics, and integrated light sources for sensing and spectroscopy.