1. Introduction

Conventional thermal emitters exhibit a blackbody-like broad emission spectrum, although various applications require only a specific spectral component. For example, in the case of infrared sensing of gases or chemical substances, which usually exhibit narrow-linewidth absorption, the required spectral range is typically less than 1/50 of the blackbody spectrum [1,2], corresponding to a Q factor of >50. The wasted emission outside the targeted spectral range decreases the power utilization efficiency, and therefore, there is strong demand for narrow-linewidth (high-Q) thermal emitters that can efficiently convert the input power into thermal emission power within the targeted spectral range.

According to Kirchhoff’s radiation law, the thermal emission spectrum of an object can be controlled by tailoring the absorptivity spectrum of the same object, and various optical resonators, such as metallic photonic crystals (PC) and metamaterials [3–8], have been utilized to achieve narrowband absorption (thermal emission) at the targeted frequency. However, these metallic resonators suffer from low-Q emission peaks (Q<10) and broadband background emission. The first experimental demonstration of efficient narrowband thermal emission was performed by our group [9]. In that demonstration, narrow-bandwidth intersubband absorption of multiple quantum wells (MQWs) was combined with the optical resonances of an airhole-type PC slab. The fabricated MQW-PC device exhibited a medium-Q (Q~30) thermal emission peak with several minor peaks, and the emission intensity of the main peak under a given heating power was found to be 4-times larger than that of a reference blackbody emitter. To further improve the power-utilization efficiency, a single-mode high-Q (Q>100) thermal emitter based on MQWs and a rod-type PC slab was also proposed [10] and experimentally demonstrated [11]. However, the peak emissivity of the fabricated device was not so high (~0.4), and a discussion of the power consumption was not possible because an external heater was used to raise the temperature of the emitter.

In this paper, we develop a single-mode high-Q (Q>100) MQW-PC thermal emitter which can be heated by current injection, and we demonstrate a drastic improvement in power-utilization efficiency compared with a reference blackbody emitter. To realize efficient conversion of the heating power into the thermal emission from the PC slab, we carefully design the PC structure and the supporting frame/wires to minimize thermal radiation loss outside the targeted spectral range and the other thermal losses such as heat conduction loss.

This paper is organized as follows. In the next section, we explain the device structure and show the measured thermal emission spectra of the fabricated device under current-injection heating. In Section 3, we compare the experimental results with the calculated results obtained by taking account of various types of thermal losses, and we discuss the power consumption of the fabricated device in detail. We summarize our results in Section 4.

2. Experimental results

Figure 1(a) shows the basic structure of a single-mode high-Q thermal emitter, which consists of a square-lattice PC slab with GaAs (6.8 nm)/Al_0.3Ga_0.7As (13.0 nm) MQW rods. The Al_0.3Ga_0.7As layers of the MQWs are doped with silicon at a density of 1.0 × 10¹⁷ cm⁻³ to induce an intersubband transition (ISB-T), which yields a narrowband absorption coefficient spectrum in the mid-infrared range. The upper panel of Fig. 1(b) shows the measured transmission spectrum of the flat MQW wafer for edge-incident light with p-polarization, normalized by the spectrum obtained with s-polarization, indicating absorption by the ISB-T at a wavenumber of 1063 cm⁻¹ with a FWHM of 75 cm⁻¹. The PC slab contains two MQW rods with different radii (r₁, r₂) in its square unit cell [10,11]. This structure supports only one (doubly degenerated) optical resonant mode at the Γ point of the photonic band within the frequency range of the ISB-T absorption spectrum, which yields single-peak polarization-independent thermal emission in the surface-normal direction. The emissivity spectrum from a resonant mode at an angular frequency ω_res is analytically given by

 

Fig. 1 (a) Schematic diagram of a single-mode high-Q thermal emitter composed of GaAs/AlGaAs MQWs and a square-lattice rod-type PC slab. (b) Upper panel: the transmission spectrum of the flat MQW wafer. Lower panel: the emissivity spectrum of a fabricated PC slab (N_w = 20, h = 2.4 µm, t = 0.8 µm, a = 7.4 µm, r₁ = 0.14a, r₂ = 0.13a) in the surface-normal direction, measured with a temperature-controlled external heater at 200 °C.

Download Full Size | PDF

The whole structure of a high-Q thermal emitter which can be heated by current injection is shown in Fig. 2(a) and consists of a 2.4 mm × 2.4 mm PC with MQWs, a GaAs substrate frame for supporting the PC slab, and electrodes/electric wires for current injection. Since the GaAs layer under the MQW rods in the PC region was undoped, we left the surrounding doped MQW regions (0.4 mm from the sample edges) unetched to ensure that electric current flowed there. The areas of the PC and the entire device are the same as those used in our previous work [9]. In order to heat this device efficiently with small input power, it is important to suppress unwanted thermal losses, such as thermal conduction loss and thermal radiation loss outside the PC slab [9,12]. To reduce the former loss, the device was suspended using two thin nichrome wires (diameter, 15 µm). We chose nichrome instead of the manganin used in the previous experiment [9] because the thermal conductivity of nichrome is smaller than that of manganin, and thinner nichrome wires are available owing to this material's larger tensile strength, enabling further reduction of the conduction loss. To decrease the latter loss, we reduced the volume of the GaAs substrate frame and the area of the electrodes as much as possible since these elements can generate broadband emission stronger than the PC emission at an elevated temperature, as we have previously confirmed [13]. Figures 2(b) and 2(c) show microscope images of the top side and bottom side of the fabricated emitter (test device without electric wires). Although a Ge/Au/Ni/Au alloy is usually used for ohmic contacts for n-doped GaAs [14], here we deposited 200 nm-thick Au because it exhibited lower emissivity than the Ge/Au/Ni/Au alloy and still enabled quasi-ohmic current injection when annealed [15]. The GaAs substrate was mechanically polished to a thickness of ~55 µm, and was then removed by chemical wet etching except for a small supporting frame with a width of ~50 µm. Finally, nichrome wires were attached to the emitter, and the device was installed in vacuum (< 0.01 Pa) to suppress thermal convectionlosses due to air. Figure 2(d) shows an infrared camera image of the fabricated device heated by current injection in vacuum. Enhanced thermal emission is uniformly obtained from the PC slab. The unwanted thermal emission from the edge of the GaAs substrate still exists, but is weaker than in the previous emitter [13] owing to the reduction of the frame volume.

 

Fig. 2 (a) Schematic diagram of a high-Q thermal emitter which can be heated by current injection. (b)(c) Microscope image of the top-side and bottom-side of the fabricated emitters. (d) Infrared camera image of the fabricated emitter heated by current injection with an input power of 1.5 mW.

Download Full Size | PDF

Figure 3(a) shows the thermal emission spectra of the fabricated emitter (N_w = 20, h = 2.6 µm, t = 0.6 µm, a = 7.3 µm, r₁ = 0.14, r₂ = 0.13a), heated by current injection with various electrical powers (P_in). As the electrical power increases, the emission intensity of the main emission peak increases owing to the increase of the emitter’s temperature. Figure 3(b) shows the central wavenumber and the Q factor of the main emission peak shown in Fig. 3(a) as a function of the electrical power. The central wavenumber gradually decreases as the emitter’s temperature increases owing to the increase of the refractive index of the material (GaAs) [11]. The Q factor of the peak, on the other hand, slightly increases from 86 to 105 as the temperature rises. According to Eq. (1), the Q factor of the peak is given by (Q_abs⁻¹ + Q_rad⁻¹)⁻¹, which becomes large as either Q_abs or Q_rad increases. Since Q_rad is temperature-independent, the result shown in Fig. 3(b) suggests that Q_abs increases (in other words, the ISB-T absorption coefficient decreases [16]) at higher temperature. As the temperature rises, the difference of the electron density in the lower and upper subbands decreases while the scattering rate of the electrons in the subbands increases. Since the ISB-T absorption coefficient at the transition wavelength is proportional to the former and inversely proportional to the latter, the increase of the device temperature causes the decrease of the ISB-T absorption coefficient (and the increase of Q). It should be also noted that the reduction of the ISB-T absorption coefficient at higher temperature causes a slight decrease in the peak emissivity due to the mismatch between Q_abs and Q_rad. For example, we experimentally confirmed that the peak emissivity of the device at 300 °C (0.56) was smaller than that at 200 °C (0.70) by using a temperature-controlled heater.

 

Fig. 3 (a) Measured thermal emission peak of the fabricated device in the surface-normal direction at various electrical input powers. (b) Central wavenumber and Q factor of the emission peak shown in Fig. 3(a) as a function of the electrical power. (c) Relationship between the device temperature and the electrical power. Red: rod-type PC, blue: airhole PC [9], black: reference blackbody [9]. The error bars show the estimation errors discussed in the main text. The three emitters have light-emitting surfaces with the same area (5.76 mm²). (d) Measured thermal emission spectra of the rod-type PC and the reference blackbody in the surface-normal direction at the same electrical power (2.26 mW).

Download Full Size | PDF

It is important to estimate the temperature of the emitter during current-injection heating. In the previous demonstration with an airhole-type PC slab [9], we coated a small area of the supporting frame with blackbody paint and directly measured the temperature with an infrared camera. It is difficult to apply the same method to our new device because the linewidth of the emission peaks of the rod-type PC is so narrow that even a small area of blackbody paint may generate thermal emission loss comparable to the emission power of the PC slab. Therefore, we estimated the device temperature from the wavenumbers of the main emission peak shown in Fig. 3(b) by using a relationship calibrated with a temperature-controlled external heater. The error of the estimation in this method was within ± 5°C, considering the wavenumber interval in the measurement (0.5 cm⁻¹) and the temperature dependence of the wavenumber (−0.048 cm⁻¹/K). The estimated device temperature is shown with a red line in Fig. 3(c). The previous results obtained with an airhole-type PC slab and a reference blackbody (both emitters have the same device size) [9] are also shown in the same figure with blue and black lines, respectively. Although the reference blackbody has a larger substrate than the new PC emitter, the comparison is still valid since the temperature of the reference blackbody is determined by the blackbody emission power, which is one to two orders of magnitude larger than the thermal losses from the substrate. Under a given electrical input power, the new emitter reaches the highest temperature among the three emitters. For example, at an input power of 2.26 mW, the temperatures of the new emitter (rod-type), the previous emitter, and the reference blackbody are 296 °C, 150 °C, and 48 °C, respectively. The reason for this is the fact that the rod-type PC exhibits single-mode high-Q emission with almost no background emission, as shown in Fig. 1(b), and the thermal losses from the substrate and electric wires are reduced, as we have mentioned above. Figure 3(d) shows the measured thermal emission spectra of the rod-type PC and the reference blackbody in the surface-normal direction at the electrical power of 2.26 mW. The rod-type PC emitter shows a single sharp emission peak at a wavenumber of 1068 cm⁻¹ with well-suppressed background emission, and its peak intensity is 12.6-times larger than that of the reference blackbody at the same wavenumber owing to the larger increase of the device temperature, as shown in Fig. 3(c).

3. Discussion

In this section, we numerically evaluate the thermal emission power and other losses of the fabricated PC emitter.

First, we confirm the validity of thermal emission enhancement in Fig. 3(d) by using Planck’s equation. The ratio between the thermal emission intensity of the PC emitter in a given direction to that of the reference blackbody in the same direction is given by

Next, we evaluate the hemispherical upward (downward) thermal emission power from the fabricated PC emitter by using the following equation:

Here, S denotes the area of the PC slab, and ${\epsilon }^{U(D)}(\omega ,\theta ,\varphi )$ denotes the emissivity for the upward (downward) emission with angular frequency ω, zenith angle θ, and azimuth angle $\varphi$, which can be calculated by rigorous coupled wave analysis. Substituting the experimental parameters into Eq. (3), the upward and downward thermal emission powers of the fabricated PC emitter are estimated to be 0.20 mW and 0.07 mW, respectively.

The rest of the input power is lost by thermal conduction through the nichrome wires and thermal emission of the substrate frame, as we have pointed out above. To evaluate the thermal conduction loss through the wires, we calculate the temperature distribution, T(x), of the electric wires by solving the following equation:

The thermal emission loss of the substrate frame can be further classified into two. One is the unavoidable thermal emission originating from lattice absorption due to multiple phonon interactions in the GaAs substrate at wavelengths longer than 18 µm [18]. It should be noted that this emission cannot be directly observed with our infrared camera since its measurement wavelength range is limited to 7.8–14 µm. According to [18], the absorption coefficient at the wavelengths range of 18 µm < λ < 28 µm is 30–50 cm⁻¹ at room temperature, and it increases as the temperature rises due to the increase in phonon density. Considering the relatively thick substrate frame (>50 µm), as well as absorption enhancement due to random light scattering [19] at the rough side-edges of the frame, here we assume that the thermal emission of the frame at these wavelengths is equal to that of the opaque bulk GaAs, which has a constant emissivity of 0.714 ( = Fresnel reflectivity of 0.286). Using Eq. (3) within the wavelength range of 18 µm < λ < 28 µm (350 cm⁻¹ < ω < 550 cm⁻¹), the estimated emission loss is 0.46 mW at 296°C. Note that the emission at the wavelengths longer than 28 µm (the reststrahlen band of GaAs) should be small due to the high reflectivity and the weak emission intensity of a blackbody; if the maximum emissivity (1.0) is assumed there for the worst case, the additional emission loss is 0.25 mW. The other loss is the unwanted thermal emission from the side-edges of the substrate frame at wavelengths shorter than 18 µm, which is strongly observed in the infrared camera image shown in Fig. 2(d). Since GaAs is basically transparent at these wavelengths, except for a small lattice absorption at 13 µm [18], the above emission probably comes from the ISB-T of the unetched MQWs and from the backside of the Au electrodes. The emissivity of the side-edges is estimated from the measurement using the infrared camera to be 0.05–0.17 (dependent on the location). The estimated emission loss at wavelengths shorter than 18 µm is 0.80 mW when an average emissivity of 0.11 at these wavelengths is assumed in Eq. (3). If the lower and higher emissivities (0.08 and 0.14) are assumed, the power is estimated as 0.58 mW and 1.02 mW, respectively.

Table 1 summarizes the estimated power consumption of the fabricated emitter. Here, not only the above-mentioned uncertainties in the various types of thermal losses but also the uncertainties in the PC emission power due to the measurement error of the device temperature ( ± 5°C) are taken into account, and the range of the estimation (smallest – most likely – largest) is included. As seen in the table, the sum of the estimated power consumption calculated with the most likely parameters (2.31 mW) is in fair agreement with the experimental result (2.26 mW). In this case, the ratio of the PC emission to the calculated total input power is about 12%. This value is still lower than the highest room-temperature wall-plug efficiency (WPE) of continuous-wave quantum cascade lasers (QCLs) developed in the lab (21%) [20], but is higher than those of some commercially available QCLs (1~5%) [21]. Our narrowband thermal emitters will be especially suitable for applications where small total power consumption (<10 mW) is demanded such as portable gas sensing, since a WPE of QCLs drops when the input power is small due to the existence of the lasing threshold (note that the highest wall plug efficiency in [20] was realized at the input power of over 10 W).

Table 1. Calculated power consumption of the emitter. For each calculation, the range of estimation (smallest – most likely – largest) is listed.

View Table

To realize more efficient thermal emitters, the frame emission, which occupies more than half of the input power in the fabricated device, should be reduced as much as possible. One simple solution is the increase of the device size; the ratio of the PC emission to the other losses should be increased by increasing the size of the PC slab, L_slab, because the former is proportional to (L_slab)², whereas the frame emission is proportional to L_slab and wire losses are unchanged. For example, if we double the size of the PC, the ratio of the PC emission will be doubled, as shown in the table. It should be noted, however, that the photonic crystal membrane will become fragile if the device size is too large. Another possible method of increasing the efficiency without losing the mechanical strength is the transfer of the photonic crystal membrane to an infrared-transparent substrate such as NaCl, which has no lattice absorption and serves as a cladding layer for PC resonances due to its low refractive index. The use of a cold-side one-dimensional PC which reflects back unnecessary emission to thermal emitters [22] is also a promising way to reduce thermal emission loss from the substrate frame. Combining these methods, monochromatic thermal emitters that have the highest power-utilization efficiency among all the existing mid-infrared light sources can be realized.

4. Conclusion

We have developed a single-mode high-Q (Q>100) mid-infrared thermal emitter based on a rod-type PC slab interacting with GaAs/AlGaAs MQWs. The emitter showed a 12-fold increase in peak emission intensity under a given electrical input power compared with a reference blackbody emitter due to the efficient increase of the device temperature. We have also numerically demonstrated that the ratio of the narrowband emission power to the electrical input power in the fabricated device was higher than that of some commercial mid-infrared light sources, and could be further increased by reducing the thermal losses from the supporting frame. We hope that our demonstration will contribute to the development of highly-efficient mid-infrared light sources for various sensing applications.