1. Introduction

Quantum-cascade lasers (QCLs) have been remarkably improved [1,2] since their first demonstration in 1994 [3] and have been become the leading type of mid-infrared light source in the areas of atmospheric detection, medical diagnosis and spectroscopic application. Most applications are based on the atmospheric windows of 3–5 µm [4,8] and 8–12 µm [9,13]. Outside these atmospheric windows, few QCLs with high continuous-wave (CW) performance have been explored. Some bending-vibration characteristic absorption peaks, such as the in-plane bending vibration of the methyl group (1430 cm⁻¹) and asymmetric deformation vibration of the aldehyde group (1450 cm⁻¹), are within 6.3–7.5 µm. The detection of these functional groups requires QCLs having both a broad-tuning range and high output power because of the strong absorption of atmospheric water.

The strategy of designing QCLs at a wavelength of ∼7 µm for high output power is different from that of designing devices at a wavelength of 4–5 µm. First, the energy spacing is reduced so that the non-radiative transition between the upper and lower laser levels is intensified. To maintain population inversion, some long-wavelength lasers adopt diagonal transition as an active region scheme. The consequence is that the transition matrix element corresponding to the stimulated emission may decrease and the gain may thus be insufficient. Second, thermal backfilling and leakage are greater than for medium-wavelength QCLs owing to the shortened energy spacing in the miniband. Third, the free carrier absorption is proportional to the wavelength, and the waveguide loss is thus larger than that for medium-wavelength QCLs. The maximum CW output power at a wavelength of 7.1 µm reported in the literature is 1.4 W at room temperature [14]. The active region is based on a three-phonon-resonance design. This active region has a shorter lifetime of carriers at the lower energy level than the traditional single-phonon extraction active region. A broadband tunable QCL, having a wavelength of 6.8 µm and an anticrossed dual-upper-state to multiple-lower-state design, has been reported to have a tuning range exceeding 240 cm⁻¹ under CW operation at 280 K [15]. The maximum CW power of this QCL in external-cavity (EC) mode is ∼100 mW. Limited by the thermal performance of the AlN heat sink, the maximum output power in the Fabry–Pérot mode is 0.7 W at 280 K.

Besides the dual-upper-state design, many other designs, such as the bound-to-continuum and continuum-to-bound designs, have been implemented for broad-gain operation. The composite well design [16], which is also used for heterogeneous QCLs, improves the accuracy of the wavelength and gain and has good tuning performance from 5.2 to 11 µm. In a standard molecular beam epitaxy reactor, changing the effusion cell temperature leads to uncertainty in the composition and thickness. For gallium sources, the uncertainty in composition and thickness directly leads to unevenness of the quantum well thickness in the gain region, which is likely to affect the effective mass of electrons and the energy band structure. However, uncertainty in the composition and thickness of InAlAs barriers has little effect on band structure for long-wavelength infrared QCLs. In addition, the aluminum beam flux is more sensitive to temperature changes. Thus, the aluminum effusion cell reaches a new beam flux and remains stable over a shorter period of time after a small temperature change. It is thus feasible to design a multi-stack-core active region with different aluminum components for long-wave infrared QCLs. Recently, a multi-stack-core design having a varying active layer thickness was investigated for its broad gain and reduced gain dispersion [17]. Symmetric heterogeneous QCLs have been used in EC-QCLs, achieving a tuning range from 7.6 to 11.4 µm [18]. Although multi-stack-core designs have been shown to have high tuning performance in recent years, high-CW-power QCLs based on such active region design have rarely been reported. The peak gain of such QCLs is insufficient, mainly owing to the number of periods of individual sub-cores being small because the number of cascade periods that can be accommodated in the QCL is limited, and the threshold current density is thus higher than that of the single-stack structure. The double-stack active region design reported in this work for different aluminum compositions allows a greater number of periods of individual sub-cores (25 stages in a sub-core) compared with the multi-stack-core design, resulting in low-threshold and high-power performance.

In this paper, we report watt-level and broad-gain QCLs emitting at 6.9 µm based on the double-stack design, realized by changing the beam flux of the aluminum effusion cell only. The tuning range of the Littrow EC mode is determined to be 195 and 104 cm⁻¹ for pulsed (1 µs, 50 kHz) and CW modes at room temperature, respectively. The maximum CW power at 293 K is 1.07 W with a threshold current density of 1.4 kA/cm². The CW output power of the EC mode exceeds 110 mW over the entire spectrum. The QCLs can thus be used to detect some functional groups.

2. Laser design and fabrication

The active region of our QCLs is based on a double-stack design with strain-balanced In_0.60Ga_0.40As/In_0.43Al_0.57As and In_0.60Ga_0.40As/In_0.41Al_0.59As (as shown in Fig. 1(a)) material systems. The layer sequence of one period of the core In_0.60Ga_0.40As/In_0.41Al_0.59As in angstroms, starting from the injection barrier, is 40.9 / 17.4 / 9.2 / 51.8 / 9.2 / 48.1 / 10.2 / 40 / 18.4 / 32.8 / 17.4 / 28.7 / 19.4 / 27.7 / 28.6 / 26.6. The layer sequence of one period of the other core In_0.60Ga_0.40As/In_0.43Al_0.57As in angstroms, starting from the injection barrier, is 40 / 17.4 / 9 / 51.8 / 9 / 48.1 / 10 / 40 / 18 / 32.8 / 17 / 28.7 / 19 / 27.7 / 28 / 26.6. The InAlAs barriers are given in bold font and the InGaAs wells in normal font. The underlined layers are doped at n = 2 × 10¹⁷cm⁻³. The conduction band of one stage is under an electric field of 70 kV/cm. The design of the active region has two advantages. First, the energy spacing between the upper laser level and the first state above it is $\Delta $E₄₃ = 65 meV. Thus, the leakage current can be further reduced. Second, multi-injection energy levels can result in lower voltage defects [19]. As shown by the last quantum well of the injection region in Fig. 1(a), the overlap of wave functions of many injection energy states is modestly greater than that of the traditional design. Carriers are injected to the upper laser level from many injection energy levels instead of the ground state only, and the voltage defects are thus lower.

 

Fig. 1. (a) Schematic of the conduction band of one stage and relevant wave functions corresponding to energy levels under an applied electric field of 70 kV/cm. The dashed and solid lines denote the In_0.60Ga_0.40As/In_0.41Al_0.59As core and the In_0.60Ga_0.40As/In_0.43Al_0.57As core, respectively. The difference value of the conduction band offset for the two cores is 40 meV. The optical transition energy difference is 8 meV. (b) Experiment and simulated X-ray diffraction spectra of a double-stack laser at a wavelength of ∼6.9 µm

Download Full Size | PDF

We use a double-stack design with different strain values to obtain a wide tuning range while maintaining high output power [20,21]. The strain values of the InGaAs wells of the two cores are the same (−0.45%) and the those of the InAlAs barriers of the two cores are 0.62% and 0.76%, respectively. After the growth of the first core, the shutters of the group III source effusion cells are closed. The temperature of the aluminum effusion cell rises by approximately 5 K such that the aluminum beam flux fits into the other core. It takes approximately 1–2 minutes to reach the new temperature. However, to prevent possible fluctuation of the Al beam flux, we wait a little longer so that the aluminum beam flux is stabilized. 5 minutes is sufficient for growth interruption. This increases the Al ratio by 0.02. The higher Al composition adds 40 meV to the conduction band offset and 8 meV to the optical transition energy, as shown in Fig. 1(a). The transition energy shift of the second core induces a gain blue-shift of 60 cm⁻¹, which contributes 30 cm⁻¹ to the gain bandwidth expansion.

The Al composition design is verified by the X-ray diffraction spectra shown in Fig. 1(b). The experimental and simulated X-ray diffraction spectra of our 50-stage strain-compensation QCL structure are in good agreement with the envelope function and indicate the existence of two cores with good periodicity. The intensity and position of the peaks relate to the layer thicknesses and material composition of the heterostructure. Although the linewidth of the peaks is modestly wider than that of the stimulation, appropriate interface roughness is beneficial to spectral broadening [22].

The epitaxial layer sequence, starting from the substrate, is an InP buffer layer (Si, ∼3 × 10¹⁶cm⁻³, 4.5 µm), two 25-stage laser cores (Si, ∼2 × 10¹⁷cm⁻³), InGaAs layer (Si, ∼4 × 10¹⁶cm^−-3, 0.3 µm) that improves the optical confinement of active regions, InP cladding layer (Si, ∼3 × 10¹⁶cm⁻³, 4.5 µm), graded doped InP layer (Si, ∼1.5 × 10¹⁸cm⁻³, 0.8 µm) and highly doped InP cap layer (Si, ∼5 × 10¹⁸cm⁻³, 0.2 µm). The active region of the two cores and InGaAs layer are grown on an (n-type) InP buffer layer through molecular beam epitaxy in a single growth step. Metal organic chemical vapor deposition is adopted to grow the InP buffer layer, InP cladding layer, graded doped InP layer and highly doped InP cap layer.

The wafer is processed into a buried ridge with a ridge width of 7.7 µm. Semi-insulated InP:Fe is grown on either side of the ridge to reduce the temperature of the active region through metal organic chemical vapor deposition. The laser is mounted epi-side down on a diamond submount with indium soft solder for CW operation. The diamond heatsink is then mounted on the copper heatsink.

3. Laser characterization

The electroluminescence spectra of our laser (where the rear cavity surface is destroyed to suppress resonance) are measured for pulsed operation (2 µs, 50 kHz) with resolution of 6 cm⁻¹ at room temperature as shown in Fig. 2(a). The spectra are corrected for atmospheric water absorption. The full-width at half-maximum of the electroluminescence spectra is 293 cm⁻¹, which is wider than that of the bound-to-continuum design at a wavelength of ∼7 µm [23]. As shown in Fig. 2(b), the sub-threshold amplified spontaneous emission (ASE) spectra are measured for CW operation at room temperature. Figure 2(c) is the local graph of ASE (from 1438 to 1447 cm⁻¹), showing that the longitudinal mode spacing is 0.362 cm⁻¹. As the cavity length is 4.13 mm, we obtain the effective group index [24] as 3.34.

 

Fig. 2. (a) Electroluminescence spectrum measured in pulsed operation (2 µs, 50 kHz) at room temperature. The rear cavity surface of the laser is destroyed. (b) ASE spectrum measured with a resolution of 0.125 cm⁻¹ in CW operation at room temperature. CL refers to the as-cleaved facet of the cavity. (c) Local ASE spectrum (from 1438 to 1447 cm⁻¹). (d) Interferogram of the ASE spectrum.

Download Full Size | PDF

In obtaining the accurate shape of the gain spectrum, we use a Fourier-transform infrared spectrometer (Nicolet 8700) to acquire the interferogram shown in Fig. 2(d). The zeroth-order peak is generated by the interference of the light emitted by the front cavity surface with itself. The high-level interference peak is generated by the interference of the light emitted from the front cavity surface and the light after circulation in the cavity. The burst is then Fourier transformed to a corresponding ASE spectrum. The ratio of two adjacent ASE spectra follows the relation [25,26]

4. Laser performance

We obtain single-mode spectra under different grating angles as shown in Fig. 3 by employing the Littrow configuration the Littrow configuration. An anti-reflective coating is applied to the laser having a cavity length of 4 mm to lower the reflectivity on the front facet. The widest tuning range of 1556 to 1361 cm⁻¹ (from 6.42 to 7.34 µm) is achieved in pulsed operation (1 µs, 50 kHz) with a current density of 2.37 kA/cm² at room temperature. The application of an EC laser at a wavelength of ∼7 µm requires high output power because of H₂O absorption. Figure 4 presents the EC mode spectrum corresponding to the CW power. The widest tuning range of 1492 to 1388 cm⁻¹ (from 6.7 to 7.2 µm) is achieved with a current density of 2.44 kA/cm² at room temperature without an anti-reflective coating. The tuning performance is better than that of the bound-to-continuum design (tuning range of 86 cm⁻¹) without an anti-reflective coating operating in CW mode reported in the literature [23]. In addition, the light power exceeds 110 mW over the entire spectral and the maximum CW power is 165 mW at a wavelength of 7 µm. Although the CW spectral coverage is narrower than the tuning range of the dual-upper-state design [15], the output power achieved in this work is higher, and the lasers can thus be used in the specific situation that high output power is required.

 

Fig. 3. Tuning performance of the Littrow EC QCL, with a 4-mm-long cavity including an anti-reflective coating, taken at room temperature with a duty cycle of 5% (1 µs, 50 kHz) at 2.37 kA/cm², with a wavelength tuning range of 195 cm⁻¹ (from 1556 to 1361 cm⁻¹). CL refers to an as-cleaved facet of the cavity. AR refers to the anti-reflective coating.

Download Full Size | PDF

 

Fig. 4. (a) Power–current–voltage characteristics at 298 K in CW operation. CL refers to an as-cleaved facet of the cavity. (b) Wavelength tuning of 104 cm⁻¹ (from 1492 to 1388 cm⁻¹) in CW operation at room temperature without an anti-reflective coating. The CW power is also plotted as a function of the emission wavelength. The black line represents the net gain as calculated from an interferogram.

Download Full Size | PDF

Figure 5(a) and (b) respectively shows CW and pulsed (1 µs, 10 kHz) power–current–voltage characteristics for a 7.7–µm–wide, 6–mm–long, highly–reflective–coated and anti–reflective–coated, buried-ridge laser mounted epi-side down on a diamond heatsink. The laser is fixed on a thermoelectric cooler for temperature control and operated at temperatures from 293 to 393 K at intervals of 20 K. The CW power is measured using a pyroelectric power meter. The maximum power at 293 K is 1.07 W, with a current density of 3.2 kA/cm².

 

Fig. 5. (a) Power–current–voltage characteristics at different temperatures in CW operation. AR refers to the anti-reflective coating and HR refers to the highly-reflective coating. (b) Power–current–voltage characteristics at different temperatures in pulsed operation (1 µs, 10 kHz). (c) Spectrum of the QCL at different current densities at room temperature in CW operation for a 4-mm-long cavity without an anti-reflective or highly reflective coating. (d) Characteristic temperature of our 6-mm-long, 7.7-µm ridge wide, anti-reflective- and highly-reflective-coated laser in pulsed operation (1 µs, 10 kHz).

Download Full Size | PDF

Watt-level CW power is mainly achieved through the design of the active region of the laser. First, high $\Delta $E₃₄ greatly reduces the leakage of shunt-type carriers [27], which is the main form of leakage within the active region. Second, in the traditional design, the temperature of active region is higher than the environment so that the electrons in the ground state of the injection region are excited to other energy states in the miniband of the injection region, resulting in lower efficiency of injection to the upper laser level. However, as shown in Fig. 1(a), carriers are injected to the upper laser level from many injection energy levels, which improves the injection efficiency.

Figure 5(c) presents the dual wavelengths of our double-stack design emitting at 6.7 and 7 µm in CW operation for a laser having a cavity length of 4 mm without anti-reflective and highly reflective coatings. T₁ in pulsed operation (1 µs, 10 kHz) is 559 K, indicating the low temperature dependence of the slope efficiency, which is mainly attributed to the high $\Delta $E₃₄. However, T₀ is only 167 K for pulsed operation. Unlike T₁, which is mainly affected by the leakage current and pump efficiency, T₀ is affected by many factors, such as the length of the cavity, the density of the threshold current, the doping concentration of the active region and diagonal transition [27]. Further optimization of T₀ requires comprehensive consideration of the above factors [28].

5. Conclusion

We reported 6.9 µm, watt-level, broad tuning QCLs with a double-stack design, achieved by changing the beam flux of the aluminum effusion cell. The maximum CW power for a 6-mm-long, highly reflective and anti-reflective laser was 1.07 W at 293 K. For a 4-mm-long laser, the maximum tuning range at room temperature was 195 and 104 cm⁻¹ in pulsed (1 µs, 50 kHz) and CW operation, respectively. The CW power of the Littrow EC mode exceeded 110 mW over the entire spectrum range. As the angle of the blazed grating changed, the external mode changed regularly without other unnecessary modes. High output power of the EC mode at approximately 7 µm can be used to detect some functional groups, such as methyl group and aldehyde group, over long distances, overcoming the absorption of water.