Quantum cascade lasers (QCLs) have been intensely researched in recent years for narrow line-width and high power mid-infrared (MIR) emissions excited by the desperate need of applications, for instance remote gas-sensing, free-space optical communication and high-resolution spectroscopy^[1-3]. The narrow ridge width is an indispensable condition for QCL high-power continuous wave (CW) operation at room temperature, as heat dissipation of the device is always a huge challenge. On the other hand, very high peak power is also required in some applications, the peak power of QCLs can be scaled up by increasing the laser gain volume, containing a straightforward way in which we could increase the ridge width, thus increasing the laser gain volume. But simply increasing the ridge width would result in poor beam quality arising from the high-order transverse mode^[4]. The spectral and far-field properties of broad area QCLs can be enhanced by the photonic crystal distributed feedback (DFB) mechanism, master-oscillator power-amplifiers, tilted facet lasers, and so on^[5-8]. Compared with single-ridged devices, coherent QCL arrays show a great advantage in increasing the peak power and improving the beam quality. Phase locking of QCL arrays has been demonstrated through evanescent waves, resonant leaky-waves, Talbot cavity and Y-junction tree-array coupling^[9-14]. In recent years, Liu et al. demonstrated a QCL based on a coupled-ridge waveguide (CRW) structure with an emitting wavelength of 7.3 µm^{[15, 16]}. Through carefully designing the CRW structure, the ridge area mainly confined the light field while the coupling ridge region has a gain that exceeded the loss. However, the CRW structure was more suitable for long-wave infrared QCL devices and applying this structure to mid-infrared QCL devices was still a challenge due to the high modal loss and the weak coupling strength among the array elements.

In this paper, we demonstrated high peak power substrate-emitting DFB QCL arrays lasing at λ ~ 5 µm by employing an optimized CRW structure. Five-element QCL arrays were simulated and fabricated through a two-step etching method to extend the CRW structure to a mid-infrared. By introducing a buried 2nd-order DFB grating, the devices could work at a stable single longitudinal mode with a linear tuning coefficient of –0.2 cm^–1/K. Lateral far-field with the main peak near the diffraction-limited intensity curve of about 10.7° was observed in a ridge-width direction even at full power. The peak output power from the substrate surpassed 1 W at 25 °C. The main-lobed far-field distribution and high output power of these devices make them very essential in real applications.

A double-phonon resonant design was introduced into the active region of this device with an emitting wavelength of 5 µm. The whole wafer was co-grown by solid-source molecular beam epitaxy (MBE) and metal organic vapor phase epitaxy (MOVPE) on an n-InP substrate (Si, 2 × 10¹⁷ cm^–3). The active region and InGaAs confined layers were grown by MBE and the InP cladding was grown by MOVPE. Along the epitaxial layers sequence, the InP bottom cladding layer (5 µm, Si, 2.2 × 10¹⁶ cm^–3), InGaAs confined layer (300 nm, Si, 2 × 10¹⁶ cm^–3), active region (40-stages In_0.6Ga_0.4As/In_0.4Al_0.6As superlattice with a strain-compensated design and a total thickness of ~2 µm), InGaAs confined layer (300 nm, Si, 2 × 10¹⁶ cm^–3), InP upper cladding layer (3 µm, Si, 2.2 × 10¹⁶ cm^–3), InP gradually doped layer (250 nm, Si, 2 × 10¹⁷ to 5 × 10¹⁷ cm^–3) and InP highly doped contact layer (500 nm, Si, 5.5 × 10¹⁸ cm^–3) was grown respectively.

Before fabrication, simulation was conducted with COMSOL to select the optimal parameters of the CRW structure. Based on our earlier research, the mode selection of the array was mainly related to the modal waveguide loss^[15]. Fig. 1(a) showed the near-field distribution of the fundamental supermode and high-order supermode. The loss difference of the modal waveguide between the high-order supermode and the fundamental supermode was simulated as the variable of the ridge width W with a constant array period S of 9 µm, as is shown in Fig. 1(b). The grating etching depth D was also fixed at 3.5 µm. In order to ensure the fundamental mode lasing, the ridge width should not be too wide, often requiring less than 8 µm for the laser with a wavelength of 5 µm. If the ridge width is too small (such as less than 5 µm), the absorption loss of silicon oxide will increase. The maximal loss difference was observed at W = 7 µm and the loss difference was about 0.0035 cm^–1. Besides, the dependence of loss difference of the modal waveguide between the high-order and fundamental supermode on etching depth with a fixed wavelength and period of array S as shown in Fig. 1(c). We observed that the loss difference did not convert positively until the etching depth reached 3.5 µm, which would promote the fundamental supermode in domination. Whereas the actual etching depth was limited by the thickness of the InP top cladding layer and the anisotropy of the etching method. Therefore, we proposed a two-step etching method to get high loss difference for fundamental supermode operation. What's more, the rationality of setting array period to 9 µm was verified, as is shown in Fig. 1(d). Since all the ridges are identical emitters, the far-field pattern will result from the ridge array diffraction. In order to limit the major far-field lobes to one, it is necessary to shorten the interval according to the equation:

$ S\,{{\sin}}\theta =\pm \left(k+\delta /2\pi \right) \lambda {,} $ ()

where S is the period of the ridges, is the diffraction angle, and is the phase shift between adjacent emitters. A smaller interval between the ridges would also lead to a negative loss difference with a fixed ridges width at 6 µm through simulation. To sum up, we successfully selected the optimal ridge width, etching depth and array period parameters through COMSOL.

Before the top cladding layer was grown, a buried 2nd-order grating (period Ʌ = 1.58 µm, duty cycle σ = 0.4, and depth d ~ 180 nm) was patterned and transformed on the InGaAs upper constrain layer by using holographic lithography and wet etching technique, as is shown in Fig. 2(a). Then the wafer was etched into a five-element CRW structure through standard photolithography techniques and reactive ion etching (RIE). The etching depth was about 2.8 µm and the top ridge width was 7 µm. This was the first etching step and a good etching profile was ensured by the excellent isotropy of RIE. After that, a wet chemical etching process was used to enlarge the etching depth. Then an inverted conical geometric structure can be obtained, where there is a larger index contrast between the etched and un-etched regions. A total etching depth of 3.5 µm was obtained, which means that we protected the integrity of the active region. Due to the isotropy of the wet etching, the top ridge width was decreased from 7 to 6 µm. The interelement space between two adjacent ridges was 3 µm. The cross-section image of the five-element device was shown as Fig. 2(b). Through this two-step etching method, the loss difference was increased considerably and the coupling strength among array elements was enhanced for mid-wave infrared QCL arrays.

After completing the etching process, a SiO₂ insulating layer with a 450 nm thickness was deposited by plasma-enhanced chemical vapor deposition (PECVD) around the ridges. We set a 3 µm wide window for electron injection on each ridge by photolithography and wet etching and ohmic contact was provided by a Ti/Au layer. To further enhance the heat dissipation, an additional 4 µm thick gold layer was electroplated. Then we deposited a Ge/Au/Ni/Au metal contact layer on the backside of the sample after thinning the substrate down to 150 µm. And then a 160 µm wide substrate-emitting window was patterned by chemical etching. Finally, the sample was high-reflectivity (HR) coated with Al₂O₃/Ti/Au/Al₂O₃ (200/10/100/120 nm) on the back facets by e-beam evaporation, and then cleaved into 2 mm long devices for testing. The QCL arrays were bonded epi-layers down to a copper heatsink with an indium solder.

The far-field pattern of the laser arrays in the ridge-width direction was most concerned. QCL arrays’ far-field measurements were carried out on a step-motor controlled rotation stage with a thermal resistor and a thermo-electrical cooler (TEC) monitoring and regulating the heatsink temperature. The devices were operated under a pulsed mode at 25 °C with a repeat frequency of 10 kHz and a pulse width of 1 µs. The mid-infrared emission was collected by a room-temperature HgCdTe detector (Viggo, PVMI-8) with a distance of 40 cm away from the substrate-emitting window of the laser array. For the substrate emitting device, a main-lobed far-field distribution with a divergence angle of 10° was observed in the ridge-width direction at 2 A, as shown in Fig. 3(a). The full width at half maximum (FWHM) of the far-field main-lobes were about 1.5 times the diffraction limit (D.L.) near the threshold. The devices did not operate in single-spatial mode causing the nearest-neighbor coupling to be inevitable due to a very low loss differential. In addition, the inevitable nearest-neighbor coupling may lead to other modes and different modes have different thresholds. Then, while the current increased from the threshold, other modes occurred, the far-field would widen and other peak intensities become obvious. But even at a high current of 7 A, the devices still kept the main-lobed far-field distribution with a divergence angle of 10.7°. The phenomenon can be concluded from the same equation:

$ D\,{{\sin}}\theta =\pm \left(k+\delta /2{{\pi }}\right) \lambda {,} $ ()

where and the calculated side mode positions (about 16°) correspond to the measured far-field graph. The strong coherent capability of this CRW structure promised a high beam quality, which was very helpful in collimation and long-distance transmission. However, the fundamental transverse mode operation can be only maintained in several number of ridges still due to poor mode discrimination, thus the maximum active ridge width is about dozens of microns. But this structure still makes sense in order to increase the peak power to a certain level. When the laser was measured as an edge emitting device, the far-field displayed a similar phenomenon. In addition, another advantage of this approach is that there is no increase in waveguide loss. Different from the surface semiconductor/Au grating structure, the lowest threshold gain mode in a buried 2nd-order semiconductor/semiconductor grating structure is the symmetric mode^[17]. Here we could observe a single-lobed pattern with a small beam divergence of 0.2° near the diffraction limit in the cavity-length direction.

The power–current characteristics of the laser arrays are shown in Fig. 4. The test was carried out at room temperature of 25 °C in the pulsed mode with a repeated frequency of 10 kHz and pulse width of 500 ns. The emitted light power was measured by a calibrated thermopile detector (Coherent, EMP1000) placed directly ahead of the device. For the five-element device, the threshold current was about 2 A and the corresponding threshold current intensity was about 2 kA/cm². A peak power of more than 1 W was obtained from the substrate with a slope efficiency of 0.3 W/A. The peak power emitted from the edge was about 3.4 W. The ratio of substrate-emitted power and edge-emitted power was about 31%.

The spectra of the QCL arrays were measured by a Fourier transform infrared (FTIR) spectrometer in a fast scan mode with a resolution of 0.25 cm^–1. The temperature of the heatsink was stabilized with a thermo-electrical cooler (TEC). The emitting spectra of the device are shown in Fig. 5. Steady single-mode emission was observed and the lasing wavelength could be tuned continuously. The spectra were measured at different temperatures of the heatsink from 10 to 60 °C with a step of 5 °C, covering a total tuning range of 9.9 cm^–1. The inset of Fig. 5 shows the dependence of the lasing frequency on the temperature of the heatsink. A linear tuning coefficient of –0.2 cm^–1/K was deduced. The wide tuning range and linear tuning ability make these devices very promising in spectroscopy applications.

We have demonstrated high-power single longitudinal mode substrate-emitting QCL arrays lasing at λ ~ 5 µm by employing a CRW structure and buried 2nd-order DFB grating. Five-element QCL arrays were simulated and fabricated to explore the coherent capability of the CRW structure. Stable main-lobed far-field distribution in the ridge-width direction with a divergence angle of 10.7° was observed even at full power. The peak output power surpassed 1 W at room temperature of 25 °C. The devices could work at a stable single longitudinal mode and could be tuned continuously with a linear temperature tuning coefficient of –0.2 cm^–1/K.

This work was supported by National Basic Research Program of China (Grant Nos. 2018YFA0209103, 2018YFB2200504), in part by National Natural Science Foundation of China (Grant Nos. 61991430, 61774146, 61790583, 61734006, 61835011, 61674144, 61774150, 61805168), in part by Beijing Municipal Science & Technology Commission (Grant No. Z201100004020006), and in part by the Key projects of the Chinese Academy of Sciences (Grant No. 2018147, Grant No. YJKYYQ20190002, Grant No. QYZDJ-SSW-JSC027, Grant No. XDB43000000, Grant No. ZDKYYQ20200006). The authors would appreciate P. Liang and Y. Hu for their help in the device fabrication processing.