Quantum cascade lasers (QCLs) have been rapidly developed and widely applied for their flexible wavelength in the midinfrared and terahertz spectrum region as portable and compact light sources ^[1–3]. A high-output power is pursued in many areas such as directed infrared countermeasures, remote sensing, and free-space optical communication ^[3]. Widening the active area is one of the most straightforward approaches to increase output power. However, simply increasing the ridge width will result in deterioration of the beam quality with the emission of a multiple-lobe far-field pattern ^[4,5]. Single-lobe emission has been obtained in the past with the methods of photonic crystal distributed feedback QCLs, angled-cavity QCLs, and master-oscillator power-amplifier QCLs ^[6–8]. Recently, phase-locked QCL arrays have been a popular approach to keep wide ridge QCL emitting with a coherent narrow beam pattern.

In fact, phase-locked arrays have been widely used for wide ridges with narrow divergence in near-infrared emitting lasers ^[9–13]. The phase-locked QCL arrays were first demonstrated by the coupling of evanescent-wave operating in out-of-phase mode with a two-lobe far-field pattern at 8.4 μm ^[14,15]. Later, Y-junction QCL arrays were demonstrated with unstable emission between in-phase and out-of-phase mode because of the spatial hole burning effect ^[16]. Resonant leaky-wave coupling QCL arrays were obtained by utilizing a complex regrowth process to form an antiwaveguide structure ^[17]. Our group has also reported evanescent wave-coupling QCL arrays with a coupled ridge waveguide and diffraction coupled QCL arrays by the Talbot effect ^[18–21]. Besides, the focused ion beam etching was used for wide ridge QCL emitting in a single-lobe far-field pattern ^[22]. However, these phase-locked QCL arrays usually either emit with high-order supermodes under high injection current ^[16,18,19] or produce extra waveguide loss resulting in a high threshold current density ^[20–22].

In this paper, we demonstrate a chirped couple ridge waveguide QCL array with a stable fundamental supermode far-field pattern at λ∼7.6μm. A near diffraction limited (DL) divergence angle of 9.4° is obtained in the whole current dynamic range, which is defined as the difference of the maximum and threshold current. Because the recent studies on the reduced stage QCLs showed attractive results with an output power up to 4 W ^[3,23], power scaling in QCL using broad-area stripes with reduced cascade number has obtained continuous wave (CW) operation ^[24–26]. Once the chirped coupled ridge waveguide structure is employed in wide ridge QCL with a reduced cascade number, a near DL far-field pattern in CW mode will have the potential to be obtained. In addition, this method has lower waveguide loss than divided active region QCL arrays because the active region of devices is not destroyed.

The QCL wafer was grown on an n-doped (Si, 2×1017cm−3) InP substrate wafer by solid-source molecular beam epitaxy (MBE) using an active-region structure similar to Ref. ^[27]. The active core structure presented in this work contains 40 periods of strain-compensated In0.58Ga0.42As/In0.47Al0.53As quantum wells and barriers. The specific layer sequence of one period, in nanometers, is as follows (layer thickness in nanometers): 4/1.7/0.9/5.06/0.9/4.7/1/3.9/1.8/3.2/1.7/2.8/1.9/2.7/2.8/2.6, where In0.47Al0.53As barrier layers are in bold, In0.58Ga0.42As well layers are in roman, and n-doped layers (2×1017cm−3) are underlined. The whole wafer structure before the fabrication is 4.5 μm lower InP cladding (Si, 3×1016cm−3), 0.3 μm thick n-In0.53Ga0.47As layer (Si, 4×1016cm−3), 40 active/injector stages, 0.3-μm-thick n-In0.53Ga0.47As layer (Si, 4×1016cm−3).

After the epitaxy in MBE, a traditional lithograph process was used to define the chirped arrays within a thin photoresist deposited on a 450 nm thick SiO2 layer used as a hard mask. The pattern was transferred from the photoresist to the SiO2 layer using inductively coupled plasma etching. Then, 2 μm thick semi-insulating InP:Fe was selectively regrown by a metal organic vapor phase epitaxy (MOVPE), for the purposes of thermal dissipation, optical coupling, and electrical insulation, as shown in Fig. 1(a), and highlighted by the region enclosed by the yellow lines in Fig. 1(b). The residual SiO2 was removed with wet etching approach, which defined the width of the active region. Next, a 3 μm InP upper cladding layer (Si, 8×1016cm−3) was grown by MOVPE, followed with a 0.15 μm InP gradually layer doped (changing from 1×1017 to 3×1017cm−3) and 0.6 μm highly doped InP cladding layer (Si, 5×1018cm−3). Next, the contact metal, a 300 nm thick gold layer was evaporated to bury the devices. An additional 5 μm thick layer of gold electroplating was deposited. Then, the wafer substrate was thinned down to 120 μm, and the Ge/Au/Ni/Au metal contacts were formed as the bottom contact. Finally, the wafer was cleaved into 2 mm long, and the devices were soldered epilayer side down onto the copper heat sink with indium solder. Figures 1(a) and 1(b) show the schematic of the device’s cross section from the facet direction and the scanning electron microscopy (SEM) image of the device facet, respectively.

The far-field distributions of the chirped coupled ridge waveguide QCL arrays at different injection currents were measured using the lock-in technique and a room temperature mercury-cadmium-telluride (MCT) detector. The MCT detector was placed ∼20cm away from the array devices, which are mounted on a rotation stage controlled by a computer with a resolution of 0.05°. Figure 2 shows measured and simulated lateral far-field radiation patterns of the devices. The far-field distributions show a strong central lobe at 0°, indicating the in-phase supermode operation according to the couple-mode theory ^[9]. For a given optical power, a DL beam has the highest brightness. Considering the light extension from the active region to side waveguides of about 6–10 μm in total, the DL far-field divergence is 9.1°–9.9°. The dashed red line shows the theoretical far-field distribution, which is almost overlapped with measured results. The measured full width at half-maximum (FWHM) of the far-field pattern is nearly diffraction limited 9.4°. Unlike the uniform coupled ridge waveguide structure in Ref. ^[20], the far-field patterns have almost no extension at higher angles in the whole current dynamic range from 2.1 to 3.3 A, which is coincident with the loss difference between high-order and fundamental supermode in Fig. 3(a). Totally, our devices show a DL single-lobe far-field pattern with negligible beam steering as in Ref. ^[28] in the whole current dynamic range.

The stable fundamental supermode emission means that the fundamental supermode has lower modal loss than the high-order ones. The supermode loss in QCL arrays consists of a modal mirror loss and waveguide loss. The difference of modal mirror loss for the fundamental and high-order modes is quite small in coupled ridge waveguide arrays because it is inversely proportional to the transverse dimension of the active region ^[29]. Therefore, waveguide loss plays an essential role in mode discrimination of the fundamental supermode and high-order supermode. The modal losses of the fundamental and high-order supermodes of different geometry were simulated with COMSOL, as shown in Fig. 3(a). For both chirped arrays and uniform array, the fundamental supermodes have the lowest modal loss. The difference between high-order supermode and fundamental supermode in the uniform array is relatively lower than that of the chirped arrays. This means that the chirped structure of interelement widths helps somewhat with intermodal discrimination. In the chirped array, the higher-order supermode can be suppressed by tailoring the lateral gain distribution across the array such that the near-field envelopes of the fundamental and the higher-order supermodes can differ appreciably ^[9]. Figure 3(b) shows the losses of the fundamental and high-order supermodes and loss difference as a function of the central ridge width w. Both the fundamental and high-order modal losses decrease when the central ridge width increases from 7 to 11 μm, so does the loss difference of the fundamental and high-order supermodes. Therefore, to achieve the fundamental supermode emitting preferentially and to make the loss not too high at the same time, a central ridge width w=8μm is selected. Figure 3(c) shows the results of a numerical 2D simulation of the near-field electric field intensity distribution (patterns) of the fundamental mode and the current distribution (red lines) of five-element chirped coupled ridge waveguide QCL arrays. The transverse electric field intensity distribution can be applied to deduce the far-field pattern as shown by the red dashed line in Fig. 2, which presents an in-phase fundamental supermode operation. The in-phase array mode peaks under the two central InP:Fe ridges, making it similar to the in-phase mode of a four-element, evanescent-wave-coupled array with somewhat low-index semi-infinite regions away from its edges ^[14]. The red lines show the nonuniform current distribution in the waveguide because the insulation InP:Fe blocks flow of the current. A good thermal dissipation is expected because the whole waveguide is an InP structure.

The emitted optical power was measured with a calibrated thermopile detector placed directly in front of the laser facet. The spectrum measurements were performed using a Fourier transform infrared spectrometer with 0.25cm−1 resolution in rapid scan mode. Figure 4 shows the power-current (P-I) characteristic under pulsed mode with the current driver maintained at 5 kHz with a duty cycle of 1%. For a 2mmlong×36μm wide QCL chirped array, a total peak power of 840 mW is obtained at 298 K with a threshold current density of 2.5kA/cm2 and a slope efficiency of 0.4 W/A, as shown by the blue line. In contrast, the single laser device with a 2mmlong×13μm wide ridge shows a maximum peak power of 360 mW, a threshold current density of 1.9kA/cm2, and a slope efficiency of 0.48 W/A. The output power of single facet per unit area in phase-locked arrays and single laser is 5.83W/mm2 and 6.92W/mm2, respectively. It can be thought that the output power scales almost linearly with the area of the active region of the arrays. The small reduction of the slope efficiency may be caused by thermal accumulation. When the cascade number of QCL active region is reduced, the chirped coupled ridge waveguide arrays may operate in CW mode. The inset of Fig. 4 shows the lasing spectrum of the phase-locked arrays at room temperature and 1.3 times threshold current. The center wavelength was measured to be 7.6 μm with a multimode nature resulting from the lack of a longitudinal-mode selection mechanism. The single-mode spectrum can be achieved by introducing a distributed feedback grating on the top cladding layer.

In summary, we demonstrate chirped coupled ridge waveguide QCL arrays of 7.6 μm operating at in-phase mode with a DL divergence. The FWHM of the far-field pattern has no extension in the whole current dynamic range from the threshold to the full power current. Further studies should be taken on the devices with the active region of reduced cascade number. We expect a CW operation to be obtained on broad area QCL arrays with DL beam quality. Once optimized, the devices have potential for high brightness diffraction-limited broad-area QCL arrays.

Acknowledgment. The authors would like to thank Ping Liang and Ying Hu for their help in device processing.