On-chip unstable resonator cavity GaSb-based quantum well lasers

Type-I quantum well (QW) lasers emitting within the 2–3.3 μm spectral band are especially valuable for applications that detect chemicals, explosives, and biological agents with pronounced absorption bands in this region. GaSb-based diode lasers are best suited for these applications, as the epitaxial growth of the 6.1 Angstrom family of alloys and heterostructures have matured in the recent years. The slope efficiency of GaSb-based lasers is still lower than GaAs-based shorter wavelength lasers, due for example to increasing carrier absorption at the longer wavelengths.¹ Consequently, requirement for high power in GaSb-based lasers is typically met by fabricating broad area (BA) devices. For example, 960 mW of continuous wave (CW) power at λ ∼ 2.95 μm,² and 500 mW at λ ∼ 3.2 μm was demonstrated in BA devices with ∼100 μm-wide stripes.³ 

The BA geometry provides an enlarged gain volume for high power operation, a large emitting facet to prevent catastrophic optical damage (COD), and ease of fabrication. However, the BA geometry also results in a degraded beam quality in lateral direction due to multimode oscillation as well as the formation of multiple filaments within the cavity. Filamentation is brought on by self-focusing as a result of gain and index changes induced by the carrier concentration in the active region and produces non-uniform and spatially incoherent light at the output facet.⁴ During multiple reflections between the parallel mirrors in a Fabry-Pérot (FP) cavity, the mode is directly counter-propagated, forming stable filaments. Several approaches have been developed to disrupt filament formation and improve lateral beam quality.^5–7 For example, laterally tapered devices use the natural divergence of the propagating mode to counteract the self-focusing.⁵ Broad area alpha-DFB devices are also shown to successfully suppress filaments.⁶ Another solution is to form an unstable resonator (UR) cavity by placing a cylindrical mirror at one or both facets of the laser cavity.⁸ A UR cavity maintains a diverging light path to avoid direct counter-propagation and can thus suppresses the formation of filaments.

While the UR cavities with cylindrical mirrors are commonly used in larger scale devices such as gas and solid state lasers, forming an on-chip UR cavity within the semiconductor laser is more difficult due to the small size. Consequently, only a few demonstrations of on-chip UR cavity semiconductor lasers have been made over the years. On-chip UR cavities were used to enhance the coherency of the output beam in both short wavelength diode lasers and mid-IR optically pumped semiconductor lasers.^9–13 One of the best achievements is an eight times brightness enhancement in short wavelength diode lasers.¹³ In this work, we demonstrate a GaSb-based diode laser with an on-chip UR cavity emitting at ∼2 μm and report on the direct comparison with the FP cavity device.

In a typical UR cavity with cavity length L and one diverging back facet mirror of radius R, there is an external focus point located behind the back facet, as shown in Figure 1. This point is located at a distance V behind the front facet and given by V = $L2+LR$⁠.¹⁰ However, due to the refraction at the flat output facet, the origin of the wavefront appears to be a focal point inside the resonator, which is referred to as the virtual source. This virtual source point is located at a refractively reduced distance of D = V/n behind the output facet, where n is the refractive index.

A critical parameter when designing the UR cavity is the round trip magnification factor M, quantified as

A large M will lead to a smaller virtual source size but incur higher round trip losses due to the more divergent path. Therefore, the slope efficiency and total power will be limited. A small M, on the other hand, may lead to lower losses but may insufficiently discriminate against high order modes or suppress filament formation.¹³ 

The laser structure in this study was grown using molecular beam expitaxy on a GaSb:Te (001) substrate. The structure nominally consists of 2 μm-thick Al_0.9Ga_0.1As_0.08Sb_0.92:Te n-cladding layer, 1 μm-thick lattice-matched Al_0.3Ga_0.7As_0.03Sb_0.97 undoped waveguide, followed and 2 μm Al_0.9Ga_0.1As_0.08Sb_0.92:Be p-cladding. At the center of the waveguide, there are three 10 nm-thick Ga_0.8In_0.2Sb quantum wells with ∼1% compressive strain that provide gain for lasing at λ ∼ 1.94 μm at room temperature. Each well is separated by 50 nm.

Following the standard fabrication procedures, gain guided devices with a stripe width of 100 μm were cleaved to an FP cavity length of 1 mm. Devices were mounted in the epi-up configuration, without facet coating, and characterized using 500 ns pulses with 0.5% duty cycle to minimize heating.

Devices with an on-chip UR cavity were fabricated for direct comparison. When designing the UR cavity, we considered the previous studies, noted that the optimum brightness was obtained at M ∼ 3,¹¹ and chose this magnification for the current study. Using Equation (1) for L ∼ 1 mm, the back facet mirror is therefore nominally designed to have a radius of ∼3 mm.

The focused ion beam (FIB) milling tool with Ga+ source was used to carve a 200 μm wide, ∼35 μm deep cylindrical mirror at the back facet to form the half-symmetric UR cavity. FIB milling was chosen because it offers a precise control of facet shape and low surface roughness. A scanning electron microscopy image of the curved mirror facet is shown in Figure 2, where the active region is also visible as a bright line 2 μm below the top surface.

Using a Zygo optical profilometer, the rms roughness of the carved mirror surface was found to be <5 nm. This is substantially smaller than 1/10 wave and should be more than adequate to prevent significant scattering losses from the reflecting surfaces. The large depth of the mirror also ensured that the uniformity of R in the range coincident with the laser optical field, which is the depth range covering the top clad, active region, and bottom clad, was high. At this depth, R is measured to remain within a range of 3.00 ± 0.05 mm. The vertical slope of the mirror is measured to be <1° from vertical.

Although a perfect tool to quickly explore new geometries and help prove the concept, FIB milling does not allow mass production. Therefore, after finding the optimal cavity geometry by FIB milling, we would investigate alternative fabrication method, plasma etching. Previous studies have shown that dry plasma etching can produce high quality mirrors and thus high-power high-brightness output.^12,13

The light vs. current behavior of the UR cavity device and the FP cavity device are compared in Figure 3. In both cases, power is collected from a single flat front facet. We observe ∼13% increase in threshold in the UR cavity with respect to the FP cavity. In addition, we also observe approximately 24% drop in slope efficiency in the UR cavity. The slope efficiency drop is expected and is due to the additional losses as the result of the beam spreading outside the confines of the current injection area in the UR cavity. The 24% drop agrees with the 14%–45% range for a 2–4 range of M observed in a similar study.¹³ 

Although the output power at a given current is reduced, the brightness generated in the UR cavity will strongly depend on the coherency, i.e., the beam quality of the output. To evaluate the beam quality in the lateral direction, we must quantify the virtual source size as well as its location. The virtual source size can be estimated by using a single lens with a focal length of f to re-image the object at the CCD plane of a camera at a distance Q away from the lens. Using the lens equation, the size of the virtual source in the camera is magnified by a factor of (Q − f)/f. In our setup, this value was 24; thus, the virtual source size is estimated as 1/24^th of the image width at the camera. The virtual-source intensity profiles collected at the camera from the UR and FP cavity lasers operated at room temperature and at ∼15× threshold are shown after the correction for magnification in Figure 4.

Emission from the FP cavity can be characterized as originating from a broad source with a non-uniform intensity distribution recorded as multiple peaks in the intensity pattern. The size of this source, w, measured at the 1/e² points is estimated to be ∼110 μm, slightly larger than the cavity width of the device, due most likely to current spreading. The presence of multiple peaks may indicate the presence of several stable filaments. An estimate of the beam parameter product (BPP), defined as the product of the source half-width (∼0.055 mm) and divergence half-angle, can be made to assess beam quality. Independent far-field measurement of the beam indicates a half-angle of ∼110 mrad. Therefore, the BPP value for the FP cavity device is estimated to be ∼6.1 mm · mrad.

In contrast to the FP cavity, the virtual source of the UR cavity is characterized by a single narrow peak in the intensity profile. The measured width of the virtual source, S, taken at the 1/e² points, is estimated to be ∼40 μm. This is approximately 2.8 times smaller than the source in the FP cavity.

In order to estimate the BPP in the UR device, an angular divergence must be associated with the virtual source. We estimate this angle using two independent methods. The first, a simple far-field measurement, indicates a divergence half-angle of ∼126 mrad. In this way, the BPP value is estimated to be ∼2.5 mm · mrad.

Alternatively, the divergence angle can be estimated by an independent measurement of D, which is the distance of the virtual source from the front facet. The divergence half-angle is estimated as tan⁻¹[w/2D], where w is the measured width of the pump stripe, ∼110 μM. D, the astigmatism difference, was experimentally quantified as the additional translation required to bring the vertical beam (fast-axis) originating at the front facet to come into focus at the camera. This technique is graphically described in Ref. 9. For the case shown in Figure 4, D was measured to be ∼0.53 mm and the divergence half-angle is therefore estimated to be ∼103 mrad. Using this alternative method, the BPP in the UR device is estimated to be ∼2.1 mm · mrad, consistent with the previous method.

Note that for the nominal values of L = 1 mm, R = 3 mm, and n ∼ 3.5, D = $L2+LR$/n is predicted to be 0.57 mm. This value is close to that experimentally measured, which indicates that the UR cavity is functioning as designed.

It is clearly evident that the UR cavity is able to enhance the coherence of the output beam, and this will result in enhanced brightness, given as power/BPP. The brightness increase is therefore estimated by the ratio of the BPPs adjusted by the 25% drop in slope efficiency. Thus, the UR cavity device, in this case, is capable of delivering an approximately 2-fold increase in brightness.

We conducted additional tests to assess the robustness of the UR cavity design with respect to injection current in the range of 1–3 Amps, corresponding to 5×–15× threshold. Measurements of D and corresponding S are reported in Figures 5(a) and 5(b), respectively. We observe D to remain relatively constant at 521–533 μm in this range. In contrast, we observe S to first slightly decrease with current and then increase with current above 2 A. Similar behavior of virtual source width was observed in a previous study.¹³ They speculated that excess spontaneous emission existing up to three times threshold is the reason for the degradation of beam quality at low current. However, this may not serve in our case since excess spontaneous emission is not expected at 5–10 times threshold range. Therefore, extra experiments need to be done in the future to discover the root of this behavior. For example, a comparison of beam qualities for a broad range of magnification may help.

To assess beam quality variation as a function of injection current, we invoke the relationship that predicts the diffraction limited waist S^DL= 2λD/w in the UR cavity, used in Ref. 10. We estimate beam quality in the lateral direction as S/S^DL and plot this as a function of injection current in Figure 6. It is evident that despite the variation in S and D, presumably due to changes in index,¹¹ the beam quality varies little in the explored range. It does appear, however, that the optimal beam quality is attained at the intermediate values of injection current. This finding is consistent with a similar trend observed in Ref. 13.

In summary, we demonstrate the formation of on-chip UR cavity on ∼2 μm GaSb-based diode lasers lasing using the FIB technique to form the convex back facet. At the UR cavity magnification factor M ∼ 3, we observe an ∼2-fold increase in brightness associated with the improvement in the beam quality relative to the FP cavity device. The beam quality improvement remains robust across the current injection range studied.

These results corroborate other demonstrations of on-chip UR cavity semiconductor lasers and help encourage the development of mass fabrication methods that adapt to incorporate UR cavity designs. They also demonstrate that the strategy will work equally well in diode lasers using GaSb based alloys that are designed for longer wavelengths.

The authors acknowledge the support from the Air Force Office of Scientific Research for this study. The devices were partly fabricated using the cleanroom facilities at the Center for Integrated Nanotechnologies, an Office of Science User Facility operated for the U.S. Department of Energy (DOE).