1. Introduction

Although interband cascade lasers (ICLs) were first proposed in 1995 and demonstrated in 1997 [1–3], only recently have critical design and growth improvements allowed their fundamental performance limits to be probed experimentally [4–7]. In particular, our latest simulations demonstrate that all standard structures based on designs from the prior ICL literature have featured hole densities in the active quantum wells (QWs) that significantly exceeded the corresponding electron densities [7]. Because excessive Auger recombination and internal losses seemed the likely consequence, this finding inspired the concept of rebalancing the hole/electron density ratio via heavy n-doping of the electron injectors [7]. Implementation of the rebalancing strategy immediately produced substantial improvements in all key ICL performance characteristics. The results include pulsed threshold current densities as low as J_th = 170 A/cm² at room temperature (RT), continuous-wave (cw) wallplug efficiencies as high as 13.5% and drive powers as low as 29 mW at RT, and a maximum cw operating temperature of 107°C for a narrow ridge emitting at λ = 3.88 μm.

These initial successes in the 3.6-3.9 μm spectral range raised the question of whether similar improvements might now be possible at somewhat longer wavelengths beyond 4.5 μm. In that region the highest reported cw lasing temperatures have been 229 K for λ = 5.1 μm [8] and 184 K for λ = 5.9 μm [9]. We report in this letter that the rebalancing scheme indeed dramatically improves the performance of ICLs emitting at λ = 4.7-5.6 μm, enabling RT cw operation. It will be seen that while the current- and power-density thresholds are somewhat higher than at shorter wavelengths, they are nonetheless substantially improved over any reported to date for state-of-the-art mid-IR quantum cascade lasers (QCLs) [10,11].

2. Design

The two designs (Samples A and B) based on rebalancing of the hole/electron density ratio are similar to those reported previously for 3.6-3.9 μm devices [7]. The main finding of that work is that while holes generated at the semimetallic interface in each stage of the ICL transfer efficiently to the active GaInSb hole QW via the very thin GaSb/AlSb hole injector, transfer from the InAs/AlSb electron injector is inhibited by its much greater thickness and is energetically unfavorable because of the need to generate electrons and holes at the semimetallic interface separating the electron and hole injectors. It was found that a slight excess of electrons over holes in the active region led to the most efficient production of gain when Auger recombination dominates and is split nearly equally between multi-electron and multi-hole processes.

In order to take advantage of the rebalancing scheme at longer wavelengths, three InAs QWs in the electron injector were heavily doped to a total sheet density of ≈5 × 10¹² cm⁻². The emission wavelength was adjusted by increasing the thicknesses of the two active InAs QWs in each period by ≈3 and 4.5 Å for λ = 4.7 and 5.6 μm, respectively, by comparison with the structure presented in [7]. The thicknesses of the two InAs QWs immediately adjacent to the active wells were increased by the same amounts, with the increase being tapered off towards the other side of the electron injector so that the thickness of the QW at the semimetallic interface [7] remained unchanged. Five rather than six InAs injector QWs for a total electron injector thickness of 210-215 Å were employed to maintain a similar electric field across the active core at longer wavelengths. Whereas the designs for the active hole QW, hole injector, and barrier at the semimetallic interface were unchanged, the GaSb separate confinement layers (SCLs) that sandwich the active regions were increased from 500 nm to 600 nm in Sample A and to 650 nm in Sample B. This was done in order to maintain a similar overlap of the optical mode with the low-doped (~1x10¹⁶ cm⁻³), low-loss SCLs as in the shorter-wavelength design [7]. The top and bottom cladding layers were also thickened to minimize optical losses due to modal overlap with the top metallization and leakage into the substrate.

3. Pulsed characterization

The two ICL wafers were grown on n-GaSb (100) substrates by molecular beam epitaxy [12]. Contact lithography and wet chemical etching were used to produce 150-μm-wide ridges with intentional lateral corrugation of the sidewalls to eliminate parasitic lasing modes. The uncoated cavities were cleaved to 2 mm length and tested at a pulse repetition rate of 3 kHz and width of 250 ns. At 300 K, the peak emission wavelengths are 4.5 and 5.5 μm for Samples A and B, respectively.

Sample A displays a threshold voltage (V_th) of 2.2 V and threshold current density (J_th) of 460 A/cm² at 300 K, while the corresponding values for Sample B are V_th = 2.1 V and J_th = 650 A/cm². The characteristic temperature T₀ for both samples above room temperature is ≈39 K. The external differential quantum efficiencies from both facets above threshold are 13.8% and 15.1% for Samples A and B, respectively. Although these RT efficiencies are much lower than the 25-30% results exhibited by carrier-rebalanced 3.6-3.9 μm devices [7], they are nonetheless higher than for any earlier ICLs emitting at λ > 4.5 μm. Assuming the internal efficiency to be independent of λ, the considerably lower slope efficiencies imply higher internal loss, possibly associated with larger free-career absorption cross sections in the SCL and active regions.

4. Continuous-wave performance

Narrow ridges were fabricated by photolithography and reactive-ion etching, using a Cl-based inductively coupled plasma (ICP) process, followed by cleaning with a phosphoric-acid-based wet etch. The etching stopped within the active core for Sample A and in the bottom SCL for Sample B. It should be noted that the threshold current densities reported below for Sample A may be slightly overestimated because there was some current spreading in the unetched part of the active region. The ridge widths were 10.9 μm for Sample A and 10.3 and 7.3 μm for Sample B. A 200-nm-thick Si₃N₄ layer was deposited by plasma-enhanced chemical vapor deposition, and a top contact window etched back using SF₆-based ICP. Next, 100 nm of SiO₂ was sputtered to block occasional pinholes in the Si₃N₄. The ridges were metallized and then electroplated with 4-5 μm of Au. Cavities of 4 mm length were cleaved, and an HR coating comprised of 46 nm Al₂O₃ topped by 100 nm Au was applied to each back facet. The micrograph in Fig. 1 shows a finished 10.3-μm-wide narrow ridge fabricated from Sample B, as seen from the front (output) facet. The devices were mounted epitaxial side up on a copper heat sink attached to a thermoelectric cooler.

 

Fig. 1 Micrograph of a 10.3-μm-wide ridge fabricated from Sample B and covered with 4 μm of electroplated gold. The positions of the top and bottom cladding layers, separate confinement layers, and active core are indicated.

Download Full Size | PDF

The inset of Fig. 2 shows the cw spectra at T = 25°C. The emission wavelength for Sample A increases from 4.75 μm at 25°C to 4.9 μm at 60°C, while that for Sample B varies from 5.6 μm at 25°C to 5.65 μm at 45°C. The main body of Fig. 2 illustrates the temperature-dependent cw threshold current densities for Samples A (10.9-μm-wide ridge, red points) and B (10.3-μm-wide ridge, blue points). For temperatures not too close to the maximum, both J_th(T) are described reasonably well by exponential fits that yield T₀ = 31 K (A) and 35 K (B). The cw threshold of 480 A/cm² for Sample A at T = 25°C is similar to the RT broad-area pulsed result reported above, while the value of 530 A/cm² for Sample B is actually lower owing to lower reflection losses associated with the narrow ridge’s longer cavity and HR-coated back facet. Because the baseline loss in broad-area ICLs is higher at longer wavelengths, the additional loss arising from roughness scattering at the narrow-ridge sidewalls has less impact on the net efficiency.

 

Fig. 2 Threshold current densities for the 10.9- and 10.3-μm-wide 4-mm-long ridges with one HR coating fabricated from Samples A and B, respectively, vs. of temperature in continuous-wave mode. The inset shows the cw emission spectra at 25°C.

Download Full Size | PDF

Light-current-voltage characteristics for the 10.9-μm-wide ridge fabricated from Sample A and the 10.3-μm-wide ridge fabricated from Sample B are shown in Figs. 3 and 4 , respectively. Both structures generate more than 15 mW of cw output power at T = 25°C. The RT threshold voltages of 1.98 V for Sample A and 1.84 V for Sample B are again lower than for the corresponding broad-area devices, because current spreading from the narrow ridge into the much wider bottom cladding layer and substrate substantially reduces the differential series resistivity (0.2-0.3 mΩ-cm² in both samples). The maximum cw operating temperatures are 60°C for Sample A and 48°C for Sample B.

 

Fig. 3 Light-current-voltage characteristics at a series of temperatures for a 10.9 μm × 4 mm ridge with one HR coating fabricated from Sample A.

Download Full Size | PDF

 

Fig. 4 Light-current-voltage characteristics at a series of temperatures for a 10.3 μm × 4 mm ridge with one HR coating fabricated from Sample B.

Download Full Size | PDF

Because single-mode cw output powers on the order of 1 mW will be sufficient for many chemical sensing systems, sources for that application can often operate very close to threshold. It is therefore noteworthy that at T = 25°C, the cw threshold power densities of 0.95 kW/cm² for Sample A and 0.98 kW/cm² for Sample B are more than an order of magnitude lower than the best values of ≈12 kW/cm² ever reported for state-of-the-art QCLs with similar cavity dimensions [10,11]. This is due in large part to the much lower operating voltage requirement of < 2.5 V for an ICL as compared to 10-15 V for most QCLs. Perhaps the most directly relevant figure of merit is threshold power, for which the two devices from Figs. 3 and 4 display RT values of 410 mW and 400 mW, respectively. The narrower ridge of width 7.3 μm from Sample B required a lower input of 320 mW at T = 25°C, due to its smaller active volume, and even lower values would have been possible were the cavity length shortened from 4 mm (down to the point where the accompanying additional mirror losses induce excessive threshold current density). Typical threshold powers for QCLs in this wavelength range from 2 to 5 W, although lower values have been demonstrated using a partially-transmissive HR coating on the front facet in addition to the HR-coated back facet [13]. The application of that approach to ICLs would also lower their required input powers from the values reported in this work. We conclude that if input power is the critical figure of merit, the ICL holds the edge even in the λ = 4-6 μm range where the QCL threshold current densities are relatively low. While the need for narrow spectral linewidth will further affect the power requirement, shorter-wavelength ICLs (processed from wafers grown before the carrier rebalancing strategy was introduced) have exhibited attractive single-mode performance [14].

5. Conclusion

To summarize, extension of the rebalanced carrier density approach to ICLs designed for longer wavelengths has led to much lower threshold current densities, as well as the first room-temperature cw operation of interband semiconductor lasers emitting beyond 4.5 μm. We find that even at wavelengths approaching 6 μm, the cw power densities needed to reach threshold are an order of magnitude lower than state-of-the-art QCL values. For fielded sensing systems, this implies enhanced battery lifetimes as well as a significant relaxation of packaging and size/weight constraints. On the other hand, owing to their lower internal loss and larger optimal number of stages, QCLs are expected to maintain their edge in longer-wavelength applications requiring high output powers.