1. Introduction

Over the past two decades, infrared quantum cascade (QC) lasers have enabled a host of sensors capable of probing mid-infrared molecular absorption lines that are difficult to access using interband semiconductor lasers [1–3]. In addition to wavelength flexibility, a key feature of mid-infrared QC lasers is the ability to achieve watt-level single-emitter output power with high wall-plug efficiency [3–6]. However, for many in situ sensing applications, milliwatt-level output is sufficient and often preferable in order to avoid detector saturation [3,7]. With the development of compact laser absorption spectrometers for environmental sensing and atmospheric science, including instruments for in situ studies on other planets [7–9], there is a growing need for mid- to long-wavelength infrared sources with low power dissipation and low-mass packaging. Continuous-wave (CW), single-mode type-II interband cascade lasers have been recently reported at wavelengths approaching 6 µm with room-temperature threshold power dissipation below 1 W [10,11]. However, for applications requiring longer infrared wavelengths, power optimization of QC devices remains one of the most promising strategies for producing compact, low-dissipation laser sources.

Recently, several reports have been published describing single-mode QC lasers with CW threshold power consumption near 1 W at room temperature. Other than an earlier report describing the approach that is expanded upon here [12], all of these devices were fabricated using epitaxial regrowth processes. Typically, in-plane distributed-feedback (DFB) gratings are etched near the QC active region and overgrown with conductive InP cladding [13]. In the case of the lowest-dissipation lasers, the sidewalls of the laser ridge are overgrown with semi-insulating InP to form a buried-heterostructure (BH) device [14,15]. The design strategy for these devices has been to both optimize the active region for low threshold current density and also minimize the ridge width and cavity length, thereby minimizing threshold current. For BH QC lasers, the active region can be extremely narrow, with a significant fraction of the lasing mode residing in the transparent InP cladding. Threshold current can be further reduced with the addition of high-reflectivity (HR) back-facet coatings [13,14] as well as front-facet partial-HR coatings [15].

Here, we describe regrowth-free, single-mode QC lasers designed for low CW threshold power consumption at high operating temperatures. These lasers can achieve sub-watt power dissipation comparable to BH devices, but with a single epitaxial deposition process followed by standard etching and deposition processes. Similar devices were recently demonstrated at wavelengths near 4.8 µm using narrow plasma-etched ridges with first-order DFB sidewall corrugations and SiN_x dielectric cladding [12]. In order to extend this approach to longer wavelengths, we have developed a process to sputter AlN cladding on the sidewalls of the laser ridge. Absorption in AlN is significantly lower than in SiN_x at wavelengths near 7 µm [16], but the real part of the refractive index is comparable (~1.9 at 7.4 µm), leading to similarly strong confinement of the lasing mode. In addition, the thermal conductivity of AlN is typically an order of magnitude higher than that of SiN_x, which facilitates more efficient heat extraction from the QC active region. For 1-mm-long devices with uncoated facets, we show mode-hop-free single-mode emission near 7.4 µm with CW threshold dissipation of 0.88 W at 20 °C and 1.07 W at 50 °C. To our knowledge, only Bismuto, et al., have reported sub-watt room-temperature operation of single-mode QC lasers emitting beyond 7 µm, and that report described buried-heterostructure, HR-coated devices with 0.8 W threshold power consumption at 20 °C with an emission wavelength of 7.82 µm [15].

2. Laser design and fabrication

For this work, the QC laser active region was designed for emission at 7.4 µm based on the non-resonant extraction approach described elsewhere [4] and grown by molecular beam epitaxy. To minimize free-carrier absorption in the laser waveguide, the active region was separated from the highly doped (3 × 10¹⁸ cm⁻³ Si) n-InP substrate by an intermediate 1.5-µm-thick n-InP layer with 5 × 10¹⁶ cm⁻³ Si doping and a 2-µm-thick n-InP layer with 2 × 10¹⁶ cm⁻³ Si doping. The laser active region was grown with strained In_0.60Ga_0.40As wells and Al_0.60In_0.40As barriers, forming a total of 40 stages with an average doping concentration of 2 × 10¹⁶ cm⁻³. On top of the active region, a symmetric 2-µm-thick n-InP layer (2 × 10¹⁶ cm⁻³ Si doping) was grown, followed by a 0.8-µm n-InP layer with 5 × 10¹⁶ cm⁻³ Si doping, a 1-µm n-InP confinement layer with 5 × 10¹⁸ cm⁻³ Si doping, and a 50-nm lattice-matched n⁺-InGaAs contact layer.

Two-dimensional mode calculations were used to determine the optimal lateral waveguide dimensions. The sidewall grating region was treated as an effective medium comprised of the epitaxial semiconductor layers and the AlN cladding, as described in Ref. 12. For a ridge width of W = 5.2 µm with 500-nm-thick AlN cladding and a thick Au contact, the waveguide cross section supports a single transverse-magnetic (TM) mode at a wavelength of 7.4 µm. For this geometry, we calculated an effective modal index of n_eff = 3.08, due in part to the low refractive index of the AlN cladding. We also calculated a modal loss of 1.7 cm⁻¹, where approximately 0.3 cm⁻¹ is attributable to absorption in the Au contact and the remainder is due to free-carrier absorption in the semiconductor layers. Figure 1(a) shows a cutaway schematic of the completed laser structure, and Fig. 1(b) shows the calculated mode profile for the fundamental TM-polarized mode assuming a grating depth of d = 450 nm.

 

Fig. 1 (a) Cutaway schematic of the QC laser structure, showing the DFB sidewall corrugations and AlN cladding. (b) Calculated electric field intensity for the fundamental TM-polarized mode supported by the fabricated waveguide geometry at a free-space wavelength of λ = 7.4 µm, with W = 5.2 µm, h = 6 µm, and d = 450 nm. The dashed lines indicate the boundaries between InP cladding layers of different doping concentration. For the structure shown, the modal effective index was calculated to be n_eff = 3.08.

Download Full Size | PDF

Laser ridges were patterned by first depositing a SiN_x hard-mask layer by plasma-enhanced chemical vapor deposition and then defining the ridge pattern by electron-beam lithography using Zeon ZEP520A resist. The pattern was transferred into the hard mask using an anisotropic SF₆/C₄F₈ inductively coupled plasma (ICP) etching process. After removing the electron-beam resist, an anisotropic, nonselective Cl₂/H₂/CH₄ ICP etching process was used to transfer the laser ridge pattern into the epitaxial structure, resulting in a vertical sidewall profile, as shown in Fig. 2(a). The laser ridges were etched to a depth of approximately 6 µm, just through the lower boundary of the QC active region. The ridge shown in the scanning electron micrographs in Fig. 2 include the residual SiN_x hard mask on top of a narrow Ti/Pt/Au contact stripe. After plasma etching, the ridges were etched in a dilute solution of H₂O₂ and HBr to remove damaged surface material from the ridge sidewalls and then immediately coated with a 500-nm-thick AlN cladding layer using high-frequency magnetron sputtering. After removing a narrow strip of the AlN and residual SiN_x from the top of the ridges by reactive-ion etching, a blanket Ti/Pt/Au top contact layer was deposited over the ridges, followed by electroplating of Au to a thickness of 5 µm. Prior to cleaving into 1-mm bars, the wafer was thinned to ~100 µm, and an ohmic n-type back contact was deposited. For testing, individual cleaved devices were attached epitaxy-side-up to submounts using Au-Sn eutectic solder.

 

Fig. 2 (a) Scanning electron micrograph of a partially fabricated DFB QC laser designed for emission near 7.4 µm. The waveguide ridge is shown after etching the first-order sidewall gratings though the QC active region and is cleaved to clearly show the position of the active region. (b) Calculated coupling coefficient, κ, as a function of sidewall grating depth, d. The calculated value of κ is indicated at the fabricated depth, d = 450 nm. The inset shows a top-down electron micrograph of the fabricated laser ridge, which has a total width of W = 5.2 µm and a grating pitch of Λ = 1205 nm.

Download Full Size | PDF

The approximation used in Ref. 12 was used to estimate the coupling coefficient as a function of depth assuming symmetric first-order DFB sidewall gratings with a sinusoidal profile. Figure 2(b) shows the calculated coupling coefficient, κ, as a function of grating depth, and the inset electron micrograph shows the final fabricated dimensions of the laser waveguide and sidewall gratings. For the fabricated grating depth, d = 450 nm, the calculated coupling coefficient is κ = 37 cm⁻¹, which indicates a moderately over-coupled DFB waveguide structure with κL ~4 for the fabricated cavity length of L = 1 mm.

3. Laser performance

After soldering the QC laser chips to submounts, the devices were tested on a temperature-controlled stage, using a calibrated thermopile detector for power measurements and a Fourier-transform infrared (FTIR) spectrometer with a cooled HgCdTe detector for spectral measurements. Figure 3 shows the measured CW voltage and light output as a function of current from one facet of a 1-mm-long DFB laser with the waveguide and grating dimensions described in the previous section. For this device, the maximum output power at 20 °C was observed to be 12 mW. This corresponds to a maximum single-facet CW output efficiency of 0.6%, which is approximately an order of magnitude lower than the efficiency of high-power single-mode QC lasers [6,17]. However, given that only milliwatt-level output power is needed for most in situ spectroscopy applications, the more critical performance parameter is power consumption, which is an order of magnitude lower for these low-dissipation devices than for the most power-efficient high-output DFB QC lasers. Also plotted in Fig. 3 is the laser output power as a function of dissipated electrical power. We observed a CW threshold current density of 2.2 kA/cm² and power consumption of 0.88 W at a heat-sink temperature of 20 °C. At 50 °C, these values increase to a threshold current density of 2.5 kA/cm² and power consumption of 1.07 W. Near maximum output power, the dissipated power reaches 2.1 W.

 

Fig. 3 (Left) CW compliance voltage and single-facet output power as a function of source current and heat-sink temperature for a DFB QC laser with cavity length, L = 1 mm, and maximum ridge width, W = 5.2 µm. The laser facets were uncoated. (Right) Single-facet output power for the same device plotted as a function of electrical power dissipated in the laser.

Download Full Size | PDF

We note that lasers were also tested after applying a metal HR back-facet coating. A 5-nm-thick Ti sticking layer and an optically thick Au reflector were evaporated onto the back facets of the lasers after sputter depositing a 100-nm-thick AlN spacer layer. For a DFB device nominally identical to the device described above, except with the addition of a metal HR back-facet coating, the measured threshold current was within 5% of the typical values measured for uncoated DFB devices, which is predictable for over-coupled DFB lasers. Similar to the uncoated devices, the HR-coated laser exhibited continuous single-mode tuning; however, we note that more devices would need to be tested to determine if the HR coating generally reduces single-mode yield. We also tested 1-mm-long, 5-µm-wide lasers without gratings that were fabricated at the same time as the DFB lasers. These lasers, which had a typically higher threshold current than the DFB devices of the same length, showed a significant decrease in threshold current (greater than 15%) with the metal HR coating deposited on the back facet. We conclude that, while the metal HR coating predictably decreases facet loss and therefore threshold current for non-DFB devices, the strong grating reflectivity limits the impact of the HR coating for the DFB lasers.

Emission spectra measured for the same uncoated DFB QC laser represented in Fig. 3 are shown in Fig. 4(a) with increasing values of input current. The measured peak wavelength is plotted in Fig. 4(b) as a function of both laser current and heat-sink temperature. Over the entire characterized range of operation, we observed continuous, mode-hop-free tuning with no evidence of side modes above the FTIR noise level of approximately 30 dB. The effective modal index corresponding to the measured peak wavelength agrees well with the value of n_eff = 3.08 calculated for the mode profile in Fig. 1(b).

 

Fig. 4 (a) Normalized emission spectra measured with source current increased by increments of 10 mA. (b) Peak emission wavelength measured as a function of current and heat-sink temperature. As a first-order DFB laser, the effective phase index of the fundamental laser mode scales with free-space emission wavelength, λ, and grating pitch, Λ, as n_eff = λ/(2Λ). The dashed contour shows the change in laser threshold current, I_th, as a function of temperature.

Download Full Size | PDF

We also measured the far-field emission profile for the same single-mode QC laser described above using a cooled HgCdTe detector mounted on a two-axis goniometer. The laser was placed at the eucentric point of the goniometer and driven with a pulsed current source using a pulse width of 0.5 µs and a period of 100 µs at 20 °C. Although the laser was operable in CW mode, pulsing enabled the use of a lock-in amplifier in order to improve the detected signal over infrared background noise. Figure 5(a) shows the far-field profile measured 13 cm from the laser emission facet over a range of horizontal and vertical angles. Only one lobe appears in the far field, confirming that the laser operates in just one lateral spatial mode. The smaller-amplitude fringes are due to reflections from the window of the HgCdTe detector.

 

Fig. 5 (a) Measured far-field emission profile for the same DFB QC laser represented in Figs. 3 and 4, collected at a heat-sink temperature of 20 °C and a peak current of 130 mA. The dashed red curves are Gaussian fits through the centroid of the far-field profile in the horizontal and vertical directions (the x- and y-directions, respectively, relative to the laser cross section shown in Fig. 1). (b) Calculated far-field power distribution for the fabricated laser geometry. The dashed black contour corresponds to the half-maximum power level.

Download Full Size | PDF

Line scans through the centroid of the far-field profile were fit with Gaussian functions to determine the full-width-half-maximum (FWHM) divergence angle in the lateral directions. Based on the waveguide cross section represented in Fig. 1(b), the far-field profile was also calculated for the fundamental TM-polarized mode at λ = 7.4 µm. The calculated optical power is plotted in Fig. 5(b), and the half-maximum contour is indicated. The measured FWHM divergence angle in the vertical direction (the y-direction in Fig. 1) is 58 ± 2°, compared with the calculated value of 54°. In the horizontal direction (the x-direction in Fig. 1), the measured FWHM divergence angle is 63 ± 1°, compared with the calculated value of 73°. This deviation could be related to the Gaussian fit being performed over the relatively limited range of experimentally measured angles as well as the spurious features caused by reflections from the detector window. We also note that this calculated value is based on the assumption that the dielectric constant in the grating region at the emission facet is the average of the semiconductor waveguide and the AlN cladding. If it is instead assumed that the emission facet was cleaved at the widest point of the grating profile, the calculated horizontal divergence angle is 70°.

It is noted that the nearly symmetric far-field divergence of these narrow-ridge QC lasers provides an advantage in terms of coupling to external optics. In particular, a single radially symmetric, high-numerical-aperture optical element (either a conventional refractive lens or a metasurface lens [18], for example) can be used to efficiently collimate or focus the laser emission. This is in contrast to interband quantum-well lasers that typically have characteristically thinner active regions and consequently exhibit more pronounced fast- and slow-axis divergence [19].

4. Summary

We have designed and fabricated single-mode QC lasers emitting near 7.4 µm with CW threshold power consumption below 1 W at heat-sink temperatures up to 40 °C. The QC active region was designed using an efficient non-resonant extraction approach and the doping levels in the semiconductor waveguide layers were optimized to reduce free-carrier absorption loss. A regrowth-free fabrication process was employed to fabricate narrow waveguides based on vertically etched DFB sidewall gratings with low-index, infrared-transparent AlN dielectric cladding. Single-mode devices exhibited continuous, mode-hop-free tuning from threshold to maximum power, and the laser far-field profile was imaged to confirm single-spatial-mode operation. Looking ahead, the availability of stable, low-power-consumption semiconductor lasers with single-mode emission at wavelengths beyond 7 µm can enable the development of portable in situ instruments for a variety of spectroscopy applications, from environmental monitoring and industrial process control to Earth and planetary science studies.