Substrate-emitting quantum cascade lasers with a distributed Bragg reflector and a spectrally detuned second-order outcoupler Open Access

Quantum cascade lasers (QCLs) offer a unique combination of compact size, high efficiency, high optical power, and design flexibility in tailoring device characteristics to various infrared laser applications employing bandgap engineering. Output optical power for QCLs in the surface-emitting configuration is extracted perpendicular to the optical cavity utilizing a second-order diffraction grating etched into the upper cladding layers of the laser waveguide. Surface-emitting QCLs offer the benefits of a narrow emission spectrum, low output beam divergence, and a much lower risk of facet damage due to dramatically reduced optical power density at the output facet. Potential applications for these devices include standoff detection, spectral imaging, spectroscopy, free space optical communications, and infrared countermeasures.

The first surface-emitting QCLs were based on an air–metal–semiconductor second-order distributed feedback grating.^1–3 In that configuration, QCLs are mounted epi-up and output optical power is extracted from the epi-side of the laser. The main disadvantages of such devices are output power loss due to partial beam diffraction into the substrate and a high thermal impedance inherent to epi-up mounting. Substrate-emitting QCLs with a metal–semiconductor grating address both these concerns.⁴ Optical power extraction from the substrate side has been successfully used, for example, in the design of epi-down mounted ring-cavity QCLs.^5,6

QCL optical power can be effectively scaled with the cavity length.⁷ The optimal facet reflectivity for long-cavity (5–10 mm) QCLs is on the order of only several percent. Therefore, even a short grating can provide sufficient optical feedback required for lasing, leaving most of the laser waveguide unperturbed. This can be leveraged in the design of surface-emitting QCLs by combining a long gain section to achieve high optical power with a short first-order distributed Bragg reflector (DBR) used for wavelength selection and a second-order outcoupler for increasing emission aperture size (Fig. 1). Since both the effective reflectivity for the first-order DBR section and outcoupling efficiency for the second-order section strongly depend on a grating profile (duty cycle, depth, and shape), it would be beneficial to spectrally detune the two gratings from each other, which would allow for their independent optimization to improve the overall laser performance. The first demonstration of such devices is reported in this work. In contrast to the earlier work on QCLs with a second-order grating sandwiched between two first-order DBR sections,^8,9 the outcoupler in the new configuration is positioned outside the laser cavity and, due to the spectral detuning, it does not contribute to optical feedback (Fig. 1). Therefore, the value of outcoupling efficiency does not influence the laser threshold, and it only determines laser slope efficiency. In addition, due to the detuning, the outcoupler can be designed to operate in either weak or strong outcoupling regimes.

An inventory laser material used in this prove-of-concept work had the following layer sequence (starting from the substrate): 1.5 μm InP doped to 4 × 10¹⁶ cm⁻³, 1.7 μm active region comprising 40 stages (λ ∼ 4.1 μm), 1.1 μm InP doped to 4 × 10¹⁶ cm⁻³, 0.5 μm InP doped to 1 × 10¹⁷ cm⁻³, 0.5 μm InP doped to 1 × 10¹⁸ cm⁻³, and a cap InGaAs doped to 1 × 10¹⁹ cm⁻³ to achieve Ohmic contact. The wafer had been grown with molecular beam epitaxy on a low doped InP substrate to reduce free carrier losses.

First- and second-order gratings’ characteristics were simulated for this waveguide layer sequence using the numerical Floquet–Bloch approach described in Refs. 10 and 11, assuming that there was no optical gain or loss associated with the active region. For modeling both gratings, the emission wavelength was taken to be equal to 4.10 μm (center of the gain curve), the grating depth was taken as 900 nm, and the duty cycle was taken as 45%. It was also assumed that both gratings had a rectangular shape.

This set of input parameters resulted in a calculated grating period of 644 nm for the first-order grating. Figure 2(a) shows calculated profiles for the forward (partial wave H₀, blue curve) and backward (partial wave H₋₁, red curve) propagating guided waves. The calculated diffraction efficiency based on these H₀ and H₋₁ profiles is equal to 79%, and material absorption efficiency is equal to 21%. The inset of Fig. 2(c) shows the calculated spectral dependence of diffraction efficiency with full width at half maximum (FWHM) equal to 1.4 cm⁻¹ The value of 1.2 × 10⁻³ calculated for the imaginary part of the effective refractive index in the DBR section results in the exponential decay of the guided mode shown in Fig. 2(c) (black curve). Since the effective reflectivity on the order of several percent is sufficient for long-cavity midwave infrared (MWIR) QCLs, a DBR section as short as 100 μm could be used in this case. However, to minimize the laser threshold in these initial experiments, we chose to use a 500 μm-long DBR section.

Figure 2(b) shows that the second-order grating with a trial 1100 nm period is projected to result in a strong outcoupling into the substrate (partial wave H₋₁, green curve). Note that the back-reflected wave (partial wave H₋₂, red curve) carries a very low energy in the outcoupler section, which is a direct consequence of the spectral detuning between the two gratings. The calculated outcoupling efficiency based on these partial wave profiles is equal to ∼85%, and material absorption efficiency is equal to ∼15%. The value of 3.4 × 10⁻³ calculated for the imaginary part of the effective refractive index in the second-order outcoupler section results in exponential decay of the guided mode shown in Fig. 2(c) (blue curve). These data show that a >500 μm-long second-order grating would provide a nearly complete outcoupling of the guided mode. Its length was chosen to be 1.5 mm in this work.

As discussed above, the spectral detuning between the two gratings also results in output beam extraction at an angle to the growth direction. Using the Floquet–Bloch formalism,^10,11 it can be shown that the angle of beam extraction, α (defined in Fig. 1), is determined by the magnitude of detuning as follows:

where neff is the effective refractive index, λ₀ is the emission wavelength, and Λ is the grating period.

From Eq. (1), the calculated angle of output beam extraction for this set of input parameters is −33°.

Based on the numerical projections presented above, the targeted device dimensions for these initial experiments were the following: a 4 mm-long unperturbed gain section, followed by a 0.5 mm-long DBR section (period = 644 nm, depth = 900 nm, and duty cycle = 45%), and a 1.5 mm-long outcoupler section (period = 1100 nm, depth = 900 nm, and duty cycle = 45%). To simplify wafer processing, all three sections shared the same electrical contact and were therefore pumped at the same current density.

The fabrication process started with the deposition of a 150 nm-thick layer of silicon nitride (Si₃N₄) hard mask using plasma enhanced chemical vapor deposition. The second- and first-order gratings were then defined by electron beam lithography on polymethyl methacrylate (PMMA) resist. The gratings were transferred into the Si₃N₄ hard mask using reactive ion etch (RIE) based on CF₄:O₂ and subsequently etched into the semiconductor using a CH₄:Ar:H inductively coupled plasma (ICP) RIE. The grating formation steps were followed by the standard double-channel ridge-waveguide QCL processing: ∼20 μm ridges were patterned with photolithography and etched through the top cladding layers of the QCL structure with CH₄:Ar:H based ICP RIE, and an active region was etched with HBr based solution to reduce sidewall roughness. The standard electrical contact formation was followed by thick (∼3 µm) gold electroplating. The processed wafer was polished down to ∼100 μm, and 1.5 mm × 200 μm openings were defined in the backside metallization to allow for substrate emission. The processed wafer was cleaved into 6 mm-long devices, and the laser chips were bonded epi-down to aluminum nitride submounts.

Figure 3 shows the measured light–current–voltage (LIV) characteristics of a 6 mm × 20 μm laser at various stages of back facet, front facet, and substrate-side coating. The red curve shows the measured substrate emission with both facets and substrate left uncoated: threshold current density, maximum peak power, and slope efficiency were 1.52 kA/cm², 0.30 W, and 165 mW/A, respectively. A high reflective coating (HRC) was subsequently deposited on the back facet, while the front facet and the substrate were left uncoated. Due to reduced mirror losses, the threshold current density decreased to 1.30 kA/cm², while the peak output power increased to 0.8 W with a slope efficiency of 420 mW/A (Fig. 3, black line). Figure 4 shows (red curve) that the emission spectrum in this case was wide (∼100 cm⁻¹) due to a strong feedback from the uncoated front facet and the fact that all three sections were pumped. Finally, an anti-reflection coating (ARC) was applied to both the front facet and the substrate. The fully coated device had a threshold current density of 1.89 kA/cm², peak power of 0.6 W, and slope efficiency of 400 mW/A (Fig. 3, green line). Figure 4 (blue curve) shows that AR-coating resulted in spectral width reduction down to ∼3 cm⁻¹, which is acceptable for many practical applications. The measured central wavelength was 4.07 μm, close to its design value of 4.10 μm. FWHM for the spectral distribution stayed within the range of several wavenumbers throughout the entire laser dynamic range. A further reduction in spectral width can be achieved by optimizing the AR-coating recipe for the front facet. Alternatively, pumping both gratings to transparency would also prevent feedback due to complete guided mode outcoupling through the substrate [see Fig. 2(c), blue trace]. This, however, would make the laser fabrication and packaging more challenging. Finally, far-field measurements were carried out for the fully coated device with a room temperature mercury cadmium telluride detector at a distance of 160 mm. The far-field profile parallel to the laser cavity [Fig. 5(a)] was a single-lobed peak with a maximum measured at an angle of −34°, very close to its predicted value of −33°. The beam pattern had a stable behavior in the entire laser dynamic range with a FWHM of ∼1°, approximately twice its projected value. The larger experimental value for the far-field angular distribution can largely be explained by the measured spectral width of ∼3 cm⁻¹ (Fig. 4) and calculated angular dispersion of 0.104°/cm⁻¹. The measured far-field pattern in the transverse direction [Fig. 5(b)] was also a single-lobed peak with a FWHM of ∼18° as opposed to its calculated value of 10.5°. The larger experimental value can be explained by the fact that the 20 μm-wide waveguide supports multiple lateral modes.

In conclusion, we have demonstrated the operation of the first substrate-emitting QCLs with a first-order DBR used for wavelength selection and a second-order outcoupler spectrally detuned from the DBR section. This configuration allows for independent optimization of the DBR and outcoupler sections. A highly reflective/antireflective (HR/AR) coated 6 mm × 20 μm device delivered a peak power of 0.6 W into a single-lobed beam with ∼1° × 18° angular FWHM. The measured emission spectrum had a FWHM of 3 cm⁻¹ and its peak was centered at 4.07 μm.

Future work on these devices will include both optimization of the laser design and improvement in wafer processing. The model has to be extended to take into account gain/loss in the DBR and outcoupler sections.

Profiles for both gratings can be subsequently independently optimized to increase device efficiency. This includes numerical analysis for different grating shapes as well as numerical sweep across different values of etching depth and duty cycle. In addition, different pumping levels should be considered for the three sections to optimize device efficiency. For example, the DBR and outcoupler sections can be left unpumped or they can be pumped to transparency. Regarding wafer processing, a better control of etch depth using, for example, an etch stop layer would help with fine-tuning of the grating depth and shape.