Output facet heating mechanism for uncoated high power long wave infrared quantum cascade lasers Open Access

High optical power quantum cascade lasers (QCLs) are required for a number of defense and commercial applications, such as infrared countermeasures and free space optical communications. However, QCLs driven in continuous-wave (CW) mode tend to fail at optical power densities on the order of 10 MW/cm²–20 MW/cm², which roughly corresponds to the total output power level of 2 W–4 W for narrow ridge (10 μm-wide) devices. There is a lack of reliable experimental data on QCL core temperature dynamics in the high-power CW regime, as prior work in this area has been primarily focused on measuring QCL thermal resistance in the low power regime.^1–3 A better understanding of the QCL core temperature dynamics at higher CW powers is required to model and fabricate new designs with a higher damage threshold.

Recently, Hu et al. experimentally studied the facet temperature distribution of the low power 4.6 μm QCLs using micro-Raman spectroscopy.⁴ This work is dedicated to the high power long wave infrared (LWIR) QCLs. The dynamics of waveguides in the midwave infrared (MWIR, 3 μm–8 μm) and long wave infrared (LWIR, 8 μm–15 μm) is expected to be different. In the MWIR, mode guidance is tighter, leading to higher peak intensities during normal operation. In contrast, LWIR waveguides have wider beam waists resulting in lower peak intensities for the equivalent power. However, oxidized species, primarily Al₂O₃, experience absorption many times greater in LWIR than in MWIR.⁵ As shown below, this leads to a significant additional laser core heating at the output facet of the high power LWIR QCLs.

The measured device was a buried heterostructure InP-based QCL with a laser core composed of an InGaAs/AlInAs superlattice, emitting at a wavelength of 8.5 μm. The waveguide layer sequence starting from a low doped InP substrate was the following: 4 μm InP, 35-stage active region, 5.5 μm InP, and 0.2 μm InGaAs cap layer. The thickness and width of the laser core were ∼1.6 μm and 9.5 μm, respectively. The tested laser had a high reflection coating on the back facet and an uncoated front facet. The front facet had been exposed to an ambient environment with ∼50% humidity for one year. The laser was mounted epi-down on an AlN submount. Figure 1 inset is a schematic of laser bar orientation.

To study output facet heating, Raman spectroscopy was employed to locally measure the output facet temperature as a non-contact thermometer.^6–9 This method tracks the linear shift of Raman peak frequency with temperature.¹⁰ Raman spectra were measured using a Horiba LabRAM HR Evolution spectrometer. A long-working distance 50× objective was used to deliver 532 nm laser excitation power of 0.6 mW focused at the center of the QCL laser core, to collect the backscattered light in the $X(YY)X¯$ configuration. Figure 1 shows an exemplary first order Raman spectrum at room temperature. InP shows sharp peaks with small intensities near ±300 cm⁻¹, while the laser core superlattice features more complicated peaks at ±226.9 cm⁻¹ due to the InAs-like transverse optical (TO) phonon frequency⁹ and a small shoulder observed at 250 cm⁻¹ is attributed to the GaAs-like TO phonon frequency,^11,12 To calibrate the dependence of the Stokes peak position with temperature, the Raman spectra of the undriven device were taken while heating the device from room temperature to a maximum temperature of 140 °C in increments of 20 °C. The mean Stokes peak frequency position was determined from Lorentzian curve fitting of the spectra.

Figure 2 shows that the Raman peak position of the InAs-like TO phonon frequency of the QCL superlattice has linear dependence with respect to the temperature, which gives a slope of (−82.8 ± 7.6 K/cm⁻¹). This downshift in the phonon frequency is attributed to the thermal expansion of the lattice causing a decrease in the energy of the vibrational mode on heating.¹¹ 

Following the linear regression in Fig. 2, the temperature of the QCL core can be calculated as follows:

Here, ω_o is the reference Stokes peak position at room temperature, ω_m is the shift in the Stokes frequency position measured from the device, and m is the slope of the linear relationship between the Stokes shift and the temperature fit from measurement.¹³ 

Figure 3 shows the voltage, optical power, and temperature change of the tested LWIR QCL as a function of the current. A point of inflection in the temperature coincides with the CW threshold current of 0.8 A. The CW output power increases to ∼0.9 W at driving currents from 0.8 A to 2.0 A. The output power causes a sharp increase in the laser core temperature at the facet after the threshold. Thermal resistance based on the input power and temperatures, as determined by the Raman shift method increased from 6.5 K/W below the threshold to 12.3 K/W above the threshold, indicating that the reabsorbed optical power at the facet is a significant contributor to heating. The peak temperature change reached at the center of the core at 2.0 A is close to 225 K.

Increased optical power absorption at the facet, relative to the bulk, can be explained by facet material oxidation due to exposure to air. The InP bulk and the AlInAs/InGaAs core give many potential candidates for oxidization products. To determine the most significant contributor to absorption, oxidation at the facet was characterized via x-ray photoelectron spectroscopy (XPS) measurements at incremental etch depths with a Thermo Scientific ESCALAB 250Xi spectrometer in an ultra-high vacuum (UHV) chamber (4 × 10⁻⁹ Torr) using Al kα monochromated x-ray source (binding energy =1486 eV) at a spot size of 200 μm with charge compensation. The C 1s peak at 284.8 eV was utilized as a reference for charge compensation. Thermo Scientific Avantage software® was utilized for data processing (smart background was used for the peak fitting) to identify the chemical composition of the AlInAs/InGaAs core. The XPS depth profile measurements were performed using the cluster Ar⁺ ion source at an ion energy of 1 kV and a beam current of 1 μA. The etching rate and the total time of etching were 0.034 25 nm/s and 780 s (39 level—20 s for each level), respectively.

The oxidation reactions of InP and Ga are seen at the surface level, where etching several nanometers into the device showed markedly reduced levels of their oxidized product. Al₂O₃ (76.4 eV), however, remained prominent throughout the measurements, as shown in Fig. 4. Because of its relative prominence and the high degree of absorption in the 8 μm–9 μm wavelength range, it was assumed that the primary cause of optical heating was due to an effective “coating” of Al₂O₃ absorbing the optical output power.

A 2D heating model built in COMSOL Multiphysics was utilized to project heating due to both Joule heating from the driving current and the optical absorption at the facet caused by the effective “coating” of the facet by the oxidized Al in the form of Al₂O₃. Based on the distribution of Al₂O₃ shown in Fig. 4 (right), a depth of 23.7 nm was assumed for the effective Al₂O₃ coating, the depth of half maximum concentration. A concentration of 63%/2 was assumed, reflecting the average relative signal strength of Al₂O₃ to all Al sources, and halved for stoichiometry. The laser core itself consists of a series of AlInAs and InGaAs layers, with the AlInAs barriers accounting for 25.6% of the volume of the laser core. The relative concentration of Al to the total laser core structure, along with the proportion previously described based on XPS measurements, was used to adjust the net absorption of a 23.7 nm Al₂O₃ layer.

The laser core itself, simulated in COMSOL Multiphysics, had its material properties adjusted to reflect a mixture of oxidized products and the original material. Previous thermal models¹⁴ assume an in plane thermal conductivity of 5 W/mK and a cross plane thermal conductivity of 0.9 W/mK. Al₂O₃ has a thermal conductivity of 30 W/mK. The net material thermal conductivity was assumed to be the weighted average of 6.4/2.5 W/mK for the in-plane/cross-plane directions, respectively. Furthermore, to take into account the temperature variation of thermal conductivity expected of III–V semiconductors, the thermal conductivities of the laser core and bulk InP were assumed to be proportional to $300 KT1.55$⁠.¹⁵ 

Simulated temperature (Fig. 3) indicates a core temperature increase of 240 K at 2.0 A in good agreement with the Raman measurement. Our results suggest that the rise in temperature at a high optical load is because of facet oxidation, which can significantly reduce the damage threshold and, therefore, lead to device failure with enough output power. Specifically, the model can be used to estimate the output power level leading to facet material melting. For a hypothetical device with approximately three times higher slope efficiency and the same electrical load, the maximum projected temperature exceeded the melting temperature of InP (1062 °C) with an output power of ∼2.6 W. In this case, nearly all the heating is contributed by the output power absorption by the oxidized material along with the temperature dependence of thermal conductivity. The injected electrical power for these projections was taken to be 17.3 W, corresponding to the rollover point in Fig. 3.

In conclusion, we have reported measured and simulated temperature dynamics of the active region of the high power LWIR QCLs. We have observed that due to facet oxidation, optical absorption occurs at the laser facet, which results in significantly higher laser core temperatures than those predicted considering only Joule heating. The oxidation problem needs to be addressed via either facet passivation immediately upon wafer cleaving or facet cleaning before anti-reflection coating to ensure LWIR QCL reliability at the high power level.