1. Introduction

Devices based on whispering-gallery-mode (WGM) micro-resonators have been used for a wide range of applications [1], such as spectroscopy, sensing [2–4], photonic integrated circuits [5–7], and nonlinear processes [8–11]. They utilize the total internal reflection of guided modes propagating close to the cavity perimeter, confining light to small volumes [12,13]. Ideal rotating symmetric cavities confine light indefinitely, with negligible tunneling losses, resulting in very high quality (Q) factors [14]. The high-quality factors, combined with small mode volumes, significantly enhance light -matter interactions, resulting in low thresholds and precise resonance frequencies [15]. Because of their compatibility with monolithic integration, micro-resonator-based lasers have been used for mid-infrared (3–12 µm) on-chip communication and sensing [16]. For these applications, continuous-wave (CW) operation is desired because of the narrow linewidths and high stabilities. This is readily achievable with microcavity lasers in the visible and telecom spectral ranges, thanks to their low threshold power densities [17,18]. However, a CW mid-infrared microcavity laser capable of operation above room temperature has not been demonstrated.

Quantum cascade lasers (QCLs) are electrically pumped semiconductor sources in the mid-infrared, with CW outputs as high as 5 W [19]. High-power CW operation of a quantum cascade ring-distributed feedback laser with vertical power outcoupling has been demonstrated [20]. Significant advances have been made using deformed microcavities that increase emission directionality and power collection efficiency. The most successful designs include spiral-shaped [21], limaçon-shaped [22], and notched elliptical microcavities [23]. Nevertheless, CW operation of microcavity QCLs has been limited to cryogenic temperature (∼125 K) [21], which confines them to a laboratory. Therefore, raising its operating temperature above that of thermoelectric coolers by optimizing the device structure is an important goal. The key challenges are the high thermal load and the threshold power density of QCLs, which are one order of magnitude higher than those of telecommunications devices. Previously, we designed a selective electrical isolation scheme for the cavity and combined it with surface passivation for high-temperature CW operation [24]. We demonstrated 8-µm CW operation of microcavity QCLs up to 5 °C. The horizontal far-field profiles of all the devices were virtually the same; hence, electrical isolation and surface passivation did not affect the WGM distribution. However, the CW operating temperature of microcavity QCLs is still not high enough for a compact, portable sensing system.

Here, by using a selective thermal dissipation scheme, together with the selective electrical isolation process, we demonstrated 4.7-µm CW operation of microcavity QCLs up to 50 °C. The low thermal conductive silicon oxide is selectively removed to improve thermal dissipation of the microcavity device. The CW output power ranged from 0.4 mW at 20 °C to 0.1 mW at 45 °C. Single-mode emission was obtained with a side-mode suppression ratio of 30 dB. The device maintained highly unidirectional lasing, with an in-plane beam divergence of 1.9°. These results are critical steps in the development of mid-infrared on-chip sensing.

2. Design and fabrication

The high thermal load of the QCL makes CW operation of microcavity QCLs rather challenging. In WGM resonators, only a small fraction of the pumped volume along the cavity periphery generates photons, and the rest produces heat [25]. We can reduce the overall injection current of the device by electrically isolating the non-contributing region, without affecting the mode intensity distribution or the mode gain. Theoretically, this design could eliminate most of unnecessary thermal dissipation by providing a horizontal dissipation channel in this region, thereby reducing the inner temperature of the device. The platform for CW operation was based on a notched elliptical resonator with unidirectional laser action, whereby light circulating along the circumference of the cavity is scattered by the notch toward the opposite boundary. Then most of the scattered light is collected from the boundary opposite to the notch to form a collimated beam. Therefore, it was very important to carefully select and design the electrical isolation medium to maintain the mode propagation characteristics. Because the effective refractive index of semi-insulating Fe-doped InP is close to that of the upper cladding layer, we used it as the electrical isolation medium with a thickness equal to that of the upper cladding layer to maintain the mode propagation characteristics.

In our previous work, we used a regular passivation approach, where silicon oxide was deposited on the surface of the device and electrical injection window was opened only on the top of the gain region [26]. However, the thermal performance of the device could be hampered by silicon oxide since it has a much lower thermal conductivity that that of Fe-doped InP. To evaluate the effect of a silicon oxide layer on thermal dissipation, we did a comparative thermal study of the devices with and without a top silicon oxide layer, using a finite-element solver for three-dimensional electromagnetic heating simulations [27–30], as shown in Fig. 1. Specific details of the modeling were reported in Ref. [24], where the drive voltage was 10 V. The simulated QCL microcavity had dimensions X = 80 µm and Y/X = 1.2, and the electrical isolation region had a 70-µm semi-minor axis and an 86-µm semi-major axis. The device was a 500-µm square bonded epilayer side down to diamond submounts (5×3×0.3 mm³) using a 5-µm-thick indium solder. The cavity boundary was 45 µm from the edge of the substrate. The bottom of the diamond is set at 273.15 K, and the Stationary study is performed using the thermal insulation and electric insulation boundary conditions. The simulated three-dimensional temperature profile of the device without the top silicon oxide layer is displayed in Fig. 1(a). As expected, the temperature in the electrical isolation region was much lower than that in the pumped region; hence, the heat from the pumped region could be conducted horizontally through the isolation region. Figure 1(b, c) shows the two-dimensional cross-sectional temperature profiles of devices with and without the top silicon oxide layer, respectively. The removal of the top silicon oxide layer significantly increased the thermal dissipation from the active region through the Fe-doped InP region to the heat sink. The dependences of the injection current and corresponding maximum temperature of the devices on the size of the Fe-doped InP region are plotted in Fig. 1(d). The simulations indicated that the removal of the silicon oxide layer did not affect the electrical isolation, but significantly improved the heat dissipation and reduced the maximum inner temperature. The thermal conductance with and without the top silicon oxide layer was calculated from the finite-element simulations using ${T_{act}} = {T_{\sin k}} + {{{J_{th}}{V_{th}}} / {{G_{therm}}}}$, where T_act and T_sink were the QCL active-region and heat-sink temperatures, respectively. In the simplified model, thermal transport between the active region and the heat sink was characterized by the thermal conductance G_therm per unit of area of the active region. V_th and J_th were the threshold voltage and current density, respectively. The dependence of the thermal conductance of the devices on the size of the Fe-doped InP region is plotted in Fig. 1(e). The calculations indicated that removal of the silicon oxide layer improved the thermal conductance by more than 60% as the size of the Fe-doped InP region increased. Besides, the electrical isolation scheme is mainly implemented in the part that does not contribute to the waveguide mode gain, and the replacement of the top cladding layer with highly resistivity Fe-doped InP should not affect the WGM formation. Therefore, the selective removal of the top silicon oxide layer will neither affect the waveguide mode transmission nor induce extra current leakage.

 

Fig. 1. Finite-element simulation results for a continuous-wave microcavity quantum cascade laser. (a) Three-dimensional temperature profiles of the device without a top silicon oxide layer. (b) Two-dimensional cross-sectional temperature profiles of the devices with, and (c) without the top silicon oxide layer. (d) Injection current and corresponding maximum temperature of the two devices as functions of the size of the Fe-doped InP region. (e) Thermal conductance of the two devices as functions of the size of the Fe-doped InP region.

Download Full Size | PDF

A high-quality layer was used to passivate surface defects in the sidewall region of the resonator to suppress leakage current caused by thermally activated surface states. Commonly used passivation dielectric layers include silicon oxide and silicon nitride which both have with much lower effective refractive indices than that of the active region to ensure light confinement. However, their high optical loss beyond 7.5 µm wavelength would lead to the decreased device performance. Here, we used a shorter-wavelength QCL to circumvent the loss issue of the passivation materials. Shorter wavelengths also correspond to smaller mode volumes. Thus, we could further optimize the dimensions of the electrical isolation region to increase the thermal dissipation.

The microcavity QCL structure was grown using solid-source molecular beam epitaxy on a n-InP substrate. The entire structure included a 5-µm-thick lower cladding layer, a 0.2-µm-thick n-In_0.53Ga_0.47As lower confinement layer, a 1.5-µm-thick active region, a 0.3-µm-thick n-In_0.53Ga_0.47As upper confinement layer, a 0.2-µm-thick grading layer, and a 0.3-µm-thick contact layer. The active region was doped to an average level of 4.5×10¹⁶ cm^–3 and consisted of 30 InGaAs/InAlAs QCL stages designed for 4.7-µm emission. The layer sequence of each period, starting from the injection barrier, was (in angstroms): 38/12/13/43/13/38/14/36/22/28/17/25/18/22/19/21/21/20/21/18/27/18, where the In_0.36Al_0.64 As barrier layers are in bold type, the In_0.67Ga_0.33 As quantum well layers are in regular type, and doped layers are underlined. Photolithography was used to define the contour of the electrical isolation region, which was wet-etched to the upper confinement layer, followed by epitaxy of Fe-doped InP via metal-organic chemical vapor deposition. The contour of the cavity was also prepared via photolithography. A 1-µm-thick silicon dioxide hard mask was grown via plasma-enhanced chemical vapor deposition to transfer the pattern, and the structure was etched using an inductively coupled plasma to form steep and smooth sidewalls. Then, a 450-nm-thick silicon dioxide layer was grown to passivate surface defects in the sidewall region of the cavity, and the top silicon dioxide layer was removed by wet etching. Ti/Au and Au/Ge/Ni/Au top and back ohmic contacts, respectively, were deposited via electron beam evaporation. As in the simulations discussed above, the devices were cleaved into squares with 500-µm sides and mounted with the epilayer side down on a diamond substrate with indium solder.

3. Results and discussion

Experimental characterization of a microcavity QCL implemented with the electrical isolation scheme, mounted on a holder containing a thermoelectric cooler, were acquired under pulsed and CW conditions. Spectra were obtained with a Fourier-transform infrared spectrometer with a resolution of 0.125 cm^–1 in rapid-scan mode. Power measurements were acquired with a pyroelectric power meter after the beam was collected with a lens.

The results of the device in pulsed-mode operation at 20 °C are shown in Fig. 2. Figure 2(a) shows a scanning electron microscope image of the device without a top silicon dioxide layer, with a semi-minor axis X = 80 µm, and a semi-major axis Y = 96 µm. The notch, located at the intersection of the minor axis and the boundary of the ellipse, had a 1.8-µm opening and a 1-µm depth. The Fe-doped InP region had a 70-µm semi-minor axis and an 86-µm semi-major axis. Figure 2(b) shows the peak power vs. current characteristics of the device operated with a 1-µs pulse width at a 10-kHz frequency. The inset shows that 4.7-µm single-mode emission was observed at 1.25 times the threshold current. Figure 2(c) shows the average power at different pulse widths. Under the test conditions of pulse width of 5 µs and repetition frequency of 100 kHz, the measured power value reaches 27% of the peak power. This was much higher than that in Ref. [24]. Figure 2(d) shows the experimental threshold current that was fitted with ${I_{\textrm{th}}} = {I_0}\exp ({{{{T_{act}}} / {{T_0}}}} )$, where T₀= 202.8 K. The heating effects were negligible due to the pulsed mode at low duty cycles. The plot of threshold current and heat-sink temperature exhibited no deviation from the exponential.

 

Fig. 2. Fabricated device and its electrical and optical characteristics in pulsed-mode operation at 20 °C. (a) Scanning electron microscope image of the notched elliptical resonator. The white line is the cavity boundary. The inset shows a detail of the cavity. (b) Power-current characteristics of the device. The inset shows the emission spectra at 1.25 times the threshold current I_th (c) Average power of the device for various pulse widths. (d) Threshold current density vs. heat sink temperature. The black line is an exponential fit.

Download Full Size | PDF

Figure 3(a) shows the measured CW power–current–voltage characteristics versus heat sink temperature. With the improved thermal dissipation, the device could be operated at 50 °C, with a CW output power ranging from 0.4 mW at 20 °C to 0.1 mW at 45 °C. At 20 °C, the device exhibited a threshold current of 280 mA. The temperature dependence of the threshold current was fit with ${T_{act}} = {T_{\sin k}} + {{{J_{th}}{V_{th}}} / {{G_{therm}}}}$, where G_therm was 1429.9 W cm^-2 K^-1. Figure 3(b) shows CW spectra for various pumping currents at 20 °C. Single-mode emission was obtained with a side-mode suppression ratio of 30 dB. The emission frequency was continuously tuned over a few cm^–1, with a current-tuning coefficient of −69.33 cm^–1A^–1. Because the gain spectrum was very sensitive to temperature and carrier concentration in the active region, changes in the injected current and heat in the cavity could cause the lasing mode to “hop” with a 5.54 cm^–1 spacing, which agreed with the calculated value of 5.53 cm^–1 (the effective refractive index n = 3.19). The output power decreased before the inflection point and increased again with increased injection current; this behavior was attributed to mode hopping.

 

Fig. 3. Electrical and optical characteristics of a device during continuous-wave (CW) operation. (a) Power-current-voltage characteristics for various heat sink temperatures. (b) CW spectra for various pumping currents at 20 °C. The corresponding current-tuning coefficient was −69.33 cm^–1A^–1.

Download Full Size | PDF

The in-plane far-field profile was measured under pulsed conditions at room temperature. The device was placed on a motorized rotation stage with 0.1° resolution, 40 cm from a mercury-cadmium-telluride detector, with a maximum scan angle of 180°. As shown in Fig. 4, the device maintained unidirectional lasing, with a full-width at half-maximum beam divergence of 1.9°.

 

Fig. 4. Measured horizontal far-field profiles of the device in pulsed-mode operation. The inset shows a beam divergence angle of 1.9°.

Download Full Size | PDF

4. Conclusions

By using a selective thermal dissipation scheme together with the selective electrical isolation process to improve the thermal dissipation, we experimentally demonstrated CW operation of microcavity QCLs up to 50 °C. The elevated operating temperature is highly desired for portable systems used outside of a laboratory. The results were a critical step for applications requiring CW microcavity QCLs, such as on-chip sensing, photonic integrated circuits, and microcavity frequency combs. Furthermore, this design is applicable to a variety of WGM resonators, especially large cavities, which will greatly increase the potentials of microcavity QCLs.