Dual comb operation of λ ∼ 8.2 μm quantum cascade laser frequency comb with 1 W optical power

There is strong interest in the development of optical frequency combs in the mid-infrared range as the implementation of dual-comb spectrometers^1,2 in this frequency range which exhibit strong fundamental lines of chemicals could revolutionize chemical sensing as well as high resolution spectroscopy. As compared to other approaches for mid-infrared comb generation,³ quantum cascade lasers (QCLs) have attractive features such as being powerful, compact, and electrically driven. In these devices, frequency comb operation can be achieved thanks to four wave mixing⁴ and dual-comb spectroscopy has already been demonstrated.^5–7 In order to further improve the sensitivity and the acquisition speed of dual-comb spectrometers, significant efforts have been dedicated to the increase in both the optical power output⁸ and the dynamical range for which the QCL operates as a frequency comb^9,10 in order to take advantage of the maximum output power and bandwidth coverage of the laser.

Even if QCL frequency combs are stable under operation, they are sensitive to small changes in optical feedback. It is therefore important to perform all characterization studies under the same optical conditions. To ensure consistent data characterization, we adopted the setup schematized in Fig. 1(a). The laser is driven with a low noise driver and a bias-T that sends the RF part of the current on a spectrum analyzer to characterize the frequency comb beatnote. The beam is collimated with a high numerical aperture lens of NA = 0.85. In order to isolate the laser from back-reflections from the Fourier Transform Infrared (FTIR) Spectrometer, a tilted neutral density filter (NDF) with 1% transmission is placed after the lens. A beam splitter can be used to monitor the evolution of the optical power after the NDF (see supplementary material, Appendix A). Finally, the spectra are measured using a FTIR spectrometer with a resolution of 0.075 cm⁻¹, and the optical power is measured by placing a thermopile sensor directly after the lens. In this configuration, we verified that no element placed after the NDF will disturb the beatnote measured on the spectrum analyzer. Figure 1(b) shows two spectra obtained under the same driving conditions but once with a narrow beatnote and once with a broad beatnote. The broad beatnote was obtained by placing the NDF perpendicular to the beam to increase the optical feedback in the laser and destabilize the frequency comb operation. The two spectra have a different bandwidth; it is reduced by more than 15 cm⁻¹ when in frequency comb operation. It is thus crucial to perform all optical spectral measurements of the frequency comb while conserving the narrow beatnote properties during the measurement. All measurements presented here were done using this method. In some cases, nevertheless, weak optical feedback can be of use such as in Ref. 11 where it allowed the generation of the self-mixing effect to confirm comb operation.

The QCL active region is a strain compensated InGaAs/AlInAs dual stack heterostructure grown by MBE (see supplementary material, Appendix B). The laser ridge is 4.5 mm long and has a high reflectivity (HR) coating on the back facet (Al₂O₃/Au). The Light-Current-Voltage (LIV) characteristics and the quantum efficiencies are presented in Fig. 2 for temperatures of 258 K and 293 K. Powers above 1 W and an efficiency of 6.85% are achieved at 258 K. Subthreshold spectral measurements were performed at 293 K using a nitrogen cooled mercury-cadmium-telluride detector. The deduced group velocity dispersion (GVD) and modal gain¹² are plotted in Fig. 3. The top cladding of the laser has been especially designed following the approach described in Ref. 13 to obtain a low GVD to favor frequency comb operation. From simulations, it is found that it lowers the group delay dispersion (GDD) of the laser by 1800 fs² on average compared to a very thick cladding. For this device, the GDD measured below the threshold is between –2000 fs² and –7000 fs² in the emission bandwidth.

At 258 K, different beatnote types were recorded across the dynamical range (the laser becomes multimode shortly before 700 mA) and are shown in Fig. 4(a). It is possible to observe these different types mostly because of the extremely high signal to noise ratio (SNR) obtained, thanks to very strong beatnote signals (up to more than 60 dB). We attribute the strong SNR to the fact that the intra-cavity power is very high, and therefore, a strong modulation of the current is induced in the laser. For most currents, a sharp beatnote was observed (showing sometimes sub-lines like for 1075 or 1375 mA but with intensities more than 40 dB below the main peak); for some currents, a pedestal was visible (at 900 and 1200 mA for example) and finally, some other currents gave broad and multipeak beatnotes (at 1330 mA for example). These observations show that many different regimes can exist in a QCL and that a careful characterization has to be conducted before using them for spectroscopy purposes for example. The corresponding optical spectra measured using a Deuterated triglycine sulfate detector are shown in Fig. 4(b). For spectroscopy purposes, it is more meaningful to characterize the quality of the beatnote with its width at –20 dB instead of –3 dB as usually presented in the literature; indeed, a very high width at –20 dB would mean a merging of neighboring lines in a dual comb multi-heterodyne signal. The beatnote width at –20 dB and the beatnote frequency as a function of current are shown in Fig. 4(c); for the beatnote width, the full circles show the frequency comb operation, while the empty circles show the high phase noise regime operation (width above several hundreds of kHz). A close up on the beatnote using a span of 200 kHz and a bandwidth resolution of 100 Hz is shown in Fig. 5(a) for a current of 1375 mA at 258 K; a sharp and strong (SNR >60 dB) beatnote is a visible signature of frequency comb operation. Neglecting dispersion and because of the similar dependence of the effective and group index on external parameters, the ratio of the optical to beatnote linewidth is given roughly by the order of the line, in our case, N = 3740. Therefore, the 3 kHz beatnote linewidth measured at –20 dB would correspond to an optical linewidth at –20 dB of 11 MHz, which corresponds approximately to 0.0003 cm⁻¹. The mode spacing and power per mode are shown in Fig. 5(b) for spectra taken at 258 K and at room temperature for a current of 1375 mA. At 258 K, a total bandwidth of 85 cm⁻¹ with a power of 1.05 W is achieved and a continuous bandwidth of 82 cm⁻¹ has a power per mode above 1 mW with an average power of 4.1 mW. At room temperature, a total bandwidth of 75 cm⁻¹ with a power of 0.664 W is achieved and a continuous bandwidth of 73 cm⁻¹ (excluding 5 modes for wavenumbers in the vicinity of 1225 cm⁻¹) has a power per mode above 1 mW with an average power of 2.99 mW. At room temperature, the beatnote is more robust and is a narrow peak in all the dynamic range (the laser becomes multimode shortly before 900 mA) and also at 1100 mA [see Fig. 5(c)]. These values are similar to Ref. 9 in terms of frequency comb power at room temperature, but here, the laser is able to operate in the frequency comb regime up to its roll-over thanks to its low GVD.

Following these results, two QCL frequency combs from the same processed epi-layer which show similar performances have been used as sources in a commercial dual-comb spectrometer.⁶ The multi-heterodyne beatnote obtained for a single shot measurement with an integration time of 20 ms is shown in Fig. 6. The shape of these multi-heterodyne spectra results both from the convolution of the optical spectra of the laser and from the optical properties of the setup such as optical element transmission spectra or beam overlap on the detector. The data processing was performed following the approach described in the supplementary material of Ref. 5. The lasers are 4.5 mm long, and one is operated at 273 K and the other at 288 K with currents of approximately 1270 and 1386 mA, respectively, without further stabilization. Even if uncoated, each laser has an optical output power above 400 mW in those driving conditions. About 215 peaks corresponding to an optical bandwidth of more than 70 cm⁻¹ are observable between 200 and 600 MHz with a spacing of approximately 1.77 MHz. This corresponds to an increase in bandwidth coverage by a factor of at least 1.5 from previous published works.^5,7,13,14 The inset shows a zoomed-in image of a few modes to pinpoint the good separation of each line of the multi-heterodyne spectra, which is of paramount importance for quantitative spectroscopy measurements. The width of each line is of few hundreds of kHz due to the drifts of the comb during the acquisition time of 20 ms. Spectroscopy measurements have been done with another pair of lasers from the same process in Ref. 15. For an integration time of 100 μs, amplitude variations of a single emission line of 2‰ rms were observed.

In summary, we demonstrated a 1 W power QCL frequency comb emitting around 8.2 μm. More than 270 modes covering a bandwidth of 85 cm⁻¹ and a continuous bandwidth of 82 cm⁻¹ showing a power per mode above 1 mW at 258 K have been obtained. At room temperature, the QCL operates in the frequency comb regime in almost the full dynamic range giving up to an output power of 664 mW with a similar bandwidth. Due to their particularly interesting properties for spectroscopy measurements, lasers from the same process showing similar performances have already been integrated by partners in dual-comb spectroscopy setups. In particular, we present broad multi-heterodyne spectra with 215 lines covering an optical bandwidth of more than 70 cm⁻¹.

See supplementary material for the details about the demonstration that the optical power remains the same when operating in the frequency comb regime or in the broad beatnote regime under the same driving conditions in the first part and layer sequence of the QCL presented here in the second part.

This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) Program Office under Contract No. W31P4Q-15-C-0083 and the Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (SNF) under Contract No. 200020_165639.