On-chip dual-comb based on quantum cascade laser frequency combs

Optical sensing by means of frequency combs¹ is seen as an attractive spectroscopy tool as it combines high accuracy and precision together with a wide spectral coverage.² Dual-comb spectroscopy^3–6 is emerging as an attractive spectroscopy technique based on frequency combs, as all the comb spectrum can be acquired in very short time scales without requiring any moving parts.

Although first demonstrated in the mid-infrared (MIR) part of the spectrum, dual-comb spectroscopy has been extensively developed in the near-infrared spectral region,^5–7 as frequency combs are mature sources in this region. Extending this technique to the MIR range,⁸ where the fundamental roto-vibrational transitions of most gas molecules are present, will allow to achieve dual-comb spectroscopy measurements with accuracies and precisions never achieved.^9–11 

Nonetheless, optical sources capable of generating MIR frequency combs are extremely complex,^8,12–15 and dual-comb spectroscopy on the MIR range is only possible in highly equipped laboratories.^10,11 Instead, a compact dual-comb spectrometer in the MIR range could bring broadband high-precision measurements to practical applications such as trace gas or breath analysis.

Quantum Cascade Lasers (QCLs)¹⁶ are semiconductor lasers capable of generating frequency combs in the MIR and Terahertz parts of the spectrum.^17–20 As the comb formation takes place directly in the QCL active region, QCL frequency combs (QCL-combs) offer the unique possibility of a completely integrated chip-based system capable of performing broadband high-resolution spectroscopy. In the meantime, a theoretical description of the comb formation has already been developed,^21,22 dispersion compensation of QCL-combs was investigated,^18,23 and dual-comb spectroscopy using QCL-combs has been demonstrated.²⁴ In this letter, we demonstrate an on-chip dual-comb source based on MIR QCL-combs. We show that control of offset and repetition frequencies of both combs is possible by integrating micro-heaters²⁵ close to the QCL-comb sources, demonstrating that this device is ideal for compact dual-comb spectroscopy systems.

An optical frequency comb is a coherent source whose spectrum consists of a set of equally spaced modes.¹ Each comb mode f_n can therefore be expressed as

where f_rep is the repetition frequency and corresponds to the spacing between the modes, and n is an integer number. The offset frequency f_ceo corresponds to a common offset shared between all modes.

Dual-comb spectroscopy is based on the generation of a multi-heterodyne beating between two frequency combs with slightly different repetition frequencies (⁠$frep,1$ and $frep,2=frep,1+Δfrep$⁠, respectively, where $Δfrep$ is the difference in repetition frequencies) on a fast detector. Each beatnote contains information regarding the sample absorption at a different optical frequency. Therefore, the full control of both offset and repetition frequencies is necessary when designing a device for dual-comb spectroscopy applications.

Our concept consists of two QCL-comb sources fabricated on a single chip, into which two independent micro-heaters²⁵ are integrated next to each QCL. A schematic representation of a QCL-comb source together with its micro-heater is shown in Fig. 1(a). The QCL-comb sources used in this study are based on a modified version of an InGaAs/InAlAs broadband QCL design previously reported.¹⁷ The micro-heater consists of a small resistor created by etching a thin layer of Si-doped InP, previously utilized for achieving spectral tuning of single frequency QCLs.²⁵ The QCL-comb source and the micro-heater are biased by two independent current sources, as shown schematically in Fig. 1(d). The source driving the QCL is used not only for achieving laser action but also for controlling the offset and repetition frequencies of the QCL-comb. In addition, the source driving the micro-heater controls the amount of heat that it generates and can also be used for controlling the comb parameters. More precisely, the repetition frequency f_rep depends on the group index n_g, whereas the offset frequency f_ceo depends on the relative difference between the group index n_g and the effective refractive index n_eff. Both n_g and n_eff can be modified by using the temperature tuning of the refractive index. The temperature is controlled directly by the micro-heater or indirectly by the heat created by ohmic losses due to the current in the active region. In addition, the laser current can also modify the voltage of the structure and therefore its transition energies. The current has thus a direct effect on the gain of the structure and, as a consequence of the Kramers-Kronig relations, an effect on both n_g and n_eff. Through their control of n_g and n_eff, current and temperature do not act in the same manner on f_rep and f_ceo. In our system, the current of the microheater has only an effect on the temperature, whereas the current of the laser induces the two effects previously described.

Fig. 1(b) shows a scanning electron microscopy image of the fabricated device containing two QCL-comb sources and two micro-heaters. The lasers are separated by 200 μm and can be electrically isolated by a deep etched section done between the two lasers.

For optical characterisation of this system, the dual-comb source is collimated by a single high-numerical aperture (0.86) aspheric lens, as shown in Fig. 1(c). Both beams are first spatially separated and then combined in a 50/50 beam-splitter, giving access to two versions of the superposition between the two beams. The first is sent to a fast detector (HgCdTe, 250 MHz 3 dB cut-off bandwidth) in order to measure the multi-heterodyne beating between the two combs. The second is sent to a Fourier Transform Infrared Spectrometer (FTIR, 0.12 cm⁻¹ resolution) for acquisition of the optical spectra. The optical spectrum of each comb can be acquired independently by blocking one of the beams. For technical reasons, both lasers are driven in parallel by using a single low noise current source (Wavelength electronics QCL2000 LAB) with a specified average current noise density of $2 nA/Hz$⁠. Also, the micro-heater is driven by using a low noise current driver source (Wavelength electronics QCL1000 OEM) with the same noise characteristics.

Fig. 2 shows the typical performance of an on-chip dual-comb source based on MIR QCL-combs. A 3 mm long device coated with a high-reflection coating on the back-facet operates at room-temperature emitting >100 mW of output power in CW operation (c.f. Fig. 2(a)). As the devices are driven in parallel, the current represents the total current circulating in both lasers. Also, the output power is the total output power of the two combs. The optical power characteristics of the individual devices were also characterized, showing similar performances for each device, with a ratio between the output power of each device P₁/P₂ = 0.83. The optical spectra of both devices (cf. Fig. 2(b)) are centered at 1330 cm⁻¹ with $≃50 cm−1$ of optical bandwidth. No micro-heater is used in this case. A zoom on the optical spectra shows that the offset frequencies of the two QCL-combs are significantly different when no micro-heater is used (⁠$Δfceo≃7 GHz$ in this case). The origin of this offset frequency difference $Δfceo$ lies on the precision of the lithography step used to define the laser ridges. Achieving devices with nearly identical offset frequencies (⁠$Δfceo<1 MHz$⁠) represents a strong requirement in terms of precision of the laser width. A high value for $Δfceo$ represents an important technological drawback for dual-comb systems, as the multi-heterodyne beating would be observed at very high frequencies (⁠$Δfceo≃7 GHz$ here), requiring high-bandwidth detectors.

For solving this problem, the micro-heaters were used to control $Δfceo$⁠. Fig. 3(a) shows the optical spectra of the two combs when the current driving the lasers I_laser and the current driving one of the micro-heaters I_heater,1 are optimized to obtain almost identical offset frequencies, i.e., $fceo,1≃fceo,2$⁠. As observed from Fig. 3(a), the two combs seem to be aligned in frequency (within the resolution of the measurement) and do not show a significant difference in offset frequencies as compared to the case when the heater is not used (see Fig. 2(a)).

Further control of the comb parameters is demonstrated by measuring the radio frequency (RF) spectrum of the laser current.²⁴ The RF spectrum contains the beatnotes at the roundtrip frequency of the two combs f_rep,1 and f_rep,2, as they are biased in parallel by a unique current driver. One can therefore precisely measure f_rep,1 and f_rep,2 with a spectrum analyzer, as shown in Fig. 1(c).

We show the control of f_rep of a single comb by using one micro-heater to tune f_rep,1 while the second micro-heater is not utilized, therefore not acting on f_rep,2. The RF spectra acquired at different values of micro-heater current I_heater,1 (all other parameters being kept fixed) are shown in Fig. 3(b). When increasing the current in the micro-heater, we observe that one RF beatnote is shifted to lower frequencies while the other RF beatnote is not shifted, effectively decreasing $Δfrep$ (observed from I_heater,1 = 80 mA up to 214 mA). Around a precise value of I_heater,1 (around 214 mA here), both repetition frequencies are similar. By further increasing I_heater,1, one RF beatnote is further shifted to lower frequencies and $Δfrep$ starts to increase. For important values of I_heater,1, we observe that both RF beatnotes start to shift to lower frequencies, one being more sensitive to the I_heater,1 than the other. This is explained by the fact that at these values of I_heater,1, the heat dissipated starts to influence the second comb. All these observations are summarized in Fig. 3(c), where f_rep,1 and f_rep,2 are represented as a function of I_heater,1. This experiment finally demonstrates the control of the repetition frequency of a single comb by employing the micro-heater. As previously discussed, although the control of f_ceo and f_rep of a single comb is possible with our system, these two quantities are yet not controllable independently from each other.

In order to finally demonstrate the capabilities of this system for compact dual-comb spectroscopy applications, we measure the multi-heterodyne beat signal created by the beating of the two combs on a high-bandwidth HgCdTe detector, as shown schematically in Fig. 1(a). In this case, both $Δfceo$ and $Δfrep$ were optimized by controlling I_laser and I_heater,1 according to the considerations detailed previously. The results shown here are achieved by employing only one micro-heater, but they were also reproduced when using two micro-heaters. Additional control of the system is possible by biasing both lasers independently, improving the system flexibility.

Fig. 4 shows the optimized optical characteristics of the on-chip dual-comb based on QCL-combs. Both comb repetition frequencies were controlled in order to obtain $Δfrep≃3 MHz$⁠, as shown in Fig. 4(a). At the same time, both comb offset frequencies were optimized to be close to each other. Fig. 4(b) shows the multi-heterodyne beating between the combs, showing that $Δfceo<50 MHz$⁠. This value could be further reduced by slightly adjusting the micro-heater current. Moreover, 64 modes are observed on the multi-heterodyne spectrum, corresponding to an optical bandwidth of 32 cm⁻¹, as the comb repetition frequency is $frep≃0.5 cm−1$ (15 GHz).

A further advantage of our system is that technical noise originating from current fluctuations is suppressed. However, Flicker noise originating from the semiconductor laser cannot be reduced by fabricating tow lasers that are not coupled. We have recently demonstrated that Flicker noise is the main source of noise in our lasers at low frequencies.²⁶ We therefore do not expect any substantial reduction in the linewidth of the laser in acquisition times up to 1 s. However, an additional advantage of our system is its robustness regarding temperature drifts, quantified by the relative frequency temperature dependence coefficient $1νΔνRFΔT$ of one multi-heterodyne beatnote. In order to demonstrate the advantage of an on-chip system compared to a system where both combs would be totally independent, we measure this coefficient by using our previously demonstrated dual-comb spectrometer,²⁴ where the two combs consist of two different devices. In this case, a temperature fluctuation of one comb will directly translate into a temperature drift of the multi-heterodyne beatnotes. However, in a system where both combs are on the same chip, any temperature drift of the substrate will influence both combs in the same manner, therefore, minimizing the temperature drift of the multi-heterodyne beatnotes. A value of −8.8 × 10⁻⁵ K⁻¹ is measured in a system where both combs are independent, compared to a value of −1.5 × 10⁻⁶ K⁻¹ for the on-chip system, demonstrating a sixty-fold increase of the robustness regarding temperature fluctuations.

In summary, we have demonstrated an on-chip dual-comb system based on MIR QCL-combs. By employing two independent micro-heaters directly integrated next to each QCL-comb source, the combs repetition and offset frequencies are controlled by utilizing the temperature tuning of the refractive index. As a first step for compact dual-comb spectroscopy, we demonstrate that a multi-heterodyne signal originating from the beating of the two combs can be observed and that the parameters $Δfceo$ and $Δfrep$ can be precisely controlled. We show a multi-heterodyne beating corresponding to an optical bandwidth of 32 cm⁻¹ with a $Δfrep$ of 3 MHz, centered at 1330 cm⁻¹. Finally, due to the proximity of the micro-heaters to the QCL active region, these devices can achieve kHz of modulation bandwidth.²⁵ Combined with the recent developments in dispersion compensation of MIR QCL-comb sources,²³ this system can be utilized for achieving a fully-stabilized compact dual-comb spectrometer by employing, for example, all-electrical frequency noise stabilization techniques.²⁷ 

We thank Dr. Pierre Jouy for fruitful discussions. This work was financially supported by the Swiss National Science Foundation (SNF200020-152962), by the ETH Pioneer Fellowship programme as well as by the DARPA programme SCOUT (W31P4Q-15-C-0083).