Introduction. Photonic generation of millimeter-wave (mm-wave) signals is attractive for numerous applications, such as radio-over-fiber, broadband wireless access networks, radar, and satellite communications, and has been intensively investigated over the past few years [1]. Due to the availability of extremely low-loss and low-cost optical fiber transmission systems, photonics technology is often preferred for generating mm-waves over conventional electronics. Normally, the approach for generating mm-wave signals in the optical domain is to beat two optical signals generated by two separate single longitudinal mode (SLM) lasers in a photodetector (PD). The beating signal has a frequency equal to the frequency difference between the two optical signals, and the beating signal quality is susceptible to external influences such as temperature and vibration, as these factors can affect the two separate lasers differently.

In comparison, utilizing a monolithically dual-wavelength laser (DWL) for beat frequency generation is more appealing. In a DWL, the two optical signals are generated in a single cavity by injecting current through a single electrode on the ridge waveguide. The impact of temperature variations on the frequency difference between the two lasing modes is negligible since the temperature drift affects the two wavelengths identically. Moreover, no external polarization controller is required before the two optical signals are input into the PD because they are generated within the same cavity.

In this paper, for the first time we have proposed and experimentally demonstrated a DWL array based on four phase-shifted sampled gratings (4PS) combined with equivalent grating chirp technology. The core of the equivalent chirp technology lies in modulating sampling periods at the micron scale to achieve complex grating responses in the ±1st channel [2]. In a linearly chirped grating cavity, the peaks in photon distributions of the two lasing modes are separated along the cavity, and mode competition is significantly suppressed [3]. However, the grating coupling coefficient κ of a conventional sampled Bragg grating (C-SBG) in the ±1st channel is only 1/π of that of a uniform Bragg grating (UBG). The 4PS structure can significantly enhance the κ value. Therefore, our new configuration not only enables equivalent chirp modulation through the adjustment of sampling periods but also achieves a κ in the +1st channel reaching 0.9 times that of a UBG while suppressing the zeroth grating reflection [4,5].

An electro-absorber (EA) section is also integrated with the DFB laser to enhance the phase stabilization between the two lasing modes [6]. By setting a series of chirp rates from 30 nm/mm to 90 nm/mm, we obtained radio frequency (RF) signals from 66.8 GHz to 73.6 GHz, with the linewidth ranging from 118 kHz to 210 kHz. It is worth noting that, compared to previously reported DWLs [3,7,8], we have achieved the narrowest linewidth with the shortest cavity length.

Device fabrication and experimental results. The epitaxial structure of the device emitting at 1.55 µm is based on an AlGaAsIn–InP heterostructure as reported in [9], and the fabrication process is the same as described in [6]. A schematic of the fabricated device is shown in Fig. 1(a). The device comprises an EA section (30 µm) and a DFB section (700 µm), separated by an electrical isolation groove (20 µm). The ridge waveguide width W is 2.5 µm and the height H is 1.92 µm; the effective refractive index of the guide mode (TE₀) is estimated to be 3.19. Both sidewalls have the same uniform seed grating with a period of 257 nm and a recess depth D of 0.6 µm. As shown in the inset in Fig. 1(a), the sampled periods are linearly modulated along the DFB laser cavity to achieve an equivalent chirp modulation, the chirp rate being defined as the ratio of the difference between the first and last sampling periods (P_n–P₁) to the length of the DFB cavity. Every sample period employs the 4PS structure. Two π phase shifts are introduced at positions 1/3 and 2/3 along the length of the DFB cavity to ensure dual-wavelength lasing. Figure 1(b) is a scanning electron microscope image of the etched ridge waveguide and 4PS sidewall grating. In the 4PS structure, the grating in each sampling period is evenly divided into four sections, with each adjacent grating section subjected to a π/2 phase shift. Compared to a C-SBG in which half of the sampling period has no grating, the 4PS structure retains a high grating coupling coefficient in the +1st-order channel, while suppressing the zeroth-order reflection [5]. Figure 1(c) is a microscope image of the fabricated four DWL arrays. The four devices (No.1 to No.4) have the same structure but different chirp rates of 30 nm/mm, 50 nm/mm, 70 nm/mm, and 90 nm/mm, respectively. After fabrication, the laser chip was cleaved into bars, with both facets left uncoated.

 

Fig. 1. (a) Schematic of the proposed device, with the inset showing the sampling period distribution along the cavity. (b) Scanning electron microscope image of the etched ridge waveguide and 4PS sidewall grating. (c) Microscope image of the DWL array.

Download Full Size | PDF

The simulation results in Figs. 2(a) and 2(b) illustrate the impact of grating chirp on the photon distribution of the two lasing modes. As shown in Fig. 2(a), in the absence of the grating chirp, the photon distribution peaks of the two lasing modes nearly coincide, concentrating at the positions of two π phase shifts. In comparison, when the equivalent chirp is applied, the peaks of photon distributions are separated, and the overlap region significantly decreases, thereby reducing a competition between the two lasing modes within the cavity. Figure 2(c) shows the simulated spectrum obtained using the traveling-wave method [10]. The two lasing modes (λ₁ and λ₂) are clearly observed, and two four-wave mixing signals (FWM 1 and FWM 2) appear on either side of them. The four-wave mixing signals result from fluctuations induced in the carrier density (and hence variations in the gain) caused by the beating of λ₁ and λ₂ [10]. Figure 2(d) illustrates the relationship between the wavelength spacing and the chirp rate. As the chirp rate increases, the wavelength spacing becomes larger; therefore, different chirp rates were designed to generate different RF frequencies.

 

Fig. 2. (a) Calculated photon distribution without a chirp. (b) Calculated photon distribution with a chirp. (c) Simulated optical spectrum. (d) Calculated wavelength spacing versus the chirp rate.

Download Full Size | PDF

Testing of the devices was carried out under continuous-wave (CW) conditions, with the laser array mounted on a copper heat sink at 20 °C. The optical outputs were measured from the DFB facet using an optical spectrum analyzer (OSA) with a 0.06 nm resolution bandwidth (RBW). Figure 3(a) depicts the operational characteristics of the device with a chirp rate of 30 nm/mm, with an EA biased at –1.7 V (V_EA= −1.7 V). It can be observed that the device exhibits a favorable dual-mode operation state when the DFB injection current (I_DFB) is within the range of 45–81 mA, and due to the presence of facet reflections, the device experiences mode hopping at 82 mA [11]. Figure 3(b) shows the power margin between λ₁ and λ₂ (PM), sidemode suppression ratio (SMSR), and wavelength spacing (Δλ) under different I_DFB. When I_DFB exceeds 55 mA, the PM remains within 3 dB, and it is less than 0.5 dB when I_DFB is between 70 mA and 80 mA; a lower PM is beneficial for improving the quality of the beating signal. When I_DFB is higher than 45 mA, the SMSR consistently remains above 30 dB. The wavelength separation, Δλ, falls in the range 0.539 nm to 0.551 nm, with a variation of 0.012 nm. Considering that this variation is smaller than the RBW of the OSA, the Δλ is independent of the injection current, implying that the beat signal’s center frequency is relatively stable. Figure 3(c) displays the measured spectrum when V_EA= −1.7 V and I_DFB= 75 mA. The results are consistent with the simulation, clearly showing the presence of two lasing modes and two FWM signals. It should be noted that significant Fabry–Perot (FP) modes are also observed in the spectrum due to the presence of facet reflections. Figure 3(d) shows the intensity variation of the larger FWM signal under a different V_EA. The FWM intensity reaches its maximum when V_EA= −1.7 V. For the proposed device, a stable dual-mode output can be achieved by adjusting the V_EA within a certain range [6]. In the subsequent experiments, the device data were obtained at the optimal V_EA where the FWM signal was the strongest.

 

Fig. 3. (a) 2D optical spectrum map of a laser with 30 nm/mm chirp rate when V_EA= −1.7 V. (b) PM, SMSR, and Δ_λ versus I_DFB. (c) Measured optical spectrum of the laser when I_DFB= 75 mA, V_EA= −1.7 V. (d) Maximum FWM signal intensity as a function of V_EA with I_DFB= 75 mA.

Download Full Size | PDF

The measured spectra and operating parameters of the array are shown in Fig. 4 and Table 1, respectively. Figure 4 shows typical spectra of the four devices with different chirp rates, each demonstrating good dual-mode characteristics with clear visibility of the FWM signals. Meanwhile, due to the facet reflection caused by cleaving, each device exhibits FP modes in the spectrum. However, these FP modes have a minimal impact on the beat frequency signal as the power ratio between the main modes and the FP modes is higher than 45 dB.

 

Fig. 4. Measured optical spectrum of the DWLs with different chirp rates.

Download Full Size | PDF

Table 1. Parameters of the Array

View Table | View all tables in this article

According to Table 1, a larger chirp rate results in a larger wavelength difference, which is consistent with the simulation results shown in Fig. 2(d). Additionally, it can be observed that devices with larger chirp rates require higher injection currents to achieve stable dual-wavelength lasing. This phenomenon can be attributed to an increase in the chirp rate leading to an increase in the threshold gain margin between λ₁ and λ₂. As a result, the device requires a higher injection to compensate for the weaker mode, thereby achieving equal output power for λ₁ and λ₂ [3].

Figure 5(a) shows typical power-injection current (P–I) characteristics from the DFB output side under a different V_EA with a scan step of 2 mA. The threshold current and output facet slope efficiency with 0 V applied to the EA section were 24 mA and 13.8% W/A, respectively. When the EA reverse voltage was increased from 0 V to –1.5 V, the threshold current was unchanged, but the slope efficiency was reduced as a result of an increase in an interband and exciton absorption. With further increase of the reverse voltage to –2.0 V, the DFB laser became optically isolated from the facet reflection and the slope efficiency increased, with the threshold current still unchanged. This indicates that the DFB laser was working as an independent cavity and the partitioning electrical slot prevented electrical coupling between the DFB and EA sections. Figure 5(b) shows the P–I curves of the four devices under the V_EA values shown in Table 1. Within the limit of 2 mA scanning steps, the threshold currents of the different devices are the same at 24 mA.

 

Fig. 5. (a) Measured P–I curves of the device with 30 nm/mm chirp rate under a different V_EA, (b) P–I curves of the devices with the different chirp rate for values of V_EA shown in Table 1.

Download Full Size | PDF

However, the four devices exhibit different slope efficiencies which can be attributed to the fact that the devices operate at different V_EA, and the different grating coupling coefficients resulting from different chirp rates also affect the efficiency [12]. It can be seen that there are some kinks in the P–I curves, which are indicative of mode hopping phenomena. For example, in the P–I curve of device No. 1 (red line in Fig. 5(b)), a kink is observed at 82 mA, corresponding to the mode hop seen in Fig. 3(a).

The RF signal was measured using a PD (u²t Photonics XPDV2020R 50 GHz) connected to an electrical spectrum analyzer (ESA). The optical signal was coupled by a lensed fiber from the device and then directly input into the PD without external phase adjustment or polarization control. A mixer (Keysight 11970V) is also utilized in the system to extend the measurement range of the ESA. Figure 6(a) shows the RF signals of the array measured under the conditions shown in Table 1; the ESA was set with an RBW of 100 kHz and a scanning range of 15 GHz. Consistent with the simulation results, the device with a higher chirp rate has an RF signal at a higher frequency. The variation of the RF linewidth with I_DFB is shown in Fig. 6(b). Since the RF signal is generated by beating λ₁ and λ₂, the linewidth of the signal is primarily influenced by three factors. One factor is the PM between λ₁ and λ₂, and when the PM is too large, the beat signal will become poor or even undetectable. The second factor is that the linewidth of the optical signal tends to decrease as the injection current increases [13]. The third factor is the facet reflection from the DFB side (up to 30%) can induce self-injection effects in the device, also leading to a reduction in linewidth [14]. Figure 6(c) shows the fitted linewidth of the device with 70 nm/mm chirp rate under V_EA= −0.6 V, I_DFB= 128 mA. During the linewidth measurement, the ESA was set with an RBW of 51 kHz and a scanning range of 80 MHz, and the –3 dB linewidth was obtained by measuring the –20 dB bandwidth and dividing by $\sqrt {99} $ to reduce the measurement error [15]. The performance of the device was also assessed as a function of the heat sink temperature over the range from 10 °C to 30 °C in steps of 2 °C. From Figs. 6(d) and 6(e), it is evident that within this temperature range, the device maintains a dual-wavelength operation. It is important to note that, despite the increase in the effective refractive index of the waveguide due to the rising temperature causing both modes to experience a redshift, the Δλ remains constant. This is because the two lasing wavelengths are generated within the same DFB cavity, and the rate of temperature-induced wavelength shift is the same for both wavelengths. The increase in temperature leads to a redshift in the gain peak, causing an imbalance in gain between the two modes and an increase in PM. In Fig. 6(f), the RF frequency decreases slightly with increasing temperature due to the overall redshift of the two lasing wavelengths [7]. The deterioration in the RF signal-to-noise ratio (SNR) is caused by the increased PM. The RF signal frequencies and lowest measured linewidths of the array at 20 °C are shown in Table 2. Table 3 presents a comprehensive comparison between the devices described in this paper and previously reported DWLs [3,7,8]. It is evident that the devices introduced in this paper (referred to as type 1) achieve the narrowest RF signal linewidth, spanning from 118 kHz to 210 kHz, while also exhibiting this outstanding performance with the shortest cavity length of 750 µm.

 

Fig. 6. (a) Measured RF signals of the DWL array. (b) RF signal linewidth of the array versus I_DFB. (c) Fitted linewidth curve of the device with a 70 nm/mm chirp rate. (d) 2D optical spectrum. (e) Wavelength spacing and PM. (f) RF frequency and SNR versus TEC temperature of the device with 70 nm/mm chirp rate under I_DFB= 75 mA and V_EA= −1.7 V.

Download Full Size | PDF

Table 2. RF Signal Frequency and Linewidth

View Table | View all tables in this article

Table 3. Comparison with Previous DWLs

View Table | View all tables in this article

Conclusion. This paper presents a dual-wavelength laser array based on a 4PS grating and equivalent chirp technology. The 4PS structure enables a chirped grating to be implemented by modulating the sampling period at the micron scale while maintaining a relatively high grating coupling coefficient. The introduction of the chirp reduces mode competition, enabling the device to produce two stable longitudinal modes. By designing different chirp rates, different wavelength spacings are achieved. The RF signals range from 66.8 GHz to 73.6 GHz with linewidths from 118 kHz to 210 kHz. Experimental results demonstrate that this device has the potential to be a highly integrated and easy-to-operate source of mm-waves.