With the rapid development of information society, the requirement for high quality of the communication network has been put forward, where a reliable light source enabling high-speed data transmission is urgently needed. Compared with externally modulated semiconductor lasers, directly modulated lasers have the advantage of small size and low energy consumption. The edge-emitting characterization of distributed feedback (DFB) lasers makes them easier to integrate with other components and gradually become the core module in optical communication systems.

The relaxation oscillation frequency is the main factor limiting the transfer rate and capacity of directly modulated laser (DML), and reducing the volume of the active region can effectively increase the relaxation oscillation frequency. Laser structures are usually divided into ridge waveguide (RWG) structures^[1,2] and buried heterojunction (BH) structures^[3,4]. In the RWG structure, the cavity length is usually compressed to less than 150 µm, and the BH structure can reduce the width of the active layer. Both methods can increase the relaxation oscillation frequency, but all have shortcomings. First, the short cavity method reduces the parasitic capacitance by compressing the cavity length to increase the relaxation oscillation frequency. In order to meet the light emission conditions, a large grating coupling coefficient is often required, which will lead to an increase in the laser threshold current and a decrease in output power. Second, reducing the cavity length will increase the contact resistance, which will aggravate the thermal effect of the current and ultimately lead to a decrease in injection efficiency. For the BH structure, multiple epitaxial wafer growth and etching processes are required, and the preparation is more complicated. Currently, there are few commercial applications.

Utilizing detuned loading^[5–7] and photon–photon resonance (PPR)^[8–10] effects to improve the modulation bandwidth of lasers has become the research frontier. These two methods usually appear in multisegment distributed reflector (DR) lasers or distributed Bragg reflector (DBR) lasers. In these structures, integration of active and passive sections is necessary and complicated, where during the butt-joint-regrowth process, the docking of the active and passive waveguides needs to be strictly controlled so that the grown waveguide layers can be on the same plane, which requires extremely precise material growth technology and greatly increases the difficulty.

In order to solve the problems above, a two-section high-speed directly modulated DFB laser with phase-shifted grating reflector is proposed. The laser consists of a DFB section and a grating reflector, and these two sections share the same active layer; the reconstruction equivalent chirp (REC) technique is utilized to design the grating structure^[11], which simplifies the production difficulty. A π phase-shift is introduced into the center of the Bragg reflection region, resulting in two steep falling edges on both sides of the reflection spectrum, where the detuned loading effect is enhanced, the PPR effect is introduced, and the modulation bandwidth is significantly improved.

Figure 1(a) plots the schematic of the sampling Bragg grating (SBG) structure; the calculated reflection and transmission spectrum of each section is shown in Fig. 1(b). Comprising a phase-shifted DFB section and a grating reflector, two steep falling edges appear in the reflection spectrum, enhancing the detuned loading effect. The phase across the interface remains continuous and both facets are antireflection (AR)-coated to avoid the effect of random phase. For ease of description, we refer to the DFB-emitting region as Section I and the DBR-reflecting region as Section II, and these two sections can be controlled by injecting current I1 and I2, respectively. The length of Section I and Section II is set to 400 µm, and the Kappa of the seed grating is designed to be 110cm−1. To inspect the enhancement of the detuned loading effect in this structure, left and right falling edges are divided into four equally spaced parts according to their widths, as shown in Fig. 1(b). We name their endpoints as positions A, B, C, D, and E, to study the modulation characteristics when the wavelength operates at different positions.

The small-signal modulation response curves when the lasing wavelength is located at corresponding positions of the left and right falling edges are calculated and presented in Fig. 2. The dashed line represents that the value of the small-signal modulation response is −3dB. With the increase of the penetration depth into the grating reflector, the modulation bandwidth increases from 17.8 to 24 GHz for the left side and from 17.6 to 22 GHz for the right side. Obviously, the steeper falling edge can provide an enhanced detuned loading effect, resulting in larger relaxation oscillation frequency and modulation bandwidth.

According to the theory described above, an eight-channel two-section DFB (TS-DFB) laser chip array is fabricated with the metal–organic chemical vapor deposition (MOCVD) technique. As shown in Fig. 3(a), the epitaxial structure of the designed chip is plotted. The physical diagram of the TS-DFB laser chip array is given in Fig. 3(b).

The TS-DFB laser consists of a front DFB section and a rare grating reflector section. The DFB laser has a circular electrode, and the grating reflector has a rectangular electrode; these two electrodes are separated from each other by electrical isolation. The length of these two sections is 400 µm. During the manufacturing, the standard length of the bar strip is 1000 µm. To meet this demand, no gratings are etched on the extra waveguide. In addition, for the efficient current injection in Section II, electrical isolation is added after the grating reflector. Thus, the third rare electrode has no practical significance and is actually not used in the test.

Then, the optical output power of the eight-array laser chip when I2=0mA is tested; the results are shown in Fig. 4. At room temperature, the overall threshold current is kept around 26 mA, as shown in Fig. 4(a). The maximum optical power is 12.29 mW, and the slope efficiency reaches 0.163 W/A. In order to study the thermal stability of the chip, the thermoelectric cooler (TEC) temperature is adjusted from 16°C to 40°C, and the data are recorded every 3°C. The results are shown in Fig. 4(b), where the slope efficiency decreases from 0.169 to 0.153 W/A, and the threshold current increases from 25.5 to 30.9 mA. The light-current (LI) curve maintains good linearity without mode hopping, but it can be seen that the slope of the curve decreases gradually with the increase of the injection current, which is the signal of thermal saturation.

The spectra of the eight-channel array laser are then tested, during which I1=100mA and I2=25mA are fixed; the results are shown in Fig. 5(a). The test results show that the structure can maintain good single longitudinal mode (SLM) operation, and the minimum side-mode suppression ratio (SMSR) in the array is still greater than 45 dB. The linear fitting analysis of the lasing wavelength demonstrates that the average wavelength spacing between neighboring channels is 1.654 nm compared with the design value of 1.6 nm, and the average error is only 0.054 nm, which accounts for 3.4%.

The effects of the injection currents I1 and I2 on the lasing wavelength and SMSR of the laser are investigated in Fig. 6 below. In Fig. 6(a), I₂ is fixed at 0 mA, and the spectra of the laser are recorded when the injection current I1 is increased from 20 to 100 mA, and in Fig. 6(b), I1 is fixed at 100 mA, and the spectra of the laser are recorded when the injection current I2 is increased from 0 to 25 mA. The data extracted from the spectra are processed and analyzed. When the injection current I1 is increased from 30 to 100 mA, the SMSR is steadily increased, and the wavelength is redshifted due to the joule heating effect from 1555.256 to 1557.096 nm, with a change rate of 0.029 nm/mA; at the same time, the SMSR tends to be saturated. In addition, it can be found that in Fig. 6(a) the lasing wavelength clearly jumps to a shorter wavelength with an interval of 0.189 nm when I1 is increased from 20 to 30 mA. This is due to the decreased effective refractive index caused by the increase in carrier concentration within the cavity when the injection current is increased. When current I2 is increased from 0 to 25 mA, its temperature change also causes the lasing wavelength to redshift, with the wavelength drifting from 1557.092 to 1557.4 nm at a rate of 0.012 nm/mA. Compared with the wavelength drift caused by the change of the current I1, the influence of the change in I2 on the working wavelength is smaller. However, an increase in the injected current I2 will change the feedback received by the main and side modes, thus changing the SMSR of the laser. In addition, there is also a quantum well in the reflection section, and the current injection will change the modes of intercavity so that the SMSR will not remain steadily increased or decreased.

In order to estimate the actual coupling coefficient of the fabricated grating, the amplified spontaneous emission (ASE) spectrum of a DFB laser with a cavity length of 400 µm is plotted with the injected current around the threshold. Thus, the grating coupling coefficient can be obtained by measuring the stopband width in the spectrum, according to Ref. [12]. From Fig. 7(a), we can see that the stopband width λs is about 2.464 nm, and the Bragg wavelength λB is about 1552.284 nm. The grating coupling coefficient κ can be expressed by κ=(πngλsλB2)2−(πLg)2,(1)where ng is the group refractive index, and Lg is the length of the DFB section. As described in Ref. [13], the coupling coefficient of +1st sampling Bragg grating can be written as κ1=κ0sin(πγ)πexp(−iπγ),(2)where κ0 is the coupling coefficient of the seed grating and γ is the duty cycle of the SBG, with a value of 0.5. The following equation can be obtained: |κ1κ0|=sin(π/2)π=1π.(3)

The coupling coefficient of the fabricated grating is then calculated to be about 28.4πcm−1, and the coupling strength of the equivalent +1st REC structure grating is 1/π of the seed grating, i.e., 28.4cm−1.

To study the effect of temperature variation on the laser operating characteristics, I1 is fixed at 100 mA and I2 at 25 mA during the test, and the spectra are measured at different temperatures. The heat sink temperature is varied from 16°C to 40°C, and the data are recorded every 3°C. From Fig. 8(b), it can be seen that the temperature change does not have much effect on the laser spectra, and the working wavelength of the laser is redshifted from 1549.44 to 1551.484 nm within the change range of 24°C, with a change rate of about 0.086 nm/°C. The SMSR decreases from 56.447 to 54.693 dB, and in general, the effect of the temperature on the laser is not drastic.

In the following, the chip is soldered to the subcarrier. The direct current (DC) and radio frequency (RF) signals are transmitted from the high-frequency ground-signal-ground (GSG) probe into the grounded-coplanar-waveguide (GCPW) and ultimately through the bonding wire to the chip. Then the spectrum is measured. In addition, in this batch of fabricated chips, we find that when the working wavelength falls on the left-falling edge of the reflection spectrum, the phenomenon of PPR may occur. Figure 9 plots the spectra of the laser at different injection currents I2 for I1=110mA when the laser is under RF signal modulation. It can be seen that when I2 exceeds 5 mA, the PPR mode starts to appear. The PPR mode is most pronounced when I2 reaches 20 mA; meanwhile, the laser always maintains SLM operation with an SMSR of more than 30 dB.

Figure 10 gives the response curve measured by varying currents I1 and I2. In Figs. 10(b) and 10(c), the internal characteristic of the laser changes with the increase of I2. During 0–5 mA, the detuned loading effect dominates, and the laser relaxation oscillation frequency and modulation bandwidth increase slightly; during 5–20 mA, the PPR effect gradually appears, and the modulation bandwidth increases significantly to 23.6 GHz. After more than 20 mA, mode hopping occurs in the cavity, and the PPR effect disappears. At this time, there is only the detuned loading effect, and the modulation bandwidth decreases to 16 GHz. Figure 10(d) plots the response curve measured by varying the injection current I1 at I2=18mA and I2=19mA. The red dashed line is the response curve when I1=120mA and I2=19mA, where the modulation bandwidth is maximized to 29 GHz. As I1 increases, the side peaks are continuously enhanced, which is reflected in the frequency response curves by the enhancement of the amplitude of the PPR peaks.

The frequency response curves of the chip with the operating wavelength located on the right falling edge are tested too, and the small signal response curves of the laser for different injection currents I1 and I2 are plotted in Figs. 11(a) and 11(b), respectively. Figure 11(a) plots the bandwidth change when fixing I2=0mA, and when continuously increasing I1, the bandwidth is increased from the initial 6.8 to 12.4 GHz. Figure 11(b) plots the bandwidth change when fixing I1=110mA and continuously increasing I2. Due to the uneven epitaxial growth processes at different locations of the wafer, the detuned loading effect starts to appear when the injected current I2 is more than 20 mA, and the modulation bandwidth increases to a maximum of 14 GHz. Comparing the results in Fig. 11 with those in Fig. 10, it can be seen that although the detuned loading effect also works when the operating wavelength falls on the right falling edge, and its improvement in modulation characteristic is much inferior to that falling on the left side with a steeper slope.

Figure 12(a) shows the relative intensity noise (RIN) measured by varying the injection current I1 with I2=0mA. As I1 increases, the increase in output power and the detuned loading effect causes the RIN to decrease from −128.63 to −145.56dB/Hz; the RIN measured by varying I2 with a fixed I1 of 100 mA is shown in Fig. 12(b). When no current is injected into Section II, the RIN is about −143.48dB/Hz. As I2 increases, the PPR effect starts to emerge, and the noise peak tends to drift to high frequencies where the peak frequency is consistent with the PPR frequency in the frequency response curve above. However, the PPR effect intensifies the dispersion of the laser, so the internal noise of the laser is not significantly suppressed. When I2 reaches 25 mA, the cavity mode jumps, and the PPR effect disappears. Then the detuned loading effect plays a dominant role in reducing the noise to −156.37dB/Hz.

Then, the 1-dB compression point of the laser is tested. Due to the limited input power of the signal source, an external 27 dB amplifier is used to test the output power curves from 2 to 19 dBm; the results are shown in Fig. 13. The curves remain linear, with no more than 1 dB of attenuation.

Chirp plays an important role in optical transmission systems. The application of DML at 1.55 µm wavelengths is limited, partly due to the large dispersion of the fiber. The poor dispersion tolerance is caused by the large chirp parameter of DML. In the following, the chirp parameters of the laser are tested under different injection currents, using a 75 km long fiber to form a time delay in conjunction with an erbium-doped fiber amplifier (EDFA) to transmit the amplified signal back to the vector network analyzer (VNA), where multiple resonance points appear in the frequency response curve. The corresponding chirp parameters can be calculated according to the method described in Ref. [14]. An example of chirp parameter measurement and calculation is given in Figs. 14(a) and 14(b). The measured chirp parameters are shown in Fig. 14(c). When I2 is fixed at 25 mA and I1 is tuned from 50 to 70 mA, the chirp parameter decreases from 4.62 to 3.08, at which time the detuned loading effect is strongest and then tends to saturate. The chirp parameter gradually stabilizes, with the wavelength drifting. However, the increase of current will also improve the relaxation oscillation frequency of the laser, so the modulation bandwidth of the laser is still rising. In addition, from the other line, it can be seen that the chirp is significantly suppressed by the appearance of the PPR effect. The chirp decreases from 5.15 to 2.58, and then the chirp increases to 4.14 after the disappearance of the PPR effect.

An integrated two-section DFB laser and an eight-channel array based on the enhanced detuning loading and PPR effect have been demonstrated. According to test results, the maximum output slope efficiency at room temperature reaches 0.163 W/A while maintaining good thermal stability. There is not much difference in the test results within the range of 16°C to 40°C. The spectral test results show good single longitudinal mode performance, with an SMSR exceeding 52 dB. The average wavelength error between adjacent channels is only 0.054 nm. The RIN can be reduced to below −156.37dB/Hz. Moreover, the modulation maintains excellent linearity, with a 1-dB compression point of more than 18 dBm. Due to the combined effect, the 3-dB modulation bandwidth increases from 13.5 to 29 GHz, and the chirp parameter decreases from 4.74 to 2.58.