Electro-optically tunable laser with ultra-low tuning power dissipation and nanosecond-order wavelength switching for coherent networks

Tunable laser diodes (TLDs) [1,2] with low-power dissipation, narrow-linewidth, and nanosecond-order tuning are key to next generation optical networks based on coherent technology that includes mobile fronthauls [3–5] and data center interconnections (DCIs) [6,7], which must be ecofriendly, low latency, and large capacity. Namely, implementing wavelength division multiplexing (WDM) using TLDs into mobile fronthauls will eliminate time-consuming procedures for dynamic bandwidth allocation. For DCI networks, high-radix optical switches using TLDs with nanosecond-order wavelength switching are expected to reduce the required number of power-hungry and large-latency electrical switches. Since a lot of network terminals are implemented in such short-reach networks, the power dissipation of TLDs is an important factor. Furthermore, a narrow linewidth is also required for TLDs to introduce digital coherent technology to increase network capacity. In addition to optical communications, a high-speed TLD with a narrow linewidth also contributes to laser sensing systems, such as tunable diode laser absorption spectroscopy (TDLAS) [8–11] and light detection and ranging (LiDAR) [12–14] with higher time resolution and signal-to-noise ratio (SNR).

However, the tuning mechanisms of conventional TLDs face a limit of further reducing the tuning power dissipation (${P_\lambda}$) and the tuning time (${\tau _\lambda}$).The recent minimum ${P_\lambda}$ seems to lie within the range of 100–300 mW, which has been achieved by using the thermo-optic (TO) effect [15,16] with a microheater placed on a laser cavity: 100 mW of heat generated from a microheater increases the power dissipation of the thermoelectric controller in the laser module by almost 200 mW (double the heat generated from a microheater). Furthermore, ${\tau _\lambda}$ of the TO effect with microheaters is limited to about 100 µs, which is much longer than the nanosecond-order wavelength switching required in DCIs. TLDs using carrier injection/depletion tuning [17–20] can provide a nanosecond-order ${\tau _\lambda}$, but they cannot be used in coherent systems because of linewidth broadening caused by the carrier effect. Moreover, Joule heating with current injection results in a microsecond-order wavelength drift, which complicates the electrical control system [19].

The electro-optic (EO) effect is a clearly superior tuning mechanism for a low ${P_\lambda}$ and short ${\tau _\lambda\!}$. Furthermore, a carrier-depletion cavity waveguide under bias voltage is suitable for narrow-linewidth lasing. However, an electro-optically TLD [20–23] with practical performance has been hardly developed due to the poor refractive-index change ($\Delta n$) of the EO effect, which results in a narrow tuning range. Even the effective $\Delta n$ of the quantum-confined Stark effect (QCSE) in a semiconductor multiquantum well (MQW), which is a relatively large EO effect, is typically about 10% of that from TO and carrier effects. To address these issues, we exploited the QCSE and used a reflection-type transversal filter (RTF) [24,25] to develop a new type of electro-optically TLD. The tunability of an RTF is also determined by the geometric design of the waveguide, while the tunability of conventional filters such as a distributed Bragg reflector depends only on the $\Delta n$ of the waveguide material.

Figure 1 shows the schematic of an RTF laser, which consists of an active waveguide (ACT) array for cavity gains and a ${5} \times {5}$ RTF as a tunable filter. Corresponding delay-line lengths of the laser are listed in Table 1.

 

Fig. 1. Schematic of RTF laser. For red and blue coarse tuning, tuning electrodes with lengths of $m{L_e}$ and $({3}\! -\! m){L_e}$ ($m = {0}$, 2, 3) are installed on a delay line whose differential length is ${0.5{\rm x}}(md{L_c} + \delta {l_m})$. Note that since light makes a round trip in a delay line with an etched facet mirror, differential “optical path” lengths become $md{L_c} + \delta {l_m}$ and $d{L_{\! f}}$. ${L_0}\geqq {3}{L_e}$ is required from a geometric condition. So, the maximum electrode length for fine tuning is ${L_0} + {0.5}d{L_{\!f}}$, which is equal to ${3}{L_e} + {0.5}d{L_{\!f}}$. A length of $\delta {l_m}$ functions as a phase compensator to change the transfer matrix of a ${5} \times {5}$ MMI to the matrix of a ${5} \times {5}$ discrete Fourier transform circuit [24,25].

Download Full Size | PPT Slide | PDF

Table 1. Delay-Line Lengths of RTF Laser

View Table

The RTF is based on a ${5} \times {5}$ multimode interference coupler (MMI) of that is 500 µm ${\rm long} \times {17}\;\unicode{x00B5}{\rm m}$ wide with a phase-tuning electrode and a delay-line array with coarse/fine tuning electrodes. Figure 2 shows the calculated reflection spectra of an RTF. When we consider one ACT port of the RTF laser, the optical-frequency-dependent complex reflectance $r(\Delta\! f)$ is

It is obvious that $\Delta \lambda$ for both coarse and fine tuning are proportional to ${L_e}$ in addition to $\Delta n$. Thus, we can obtain a large tunability of an RTF with a small $\Delta n$, such as that from the QCSE, by employing a long ${L_e}$. It should be noted that the value of ${L_0}$ does not affect the “differential” lengths of delay lines, which means that the reflection spectrum of an RTF is independent of ${L_0}$. The geometrical approach is impossible for conventional tunable filters.

 

Fig. 2. Calculated tuning spectra of RTF.

Download Full Size | PPT Slide | PDF

Figure 3 shows a fabricated RTF laser chip. The laser was fabricated on an InP wafer by butt-joint regrowth consisting of two types of MQW: one for optical gain (ACT section) and the other for the QCSE (RTF section including the MMI, the delay-line array, and the output port).

 

Fig. 3. Photograph of fabricated RTF laser chip. ${\lambda _{\text{PL}}}$ represents a photoluminescence peak wavelength of each MQW. See Supplement 1 for a description of the MQW for the QCSE.

Download Full Size | PPT Slide | PDF

After preparing the wafer, we formed ACT sections with a ridge waveguide and RTF sections with a deep-ridge waveguide by reactive ion etching. Next, we covered the waveguides with dielectric thin layers. Then, we used an electron beam process to deposit Au-based metal layers for electrodes and the etched facet mirrors of the RTF. Finally, we chipped out RTF lasers from the wafer by cleaving and coated the cleaved facets with antireflection (output port) and high-reflection (ACTs) film by sputtering. All layer patterns in the wafer process were defined by conventional photolithographic processes.

 

Fig. 4. Laser characteristics of an RTF laser. (a) Lasing spectra. (b) Corresponding applied voltage for ACT5. (c) Tuning power dissipation and laser linewidth. (d) Beat spectrum in delayed self-heterodyne measurement corresponding to a linewidth plot allowed in (c).

Download Full Size | PPT Slide | PDF

First, we measured the static lasing characteristics of a fabricated RTF laser. Figures 4(a) and 4(b) show lasing spectra of an RTF laser chip and bias voltages for successive longitudinal modes (${\rm FSR}\, =\, \sim{0.25}\;{\rm nm}$) for ACT5. All the experimental data below, including what is shown in Fig. 4, were measured at a bias current of 40 mA (threshold ${\rm current} = {15}\;{\rm mA}$) for one ACT in five, a chip temperature of 45ºC, and a fixed phase-tuning voltage of ${-}{4}\;{\rm V}$ while voltages from 0 to ${-}{9}\;{\rm V}$ are applied for coarse and fine tunings. The laser covers a 35 nm range (the full C-band) by individually biasing each ACT and applying voltage to the tuning electrodes of the RTF. Even for one ACT port, a tuning range of approximately 20 nm is achieved with the above voltage range. The TLD chip output power is about ${+}{0}\;{\rm dBm}$ for each wavelength (see Supplement 1). To the best of our knowledge, it is the first electro-optically TLD with a practical tuning range that can be further extended by using a longer ${L_e}$; namely, a longer ${L_0}$ as described above.

Next, we confirmed the ${P_\lambda}$ of the same RTF laser. The plots of ${P_\lambda}$ in Fig. 4(c) show the totals of the products of the bias voltages and corresponding currents (mainly caused by photocurrent) from the coarse-, fine-, and phase-tuning electrodes.

We obtain an extremely low ${P_\lambda}$ of less than 10 mW. We also measured the laser linewidth of the RTF laser using the delayed self-heterodyne method. The results are also plotted in Fig. 4(c), and one of the measured beat spectra is shown in Fig. 4(d), whose full width at half-maximum (FWHM) is about 700 kHz, which corresponds to the laser linewidth of less than 350 kHz. To the best of our knowledge, ${P_\lambda}$ of less than 10 mW is the best reported performance for a TLD that can cover the full-C band with linewidths available for conventional coherent systems.

In addition to the above extremely low ${P_\lambda}$ with a practical narrow linewidth, a further feature of the RTF laser based on the EO effect is a very short ${\tau _\lambda}$ and small wavelength drift. Using the experimental setup shown in Fig. 5(a), we measured the dynamic wavelength switching performance of an RTF laser. In the experiment, we biased ACT3 of the five ACTs and set the tuning voltages so that the lasing wavelength was ${\lambda _1} = {1532.5}\;{\rm nm}$. We applied a 10-kHz and 50-MHz square wave with an amplitude of ${-}{1}\;{\rm V}$ to the red-coarse electrode to change the lasing wavelength ${\lambda _1}$ to ${\lambda _2} = {1534.0}\;{\rm nm}$. As shown in Fig. 5(b), we confirmed a very short ${\tau _\lambda}$ of ${\sim}{500}\;{\rm ps}$. Although we need to simultaneously turn ACT ports on and off (causes a slow tuning due to thermal drift) for wavelength switching across a range wider than 20 nm [Fig. 4(b)], Fig. 5(b) shows the fundamental performance of very-high-speed wavelength switching based on the EO effect. Moreover, we also confirmed a stable output intensity during switching for 10 kHz modulation, which indicates that there is little of the wavelength drift in burst switching typically observed in carrier-injection TLDs. The suppressed wavelength drift, an unprecedented achievement that, to the best of our knowledge, is a practically important characteristic for simple laser control in optical switching with burst signals and also in sensing with adaptive tuning speeds, such as for foveated imaging in time-stretch LiDAR [14].

 

Fig. 5. Dynamic wavelength switching performance of an RTF laser. (a) Experimental setup (PPG, pulse patter generator; EDFA, erbium-doped fiber amplifier; OBF, optical bandpass filter). Inset: Lasing spectra of a pair of switching wavelengths of ${\lambda _1}$ and ${\lambda _2}$. (b) 10 kHz and 50 MHz modulation waveform from an oscilloscope. For 50 MHz modulation, we directly input electrical signals to the laser chip while putting 2200 pF capacitors beside the laser chips to reduce intrinsic electrical noise for all other measurements.

Download Full Size | PPT Slide | PDF

Using the laser, we also demonstrated high-speed optical switching of coherent modulation signals. The experimental setup is shown in Fig. 6(a). The bias conditions were the same as in the experiment shown in Fig. 5. We chose 100 Gb/s optical signals with 32 Gbaud dual polarization quadrature-phase shift keying (DP-QPSK) as the coherent modulation format, which is a mature format that has been used for recent digital coherent systems. We successfully achieved a ${\lt}{50\! - \!{\rm ns}}$ optical switching operation and a dynamic bit error rate (BER) of ${\lt}{1} \times {{10}^{- 10}}$, evaluated from the error vector magnitudes (EVMs) for the 100 Gb/s signals. The switching time was limited to evaluate the BER from the EVMs from the constellation diagrams [1000 plots/symbol corresponds to the 35 ns timeslot in Fig. 6(b)]. Moreover, it should be noted that, in spite of the slow repetition switching of 10 kHz, we observed no degradation of the dynamic BER. The results indicate that the RTF laser exhibits no wavelength drift, as predicted in Fig. 5(b). The demonstration confirmed the practical performance of the RTF laser for high-speed optical switching, even in coherent systems.

 

Fig. 6. Dynamic optical switching demonstration for coherent modulation signals. (a) Experimental setup (LN mod., Lithium niobate modulator; AWG, arbitrary waveform generator; OMA, optical modulation analyzer). (b) Dynamic BER characteristics with corresponding constellation diagrams for ${X}$ and ${Y}$ polarization (pol.) with optical signal-to-noise ratio (OSNR) of ${\sim}{30}\;{\rm dB}$.

Download Full Size | PPT Slide | PDF

We developed what we believe, to the best of our knowledge, is a novel electro-optically TLD based on the QCSE of an MQW. The laser exhibited a record extremely low ${P_\lambda}$ of ${\lt}{10}\;{\rm mW}$ for 35 nm in the full C band range with a linewidth of ${\lt}{350}\;{\rm kHz}$. We also confirmed ${\tau _\lambda}$ of ${\sim}{500}\;{\rm ps}$ for an ACT port, which, to the best of our knowledge, is the highest operation speed yet reported. Our results prove that, in spite of the small $\Delta n$, we can use the EO effect for a TLD with a practical tuning range by employing an RTF. This electro-optically TLD meets future demands for coherent technology-based WDM systems, including low-latency mobile fronthauls and optical switching in DCI networks.