1. Introduction

Advanced modulation formats such as a high-order in-phase quadrature modulation (IQM) format and a multifunctional photonic integrated circuit (PIC) are playing an important role in terabit-class signal generation [1, 2]. A LiNbO₃ (LN) based Mach-Zehnder optical modulator (MZM) has been used to produce advanced modulation formats [1] and offer the advantages of a low optical loss and a low-chirp drive. Although, the LN-based modulator is suitable for the next generation of transmission systems, its large chip size and the difficulty of active device integration have hindered its progress. On the other hand, InP-based MZMs have been studied intensively as an alternative to LN-based modulators because of their small chip size and potential for realizing a compact transmitter [2–4]. Although an InP-based MZM chip is very small the module size tends to be larger than expected. One factor preventing the realization of a compact module is the spaces needed for an RF interface and discrete thermoelectric components. The number of RF connectors required for a single-RF electrode MZM is half that needed for a conventional dual-RF electrode MZM. Therefore, a single-RF electrode is preferable as regards reducing the module size [4]. However, the structure and fabrication process of a single-RF MZM on a conventional InP substrate are complicated. Furthermore, a thermoelectric cooler (TEC) control and adjustment of the bias voltage to realize a constant drive voltage are essential for wideband operation because most InP MZMs employ a band-edge effect which induces the temperature and wavelength sensitive quantum-confined Stark effect (QCSE) and the Franz-Keldysh (FK) effect.

We have already proposed a wavelength and temperature insensitive single-RF drive MZM with simple planar structure on an InP (110) substrate. Although the (110) substrate is usually known to be a difficult facet for epitaxial growth due to the difficult nucleation [5–7], we have obtained a device quality epitaxial layer with a smooth surface by optimizing the crystal growth and fabrication processes, and demonstrated high-speed operation [8, 9]. On the other hand, high-speed characteristics such as bit-error-rate for non-return-to-zero (NRZ) modulation and IQM have been insufficiently studied.

In this paper, we report the static and dynamic characteristics of the InP (110) MZM in detail. The MZM exhibits a 40-Gb/s error-free and zero-chirp operation over the entire C-band and between temperatures of 10 and 50 °C at constant drive and bias voltages. Furthermore, we also investigated a twin-IQM in which MZMs are embedded. The twin-IQM exhibits 56-Gb/s × 2 quad phase shift keying (QPSK) modulation between temperatures of 20 and 80 °C at constant drive and bias voltages.

2. Device description

Figure 1 shows a cross-sectional diagram of a (110) InP-based MZM. A (110) InP substrate with a vicinal surface is widely used to obtain a smooth epitaxial layer. In this report, we used a 3° off (110) towards [111] B substrate and obtained an InP/ InGaAsP epitaxial layer with good surface morphology. The entire layer structure was grown at 570 °C using horizontal low-pressure metal-organic vapor-phase epitaxy (MOVPE). A feature of the structure is its single-RF drive operation with a simple buried waveguide and electrode. MZ arm waveguide stripes were formed along the [1$\overline{1}$0] direction on a (110)-oriented substrate. And the electrodes were arranged so that the electrical field could be applied in a lateral direction. In this configuration, when a voltage is applied to the center signal electrode, the direction of the electrical field in each MZ arm reverses as shown in Fig. 1. Since the change in the sign of the refractive index due to the Pockels effect depends on the direction of the electrical field, a single-RF drive push-pull operation is achieved. In this structure, as the amplitude of the electric field is always the same for both MZ arms, any undesired FK effect is automatically compensated for and the result is wavelength and temperature insensitivity. To apply the electrical field effectively, we employed a semi-insulating (SI) core layer buried in an n-type InP cladding layer. In this structure, the optical and electrical losses are expected to be smaller than with the conventional p-i-n type InP-based MZM, because it does not employ a p-type cladding layer. Figure 2(a) shows a cross-sectional SEM image of an InP (110) MZM in which an SI bulk core waveguide was introduced to act as a current blocking layer. The MZM was composed of two waveguide structures where a deep-etched mesa waveguide and a buried n-SI-n symmetric structure were introduced in the I/O passive waveguide regions and the electro-optic control region, respectively, as shown in Fig. 1. An Fe-doped SI-InGaAsP mesa waveguide on a SI-InP (110) substrate was buried in Si-doped n-InP cladding layers and Si-doped n-InGaAsP contact layers. The buried n-SI-n structure was 0.65 μm thick. A 10 μm thick Au traveling-wave electrode was designed to match the velocity difference between a lightwave and a microwave. To reduce the driving voltage, the electrode length and the entire chip size were set at 6.0 mm and 0.35 mm x 9.0 mm, respectively, as shown in Fig. 2(b). These sizes are relatively long compared with a conventional InP-based MZM utilizing the QCSE and the FK effect.

 

Fig. 1 Schematic diagram of InP(110) MZM structure.

Download Full Size | PDF

 

Fig. 2 (a) Cross-sectional SEM image and (b) microscope image of MZM.

Download Full Size | PDF

3. Static characteristics of MZM

First, we measured the optical losses of the MZM. The input polarization state was the TM mode. The entire insertion loss was 9.1 dB (including the lensed fiber coupling losses). The propagation losses of the I/O and n-SI-n waveguides were < 1.0 and < 1.5 dB/cm, respectively. These values are nearly the same as that of an undoped intrinsic core waveguide, and we found that Fe-doping has little effect on the propagation loss. Then, we measured the extinction ratio (ER) characteristics of the MZM for operating wavelengths of 1530, 1540, 1550, and 1560 nm as shown in Fig. 3(a). The ER was more than 25 dB for the entire C-band and as high as 37 dB for a wavelength of 1550 nm. The result indicates that the optical imbalances in the two MZ arms and excess loss are sufficiently small, even when a voltage is applied. The half-wavelength voltage (V_π) for all the measured wavelengths was 8.1 V (DC). We also measured the wavelength and temperature characteristics at a modulation speed of 1MHz. Figure 3(b) shows the V_π deviation at wavelengths of 1500 to 1580 nm, and at temperatures of 20 to 80°C. The V_π deviation over the entire range was less than 6%, which is about the same as that for a LN modulator. Furthermore, this value is sufficiently small compared with that of a conventional InP MZM in which the QCSE and FK effect are highly sensitive to the operating wavelength and temperature [10]. The V_π (1MHz) value was 5.0 V, which is lower than that for static DC operation. The reason is that the drift velocities of the trapped electron and hole in the SI-core layer are very slow. Specifically, when applying a DC voltage, the electrical field is concentrated at the edge of an n-SI heterojunction due to the transformation of the carrier distribution. Thus, the interaction between the lightwave and the electrical field is decreased. However, the asymmetric distribution becomes symmetric with respect to the center of the SI layer as the frequency is increased (>MHz) because we can disregard carrier transformation. As a result, the interactions when applying a high-frequency voltage become larger than when applying a DC voltage.

 

Fig. 3 (a) ER characteristics of MZM and (b) V_π deviation at wavelengths of 1500 to 1580 nm, and temperatures of 20 to 80°C.

Download Full Size | PDF

4. Small-signal responses and dynamic characteristics of MZM

Figure 4(a) shows the small-signal electrical and electro-optical (EO) S-parameters of the MZM. The electrical and EO responses were measured by a vector network analyzer and an Agilent N4373B lightwave component analyzer (LCA), respectively. A 6-dB electrical-electrical (E/E) S21 bandwidth of 30 GHz and a 3-dB electro-optical (E/O) bandwidth of 30 GHz were achieved. The results indicated that the MZM satisfied the lightwave - microwave velocity matching condition. In addition, the electrical reflection (S11) was below −20 dB, which is sufficiently small for practical use. One feature of the n-SI-n structure is that the RF responses are insensitive to the applied bias voltage. In particular, since the electron diffusion is blocked by the n-SI heterojunction, the depletion layer remains constant regardless of the applied voltage. We also measured the characteristic impedance of the MZM using time-domain reflectometry (TDR) with a Tektronix 80E10 TDR module as shown in Fig. 4(b). The input and output impedances of the modulation region were 45 and 50 Ω, respectively. Therefore, we obtained the S11 characteristics by using a conventional 50 Ω termination.

 

Fig. 4 (a) RF (E/E) response and small signal optical (E/O) response, (b) TDR characteristics of MZM.

Download Full Size | PDF

We also investigated the dynamic performance of the MZM. A 40-Gb/s NRZ signal with a pseudo random binary sequence (PRBS) of 2³¹-1 was applied to a single-RF electrode. The operating wavelength, input optical power and driving voltage were 1550 nm, + 13 dBm, and 5.5 V_pp, respectively. Figure 5(a) shows the operating-wavelength dependence and Fig. 5(b) shows the temperature dependence. Clear eye openings and error-free operation were obtained over the entire C-band and between temperatures of 10 and 50 °C at constant drive and bias voltages. In addition, a small signal chirp parameter α was estimated by utilizing a dispersive fiber and an LCA [11]. We found that the device had zero chirp (α = 0.02) at the quadrature bias point.

 

Fig. 5 Measured RF (E/E) response and small signal optical (E/O) response of MZM.

Download Full Size | PDF

5. MZM integrated twin-IQM

We described the athermal characteristics of a single MZM in the previous sections. In this section, we describe an MZM integrated twin-IQM for polarization-multiplexed (Pol-Mux) IQM that we fabricated and experimentally obtained single-polarization QPSK modulations. The chip, whose size was set at 1.2 mm x 13.0 mm, as shown in Fig. 6(a), was mounted on a ceramic carrier and terminated with a 50 Ω chip resistor. The electrical RF and DC interface is compatible with that of a LiNbO₃ IQM. The insertion loss for each IQM was 13.1 dB (including the lensed fiber coupling losses). The on-chip loss was estimated to be less than 4 dB, and this value is almost the same as that of a LiNbO₃ modulator. We also measured the optical E/O responses as shown in Fig. 6(b). We achieved a 3-dB bandwidth of over 25 GHz for all channels, confirmed the device to be suitable for 28-Gbaud operation. Figure 7 shows the experimental setup for the QPSK modulations. We used an external-cavity laser (ECL) mounted in an Agilent N4391A optical modulation analyzer (OMA). The electrical signal was generated by an Anritsu MP1800 pulse-pattern generator (PPG). To generate a 112-Gb/s Pol-Mux QPSK signal, the modulator was driven with four 28-Gb/s NRZ signals with different delay PRBSs of 2¹⁵-1 applied to each MZM. The signal was then amplified by four TriQuint electrical amplifiers to a peak-to-peak voltage of 7.5 V. The operating wavelength and input optical power were 1550 nm and + 13 dBm, respectively. To control the optical signal-to-noise ratio (OSNR), the optical output signal was passed through a variable optical attenuator (VOA), and amplified by an erbium doped fiber amplifier (EDFA). Then, the signal was passed through a 100-GHz optical filter and a VOA, and fed into the OMA for coherent detection.

 

Fig. 6 (a) Microscope image and (b) E/O (S21) responses of twin-IQM.

Download Full Size | PDF

 

Fig. 7 Experimental setup for each QPSK modulation.

Download Full Size | PDF

Figure 8(a) shows the constellation diagrams generated by lower and upper IQMs. Clear QPSK constellations were obtained for both IQMs. The Q values as a function of the OSNR at temperatures of 20 to 80 °C were also measured at constant drive and bias voltages as shown in Fig. 8(b). The Q values were well below the hard-decision threshold limits for forward-error correction. The Q-value deviation over the entire range was less than 1.0 dB, which is sufficiently small for practical use. These results indicate that the InP IQM is capable of TEC-free operation.

 

Fig. 8 (a) Constellation diagrams generated by lower and upper IQM and (b) OSNR characteristics at temperatures of 20 to 80 °C at constant drive and bias voltages.

Download Full Size | PDF

6. Conclusion

We reported an athermal and low optical loss InP(110) modulator with a simple planar n-SI-n heterostructure. The MZM exhibits 40-Gb/s error-free and zero-chirp operation over the entire C-band and between temperatures of 10 and 50 °C at constant drive and bias voltages. Furthermore, we investigated a twin-IQM in which the athermal MZMs were embedded. The twin-IQM exhibits 56-Gb/s × 2 QPSK modulation between temperatures of 20 and 80 °C at constant drive and bias voltages. We believe the (110)-oriented planar MZM to be suitable for a high-speed, high-order, and compact IQM including nested parallel MZMs.