Photo-induced DC drift in Mach-Zehnder modulators using lead lanthanum zirconate titanate thin films

Lead lanthanum zirconate titanate (PLZT) thin films have attracted much attention as an optical material because of their large electro-optic (EO) coefficient. The EO coefficient of PLZT is three to four times higher than that of lithium niobate (LiNbO₃).¹ Lithium niobate Mach-Zehnder (MZ) optical modulators are widely used in optical communications because of their excellent optical performance. However, they are about 10 cm in size. PLZT film is a promising candidate for miniaturizing optical devices such as high-speed optical modulators. We have successfully developed MZ optical switches and high-speed optical modulators by using PLZT thin films operating at 40 Gb/s.^2,3 However, direct current (DC) drift deteriorates the optical properties of PLZT MZ modulators; thus, the suppression of DC drift is crucial for ensuring the long-term stability of devices. Thapliya et al. reported on the DC drift of a PLZT MZ-type modulator.⁴ They argued that the DC drift is suppressed by using hydrogen-deficient dry etching. The DC drift in LiNbO₃ MZ optical modulators has been studied extensively.^5–7 These studies successfully explained the mechanism of DC drift, which can be suppressed by using a slight conductivity material as the buffer layer of the modulator to reduce the charge accumulation due to the DC bias voltage. Adjusting the resistance of the buffer layer, which involves adding conductive indium oxide (In₂O₃) in silicon dioxide (SiO₂), was proposed;⁵ thus, the DC drift of a LiNbO₃ MZ modulator can be drastically reduced, and the long-term operation of more than 10 000 h can be achieved with these improvements.⁸ The DC drift of a PLZT MZ modulator with low optical input power of less than 20 mW can be explained in the same manner as that of a LiNbO₃ MZ modulator. However, when the optical input power exceeds 100 mW, photo-refractive effects in ferroelectrics need to be considered. We experimentally investigated photo-induced DC drift in a PLZT MZ modulator with large optical inputs at the 1300-nm wavelength.

The PLZT MZ modulator chip we fabricated is shown in Fig. 1(a) and the cross-sectional structure of the modulation section is shown in Fig. 1(b). The modulator has 2-μm-wide ridge waveguides and the MZ interferometer consists of a pair of 2 × 2 multi-mode interferometer (MMI) couplers. The ridge waveguide structure was fabricated by using electron cyclotron resonance etching with fluorinated carbon-based chemistry. The coplanar waveguide (CPW) with a length of 10 mm was used as the electrode to apply the electric field to the waveguides. CPW-type electrodes are suitable for the high-speed operation of the modulator because of their large electrical bandwidth. The DC bias voltage was simultaneously applied to control the operating point of the modulator by a bias tee. The half-wave voltage (Vπ) of the PLZT MZ modulator was about 6.6 V, and the total optical insertion loss including the coupling loss of the fibers was about 8 dB.

The experimental setup for measuring DC drift is shown in Fig. 2. The DC bias voltage of the PLZT MZ modulator is set at the midpoint between the peak and the bottom of the modulation curve. The DC bias voltage is varied to maintain the midpoint of the modulation curve during measurements. The magnitude of the DC drift is expressed as

where V₀ and V(t) denote the initial bias voltage and the bias voltage at time t. Since the magnitude of the DC drift depends on V₀, it is normalized by dividing the V(t) shift by V₀.

One possibility for DC drift when 1310-nm light (photon energy of about 0.95 eV) is irradiated on a PLZT film is the photo-carriers generated in the PLZT film. To clarify the impurity level in PLZT and SiO₂ films as a buffer layer, we used the thermally stimulated current (TSC) method. The procedure of the TSC method is shown in Fig. 3. When the temperature of the sample increases by applying Vt, the thermally excited carriers are trapped in the impurity level E_t. The temperature of the sample is reduced to room temperature, and the trapped carriers are retained. When the sample is heated at a constant heating rate β, the trapped carriers are released at the temperature corresponding to the impurity level E_t, and the TSC corresponding to the level is observed. The TSC I is expressed as⁹ 

where k, ν, and T denote Boltzmann’s constant, the escape frequency factor, and temperature, respectively. The trap level can be determined by fitting the TSC results using Eq. (2). The experimental conditions were as follows: the sample was heated to 200 °C for 20 min to generate the thermal carriers, and then 100 °C for the next 20 min. The current when Vt is applied was kept constant at 100 pA so that all the existing impurity levels could be trapped. The TSC was measured at a collection voltage V_c of 0 V and β at a temperature increase of 8 °C in 1 min. To subtract background noise in the experimental system, the sample was heated again after the TSC current measurement to evaluate the background noise.

Figure 4 shows the normalized DC drift of a PLZT MZ modulator with varying optical input power applied to the modulator from 6 to 140 mW. The DC drift was kept constant at a low optical input power. For relatively large optical inputs of more than 20 mW, the DC drift was short-term, which changes abruptly in a few hours, and long-term, which gradually increases after a few hours. These results indicate that the DC drift of the modulator strongly depends on the optical input power.

We measured the TSC properties of PLZT and SiO₂ films, which were used in the modulator. The PLZT(8/65/35) films with a thickness of 1 µm were prepared on a SrTiO₃ substrate by using the sol–gel method, while 1-μm-thick SiO₂ film was deposited on a sapphire substrate by using a radio-frequency sputtering method. Figure 5 shows the TSC spectrum of the PLZT film. Broad peaks were observed at 80, 240, and 340 °C. These peaks correspond to 0.19, 0.83, and 1.1 eV, respectively, by fitting to Eq. (2). Nishida et al. reported that the impurity level of 0.95 eV in lead zirconate titanate (PZT) was caused by PbO_x defects through the TSC evaluation.¹⁰ In our measured TSC spectrum of PLZT, the peaks at 0.75 and 0.95 eV overlapped, which indicates 0.83 eV. The peaks of 0.19, 0.83, and 1.1 eV obtained from the TSC spectrum correspond to oxygen deficiency (V_O), lead deficiency (V′_Pb), and lead deficiency(V″_Pb), respectively. In other words, the incident light at 1310 nm (0.95 eV) is considered to excite the impurity levels of 0.19 and 0.83 eV in PLZT film and generate photo-carriers. The TSC spectrum of the SiO₂ film is shown in Fig. 6. The TSC current of the SiO₂ film without annealing could not be detected even when increasing temperature from 25 to 300 °C, while the SiO₂ film annealed at 300 °C for 2 h in the atmosphere showed a distinct peak at 0.57 eV. This indicates that this peak is attributed to the oxygen deficiency in the SiO₂ film caused by atmosphere annealing.¹¹ 

We discuss the short-term DC drift of V(t) in the PLZT modulator. The cross-sectional structure of the PLZT MZ modulator and the equivalent circuit model of short-term DC drift are illustrated in Fig. 7. The short-term DC drift of the optical modulator is usually explained by the following mechanism.⁷ When operating the modulator, high-speed radio-frequency signals and V(t) are simultaneously applied to the CPW electrodes to maintain the optimum bias point. The injected carriers by V(t) are transferred by the electric fields and accumulated in the high-resistivity buffer layer. In this case, the electric field applied to the optical waveguide decreases because of the opposite electric fields generated by the accumulated carriers. Therefore, V(t) is increased by these opposite electric fields to maintain the optimum bias point. However, the optical-input power-dependent DC drift in the modulator shown in Fig. 4 could not be fully explained with this mechanism, so further discussion on this phenomenon is needed. We assume that the refractive index change in the PLZT waveguide occurs from the photo-refractive effect due to the photo-carriers generated by exciting the impurity level of the PLZT film at large optical inputs. The increase in optical carriers is represented by a decrease in the resistance (R_wg) of the optical waveguide in the equivalent circuit shown in Fig. 7. The generated carriers are transferred to the electrodes through the buffer layer or PLZT thin film by an external DC electric field; thus, the time constant of the equivalent circuit depends on the resistance of the buffer layer (R_Buffer) and PLZT (R_PLZT). When the buffer layer is non-dope SiO₂, R_Buffer is about 10¹⁶ Ω cm and R_PLZT is about 10¹⁴ Ω cm. The magnitude of the DC drift can then be obtained by calculating V(t) such that the voltage across the waveguide of the equivalent circuit is constant. Thus, the short-time DC drift can be expressed as

From Eq. (3), the short-time DC drift can be suppressed by adjusting the resistance of the buffer layer to reduce R₁C₁ − R_wgC_wg. Then, we investigated the short-term DC bias drift for about 1 h. Figure 8 illustrates the relationship between the optical input power and the magnitude of the normalized DC drift after 1 h. As can be seen in Figs. 4 and 8, the short-term DC drift strongly depends on the optical input power. This result is consistent with the DC drift mechanism, in which the decrease in R_wg depends on the number of photo-carriers excited by the incident light. Figure 9 shows the relationship between the magnitude of short-term DC drift and R_Buffer when the optical input power is 130 mW. With increasing R_Buffer, the magnitude of the short-term DC drift also increases. From Eqs. (3)–(5), the magnitude of the normalized DC drift decreases with decreasing R_Buffer. Hence, the short-term DC drift can be suppressed by adjusting the resistance of the buffer layer.

The model of long-term DC drift is shown in Fig. 10. The photo-carriers move to the weak part of the light due to the external electric field and are trapped in the impurity level of the buffer layer or PLZT. Long-term DC drift is caused by the accumulation of the charge trap. We investigate the long-term DC drift more than 1 h later. Figure 11 shows the results of DC drift with different buffer-layer annealing conditions. Long-term DC drift drastically decreased without annealing the buffer layer. From the TSC results shown in Fig. 6, we argue that the impurity level of 0.57 eV due to oxygen deficiency caused by annealing in the atmosphere is related to long-term DC drift. Thus, the long-term drift is caused by carrier trapping in the impurity level.

The normalized drift magnitude of a PLZT MZ modulator with suppressed short- and long-term DC drift is shown in Fig. 12. The normalized drift of the modulator was found to be less than 1.3 at 100 h when the optical input power was 200 mW. This indicates that short-term DC drift can be suppressed by adjusting the buffer-layer resistance. Long-term drift was suppressed by reducing the defects in the buffer layer; thus, a PLZT modulator with stable operation can be achieved.

We discussed and verified the mechanism of DC drift at various optical input power. Short-term drift can be explained by using an equivalent circuit model. We revealed that a decrease in the resistance of the waveguide due to the generation of photo-induced carriers leads to larger short-term DC drift, and the short-term DC drift is suppressed by adjusting the resistivity of the buffer layer. We also revealed that the long-time DC drift is attributed to impurity-level trapping in the buffer layer. Thus, we achieved a PLZT MZ modulator with long-term stable performance.

We thank Takashi. Nishida of Fukuoka University for his useful advice on the evaluation of TSC.

The authors have no conflicts to disclose.

Hideo Hara: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Tomoki Joichi: Data curation (equal); Formal analysis (equal); Methodology (supporting). Shunsuke Abe: Data curation (equal); Formal analysis (supporting); Investigation (supporting); Methodology (supporting). Shin Masuda: Conceptualization (equal); Supervision (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.