1. Introduction

Over the past decade, the demands of modern telecommunications engineering have exponentially developed the technology of high speed electro-optic devices. In particular, the electro-optic (EO) modulator is one of the vital building blocks of these devices. The translation of the electric and optical signal is attractive because the modulator linearly responds to the applied voltage with pure phase modulation. Among the different types of materials used in the modulators, the organic polymer offers intrinsic advantages such as a large EO coefficient (r₃₃), low dielectric constant and loss, and widespread compatibility with other materials and substrates [1,2]. Therefore, the recent progress of polymer modulators opens up the variety of the devices and demonstrations the low driving voltage and high-speed operation [3–5]. Although current modulators realize the broadband modulation rate over 100 Gbps with common methods such as dual-polarization quadrature-phase-shift-keying (DP-QPSK, 4 × 25 Gbaud) [6–8]. The development of higher baud rates of the device is still an objective in communications technology. For instance, toward optical interconnection at short-reach application serial 100 Gbps transmission might be a cost effective solution. To date, traveling-wave EO polymer modulators demonstrate the frequency response of 100 GHz and broadband data modulation [8,9]. Such polymer modulators should play an important role in the versatile optical technology that cannot be practically handled by using previous high-cost modulator. Based on these advantages, low-cost high-throughput polymer waveguides are under investigation for future application [10]. Obviously, the long-term stability and photo-stability of polymer modulators becomes the important issue in the field of material development.

The EO polymer modulators have been studied for the last few decades to realize the characteristic low half-wavelength voltage (V_π) and high modulation bandwidth. There is no doubt that the optimum reported polymer modulators are mainly derived from the materials’ properties [11–13]. Recently, a significant enhancement of the EO activities have been demonstrated by using silicon slot and photonic crystal waveguides [14,15]. On the other hand, there is still an increasing interest in the unique and realistic technique toward commercial viability with a small fabrication footprint. While previous silicon and polymer hybrid modulators performed small V_π with TE-polarized light operation, the measured effective EO coefficients in device were rather smaller than the values in bulk films. This is mainly due to the poling difficulty in such silicon waveguides except using specially designed chromophores [16]. In this study, we fabricated the hybrid EO modulator, which consists of a waveguide structure with an ultra-thin silicon, EO polymer, and traveling-wave electrodes. In this waveguide, the optical quasi-TM mode is guided by an ultra-thin silicon core, while the EO effect is provided by the polymer cladding with a high optical nonlinearity. The devices are obtained via simple and facile process with a production yield better than the standard inverted-rib polymer waveguides. Since the thickness of silicon is only 50 nm in layer-by-layer structure, the resulted EO modulator performed the high EO coefficient under the best poling condition. The thermal stability of the EO modulator to maintain the high-EO activity is a key aspect beyond the laboratory investigation. We used the EO polymer having an enhanced glass transition temperature (T_g) up to 172°C, which was prepared according to our previous work with modified synthetic technique [17]. Such a high temperature property is advantageous as it allows thermal insensitive processes, and also offers excellent device stability at elevated temperatures [18]. For high frequency response of the modulator, the electric signal up to 40 GHz was applied by utilizing traveling-wave-electrodes onto the waveguide. The modulated sideband spectra were measured to evaluate the frequency dependence and the modulation index property.

2. Hybrid EO modulator design and fabrication

The silicon and EO polymer hybrid modulator is based on the ultra-thin silicon core, EO polymer and sol-gel SiO₂ claddings, and electrodes as illustrated in Fig. 1(a). The electro-optic interaction occurs at the interface of silicon and EO polymer. To achieve the high EO efficiency property and low propagation loss, it is essential for the light to be appropriately confined between these two layers. Figure 1(b) shows the simulated TM₀ mode distributions for the waveguide using 50 nm-thick and 5 μm-wide silicon. In the calculation, we set the refractiveindices of the EO polymer and amorphous silicon as 1.67 and 3.50, respectively.

 

Fig. 1 (a) The cross section of the hybrid silicon and EO polymer waveguide with 50 nm-thick silicon core, (b) the calculated mode distribution of the TM polarization, (c) the schematic top view of the silicon MZI with the traveling-wave electrode (EO polymer is not shown).

Download Full Size | PDF

It can be observed that the optical field is tightly concentrated around the silicon and partly penetrates into the EO polymer layer. The mode calculation indicated that the confinement factor (Γ) defined by the ratio of the optical field intensity in the EO polymer to the total intensity was 73%. Since such a field distribution is attributed to both the high refractive index contrast and the ultra-thin silicon, the confinement varies with different silicon thicknesses. According to the calculation, the Γ increased up to 83% when the thickness of the silicon decreases as 10 nm. However, the excessive modal expansion may lead to increased propagation loss due to the absorption from the metal electrode. A larger confinement is preferred to allow a higher EO activity, while avoiding such a loss we chose the thickness of silicon as 50 nm.

In fabrication, the bottom SiO₂ cladding was firstly prepared on the patterned electrode of 1 μm-thick aluminum from the solution of a sol-gel trimethoxysilyl derivatives [19,20]. After baking the spin-coated film at 120-140°C for 2 hours, the 3 μm-thick film was obtained. A 50 nm-thick silicon layer was deposited by using PECVD (PD-220NL, SAMCO Inc.). The waveguide was then patterned as the Mach-Zehnder interferometer (MZI, Fig. 1(c)) onto the silicon layer by the conventional photolithography technique followed by the ICP etching with SF₆ gas (RIE-400iPB, SAMCO Inc). Figures 2 show the SEM images of the silicon waveguide. The cross-section indicates that the fabricated waveguide has the identical dimension to the designed structure. The MZI consists of the moderate Y-junction and two-arm interferometer. The 5 mm-long junction is long enough to minimize the optical loss, and the distance between two arms is 300 μm. The time for ICP was only 40 seconds under our experimental condition, so that the fabrication tolerance is insensitive to the etching process. In such a shallow silicon strip, the optical TM mode near to the sidewall occupies a small fraction of the total available optical field. Consequently, the propagation loss can be expected to be quite low. In comparison, as for common polymer trench waveguide [21], a relatively deep trench results in a scattering loss due to the sidewall roughness.

 

Fig. 2 The SEM images of the silicon core indicating (a) 5 μm width and (b) 50 nm thickness.

Download Full Size | PDF

The EO polymer was then spin-coated, and baked at 105°C for 24 hours to form a 1.5 µm-thick film. Finally, a 1 µm-thick sol-gel SiO₂ layer was spin coated onto the EO polymer as the buffer to protect the device during the poling and modulating processes. The structure of the EO polymer is shown in Fig. 3(c). The EO polymer was prepared according to our previous work [17], which was modified to achieve the higher T_g by using the adamantly unit. Using such a polymeric bulky unit is effective to improve the thermal resistance. The measured T_g of the EO polymer was 172°C, which can be further increased up to 192°C by changing the polymeric unit ratio. Such a high-T_g property promises excellent EO stability performance. Conventionally, the EO polymers have been prepared by mixing EO chromophores in the polymer hosts. However, such guest-host materials showed T_g around 120-145°C or lower [3,4,8]. The traveling-wave electrodes were fabricated on the modulator to test the GHz frequency response. The 10 nm-thick gold seed layer with 50 nm-thick titanium adhesion layer was deposited by vacuum evaporation. The 10 μm-thick photoresist was spin coated onto the metal surface, and the mask was patterned by photolithography. After deposition of the 5 μm-thick, 20 μm-wide, and 10 mm-long gold electrode by an electroplating technique, photoresist and seed layers were removed by etching. The terminals of the traveling-wave electrode are designed in a similar manner to the literature’s geometry for the smooth transformation of the electric field of GSG-coplanar electrode to the micro-strip [22,23]. The ground pads were connected to the bottom electrode with silver epoxy though holes.

 

Fig. 3 (a) Poling of the hybrid MZI modulator, (b) measurement of V_π by differential voltage operation, (c) the EO polymer structure used in this study, and (d) the measured transfer function of the modular at 10 KHz.

Download Full Size | PDF

3. Experimental setup and measurements

To realize a modulator, the waveguide was poled at 167-170°C by applying a DC electric field though the electrodes as shown in Fig. 3(a). Since the poling temperature is around at the T_g of the EO polymer, the dipoles of the chromophore align in the direction of the electric field, which is also parallel to the optical field of the TM mode. The temperature was rapidly reduced to room temperature to lock the molecular orientation, and the electric field was then removed. In our experiment, an 100 V/μm electric field was applied onto the two-arm electrodes on the MZI waveguide to control the modulator under the differential electric signals.

The low-frequency modulation was performed on the waveguide to measure the half-wave voltage (V_π). The input light from a 1550 nm laser (81689A, Agilent) was coupled into the modulator by a polarization maintaining lensed fiber in TM polarization. The output light was collected by another fiber and detected by a photo detector (PDA10CS, Thorlabs). The differential voltages with 10 KHz triangular waveform from the function generator (AFG1022, Tektronix) were applied onto two electrodes of the modulator as shown in Fig. 3(b). Figure 3(d) shows the typical modulation curve of the output optical signal with V_pp = ± 4 V. By using the measured V_π of 0.9 V, the in-device EO coefficient was found to be r₃₃ = 165 pm/V under two-arm electrodes operation using $r_{33}=\lambda d/2n_{eff}^3{V}_{\pi }L\Gamma$ with λ = 1550 nm, d = 6.7 μm, $n_{eff}^3$ = 1.63, L = 10 mm, and Γ = 0.73. Here λ is the laser wavelength, d is the thickness of the waveguide, is $n_{eff}^3$the effective refractive index, and L is the length of the electrode. The obtained high EO activity originates from the materials r₃₃ and the effective poling realized in the hybrid ultra-thin silicon and EO polymer structure. The measured in-device EO coefficient was higher compared to the values measured in the analogous polymer modulators having standard inverted-rib structure [5,21]. For this comparison, the same EO polymer was used as the core and sol-gel SiO₂ as the top and bottom claddings, and the EO coefficient was measured to be 87 pm/V. Obviously, higher EO coefficient of the hybrid device is attributed to the higher poling efficiency. Since the thickness of silicon core is only 50 nm, the cross-section of the waveguide is rather flat, which presumably spreads the homogeneous electric field between bottom and top electrodes to induce the orderly molecular arraignment. Further, sol-gel SiO₂ claddings may become the electric buffer to avoid the dielectric breakdown of the EO polymer layer during the high-voltage poling.

The polymer EO modulator has one obvious advantage having the relatively small velocity mismatch between RF and optical waves [22–24]. These properties realize a high-frequency photonics device with large gain and small noise. The RF response of the fabricated hybrid modulator is characterized by measuring the S21 parameter with a vector network analyzer (VNA MS4644B, Anritsu) up to 40 GHz. The RF signal was applied onto the one arm of the electrodes via a picoprobe (40A-GSG-250-DS, GGB Industries Inc.). The electrode was terminated with an external 50 Ω load. The modulated signal was received by the photo detector (MN4765B O/E calibration module, Anritsu). The measurement system was calibrated using an automatic VNA calibrator (36585 Auto Cal, Anritsu). The measured 1-40 GHz range S21 responses for the hybrid modulator are shown in Fig. 4. The signal intensities are normalized at 1 GHz. The measured 3 dB modulation bandwidth of the hybrid modulator was 23 GHz, and a 9 dB reduction was found in response at 40 GHz.

 

Fig. 4 Measured normalized frequency response of the hybrid EO modulator.

Download Full Size | PDF

In order to discuss the detail of the modulation index of the hybrid EO modulator, we performed another measurement by using the sideband detection technique at different high frequencies. In the experiments, the output light from the modulator was directly connected to an optical spectrum analyzer (AQ6370, Yokogawa), and transmission spectrum was measured under the RF signals. The RF signals were provided by a signal generator (HMC-T2270, Hittite Microwave Corp.) and applied onto the traveling-wave electrode. Figures 5 show the transmission spectra of the output light, which is modulated using sinusoidal signal with frequencies of 10, 20, 30, and 40 GHz. The optical spectra consist of a main peak of the carrier light at 1550 nm and side bands equally spaced around the center with the applied frequencies.

 

Fig. 5 The measured optical transmission spectra of the hybrid modulator at (a) 10 GHz, (b) 20 GHz, (c) 30 GHz, and (d) 40 GHz

Download Full Size | PDF

Generally, the power ratio of main peak (J₀) and first sideband (J₁) is proportional to the square of zero-order and first-order Bessel function of the first kind [ΔP = J₀/J₁ = [J₀(η)/J₁(η)]²], where η is the modulation index [14,26]. Figure 6(a) shows the change of η as increasing the RF power, which clearly sustains the linear modulator response to the applied field with pure phase modulation. Figure 6(b) shows the frequency dependent of the η with the RF powers of 1.0 and 2.0 V_pp. The measured frequency response of the modulation index is identical to the bandwidth result shown in Fig. 4. The half-value of η was observed at 23 GHz and 25% at 40GHz. In reference, the commercialized LiNbO₃ intensity modulator (MX-LN-10, Photoline Tech.) was tested to compare the modulation index to the hybrid modulator. The measured η of the reference was 0.13 with the RF power of 2.0 V_pp at the frequency of 10 GHz. Clearly, our hybrid silicon and EO polymer modulator showed a higher modulation index due to a larger EO activity of the waveguide. Since the EO activity of the polymer is independent on the signal frequency, the high frequency response is limited primarily by the RF power loss in the device electrode. In the simple S21(E/E) electric experiment, we measured the increasing power loss of the fabricated electrode from 10 dB to 27 dB as increasing the frequency up to 40 GHz. Such a frequency dependent power loss may be attributed to the conductor loss of the electrode due to the imperfection of the electrode design and fabrication. Therefore, furtheroptimized technique of the traveling-wave-electrodes may yield a larger modulation index and enhanced bandwidth properties.

 

Fig. 6 (a) Change of the modulation index for different voltages for 10 GHz, 20 GHz, and 40 GHz operations. (b) Change of the modulation index at different frequencies for 1 V_pp and 2 V_pp

Download Full Size | PDF

4. Summary

We have demonstrated a hybrid silicon and EO polymer modulator for low-driving voltage and large bandwidth applications. The 50 nm-thick and 5 mm-wide silicon was used to form the MZI modulator. After poling the EO polymer, the in-device EO activity was estimated to be r₃₃ = 165 pm/V from the measured V_π of 0.9 V. A modulation response up to 40 GHz was observed, with the measured 3 dB bandwidth of 23 GHz. The fabrication simplicity of the waveguide using the ultra-thin silicon offers further increase in the bandwidth property and low-power consumption of the modulator. Besides the low-driving voltage and bandwidth properties of the modulator, the waveguide consisted of the high temperature stable EO polymer. The T_g of the EO polymer was 172°C, which can be enhanced over 190°C. Such a thermal stability overcomes the conventional EO materials based on guest-host molecules. The waveguide modulator indicates the temporal stability for reliability testing such as Telecordia requirements expressed as 85°C/2000 hours.