1. Introduction

Active silicon photonic devices play critical roles in future silicon photonic integrated circuits including light emitting devices, optical modulator and detector [1–4]. Generally speaking, in order to integrate active functions into a passive device system, it is necessary to dope the silicon waveguide with either donor or acceptor ions. In this process, the introduced free carriers induce unavoidable absorption loss under optical communication wavelength following free carrier’s effect [5,6] which degrades the performance of photonic devices. In traditional p-i-n silicon modulator with lager rib waveguide dimensions, this problem is not that distinct especially after enlarging the distance between doping regions and optical waveguide. However, when the modulator’s dimension getting smaller (450 × 250 nm²), diffusion of free carriers into non-intentional doped region has to be considered. How to reduce the absorption loss coming from the diffused carriers has being one of the most urgent problems. For example, the recently reported silicon optical modulator with highest switching speed is realized with a reversed bias PN junction lied near the center of the waveguide [7]. With a useful extinction ratio (~6.1 dB), the switching speed reached 10 Gb/s with 2mm-long phase shifter and the corresponding loss caused by two arms is 1 dB/mm [8]. Another 1mm-long modulator can operate at 12 GHz modulation speed with a 6 dB extinction ratio under −8 V bias with 1.9 dB/mm phase shifter absorption loss [9]. The high additional loss mainly comes from the free carriers’ absorption in the doped regions [9]. A p-i-p-i-n diode is presented to achieve low loss optical modulator with the same cutoff frequency with PN diode [10,11]. However, both of the intrinsic regions will be suffered by the lateral diffused electrons and holes from neighbor p-doped and n-doped regions. For boron ions implant with 80 keV ion energy, the lateral diffused length reaches 75 nm [12] which made the neighbor intrinsic region actually a p-type semiconductor.

On the market of optical communication, the insertion loss of 10 Gbps lithium niobate modulator is nearly 4~5 dB. Right now, the total loss of silicon modulator with 4mm-long phase shifter is about 10 dB which included 2.5 dB/facet fiber-waveguide coupling loss, 3 dB phase shifter absorption loss and 2 dB waveguide loss. For the coupling loss, new methods have been proposed which can reduce the value to 1.5 dB/facet [13]. If we can further reduce the absorption loss of the silicon modulator to less than 3 dB and the other excess loss coming from the passive waveguide to less than 1 dB, total loss of the silicon modulator can be reduced to less than 6 dB. Thereafter, we can get the performance of silicon modulator comparable with that of lithium niobate. In order to reduce the absorption loss of the phase shifter, lots of methods have been reported. However, reduction of the absorption loss always comes along with the degradation of the other two characteristics including modulation efficiency or switching speed. In this paper, we present a novel silicon modulator in which the absorption loss is reduced while keeping the high performance of modulation efficiency and switching speed.

2. Structure

Main performance parameters of optical modulator include switching speed, modulation efficiency and insertion loss in which absorption loss takes a large part. As the high additional loss of general PN junction modulator mainly comes from the free carriers’ absorption in the doped regions. Our approach is to optimize this part of absorption loss without degrading the switching speed and modulation efficiency. We realize this through compensated doping the active region with donor (acceptor) ions. We optimized the doping profile of PN junction cross section considering two factors: (a). Reduce the absorption loss of the waveguide; (b). Keep the high doping level at the bottom of the waveguide in order to avoid degrading other characteristics of the modulator such as the switching speed. The general and optimized PN junction cross section of optical modulator is shown in Fig. 1(a) and Fig. 1(b). In the optimized structure, the width of P-type and N-type doping region is W_p and W_n. As the PN junction was set at the central of the waveguide, we choose W_p = W_n as an example in the experiment. Actually better performance can be achieved through optimize the offset between W_p and W_n which is not the focus in this paper.

 

Fig. 1 The structures and doping level on the cross sections of optical modulator with (a) a general reversed biased PN junction (b) The present structure in which we choose W_p = W_n.

Download Full Size | PDF

3. Experimental results and discussion

The SOI wafer has a 220 nm top silicon layer and 2 µm buried oxide layer. Channel waveguide with (600 × 220 nm²) cross sections and corresponding ring resonators were fabricated with standard CMOS processes as shown in Fig. 2 . Uniform nano-tips [8] were integrated on the input and output ends of the waveguides as the fiber-waveguide couplers. We do the compensated ion implant on the surface of SOI. Then, original p-type and n-type ion implant were done after the channel and rib waveguide etching. Here we choose to do the compensated doing first before the waveguide etching. It can also be done together with the active region doping after waveguide etching, however, the doping conditions have to be optimized. The waveguides were covered with oxide layer after doping process and via holes are opened for electrode contacting. After dicing, the modulators were tested under a 6 axis optical fiber to waveguide alignment system. All of the modulators show good electrical and optical performance. The measured free spectrum range (FSR) of the asymmetric Mach Zehnder Interferometer is 6.015 nm corresponding to a group index N_g = 4.0. We measured the phase shifter efficiency under different applied voltage and corresponding absorption loss through cut back method. The results are analyzed below.

 

Fig. 2 The fabrication processes of the compensated doping optical modulator.

Download Full Size | PDF

3.1 Phase shift efficiency and absorption Loss

The measured phase shift of the compensated doping silicon waveguide under different W_p is shown in Fig. 3 . In the reverse-biased PN junction, phase shift Δφ = 2πΔn_effL/λ coming from the overlap between optical mode distribution and the depletion region on the cross section of the phase shifter. If we define N_a the free carriers’ concentration and V_D the flat band voltage, the relationship between the PN junction depletion widths W_d and applied voltage follows W_d = [2ε₀ε_r (V-V_D)/ qN_a] ^1/2, which explain the nonlinearity of the measured curves in Fig. 3. Hereafter, the phase shifter efficiency V_πL_π(V·cm) also experience increase according to the applied voltage. For example on the no compensated case, phase shifter efficiency changes from V_πL_π = 2.18 (V·cm) under V_dc = −2 V to V_πL_π = 3.63 (V·cm) under V_dc = −10 V which means given a fixed ac peak-to-peak voltage V_pp, smaller bias voltage V_dc results in a higher phase shift efficiency. On the other hand, it is clear in Fig. 3 that for a fixed applied voltage, the phase shift and phase shift efficiency are also reduced according to W_p. The reason is consider the effective doping concentration N_a, when the phase shifter is compensated doped, smaller W_p means a lower effective doping concentration N_a in the depletion region which results in the lower level of the available depleted free carriers’ concentration. As a result, given a fixed applied voltage, the phase shift and phase shift efficiency are lower for smaller W_p. Same conclusion has been drawn in Ref. [9].

 

Fig. 3 The phase shift of the 4mm-long phase shifter with different Wp.

Download Full Size | PDF

We study the phase shift efficiency and absorption loss of the compensated doping silicon waveguide under different W_p as shown in Fig. 4 . Silicon waveguide loss depends on both the doping density profile and interface/surface roughness. The part from surface roughness is about 1.2 dB/cm which is independent with W_p. This part of loss can be reduced through optimizing the ICP etching process and thermal smoothing method. The absorption loss of phase shifter as shown in Fig. 4(a) is simulated through the overlap between free carriers’ concentration and optical mode distribution [14]. The measured result is shown in Fig. 4(b). Here we choose the phase shifter efficiency of different W_p when V_dc = −9 V. Measurement results of cut-back method shows that as the free carriers’ concentration is reduced after compensated doping of the active region, absorption loss of the phase shift is reduced from 0.75 dB/mm under no compensation case to 0.45 dB/mm when W_p = 30 nm accordingly. However, the phase shift efficiency also experiences a decrease along with W_p. In order to estimate the performance of the phase shifter, we present figure-of-merit (FOM) F = Loss·Efficiency (dB·V). Lower F value means higher performance of the phase shifter. The F value of the phase shifter according to different W_p is shown in Fig. 5 . After compensated doping the active region, the Loss·Efficiency value is greatly reduced which demonstrate the higher performance of the new structure.

 

Fig. 4 Simulated (a) and Experimental (b) absorption loss and phase shifter efficiency of phase shift with different W_p. The phase shifter efficiency of different W_p were chosen at V_dc = −9 V.

Download Full Size | PDF

 

Fig. 5 The FOM of silicon modulator under different doping conditions with different W_p.

Download Full Size | PDF

6.2 Switching speed

We measured the rise and fall time of the silicon optical modulator under 10 Gbit/s electrical signals with V_pp = 5 V and V_dc = −3 V coming from pattern generator. The optical response of the compensated doping optical modulator with W_p = 90 nm is shown in Fig. 6 . The yellow and blue curves denote the output optical response and input electrical signals. The rise and full time of the optical modulator defined as the time for phase shift changing from 10% to 90% or 90% to 10% of its maximum amplitude reaches 35.7 ps and 29.3 ps respectively. If the 3 dB bandwidth is defined as BW_3dB = 0.35/ t_MAX [10], where t_MAX is the longer one of the fall time and the rise time, the corresponding 3 dB bandwidth still localized near 10 GHz after compensated doping.

 

Fig. 6 The optical response of the phase shifter with W_p = 90 nm under 10 Gbit/s pattern generator. The 5V_pp RF signal is added with a dc bias tee V_dc = −3 V.

Download Full Size | PDF

The 3 dB bandwidth response was measured using Agilent N4373C lightwave component analyzer. The high speed electrical signal after amplified was applied on the modulator through a DC bias tee and a 40 G probe. The output optical signal of the modulator was collected into the lightwave component analyzer directly. The devices are driven through traveling wave electrodes. A 50 ohm termination load is added on the other end of the travelling wave electrodes. The RF power is 15 dBm with scan frequency from 100 MHz to 40 GHz. The RF signal is loaded on the modulator with V_bias = −3.5 V. We normalized the frequency response of the set up (including connectors and electrical wires) using a commercially used lithium niobate optical modulator before the measurement of the DUT. The measured frequency response of the 40 G lithium niobate modulator fits well with the data of the product datasheet. After optical and electrical calibration, the optical response of the MZI modulator with 4mm-long phase shift is shown in Fig. 7 . A 10 GHz 3 dB-bandwidth was achieved based on our devices which demonstrated the high speed performance of our devices. After optimizing the doping profile, the bandwidth of the 4 mm-phase shifter with doping compensation reaches 17 GHz. The corresponding Loss·Efficiency value still localized at 25.4 dB·V as shown in Fig. 5. For 17 GHz bandwidth, the doping density of P and N type region is higher and the corresponding compensated doping density is also higher. The extinction ratio reaches ~8 dB on the 10 G eye diagram shown in Fig. 7. An electrical driver is utilized to amplify the electrical signal added on the modulator in which, the performance of the electrical driver is still under optimization. From the measured results, we demonstrated that the Loss·Efficiency FOM of the modulator can be reduced through doping compensation method while keeping the high switching speed performance. Through this method, higher switching speed of the modulator can also be achieved while keeping the low value of Loss·Efficiency FOM.

 

Fig. 7 Frequency response of MZI optical modulator with 4mm long phase shifter and W_p = 90 nm (a) 3 dB bandwidth and (b) 10 Gbit/s eye-diagram of the modulator with compensation and optimized doping level.

Download Full Size | PDF

4. Conclusion

Utilizing compensated doping method, we reduced the absorption loss of the phase shifter in silicon optical modulator from 0.75 dB/mm to 0.45 dB/mm while keeping the high performance of phase shifter efficiency and switching speed. The phase shifter efficiency reaches V_πL_π = 2.18 (V·cm) under V_dc = −2 V and V_πL_π = 3.63 (V·cm) under V_dc = −10 V. The figure of merit F = Loss·Efficiency (dB·V) is reduced from 25.8 dB·V to 19.4 dB·V while keeping the 3 dB bandwidth about 10GHz which demonstrated the high performance of the presented devices. The measured 3dB bandwidth can be further improved to 17 GHz though optimizing the doping profile on the cross section of the phase shifter while keeping the Loss·Efficiency FOM still localized at a low level.