We discuss an optical isolator design based on tandem phase modulators in a long interferometer. It provides low-loss, broadband isolation in a photonic integrated circuit without requiring special materials or fabrication steps. It was demonstrated in silicon photonics.
© 2014 Optical Society of America
Integrated optical isolators are needed to isolate lasers and optical amplifiers from back reflections, prevent multi-path interference, and make ring lasers oscillate in one direction. Isolation is especially important when a photonic integrated circuit (PIC) contains lasers or optical amplifiers and is connected to a fiber plant. The traditional method to make an isolator is to use magneto-optic materials [1, 2]. However, this generally requires either deposition of poly-crystalline films , which is challenging due to the large number of elements in garnets, or bonding of crystals [4, 5], which is challenging to do on a wafer scale. Many of the demonstrations of integrated isolators that use magneto-optic materials have used ring resonators and thus are narrow band. Also, integrated magneto-optic solutions usually work only for transverse-magnetic polarized light, whereas PICs usually work with transverse-electric polarized light. Also, magneto-optic solutions usually require placing a magnet on the PIC.
A very different approach is to use electro-optic modulation. It can be integrated in a PIC without needing any special materials or processing steps. One electro-optic method is to employ traveling-wave modulators, which give a different modulation depending on the direction of optical propagation. Such designs have either a relatively high intrinsic loss (> 6 dB) and residual frequency shift (unless two stages are employed, the second stage undoing the frequency shift of the first)  or require long modulated sections . For a traveling-wave solution to be effective, the length must be on the order of the wavelength of the modulation frequency. Thus to have a low modulation frequency (< 5 GHz), the modulation length must be long (> 2 cm), which can result in high optical losses in semiconductor modulators due to doping and metal absorption. Another electro-optic method is to use a tandem arrangement of two phase modulators with a long passive waveguide in between the modulators . This requires only two short modulators and has no intrinsic insertion loss. Ref.  demonstrated it in InP, providing 11 dB of narrowband isolation. In , the design was expanded to be a tandem arrangement of two push-pull phase modulators in a Mach-Zehnder interferometer. It was demonstrated in Si, providing 3 dB of broadband isolation.
2. Isolator design
The tandem phase modulator scheme is shown in Figs. 1 and 2. The most basic configuration is shown in Fig. 1(a). One modulator is driven by a sine wave at frequency f and the other by a cosine wave also at frequency f. The two phase modulators are separated by a waveguide propagation distance of vg/(4f), where vg is the optical group velocity in the waveguides. When the signal passes from left to right, the transmission is
Equation (1) is equal to 1, and Eq. (2) is equal to ej2Asin(2πft). Thus there is no effect on the forward signal, and when J0(2A) = 0 (peak-to-peak modulation of 138°), where J is the Bessel function of the first kind, the carrier is fully depleted from the backward signal, and all the backward energy appears at other wavelengths. This can be seen in the simulated performance shown in the top row of Fig. 3.
The drawback to this simple design is that the isolation is narrow band. The backwards propagating light is distributed to other frequencies, and thus if a broadband input is applied to this device, no isolation is observed, as shown in the top row of Fig. 4.
The broadband design is shown in Fig. 2, and cases for N = 2 and N = 4 are shown in Fig. 1(b) and (c), respectively. It comprises placing a narrow-band isolator in each arm of a N-arm interferometer. The N narrow-band isolators are identical except that each is driven with a different overall RF phase. The relative RF phase in arm n is 2π(n − 1)/N. The optical phase is equal in all arms. In the forward direction, there is no effect on the signal from each narrow-band isolator, and thus the final combined signal also experiences no effect and 100% transmission. In the backward direction, however, because each narrow-band isolator is driven with a different overall RF phase, the generated sidebands interfere destructively, and thus there is broadband isolation. The broadband isolation can be seen in the simulations in the 2nd and 3rd rows of Fig. 4. The larger N is, the higher the isolation.
The amplitude of the mth sideband of a CW signal when propagated backward through the isolator is given by Jm(2A). All the sidebands experience destructive interference except for sidebands nN, where n is an integer. For example, for N = 2, every other sideband experiences destructive interference. Thus the theoretical broadband isolation vs the number of interferometer arms, N, is given by Eq. (3)Fig. 5 for A = 1.203, which is what makes J0(2A) = 0 and gives the highest isolation. One can see that a practical isolator, requiring >20 dB isolation, would require 4 or 5 interferometer arms.
To facilitate implementation, one would like f to be as low as possible. This can be done by lengthening the waveguide propagation distance between the two modulators, vg/(4f). Regardless of how low f is, the isolation will remain broadband.
3. Experimental demonstration
A two-arm version was made in silicon photonics. A schematic and photograph of the device are shown in Fig. 6. There are two pairs of push-pull current-injection plasma-effect modulators in a two-arm interferometer. The modulators are 600 μm long. The width of the intrinsic layer is 1.75 μm. Each push-pull pair is connected n-type to n-type in series with a bias control connected to the n-type terminals. The waveguide length between the modulator pairs is 7.8 mm, and thus the ideal modulation frequency is 2.4 GHz. There is a thermooptic phase shifter to control the relative phase between the arms of the long interferometer. The input/output coupling to the PIC is via two 1-D grating couplers.
We aligned a fiber assembly to the PIC to couple light in and out of the isolator. We drove each modulator pair with a 50-Ω ground-signal probe. The bias voltage was −0.5V. The modulator bandwidth was 300 MHz. Because of the low bandwidth, the modulator response at 2.4 GHz was weak. To compensate, we drove the modulator pairs at 2.0 GHz with 10 V peak-to-peak (each individual phase modulator receiving 5 V peak-to-peak). Despite this, the peak-to-peak modulation was significantly less than the desired 138°, limiting the achievable isolation.
We first launched a cw laser through the device. The fiber-to-fiber insertion loss at 1550 nm was 11.1 dB. The estimated loss breakdown is the following: 7 dB from fiber coupling in and out, 2 dB waveguide propagation loss, 1 dB from the 1 × 2 multimode interference couplers, and 1 dB from the two modulators. We drove both modulator pairs with 2.0-GHz sine waves. We adjusted the RF phase between the drives such that in the forward direction there was no effect on the laser signal, and we adjusted the optical phase between the interferometer arms for maximum forward transmission. We swapped the input and output connections (without readjustment of the RF and optical phases) to measure the backward signal. The measured performance for the cw laser input is shown in Fig. 7 using an optical spectrum analyzer. The spectrum analyzer cannot fully resolve the 2-GHz features, but one can see that the narrow-band isolation is ∼5 dB, limited by insufficient modulation amplitude. The signal when the modulation is turned off is ∼3 dB higher than the forward direction. This loss is due to residual amplitude modulation due to free-carrier absorption that accompanies the phase modulation, which also limits the achievable isolation.
The achievable isolation is limited when the attenuation changes with phase because one can no longer achieve zero energy in the backward direction at the center frequency. This is because the backward-direction modulation “constellation” (which instead of being part of a circle is now part of a spiral) center of mass is no longer at the origin. However, this residual amplitude modulation could be compensated by adding an extra interferometer arm carrying a small amount of power to cancel out the center-of-mass shift.
We then launched amplified spontaneous emission from an Er-doped fiber amplifier into the device. The measured spectra are shown in Fig. 8. 3.0 dB of isolation over the C-band was achieved. While this isolation is less than the theoretical value of 4.2 dB for a two-arm design, it does prove the principle and can be improved significantly by using a four- or five- arm design.
We described a broadband modulation-based integrated optical isolator. It can use low modulation speeds, does not require any special materials or fabrication, and has low intrinsic insertion loss. We achieved 3 dB of broadband isolation. The main limitations in the isolation were low modulator bandwidth and the use of only two interferometer arms.
We acknowledge S. Stulz, X.-M. Xu, T. Nielsen, S. Azemati, G. McBrien, B. Mikkelsen, and C. Rasmussen for their support and discussions.
References and links
1. M. Levy, J. R.M. Osgood, H. Hegde, F. J. Cadieu, R. Wolfe, and V. J. Fratello, “Integrated optical isolators with sputter-deposited thin-film magnets,” IEEE Photon. Technol. Lett. 8, 903–905 (1996). [CrossRef]
2. H. Shimizu and Y. Nakano, “Monolithic integration of a waveguide optical isolator with a distributed feedback laser diode in the 1.5-μm wavelength range,” IEEE Photon. Technol. Lett. 19, 1973–1975 (2007). [CrossRef]
3. L. Bi, J. Hu, P. Jiang, D. Kim, G. Dionne, L. Kimerling, and C. Ross, “On-chip optical isolation in monolithically integrated non-reciprocal optical resonators,” Nat. Photonics 5, 1–5 (2011). [CrossRef]
4. Y. Shoji, T. Mizumoto, H. Yokoi, I.-W. Hseih, and R. Osgood, “Magneto-optical isolator with silicon waveguides fabricated by direct bonding,” Appl. Phys. Lett. 92, 071117 (2008). [CrossRef]
6. S. Bhandare, S. K. Ibrahim, D. Sandel, H. Zhang, F. Wust, and R. Noe, “Novel nonmagnetic 30-dB traveling-wave single-sideband optical isolator integrated in III/V material,” IEEE J. Sel. Top. Quantum Electron. 11, 417–421 (2005). [CrossRef]
7. Z. Yu and S. Fan, “Complete optical isolation created by indirect interband photonic transitions,” Nat. Photonics 3, 91–94 (2009). [CrossRef]
9. C. Doerr, L. Chen, and D. Vermeulen, “Tandem-phase-modulator-based optical isolator in silicon,” European Conf. Opt. Comm., postdeadline (2013).