Lorentz reciprocity is a direct consequence of Maxwell equations governing the propagation of light in passive linear media with symmetric permittivity and permeability tensors. Here, we demonstrate the first active optical isolator and circulator implemented in a linear and reciprocal material platform using commercial Mach-Zehnder modulators. In a proof-of-principle experiment based on single-mode polarization-maintaining fibers, we achieve more than 12.5 dB isolation over an unprecedented 8.7 THz bandwidth at telecommunication wavelengths, with only 9.1 dB total insertion loss. Our architecture provides a practical answer to the challenge of non-reciprocal light routing in photonic integrated circuits.
© 2014 Optical Society of America
As was recently emphasized [1,2] the realization of a true optical isolator requires breaking Lorentz reciprocity. The advent of photonic integrated circuits (PIC) has made urgent the need for non-reciprocal systems that do not rely on magneto-optic materials [3–9], which are difficult to integrate in large-scale photonic platforms [10–16]. The achievement of optical isolation without magneto-optical materials has been a long-standing challenge that has regained interest with the development of PIC in silicon and III/V material platforms. Magneto-optical materials break Lorentz reciprocity because of their asymmetric magnetic permeability and are ubiquitously employed in current bulk and fibered optical systems to achieve the key functionalities of optical isolation and circulation. Isolators are in particular necessary to protect laser sources and amplifiers from back-reflections. Circulators are devices where light injected from a first input port is transmitted to the output port, while light incoming into the output port is redirected to a third port (called the “circulated” port below). They are used to perform tasks such spectral filtering in combination with fiber Bragg gratings, and more generally they allow for versatile light routing . Unfortunately, it has proved difficult to integrate magneto-optic materials in a process compatible with the scalable fabrication of monolithic PICs [10–16], leaving open the problem of how to implement isolators and circulators in these emerging platforms. This situation has triggered intense research efforts toward alternatives to magneto-optic materials, one of them being the use of optical nonlinearities [7–9] that can be enhanced in nanoscale cavities such as ring resonators. However, this approach is intrinsically narrow-band (due to the resonance condition) and the performance is power-dependent, limiting the widespread applicability of such devices. An overview of some recent results obtained with the above-discussed techniques is presented in the Appendix, Table 2 (the reader is referred to  for a more comprehensive review).
Probably the most promising route toward a practical non-reciprocal system entirely compatible with existing PIC technologies is to use a time-dependent modulation of the refractive index to break time-reversal symmetry, which can be interpreted as a photonic Aharonov-Bohm effect [19–21]. Following the pioneering work of Noé and associates , a few recent experimental results [4–6] have shown promise along this line of research, but still suffer from either low isolation, prohibitively high insertion loss, or narrow-band operation, and can be complex to implement.
2. Isolator and circulator system design
In this paper, we introduce a new architecture for a time-modulated optoelectronic system that performs the key functions of optical isolator and circulator with a minimum level of complexity. Although, at present, the system is realized with discrete optoelectronic devices, it is a first proof of principle of a design that is compatible with existing PIC technologies such as integrated silicon photonics. In previous approaches based on traveling-wave modulators [3,5] reciprocity could only be broken if the light travel time through the modulators was larger than the period of the modulation signal, requiring both long modulators and high (>1 GHz) driving frequencies. We overcome these requirements by using a two-stage architecture in which two Mach-Zehnder modulators are separated by a passive optical delay line [6,22]. This strongly relaxes the demand on modulator speed and geometry, as the modulation frequency is inversely proportional to the light propagation time inside the delay line instead of inside the modulator itself. Compared to the scheme used by Doerr et al. , we achieve efficient and broadband isolation over the telecommunication band with only two modulators, without the need for a multiplicity of active systems in parallel. The main characteristics and performances of our system are compared to previous results on modulation-based isolators in Table 1. Last, but not least, we are the first to demonstrate a device that is also an optical circulator – a vital component in large-scale photonic circuits – without relying on magneto-optical materials [23,24].
By making use of a non-sinusoidal drive signal, a time-modulated optical system can be constructed that violates reciprocity in what we conjecture is the simplest possible architecture for broadband isolation. With a square-wave signal, perfect isolation can be achieved in theory. Moreover, by using passive delay lines, the modulation speed can be lowered to 20 MHz instead of several GHz. Figure 1(a) shows a block diagram of the system. Due to the very low speeds involved, we note that the modulators are expected to function essentially as lumped elements.
As shown in Fig. 1(a), the signals driving the two modulators are shifted by a time lag matching the propagation time in the optical delay line, which is equal to one quarter of the modulation period. As a result, when each modulator is driven with an amplitude Vπ (the voltage required to apply a π optical phase shift, Fig. 1(b)) the forward-propagating mode experiences an accumulated optical phase shift of 0 or 2π radians (Fig. 1(c)), while the reverse propagating mode sees a net optical phase shift of π radians (Fig. 1(d)). The offset bias of the system has been chosen so that the forward propagating mode is coupled from p1 to p4, while the reverse mode couples from p4 to p2. Thus, non-reciprocal coupling and both isolator and circulator functionalities are achieved. A detailed analytical derivation of the port-to-port transmission of the system is presented in Section 2 of the Appendix. We note that if an ideal square-wave is provided, perfect isolation is expected. In practice, the condition for isolation is not satisfied at the rising and falling edges of the square-wave if these are not perfectly sharp. As long as the modulation frequency f = 1/T is small compared to the bandwidth of the modulators, the duration over which the circulator function is impaired will be relatively small as shown in the Appendix. This condition is well satisfied in our system, with a measured rise and fall time of 4 ns (see Fig. 8), corresponding to a bandwidth of 250 MHz. It is therefore advantageous to use a long delay line for optimal isolation, while a trade-off with increasing insertion loss must be considered.
3. Experiment results and discussion
Our isolator/circulator system was realized in polarization-maintaining (PM) single-mode optical fibers by means of two EOSpace X-cut Lithium Niobate 2x2 modulators (M1 and M2) configured for push-pull operation. For simplicity and without loss of generality, we replaced push-pull configurations in the schematics by the equivalent system as shown in Fig. 1(a). Detailed analysis is presented in the Appendix, Section 2. The Vπ was approximately 4.6 V. In addition to the modulation signal, a d.c. offset on each modulator was applied to properly bias the phase imbalance of the interferometers. Two tunable optical delay lines were used to balance the path between M1 and M2 with an accuracy of better than 100 ps, a key requirement for broadband performance. An arbitrary function generator (AFG, Tektronix AFG3252C) provided the modulators’ RF drive voltage with 4.5 V peak-to-peak amplitude. Given the 2.5 m long optical path of the delay line, the corresponding time-shift of the two driving signals was T/4 = 12.4 ns, corresponding to an optimal modulation frequency f = 20.2 MHz.
All components in the system were PM and the laser was linearly polarized. Thus, only a single optical mode existed, common to all ports, which is an essential requirement to prove unambiguously the non-reciprocal nature of our system [1,2]. The delay lines were adjusted to give as close to a balanced path as possible, which was correlated with the optical bandwidth seen in a spectral sweep of the circulator transmission from p1 to p4. In the course of our experiments as shown in Fig. 2, the d.c. offset on the modulators usually had to be adjusted slightly between measurements every 10 minutes on average. This was due to the fluctuating phase imbalance between the optical paths in the delay lines and could easily be mitigated by an active feedback stabilization loop, which would however add unnecessary complexity to our proof-of-principle experiment. Moreover, it is expected that an integrated version of our architecture would be intrinsically much more stable. The delay lines and modulation frequency were not adjusted within each set of steady state and time-domain measurements shown below, but adjustment was performed between them. In particular, the frequency was adjusted by around 2%. This is likely due to a combination of fluctuations in effective path lengths and possible instabilities in the arbitrary function generator’s internal clock.
To balance the two optical delay lines, we first maximized the Free Spectral Range (FSR) of the system’s transmission without modulation. Accurate balance is essential for optimal optical bandwidth. In forward configuration, when 0 dBm of continuous wave (CW) laser beam was sent to port p1, port p4 output was measured. In reverse configuration, keeping all system’s parameters unchanged, we sent the same laser beam into port p4 and measured port p1’s output power. The laser was swept from 1470 to 1570 nm. Overall power conservation in forward versus reverse propagation was verified by measuring the ports p2 and p3 outputs.
The main results of isolation are presented in Fig. 3. The maximum isolator/circulator system excess loss is 9.2 dB with an isolation ratio of 12.5 dB or more across the wavelength range 1500 – 1568 nm, which corresponds to a record optical bandwidth of 8.7 THz. The total loss during active operation was only 1 dB more than when the modulators were biased at full transmission and the modulation signal was powered off.
The real-time output powers of the system in forward and reverse configurations were measured as shown in Fig. 4. Time domain characterization of the system transmission was performed with a 1.4 GHz bandwidth avalanche photodetector (APD) and an optical power meter. We could obtain the average power sent to the APD while recording the real-time output signal with a fast oscilloscope (Agilent DSO7014A). A CW laser beam (wavelength = 1555.51 nm, power = −8 dBm) was sent successively through the path p1 to p4 (forward path), through p4 to p1 (isolated path), and p4 to p2 (circulated path) to observe any possible fluctuations in the time domain transmission. Experimental results showed that typical amplitudes fluctuations were lower than 1.0% for the forward path, while fluctuations of 23.4% were seen for the isolated path and 1.6% for the circulated path as shown in Fig. 6(a). The fluctuations on the isolated path were low in absolute terms, given the high extinction on this port (Inset in Fig. 6(a)). The isolation performance was weakened by the non-ideal square-wave with a rise and fall time of 4 ns as shown in the Appendix, Section 2. But the averaged isolation ratio is still better than 10 dB.
In view of the important applications of PICs in high-speed digital telecommunications, we characterized the impact of our system on the data transmission fidelity. Due to the system’s working principle, the optical phase is flipped twice per period, which could be problematic if a phase-modulated signal format was sent to the system. Therefore we limit our study here to On-off keying modulation. The testing setup is shown in Fig. 5. We sent an optical signal encoding a 25 Gb/s non-return-to-zero (NRZ) pseudo-random bit sequence (PRBS was generated by Tektronix Arbitrary Waveform Generator AWG70001A) through the circulator from port p1 to p4, and measured the output on a sampling scope (Tektronix DSA8200 Digital Serial Analyzer, 10 GHz bandwidth). Compared to the signal without the circulator (Fig. 6(b)), some additional dispersion in the level of the “1” bit was observed, but the so-called “eye” of the diagram remained well open (Fig. 6(c)), showing that faithful digital data transmission is not impaired by the isolator/circulator system.
One key question is whether the circulator system presented here could indeed be implemented in integrated silicon photonics. Gathering some of the most recent results from the literature [25–32], we predict that a 1.5 mm long modulator could provide the needed π phase shift (if both arms are driven) with only 3.5 dB of loss , while the needed delay line could be implemented with 1.2 µm wide silicon ridge waveguides . In such waveguides, group indices of around 4.3 and losses of 0.3 dB/cm are obtained. Therefore, a 4 cm long on-chip delay line, which entails only 1.2 dB of loss, could enable a modulation frequency around 400 MHz. At such a low speed, a high quality square-wave can be generated and imparted onto the optical phase, considering the reported silicon modulator speeds of 30 GHz [30–32]. The total insertion loss of this system would be 8.2 dB, while its extinction ratio is expected to exceed the values reported here thanks to the intrinsic stability of integrated Mach-Zehnder interferometers compared to fiber optics. Due to group velocity dispersion, the optical delay between M1 and M2 is actually wavelength dependent. Considering a 500 x 220 nm2 silicon wire waveguide surrounded by SiO2 as the delay line, the group dispersion is about 1.2 x 10−4 /nm around 1550nm. We estimate a delay variation of less than 1ps over more than 100 nm bandwidth around 1550 nm . This is much smaller than the modulation period (2500 ps here) and has therefore negligible impact on performance.
Based on the concept introduced in Ref , we have presented a simple architecture to break Lorentz reciprocity using neither magneto-optical effects nor non-linear effects but relying on commercial optical modulators. We have built a fiber-based optical circulator that achieves a level of performance approaching the needs of practical applications. Our approach can be adapted to silicon photonic circuits and other integrated optics platform with equal or better performance assuming existing state of the art technologies. Our design is a practical approach to optical isolation in future integrated photonics systems and can become an enabling component for further developments in this field.
A.1. Overview of on-chip isolators
To the best of our knowledge, reported on-chip isolators with state-of-the-art performance are highlighted in Table 2, together with their working principles.
A.2. Theoretical analyzing of the isolator/circulator system
In this section, we derive analytical expressions for the port-to-port, time-dependent transmission coefficients of our system, using the Transfer-matrix method. The complex electric field amplitudes at the input and the output of a four-port network as shown in Fig. 1(a) can be written in vector notation by
For simplicity, we first neglect the loss of each component in Fig. 1(a). Later, a model will be established considering the device loss, the non-idealities in the modulators and the unbalance of optical paths between M1 and M2. The scattering matrix of an ideal 50/50 directional coupler (DC) writes
The scattering matrices of phase modulators shown in Fig. 1(a) write
The optical phase shifts (,) are created by modulators M1 and M2. They are time dependent as shown in Fig. 1(b). The scattering matrices of the two optical delay lines between M1 and M2 are same and write
Therefore the total scattering matrix of the system in forward propagation is
If light is sent into port p1, the input light field’s complex amplitude is
The output light at port p3 and p4 are
Equation (7) shows that all the light sent from port p1 is directed to port p4 only with an imprinted phase modulation.
The system’s scattering matrix in reverse operation is
The complex amplitude of input light to M2’s p4 is
Referring to Fig. 1(d), the output light from M1’s ports should be
Therefore, in reverse direction, light sent from port p4 will only come out at port p2 and Lorentz reciprocity is broken. As we take into consideration of the real waveform of the drive voltage, the optical losses, the optical paths’ unbalance and d.c. bias drift from the 3 dB point in the Mach-Zehnder modulators, the model can be improved as below. The directional coupler’s scattering matrix is modified to
r is the power coupling ratio and k is the insertion loss (unit: dB). The scattering matrices of the phase modulators can be modified to
Let M1 insertion loss = 3.4 dB and M2 insertion loss = 3.2 dB. Let ki (i=1,2) to be the insertion loss (unit: dB) of Mi. The optical delay lines’ transfer matrices are modified to
The drive voltage is simulated by a combination of square-wave, raised-cosine wave and decayed sine wave as shown in Fig. 7 below.
The decayed sine pulse that overlaps on the square-wave can be expressed as
A1 = 0.4, A2 = 16, T is the period of drive voltage, f1=5*fRF_drive. These parameters are obtained from the experimental results as shown in Fig. 8 below.
A raised-cosine function is adopted to connect the high level and low level of the square-wave in order to simulate the real signal. Based on Fig. 5, the rise and fall time is about 4 ns, so the raised cosine factor (β) is 0.25 in the raised-cosine function as shown below.
fsa=2*fRF_drive, Vpp is the RF drive signal’s peak-to-peak voltage. The time domain simulation of the port-to-port transmission is shown in Fig. 9. Simulation parameters are summarized in the Table 3 and the simulated time averaged extinction ratio is about 23.7 dB with insertion loss of 9.5 dB. Because the losses of the components (e.g. DCs and delay lines) in the system’s top and bottom arms are different, as listed in Table 3, the successive peaks and dips of output signal are not identical, as shown in Fig. 9. This simulation is verified by our experimental results as shown in Fig. 6(a). Our simulation also proves that when the first modulator’s driving voltage is not exactly T/4 delayed compared to the second one the ER remains higher than 18dB for a time delay deviation from T/4 less than T/10.
C. G. would like to acknowledge the financial support of the Swiss National Science Foundation through an Ambizione fellowship. All authors gratefully acknowledge support from Portage Bay Photonics. The authors would like to thank Gernot Pomrenke, of AFOSR, for his support of the OpSIS effort, though both a PECASE award (FA9550-13-1-0027) and ongoing funding for OpSIS (FA9550-10-1-0439), as well as Brett Pokines and the AFOSR SOARD office, for their support of the development of high-speed modulators under grant FA9550-13-1-0176. The authors are grateful for support from an MOE ACRF Tier-1 NUS startup grant and NRF NRF2012NRF-NRFF001-143. The authors would like to gratefully acknowledge AT&T for the loan of critical equipment.
References and links
1. S. Fan, R. Baets, A. Petrov, Z. Yu, J. D. Joannopoulos, W. Freude, A. Melloni, M. Popović, M. Vanwolleghem, D. Jalas, M. Eich, M. Krause, H. Renner, E. Brinkmeyer, and C. R. Doerr, “Comment on “Nonreciprocal Light Propagation in a Silicon Photonic Circuit”,” Science 335(6064), 38 (2012). [CrossRef] [PubMed]
2. D. Jalas, A. Petrov, M. Eich, W. Freude, S. Fan, Z. Yu, R. Baets, M. Popovic, A. Melloni, J. D. Joannopoulos, M. Vanwolleghem, C. R. Doerr, and H. Renner, “What is - and what is not - an optical isolator,” Nat. Photonics 7(8), 579–582 (2013). [CrossRef]
3. 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(2), 417–421 (2005). [CrossRef]
7. M. Krause, J. Muller, and E. Brinkmeyer, “Measurement of nonreciprocal stimulated Raman scattering in silicon photonic wires,” in Proceedings of 2012 IEEE 9th International Conference on.Group IV Photonics (GFP) (San Diego, 2012), pp. 6–8. [CrossRef]
10. Y. Shoji, T. Mizumoto, H. Yokoi, I. W. Hsieh, and R. M. Osgood, “Magneto-optical isolator with silicon waveguides fabricated by direct bonding,” Appl. Phys. Lett. 92(7), 071117 (2008). [CrossRef]
11. T. R. Zaman, X. Guo, and R. J. Ram, “Semiconductor Waveguide Isolators,” J. Lightwave Technol. 26(2), 291–301 (2008). [CrossRef]
12. L. Bi, J. Hu, P. Jiang, D. H. Kim, G. F. Dionne, L. C. Kimerling, and C. A. Ross, “On-chip optical isolation in monolithically integrated non-reciprocal optical resonators,” Nat. Photonics 5(12), 758–762 (2011). [CrossRef]
13. M. C. Tien, T. Mizumoto, P. Pintus, H. Kromer, and J. E. Bowers, “Silicon ring isolators with bonded nonreciprocal magneto-optic garnets,” Opt. Express 19(12), 11740–11745 (2011). [CrossRef] [PubMed]
14. Y. Shoji and T. Mizumoto, “Magneto-optical non-reciprocal devices in silicon photonics,” Sci. Technol. Adv. Mater. 15(1), 014602 (2014). [CrossRef]
15. L. Bi, J. Hu, P. Jiang, H. S. Kim, D. H. Kim, M. C. Onbasli, G. F. Dionne, and C. A. Ross, “Magneto-Optical Thin Films for On-Chip Monolithic Integration of Non-Reciprocal Photonic Devices,” Materials 6(11), 5094–5117 (2013). [CrossRef]
16. S. Ghosh, S. Keyvavinia, W. Van Roy, T. Mizumoto, G. Roelkens, and R. Baets, “Ce:YIG/Silicon-on-Insulator waveguide optical isolator realized by adhesive bonding,” Opt. Express 20(2), 1839–1848 (2012). [CrossRef] [PubMed]
17. B. Saleh and M. Teich, Fundamentals of Photonics, Wiley Series in Pure and Applied Optics (John Wiley, 1991).
18. D. Dai, J. Bauters, and J. E. Bowers, “Passive technologies for future large-scale photonic integrated circuits on silicon: polarization handling, light non-reciprocity and loss reduction,” Light Sci. Appl. 1(3), e1 (2012). [CrossRef]
19. Z. Yu and S. Fan, “Complete optical isolation created by indirect interband photonic transitions,” Nat. Photonics 3(2), 91–94 (2009). [CrossRef]
20. K. Fang, Z. Yu, and S. Fan, “Realizing effective magnetic field for photons by controlling the phase of dynamic modulation,” Nat. Photonics 6(11), 782–787 (2012). [CrossRef]
22. C. Galland, R. Ding, N. C. Harris, T. Baehr-Jones, and M. Hochberg, “Broadband on-chip optical non-reciprocity using phase modulators,” Opt. Express 21(12), 14500–14511 (2013). [CrossRef] [PubMed]
23. T. Mizumoto, H. Chihara, N. Tokui, and Y. Naito, “Verification of waveguide-type optical circulator operation,” Electron. Lett. 26(3), 199–200 (1990). [CrossRef]
25. Y. Zhang, S. Yang, Y. Yang, M. Gould, N. Ophir, A. E.-J. Lim, G.-Q. Lo, P. Magill, K. Bergman, T. Baehr-Jones, and M. Hochberg, “A high-responsivity photodetector absent metal-germanium direct contact,” Opt. Express 22(9), 11367–11375 (2014). [CrossRef] [PubMed]
26. S. Yang, Y. Zhang, D. W. Grund, G. A. Ejzak, Y. Liu, A. Novack, D. Prather, A. E.-J. Lim, G.-Q. Lo, T. Baehr-Jones, and M. Hochberg, “A single adiabatic microring-based laser in 220 nm silicon-on-insulator,” Opt. Express 22(1), 1172–1180 (2014). [CrossRef] [PubMed]
27. Y. Zhang, S. Yang, A. E.-J. Lim, G.-Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]
28. Y. Liu, S. Dunham, T. Baehr-Jones, A. E. Lim, G. Lo, and M. Hochberg, “Ultra-Responsive Phase Shifters for Depletion Mode Silicon Modulators,” J. Lightwave Technol. 31(23), 3787–3793 (2013). [CrossRef]
29. C. Galland, A. Novack, Y. Liu, R. Ding, M. Gould, T. Baehr-Jones, Q. Li, Y. Yang, Y. Ma, Y. Zhang, K. Padmaraju, K. Bergman, A. E. Lim, G. Lo, and M. Hochberg, “A CMOS-compatible silicon photonic platform for high-speed integrated opto-electronics,” Proc. SPIE 8767, 87670G (2013). [CrossRef]
31. M. Streshinsky, R. Ding, Y. Liu, A. Novack, Y. Yang, Y. Ma, X. Tu, E. K. Chee, A. E. Lim, P. G. Lo, T. Baehr-Jones, and M. Hochberg, “Low power 50 Gb/s silicon traveling wave Mach-Zehnder modulator near 1300 nm,” Opt. Express 21(25), 30350–30357 (2013). [CrossRef] [PubMed]
32. D. J. Thomson, F. Y. Gardes, J. Fedeli, S. Zlatanovic, Y. Hu, P. Kuo, E. Myslivets, N. Alic, S. Radic, G. Z. Mashanovich, and G. T. Reed, “50-Gb/s Silicon Optical Modulator,” IEEE Photon. Technol. Lett. 24(4), 234–236 (2012). [CrossRef]
33. T. Shintaku, “Integrated Optical Isolator Based on Efficient Nonreciprocal Radiation Mode Conversion,” Appl. Phys. Lett. 73(14), 1946–1948 (1998). [CrossRef]
34. J. Hwang, M. H. Song, B. Park, S. Nishimura, T. Toyooka, J. W. Wu, Y. Takanishi, K. Ishikawa, and H. Takezoe, “Electro-Tunable Optical Diode Based on Photonic Bandgap Liquid-Crystal Heterojunctions,” Nat. Mater. 4(5), 383–387 (2005). [CrossRef] [PubMed]
35. H. Shimizu and Y. Nakano, “Fabrication and Characterization of an InGaAsp/InP Active Waveguide Optical Isolator with 14.7 dB/mm TE Mode Nonreciprocal Attenuation,” J. Lightwave Technol. 24(1), 38–43 (2006). [CrossRef]
36. W. V. Parys, B. Moeyersoon, D. V. Thourhout, R. Baets, M. Vanwolleghem, B. Dagens, J. Decobert, O. L. Gouezigou, D. Make, R. Vanheertum, and L. Lagae, “Transverse Magnetic Mode Nonreciprocal Propagation in an Amplifying AlGaInAs/InP Optical Waveguide Isolator,” Appl. Phys. Lett. 88(7), 071115 (2006). [CrossRef]
37. Y. Shoji, T. Mizumoto, H. Yokoi, I. Hsieh, and R. M. Osgood Jr., “Magneto-optical isolator with silicon waveguides fabricated by direct bonding,” Appl. Phys. Lett. 92(7), 071117 (2008). [CrossRef]