Open Access
Add to BibSonomy
Abstract
We propose a low-power all-optical switch based on the structure of a graphene-buried balanced Mach-Zehnder interferometer (MZI), where the signal light is switched between the two output ports of the MZI by the heat generated from graphene’s absorption of the pump light. We use orthogonal polarizations for the pump and the signal light to maximize pump absorption and minimize graphene-induced signal loss. Our experimental device fabricated with polymer waveguides buried with 5-mm long graphene shows a pump absorption of 10.6 dB (at 980 nm) and a graphene-induced signal loss of 1.1 dB (at 1550 nm) and can switch the signal light with a pump power of 6.0 mW at an extinction ratio of 36 dB. The actual pump power absorbed by graphene for activating switching is estimated to be 2.2 mW. The rise and fall times of the switch are 1.0 and 2.7 ms, respectively. The switching characteristics are weakly sensitive to ambient temperature variations. Our device can be butt-coupled to single-mode fibers and could find applications in fiber-based and on-chip all-optical signal processing.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
All-optical switching is a key function in all-optical signal processing [1,2]. In general, nonlinear media are required for the realization of all-optical control functions and power consumption in fulfilling such functions is always a major consideration. Early all-optical switches based on the Kerr effect in silica fibers require high pump powers (∼10 GW/cm2) (see, for example, [3]). Over the years, many nonlinear materials for the implementation of all-optical switching have been explored, such as III-V semiconductors [4], chalcogenide glass [5], nanoparticles [6], and highly doped glass [7]. It remains a challenge to achieve milliwatt all-optical switching at a high extinction ratio with conventional nonlinear materials.
The availability of graphene as a nonlinear optical material has opened up many new possibilities for the realization of all-optical control devices [8–23]. Different nonlinear effects in graphene can be explored for achieving all-optical switching, such as the Kerr effect [11], the photothermal effect [12–19], the Pauli blocking effect [20–22], and the saturable absorption effect [23]. In particular, the photothermal effect, where heat is generated from graphene’s absorption of light, can be effectively used for the construction of low-power all-optical switches. For example, a switching power of 11 mW at an extinction ratio of 20 dB has been demonstrated with a graphene-wrapped microfiber Mach-Zehnder interferometer (MZI), but the graphene coating induces a signal loss of 5.4 dB [12]. By using an externally pumped fiber MZI coated with 20 layers of graphene, the switching power can be lowered to ∼5 mW at an extinction of 19 dB with negligible graphene-induced signal loss, but the response time is long (several tens of milliseconds and above) [13]. On the other hand, waveguide-based all-optical switches can respond much faster than fiber switches, but require much higher pump powers. For example, a graphene-attached Si3N4 waveguide MZI can achieve a switching time of 0.57 µs with a switching power of 109 mW [16], and a graphene-on-silicon nanobeam cavity can achieve a switching time of 3.24 µs with a lower switching power of 47 mW [18]. In these waveguide switches, the graphene film absorbs both the pump and the signal light, which limits the useful length of graphene and hence further reduction in the switching power. Another fundamental issue with such high-index-contrast waveguide devices is their high fiber coupling losses, which hinders their application in fiber-based systems.
In this paper, we propose an efficient all-optical switch based on the structure of a graphene-buried polymer waveguide balanced MZI. Our all-optical switch offers a number of distinct advantages. First, the polymer waveguide technology based on the spin-coating process allows a graphene film to be placed directly on the waveguide core to greatly facilitate both pump absorption and heating. Second, the low-index-contrast material system allows the use of pump light polarized in parallel with the graphene surface to maximize its absorption and the signal light polarized in perpendicular to the graphene surface to minimize its absorption, so that a sufficiently long graphene length can be used to generate enough heat from a low pump power without inducing a significant loss to the signal power. Third, the large thermo-optic coefficient of polymer material can produce a large heat-induced phase change in the signal light. Fourth, compared with a glass fiber, our waveguide device can respond much faster for its much smaller heat capacity. Last, the polymer waveguides can be butt-coupled to single-mode fibers with low losses, which allows our device to be readily used in fiber-based systems. Our typical experimental device, which employs a 5-mm long buried graphene film, absorbs 10.6 dB of the pump light at 980 nm, while induces only 1.1 dB loss to the signal light at 1550 nm. The signal light can be switched between the two output ports of the MZI at an extinction ratio of 36 dB with an input pump power variation of 6.0 mW. The actual pump power absorbed by graphene for activating switching is only 2.2 mW. The rise and fall times of our all-optical switch are 1.0 and 2.7 ms, respectively, which are as good as those of electrically driven thermo-optic polymer waveguide switches [24–26]. Our all-optical switch could find applications in fiber-based or on-chip all-optical signal processing, where low-power consumption is desired and fast response is not required, such as optical circuit switching.
2. Operation principle and design
Figure 1(a) shows a graphene-buried polymer waveguide, which is a low-index-contrast waveguide. A recent study [27] shows that the graphene film buried in such a waveguide strongly absorbs the transverse-electric (TE) mode whose dominant electric field is parallel with the graphene surface, while exhibiting little absorption to the transverse-magnetic (TM) mode whose dominant electric field is perpendicular to the graphene surface. This property of graphene has been applied to the construction of polarizers [28,29], mode filters [29,30], and electrodes for electro-optic [31,32] and thermo-optic [25,26] devices. In the present study, the pump light is launched into the fundamental TE mode to maximize its absorption by graphene, while the signal light is launched into the fundamental TM mode to minimize its absorption, as shown in Fig. 1(a). The pump light absorbed by graphene is converted into heat, which can change the phase of the signal light through the thermo-optic effect of the polymer material. To realize all-optical switching, we employ the structure of a balanced MZI, as shown schematically in Fig. 1(b), where all the waveguides are polymer waveguides. The MZI is formed with two 3-dB directional couplers (DCs) designed for the signal wavelength. Graphene films of a suitable length are placed on the cores of the two arms of the MZI. Both the pump and the signal light are launched together into the MZI from one end. The signal light, which has a wavelength of 1550 nm, is split equally into the two arms of the MZI by the first DC and combined at the output ports by the second DC, which are labeled as the bar port and the cross port, respectively. The pump light, which has a wavelength of 980 nm and a tight confinement, is not split by the DCs and passes through the MZI along one arm. The operation of our device is based on converting the power of the TE-polarized pump light absorbed by graphene into heat, which induces a change in the refractive index along one arm of the MZI and hence a relative phase shift of the TM-polarized signal light between the two arms of the MZI. With a phase shift of π, the TM-polarized signal light is switched completely from the cross port to the bar port. The choice of 980 nm as the pump wavelength is for measurement convenience. The broadband absorption characteristics of graphene [33] allows the use of any other wavelength as the pump wavelength, provided that the DCs are designed not to significantly split the pump light. Our device offers much flexibility not only in the choice of the pump wavelength but also in the choice of the signal wavelength, as a balanced MZI fabricated by polymer waveguides can operate over a broad range of signal wavelengths [24].
Fig. 1. (a) Schematic diagram of a graphene-buried polymer waveguide showing the propagation of TE-polarized pump light and TM-polarized signal light; (b) schematic diagram of the proposed graphene-buried balanced MZI for all-optical switching; (c) cross-section of the graphene-buried polymer waveguides in the MZI; and (d) intensity profiles of the TE-polarized pump light at 980 nm and the TM-polarized signal light at 1550 nm, where the white arrows indicate the directions of the dominant electric-field components of the modes.
The polymer materials used for the core and the cladding of the waveguides are EpoCore and EpoClad (Micro Resist Technology), respectively [22,24–27,29,30], whose refractive indices measured at 1530 nm are 1.569 and 1.559, respectively. The waveguide core has a size of 4 µm × 4 µm and is covered with an 8-µm wide graphene film along each arm of the MZI, as shown in Fig. 1(c). The waveguide can provide single-mode operation at 1550 nm. The intensity profiles of the pump and the signal light are shown in Fig. 1(d). The pump light has a much tighter confinement than the signal light. The core separation and the coupling length of the 3-dB DCs are S = 5 µm and Lc = 51 µm, respectively. To choose a suitable graphene length, we calculate the graphene-induced optical losses with the mode solver in COMSOL, where the graphene film is modeled as a conductive boundary with the complex surface conductivities 6.0840 × 10−5−1.5959 × 10−6i for 980 nm and 6.0792 × 10−5−8.6160 × 10−6i for 1550 nm [27]. The calculated graphene-induced losses to the TE mode are 48 dB/cm at 980 nm and 63 dB/cm at 1550 nm, while those to the TM mode are 0.44 dB/cm at 980 nm and 0.36 dB/cm at 1550 nm. As the evanescent field along the graphene film at 1550 nm is stronger than that at 980 nm, the TE absorption at 1550 nm is larger than that at 980 nm. The small absorption to the TM mode is caused by the minor electric-field component of the mode along the propagation direction, which is parallel with the graphene surface [27]. As expected, the TE absorption losses are much larger than the TM absorption losses. In our design, we choose a graphene length of Lg = 5 mm, which should provide a pump (TE) absorption of 24 dB and a signal (TM) absorption of 0.18 dB. The very sharp contrast in the absorption losses between the pump and the signal light owes much to the use of a low-index-contrast waveguide structure, whose guided modes are almost linearly polarized. For the TM-polarized signal, the phase shift induced by the complex conductivity of graphene is only ∼6.0 × 10−4 π at 1550 nm, which is negligible. Besides, as graphene is embedded in both arms of the MZI, the graphene-induced phase shifts for the signal light in the two arms are cancelled out and, therefore, have no influence on the operation of the device.
Our all-optical switch can be regarded as a thermo-optic switch activated by the pump light absorbed by graphene. The heat conversion from light absorption by graphene is highly efficient and the large thermal conductivity of graphene (∼5000 W/m·K [34]) allows the heat generated locally to be distributed along the entire graphene area quickly. To estimate the switching power, we treat graphene as a boundary heat source (i.e., a uniform sheet of heat source without thickness) [13] in applying the heat transfer module in COMSOL, where the thermal conductivities of the polymer material and the silicon substrate are set at 0.12 W/m·K, and 130 W/m·K, respectively. As shown by the simulation results in Fig. 2(a), the average temperature of the core increases by about 2 K at a pump power of 3 mW. Using the calculated temperature distribution and assuming a thermo-optic coefficient of –1.0 × 10−4/K for the polymer material, we can estimate the change in the effective index Δneff of the signal light and then its phase shift Δφ from the following expression:
Fig. 2. (a) Calculated temperature distribution in the graphene-buried polymer waveguide at a pump power of 3.0 mW; (b) calculated phase shift of the signal light as a function of the pump power for two graphene lengths.
3. Device fabrication
We fabricated the device with our in-house microfabrication facility based on spin-coating and photolithography [22,24–27,29,30]. The fabrication steps are shown in Fig. 3(a). First, a thick (∼11 µm) EpoClad film was spin-coated on a silicon substrate as the lower cladding. An EpoCore film was next spin-coated on the EpoClad film to a thickness slightly larger than 4 µm and patterned into cores by photolithography and wet-etching. A middle cladding of EpoClad was spin-coated onto the cores and then etched down to the cores by reactive ion etching (RIE) to form a flat surface for graphene transfer. The RIE process also served to trim the cores into a square shape. A monolayer graphene film grown on a copper foil and attached to a polymethylmethacrylate (PMMA) buffer, a commercial product from Hefei Vigon Material Technology, was next wet-transferred to the sample and the PMMA buffer was removed with acetone. Figure 3(b) shows a top view of the sample after graphene transfer, where the boundary of the graphene film and the waveguides of a DC can be seen. The graphene film was then covered with a thin layer of EpoClad and etched into the desired pattern with a mask by photolithography and O2 plasma etching. Figure 3(c) shows the EpoClad/graphene strips on the two arms of the MZI after the etching process. The etched graphene strips were 5-mm long. Finally, a 6.0-µm thick EpoClad film was spin-coated on the sample as the upper cladding. The total length of the fabricated device was about 22 mm. Figure 3(d) shows a microscopic image of an end face of the fabricated device. In addition to the all-optical switch, we fabricated reference straight waveguides and DCs of the same dimensions as those used in the switch on the same substrate to facilitate the characterization of the fabricated device.
Fig. 3. (a) Steps for the fabrication of the device and microscopic images of (b) the sample after graphene transfer, showing two waveguides branching out from one of the DCs and covered by the graphene film, (c) the sample after the EpoClad/graphene layer etched into two narrow strips on the two arms of the MZI, and (d) an end face of the fabricated device.
4. Experimental results
The pump and the signal light used in our measurement were generated by a 980-nm laser (Lumentum, S27-7402-360-AL) and a 1550-nm laser (Apico, CMP-LD-1550-BTF), respectively. In our experiments, both the pump and the signal light were launched together into the fabricated device with a 980/1550 wavelength-division multiplexer (WDM). Separate polarization controllers (PCs) were used for independent control of the input polarization states of the pump and the signal light.
We first characterized the MZI switch by turning on only the pump light or the signal light launched into the MZI. The output near-field images taken with an infrared camera (Hamamatsu, C2714) for different polarization states of the pump and the signal light are shown in Fig. 4. For the TM polarization of the pump light, as shown in Fig. 4(a), most of the pump light exits from the bar port as the fundamental mode and some weak pump light exits from the cross port as the higher-order mode. The higher-order mode could be generated from any non-uniformities along the waveguides (including the waveguide bends). As the higher-order mode has a weak confinement, it can be coupled to the cross port of a DC and hence emerge from the cross port of the MZI. For the TE polarization of the pump light, as shown in Fig. 4(b), only weak pump light exits from the bar port as the fundamental mode and no light exits from the cross port. This is expected, as the TE-polarized pump light should be largely absorbed by the buried graphene strips in both arms. On the other hand, the TM polarization of the signal light mainly exits from the cross port, as shown in Fig. 4(c), while the TE polarization is almost completely absorbed, as shown in Fig. 4(d).
Fig. 4. Output near-field images taken from the bar port and the cross port of the MZI switch for different polarization states of 980-nm and 1550-nm light: (a) the TM-polarized light at 980 nm, (b) the TE-polarized light at 980 nm, (c) the TM-polarized light at 1550 nm, and (d) the TE-polarized light at 1550 nm.
To confirm that the observed polarization-dependent losses shown in Fig. 4 were caused by graphene’s absorption, we repeated the measurements with a reference straight waveguide without graphene and observed little difference in the transmission between the TE and the TM polarization at either 980 nm or 1550 nm. By comparing the output powers of the TE and the TM polarization for the fabricated MZI switch, we deduced that the graphene-induced losses to the TE-polarized 980-nm and 1550-nm light were approximately 10.6 dB and 18.0 dB, respectively. A 10.6-dB pump loss corresponds to 92% of the pump power absorbed by graphene, which represents a high utility of the pump power. We also compared the output powers of the TM-polarized signal light between a straight waveguide with a 5-mm long graphene film and a straight waveguide without graphene, from which we deduced that the graphene-induced signal loss was approximately 1.1 dB. The measured graphene-induced losses agree qualitatively with the theoretical calculation. The discrepancies between the measurement data and the theoretical values could be attributed to the slight unevenness of the graphene film (caused by the slight unevenness of the core surface).
We next measured the switching characteristics of the MZI switch by turning on both the pump and the signal light and varying the pump power, where the pump light was set to be TE-polarized and the signal light TM-polarized. By varying the input pump power, we monitored the output powers from the bar port and the cross port, respectively, by using an optical spectrum analyzer (OSA) (Agilent, 86140B). The switching characteristics measured at 20 °C are shown in Fig. 5(a). As shown in Fig. 5(a), the signal powers from the bar port and the cross port vary with the input pump power in a complementary manner. Here, the pump power refers to the output power from the input launching fiber (the actual amount of pump power coupled into the waveguide is smaller). As the input pump power increases from 0 mW to 0.8 mW, the signal power from the cross port increases, while the signal power from the bar port decreases. As the pump power varies from 0.8 mW to 6.8 mW, the signal power from the cross port drops by 36 dB, which suggests that the signal from the cross port can be switched at an extinction ratio of 36 dB with a pump power variation of 6.8–0.8 = 6.0 mW. The corresponding extinction ratio for the bar port is 21.5 dB. By comparing the output pump powers for the TE and TM polarizations from the MZI switch, we deduce that the power of the TE-polarized pump absorbed by graphene is approximately 2.2 mW, which is the actual amount of the pump power converted into heat for achieving all-optical switching. Given the uncertainties in our power measurements (about ±0.3 dB) and the assumed parameters in the theoretical model, the measurement result 2.2 mW is close to the theoretical value 2.4 mW shown in Fig. 2(b).
Fig. 5. (a) Measured variations of the output signal powers at the two ports of the MZI switch with the input pump power; and (b) theoretical switching characteristics of a balanced MZI where the two DCs have the same splitting ratio 49:51.
The measured switching characteristics shown in Fig. 5(a) differ from the ideal switching characteristics of a balanced MZI in two aspects: the existence of a pump power offset (0.8 mW) and different extinction ratios for the two ports. We believe that the non-ideal performance of our fabricated switch is caused by the non-ideal splitting ratio of the DC. To confirm that, we measured the splitting ratio of a reference DC at 1550 nm and found that the splitting ratio of the DC was 49:51 instead of 50:50. Using this splitting ratio, we calculate the switching characteristics and show the results in Fig. 5(b). Indeed, the simulation results show a phase bias and a larger extinction ratio at the cross port, which agrees qualitatively with the experimental results. The non-ideal splitting ratios of the two DCs in the MZI break the balance of the MZI and thus leads to unbalanced switching characteristics. The switching characteristics of the device could be improved by improving the fabrication process or by applying a post-tuning or post-processing technique to tune or adjust the splitting ratio of the DC (e.g., by thermal tuning of the DC with heat electrodes deposited on the two arms of the DC). In addition, with the help of the reference DC, we find that ∼4% of the pump light leaks to the cross port of the DC as the higher-order mode, which slightly reduces the pumping efficiency of the switch.
The insertion losses of the device (excluding fiber-waveguide coupling losses at the two ends) measured for the TM-polarized pump light and the TM-polarized signal light are 2.1 dB and 5.3 dB, respectively. As the polymer system is developed for applications at 850 nm [35,36], the material loss at 980 nm is smaller than that at 1550 nm. The tight mode confinement of the 980-nm pump light also helps to reduce the bending loss. The measured fiber-waveguide coupling losses at 980 nm and 1550 nm are 2.0 dB and 2.4 dB, respectively, which are polarization-insensitive. With the fiber-waveguide coupling loss deducted, the actual pump power launched into the waveguide for switching is 3.8 mW.
We repeated the measurements at different temperatures by placing a semiconductor heater at the bottom of the device. For the temperature range from 20 °C to 50 °C, the fluctuations in the switching power was within ∼0.6 mW and the changes in the extinction ratios at the cross port and the bar port were within ∼4 dB and ∼1 dB, respectively. The device had a weak temperature sensitivity, thanks to the balanced MZI structure.
We finally measured the response times of the device by modulating the pump laser with a 100-Hz square-wave electrical signal generated from a function generator (GW Instek, AFG-2225). Another WDM was placed at the output end to filter out any surplus pump light and the output signals were detected by a photodetector and displayed on an oscilloscope (MCP Lab Electronic, DQ8304C). As shown in Fig. 6, the rise and fall times of the switch measured at the cross port are 1.0 ms and 2.7 ms, respectively, and those measured at the bar port are 0.9 ms and 2.5 ms, respectively. The corresponding switching frequency is several hundred Hz. The switching times measured for both ports are more or less the same, as expected. The switching times of our device are comparable to those of polymer waveguide thermo-optic switches driven with graphene heater electrodes [25,26]. The fact that the rise time is shorter than the fall time can be explained by the large thermal conductivity of graphene, which provides a fast transfer of heat to the waveguide, and the good thermal insulation of the polymer material, which tends to slow down the process of heat dissipation.
5. Comparison of devices and discussion
We summarize in Table 1 the performances of representative experimental all-optical switches based on graphene’s photothermal effect. We may divide the devices into two groups: fiber devices and on-chip devices. Fiber devices are formed with fibers and can be readily connected to fiber-based systems, while on-chip devices are formed with high-index-contrast waveguides and require special waveguide structures, such as grating couplers, for achieving fiber coupling. The fiber-coupling losses of on-chip devices are usually very high (e.g., 7 to 9 dB for one end [17,18]). Our low-index-contrast polymer waveguide switch is a kind of its own. On one hand, our switch can be butt-coupled to fibers with much lower losses and, therefore, is fiber-compatible. On the other hand, planar lightwave circuits for on-chip signal processing can be fabricated on the polymer waveguide platform using the well-established microfabrication technology and the fabrication cost is low.
Table 1. Comparison of reported all-optical switches based on graphene’s photothermal effect
As shown in Table 1, the switching power of our device (6.0 mW) is among the lowest and the extinction ratio achieved is the highest (36 dB). The switching time of our device (1.0/2.7 ms) is shorter than most of the fiber devices. The switching speed is limited by the relatively poor heat conductivity of the polymer material. On-chip devices can provide much faster response, thanks to their small sizes and superior heat conductivity, but the switching powers required are high. Apart from the problem of high fiber-coupling loss, another problem with on-chip devices formed with high-index-contrast waveguides is the high graphene-induced signal loss, which not only limits the length of graphene and the number of graphene layers that can be used for pump absorption, but also further increases the (already large) insertion loss of the device. This constraint can be relaxed by using low-index-contrast waveguides, as demonstrated by our device. The graphene-induced signal loss of our device (0.22 dB/mm) is lower than those of on-chip devices by more than an order of magnitude. Only an externally pumped fiber MZI switch, which has an almost negligible graphene-induced signal loss [13], can outperform our device in this aspect. Furthermore, the grating couplers used for fiber coupling in those on-chip devices limit the choice of the pump wavelength.
Most all-optical switches, including our proposed switch, employ an interferometer structure, such as the MZI structure. All-optical switches implemented with different forms of resonators, such as microrings, are available. Being narrow-band devices, resonators can be more sensitive to heat-induced phase shifts, which could be translated into lower switching powers or faster responses. However, the requirement that both the pump and the signal wavelength must be aligned with the resonance wavelengths of the resonator poses much restriction on the choice of wavelengths, and thermal stability can also be an issue. On the other hand, MZI-based all-optical switches are broadband devices, which offer much flexibility in the choice of the pump wavelength and the signal band. Thanks to the use of a balanced MZI, our proposed all-optical switch has good thermal stability.
6. Conclusion
We have proposed and experimentally demonstrated an efficient all-optical switch that explores graphene’s photothermal effect with a graphene-buried polymer waveguide balanced MZI. Thanks to the low-index-contrast material system, we can use the TE polarization for the pump light (980 nm) and the TM polarization for the signal light (1550 nm) to ensure maximum pump absorption by graphene and minimum graphene-induced signal loss. The input pump power change required for achieving an extinction ratio of 36 dB is 6.0 mW and the rise/fall switching times are 1.0 and 2.7 ms, respectively. The actual pump power absorbed by graphene for activating switching is estimated to be 2.2 mW. The graphene-induced signal loss is 1.1 dB. Our all-optical switch shows superior performance in many aspects, compared with reported fiber and waveguide all-optical switches based on graphene’s photothermal effect (Table 1). Our switch can be easily butt-coupled to fibers for application in fiber-based systems or further integrated with waveguide devices for on-chip all-optical signal processing, where low-power consumption is important and high switching speed is not required, such as optical circuit switching. Graphene-buried polymer waveguide can be designed to optimize the interaction of graphene with light and thus serve as an effective platform for the development of all-optical control devices based on various nonlinear effects of graphene.
Funding
City University of Hong Kong (7005224).
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. L. Yan, A. E. Willner, X. Wu, A. Yi, A. Bogoni, Z. Y. Chen, and H. Y. Jiang, “All-optical signal processing for ultrahigh speed optical systems and networks,” J. Lightwave Technol. 30(24), 3760–3770 (2012). [CrossRef]
2. A. E. Willner, S. Khaleghi, M. R. Chitgarha, and O. F. Yilmaz, “All-optical signal processing,” J. Lightwave Technol. 32(4), 660–680 (2014). [CrossRef]
3. B. J. Eggleton, R. E. Slusher, J. B. Judkins, J. B. Stark, and A. M. Vengsarkar, “All-optical switching in long-period fiber gratings,” Opt. Lett. 22(12), 883–885 (1997). [CrossRef]
4. K. Nozaki, T. Tanabe, A. Shinya, S. Matsuo, T. Sato, H. Taniyama, and M. Notomi, “Sub-femtojoule all-optical switching using a photonic-crystal nanocavity,” Nat. Photonics 4(7), 477–483 (2010). [CrossRef]
5. M. Asobe, “Nonlinear Optical Properties of Chalcogenide Glass Fibers and Their Application to All-Optical Switching,” Opt. Fiber Technol. 3(2), 142–148 (1997). [CrossRef]
6. Y. Bo, C. Guan, R. Chu, T. Cheng, L. Xu, L. Liu, J. Yang, J. Shi, P. Wang, J. Yang, and L. Yuan, “Compact all-fiber thermo-optic modulator based on a Michelson interferometer coated with NaNdF4 nanoparticles,” Opt. Express 29(5), 6854–6862 (2021). [CrossRef]
7. M. J. Holmes, D. L. Williams, and R. J. Manning, “Highly nonlinear optical fiber for all optical processing applications,” IEEE Photonics Technol. Lett. 7(9), 1045–1047 (1995). [CrossRef]
8. T. Gu, N. Petrone, J. F. McMillan, A. Van Der Zande, M. Yu, G. Q. Lo, D. L. Kwong, J. Hone, and C. W. Wong, “Regenerative oscillation and four-wave mixing in graphene optoelectronics,” Nat. Photonics 6(8), 554–559 (2012). [CrossRef]
9. H. Chen, C. Wang, H. Ouyang, Y. Song, and T. Jiang, “All-optical modulation with 2D layered materials: Status and prospects,” Nanophotonics 9(8), 2107–2124 (2020). [CrossRef]
10. C. Zhong, J. Li, and H. Lin, “Graphene-based all-optical modulators,” Front. Optoelectron. 13(2), 114–128 (2020). [CrossRef]
11. S. Yu, X. Wu, K. Chen, B. Chen, X. Guo, D. Dai, L. Tong, W. Liu, and Y. Ron Shen, “All-optical graphene modulator based on optical Kerr phase shift,” Optica 3(5), 541–544 (2016). [CrossRef]
12. X. Gan, C. Zhao, Y. Wang, D. Mao, L. Fang, L. Han, and J. Zhao, “Graphene-assisted all-fiber phase shifter and switching,” Optica 2(5), 468–471 (2015). [CrossRef]
13. T. Hao, Z. Chang, and K. S. Chiang, “Externally pumped low-loss graphene-based fiber Mach-Zehnder all-optical switches with mW switching powers,” Opt. Express 27(4), 4216–4225 (2019). [CrossRef]
14. Y. Wang, X. Gan, C. Zhao, L. Fang, D. Mao, Y. Xu, F. Zhang, T. Xi, L. Ren, and J. Zhao, “All-optical control of microfiber resonator by graphene’s photothermal effect,” Appl. Phys. Lett. 108(17), 171905 (2016). [CrossRef]
15. R. Chu, C. Guan, Y. Bo, J. Liu, J. Shi, J. Yang, P. Ye, P. Li, J. Yang, and L. Yuan, “Graphene decorated twin-core fiber Michelson interferometer for all-optical phase shifter and switch,” Opt. Lett. 45(1), 177–180 (2020). [CrossRef]
16. C. Qiu, C. Zhang, H. Zeng, and T. Guo, “High-performance graphene-on-silicon nitride all-optical switch based on a Mach-Zehnder interferometer,” J. Lightwave Technol. 39(7), 2099–2105 (2021). [CrossRef]
17. C. Qiu, Y. Yang, C. Li, Y. Wang, K. Wu, and J. Chen, “All-optical control of light on a graphene-on-silicon nitride chip using thermo-optic effect,” Sci. Rep. 7(1), 1–7 (2017). [CrossRef]
18. T. Guo, S. Gao, H. Zeng, L. Tang, and C. Qiu, “All-optical control of a single resonance in a graphene-on-silicon nanobeam cavity using thermo-optic effect,” J. Lightwave Technol. 39(14), 4710–4716 (2021). [CrossRef]
19. Z. W. Li, Q. Liu, H. Wang, M. Deng, Q. Huang, and Y. Wang, “Photo-induced thermo-optical refraction switching by a graphene-assisted silicon microring resonator,” J. Lightwave Technol. 39(11), 3471–3477 (2021). [CrossRef]
20. W. Li, B. Chen, C. Meng, W. Fang, Y. Xiao, X. Li, Z. Hu, Y. Xu, L. Tong, H. Wang, W. Liu, J. Bao, and Y. R. Shen, “Ultrafast all-optical graphene modulator,” Nano Lett. 14(2), 955–959 (2014). [CrossRef]
21. M. Ono, M. Hata, M. Tsunekawa, K. Nozaki, H. Sumikura, H. Chiba, and M. Notomi, “Ultrafast and energy-efficient all-optical switching with graphene-loaded deep-subwavelength plasmonic waveguides,” Nat. Photonics 14(1), 37–43 (2020). [CrossRef]
22. Z. Chang and K. S. Chiang, “All-optical loss modulation with graphene-buried polymer waveguides,” Opt. Lett. 44(15), 3685–3688 (2019). [CrossRef]
23. G. Xing, H. Guo, X. Zhang, T. C. Sum, and C. H. A. Huan, “The physics of ultrafast saturable absorption in graphene,” Opt. Express 18(5), 4564–4573 (2010). [CrossRef]
24. Q. Huang, W. Jin, and K. S. Chiang, “Broadband mode switch based on a three-dimensional waveguide Mach–Zehnder interferometer,” Opt. Lett. 42(23), 4877–4880 (2017). [CrossRef]
25. X. Wang, W. Jin, Z. Chang, and K. S. Chiang, “Buried graphene electrode heater for a polymer waveguide thermo-optic device,” Opt. Lett. 44(6), 1480–1483 (2019). [CrossRef]
26. Q. Q. Song, K. X. Chen, and Z. F. Hu, “Low-power broadband thermo-optic switch with weak polarization dependence using a segmented graphene heater,” J. Lightwave Technol. 38(6), 1358–1364 (2020). [CrossRef]
27. Z. Chang and K. S. Chiang, “Experimental verification of optical models of graphene with multimode slab waveguides,” Opt. Lett. 41(9), 2129–2132 (2016). [CrossRef]
28. Q. Bao, H. Zhang, B. Wang, Z. Ni, C. H. Y. X. Lim, Y. Wang, D. Y. Tang, and K. P. Loh, “Broadband graphene polarizer,” Nat. Photonics 5(7), 411–415 (2011). [CrossRef]
29. Z. Chang and K. S. Chiang, “Ultra-broadband mode filters based on graphene-embedded waveguides,” Opt. Lett. 42(19), 3868–3871 (2017). [CrossRef]
30. Q. Huang and K. S. Chiang, “Polarization-insensitive ultra-broadband mode filter based on a 3D graphene structure buried in an optical waveguide,” Optica 7(7), 744–745 (2020). [CrossRef]
31. Z. Chang, W. Jin, and K. S. Chiang, “Graphene electrodes for lithium-niobate electro-optic devices,” Opt. Lett. 43(8), 1718–1721 (2018). [CrossRef]
32. W. Wang, J. Wu, K. Chen, Q. Huang, J. Luo, and K. S. Chiang, “Graphene electrodes for electric poling of electro-optic polymer films,” Opt. Lett. 45(8), 2383–2386 (2020). [CrossRef]
33. M. Liu, X. Yin, E. Ulin-Avila, B. Geng, T. Zentgraf, L. Ju, F. Wang, and X. Zhang, “A graphene-based broadband optical modulator,” Nature 474(7349), 64–67 (2011). [CrossRef]
34. A. A. Balandin, S. Ghosh, W. Bao, I. Calizo, D. Teweldebrhan, F. Miao, and C. N. Lau, “Superior thermal conductivity of single-layer graphene,” Nano Lett. 8(3), 902–907 (2008). [CrossRef]
35. W. Jin, K. S. Chiang, K. P. Lor, H. P. Chan, J. T. L. To, and R. H. M. Leung, “Industry compatible embossing process for the fabrication of waveguide-embedded optical printed circuit boards,” J. Lightwave Technol. 31(24), 4045–4050 (2013). [CrossRef]
36. V. Prajzler, M. Neruda, P. Jašek, and P. Nekvindová, “The properties of free-standing epoxy polymer multi-mode optical waveguides,” Microsyst. Technol. 25(1), 257–264 (2019). [CrossRef]
