We experimentally demonstrate waveguide and microring resonator (MRR) polarizers by integrating 2D graphene oxide (GO) films onto silicon (Si) photonic devices. The 2D GO films with highly anisotropic light absorption characteristic are on-chip integrated with precise control over their thicknesses and sizes. Detailed measurements are performed for the fabricated devices with different GO film thicknesses, coating lengths, and Si waveguide widths. The results show that a maximum polarization-dependent loss of ∼17 dB is achieved for the hybrid waveguides, and the hybrid MRRs achieved a maximum polarization extinction ratio of ∼10 dB. We also characterize the wavelength- and power-dependent response for these polarizers. The former demonstrates a broad operation bandwidth of over ∼100 nm, and the latter verifies performance improvement enabled by photothermal changes in GO films. By fitting the experimental results with theoretical simulations, we find that the anisotropy in the loss of GO films dominates the polarization selectivity of these devices. These results highlight the strong potential of 2D GO films for realizing high-performance polarization selective devices in Si photonic platforms.
Optical polarizers, which selectively transmit light with a specific polarization orientation while blocking light of the orthogonal polarization, are essential components underpinning modern optical systems.1–4 A variety of optical polarizers have been implemented based on different bulk material device platforms, such as refractive prisms,5,6 birefringent crystals,7,8 fiber components,9,10 and integrated photonic devices.11,12 Among these, integrated photonic polarizers, particularly those based on the well-developed silicon (Si) photonic platform, offer attractive advantages of compact footprint, low power consumption, and the capability for large-scale manufacturing.13 Nevertheless, the rapid advancement of photonics industries drives the demand for high-performance optical polarizers across broad wavelength ranges, which is usually challenging for optical polarizers based on bulk materials.14–16
Recently, owing to their high anisotropy in light absorption and broadband response, 2D materials have been on-chip integrated to realize high-performance optical polarizers.17–25 As a common derivative of graphene, graphene oxide (GO) with facile fabrication processes for on-chip integration shows a high compatibility with integrated device platforms.26–28 Previously, we reported high performance waveguide and microring resonator (MRR) polarizers by integrating 2D GO films onto high index doped silica devices.20 In this work, we further integrate 2D GO films onto the more widely used Si photonic devices to realize waveguide and MRR polarizers. Precise control of the GO film thicknesses is achieved by using a transfer-free and layer-by-layer film coating method, together with window opening on the upper cladding of Si photonic devices to control the film coating lengths. We perform detailed measurements for devices with different structural parameters, achieving a maximum polarization-dependent loss (PDL) of ∼17 dB and a maximum polarization extinction ratio (PER) for the waveguide and MRR polarizers, respectively. The wavelength- and power-dependent response of these polarizers is also characterized, showing a broad operation bandwidth over ∼100 nm and improvement in the polarization selectivity enabled by photothermal changes in GO films. Finally, we perform theoretical analysis by fitting the experimental results with simulations, which reveals that the anisotropy in the loss of GO films dominates the polarization selectivity of these polarizers. These results verify the effectiveness of integrating 2D GO films onto Si photonic devices for implementing high-performance optical polarizers.
Figure 1(a) illustrates the atomic structure of the monolayer GO, in which the carbon network is decorated with diverse oxygen functional groups (OFGs) such as hydroxyl, epoxide, and carboxylic groups.29,30 Unlike graphene that has a zero bandgap and exhibits metallic behavior,31 GO is a dielectric material with an opened bandgap typically ranging between ∼2.1 and ∼3.6 eV.27,29 This is larger than the energy of two photons at 1550 nm (i.e., ∼1.6 eV), which yields both low linear light absorption and two photon absorption at infrared wavelengths. In this work, we choose GO because it has several advantages for implementing high-performance optical polarizers. First, GO exhibits strong anisotropic light absorption over a very broad spectral bandwidth,20 which is challenging to achieve for optical polarizers based on bulk materials. Second, GO has relatively low material absorption at infrared wavelengths as compared to other 2D materials (e.g., its extinction coefficient is about 2 orders of magnitude lower than that of graphene32,33). Finally, in contrast to the cumbersome transfer processes employed for other 2D materials such as graphene and transition metal dichalcogenides (TMDCs),32,34 GO has facile fabrication processes for on-chip integration and shows a high compatibility with integrated device platforms.35,36
(a) Schematic of GO's atomic structure. (b) Schematic illustration of a waveguide optical polarizer based on a Si waveguide integrated with a three-layer GO film. Inset illustrates the layered structure of GO films fabricated by the self-assembly method. (c) TE and TM mode profiles for the hybrid waveguide in (b). (d) Microscopic image of the fabricated devices on a silicon-on-insulator (SOI) chip coated with a monolayer GO film. (e) Measured Raman spectra of the SOI chip in (d) before and after coating the GO film.
(a) Schematic of GO's atomic structure. (b) Schematic illustration of a waveguide optical polarizer based on a Si waveguide integrated with a three-layer GO film. Inset illustrates the layered structure of GO films fabricated by the self-assembly method. (c) TE and TM mode profiles for the hybrid waveguide in (b). (d) Microscopic image of the fabricated devices on a silicon-on-insulator (SOI) chip coated with a monolayer GO film. (e) Measured Raman spectra of the SOI chip in (d) before and after coating the GO film.
Figure 1(b) shows the schematic of a Si nanowire waveguide integrated with a three-layer GO film. The corresponding transverse electric (TE) and transverse magnetic (TM) mode profiles for the hybrid waveguide are shown in Fig. 1(c), which were simulated using a commercial mode solving software (COMSOL Multiphysics). The width and height of the Si waveguide are 400 and 220 nm, respectively. Due to the interaction between the evanescent field from the Si waveguide and the 2D GO film with strong anisotropy in its light absorption, the hybrid waveguide exhibits much stronger light absorption for TE (in-plane) polarization than TM (out-of-plane) polarization. This enables it to function as a TM-pass optical polarizer.
Figure 1(d) shows a microscopic image of the fabricated devices on a silicon-on-insulator (SOI) chip. The detailed fabrication processes are introduced in Note 1 of the supplementary material. We fabricated devices with five different lengths (ranging between ∼0.1 and ∼2.2 mm) for the opened windows, which correspond to different GO film coating lengths in the hybrid waveguides. As shown in Fig. 1(d), the good morphology of the coated GO films provides evidence for the high film uniformity achieved by our coating method.
Figure 1(e) shows the measured Raman spectra of the SOI chip in Fig. 1(d) before and after coating a monolayer GO film. The GO film had a thickness of ∼2 nm, which was characterized by atomic force microscopy measurement. In the Raman spectrum for the GO-coated chip, the presence of the representative D (1345 cm−1) and G (1590 cm−1) peaks verifies the successful on-chip integration of the GO film.
In Fig. 2, we show the results for the measured insertion losses (ILs) of the fabricated devices for input light with different polarization states. We measured devices with different GO layer numbers (N), GO coating lengths (LGO), and Si waveguide widths (W). For all the devices, the total length of the Si waveguides was ∼3.0 mm. In our measurements, a continuous-wave (CW) light at ∼1550 nm was butt coupled into/out of the devices via lensed fibers. The fiber-to-chip coupling loss was ∼5 dB per facet. For comparison, the input power was kept the same as Pin = ∼0 dBm. Unless otherwise specified, the values of Pin and IL in our discussion refer to those after excluding the fiber-to-chip coupling loss.
(a) Measured (i) TE- and (ii) TM-polarized IL vs GO coating length LGO for the hybrid waveguides with GO layer number N = 1–5. (iii) PDL calculated from (i) and (iii). (b) Polar diagrams for the measured IL of devices with (i) N = 1, (ii) N = 3, and (iii) N = 5. The polar angle represents the angle between the input polarization plane and the substrate. In (a) and (b), the input CW power and wavelength were ∼0 dBm and ∼1550 nm, respectively. In (a), the data points depict the average of measurements on three duplicate devices, and the error bars illustrate the variations among the different devices, the waveguide width W = ∼400 nm. In (b), LGO = ∼2.2 mm and W = ∼400 nm.
(a) Measured (i) TE- and (ii) TM-polarized IL vs GO coating length LGO for the hybrid waveguides with GO layer number N = 1–5. (iii) PDL calculated from (i) and (iii). (b) Polar diagrams for the measured IL of devices with (i) N = 1, (ii) N = 3, and (iii) N = 5. The polar angle represents the angle between the input polarization plane and the substrate. In (a) and (b), the input CW power and wavelength were ∼0 dBm and ∼1550 nm, respectively. In (a), the data points depict the average of measurements on three duplicate devices, and the error bars illustrate the variations among the different devices, the waveguide width W = ∼400 nm. In (b), LGO = ∼2.2 mm and W = ∼400 nm.
Figures 2(a-i) and 2(a-ii) show the measured IL vs LGO for TE- and TM-polarized input light, respectively. Here, we show the results for devices with N = 1–5. It can be seen that the IL increases with LGO and N for both polarizations, with the TE polarization showing a more significant increase than the TM polarization. In Fig. 2(a-iii), we further calculated the PDL (dB) by subtracting the TM-polarized IL (dB) from the TE-polarized IL (dB). As can be seen, the PDL increases with LGO and N. For the five-layer device with LGO = ∼2.2 mm, a maximum PDL value of ∼17 dB was achieved. In contrast, the uncoated Si waveguide did not show any significant polarization dependent IL, with the PDL being less than 0.2 dB. We also characterize the measured IL vs W for TE- and TM-polarized input light, the results are shown in Note 2 of the supplementary material. For optical polarizers based on 2D materials, there is usually a trade-off between achieving a high PDL and maintaining a low IL. The difference in the TM-polarized IL values between the uncoated and hybrid waveguides reflects the minimum excess IL induced by the GO film. For the five-layer device with LGO = ∼2.2 mm, the difference was ∼10 dB. This value is comparable to those reported for other 2D-material-based optical polarizers.19,20,24 It is also worth noting that the excess IL induced by GO is not a fundamental limitation. It can be further reduced by either designing silicon waveguides with different widths to adjust the relative overlap of the TM mode compared to the TE mode or optimizing the fabrication process to minimize the loss of GO films.
Figure 2(b) shows the polar diagrams for the measured IL of devices with N = 1, 3, and 5. In the polar diagrams, the variations in the IL values across various polarization angles further confirm the polarization selection capability for the hybrid waveguides. Compared to the five-layer device that achieved a PDL of ∼17 dB, the one-layer and three-layer devices achieved lower PDL values of ∼5 and ∼9 dB, respectively.
Figure 3(a) shows the measured PDL vs input CW wavelength. For all the devices, the PDL shows a very small variation that was less than 1 dB within the measured wavelength range of ∼1500–1600 nm. This reflects the broad operation bandwidth for these waveguide polarizers. We note that there was a slight increase in the PDL as the wavelength increased, which can be attributed to a minor change in GO's mode overlap induced by dispersion. In our measurements, the wavelength tuning range was limited by the employed tunable CW laser. The light absorption bandwidth for GO is in fact quite broad, which covers infrared wavelengths and extends into the visible and terahertz (THz) ranges.26,37 Such broadband response enables a much broader operation bandwidth (potentially spanning several hundreds of nanometers, as we demonstrated in Ref. 20) than that demonstrated here. This represents a significant advantage of 2D-material-based polarizers, which is usually challenging to achieve for integrated photonic devices based on only bulk materials.2,38
(a) Measured PDL vs wavelength of input CW light for (i) devices with different N but the same LGO of ∼2.2 mm and (ii) devices with different LGO but the same N = 5. (b-i) Measured TE and TM polarized IL vs input power Pin for devices with different N = 1, 3, and 5 but the same LGO = ∼2.2 mm. (b-ii) PDL calculated from (b-i). In (a) and (b), W = ∼400 nm. In (a), Pin = ∼0 dBm. In (b), the input CW wavelength was ∼1550 nm.
(a) Measured PDL vs wavelength of input CW light for (i) devices with different N but the same LGO of ∼2.2 mm and (ii) devices with different LGO but the same N = 5. (b-i) Measured TE and TM polarized IL vs input power Pin for devices with different N = 1, 3, and 5 but the same LGO = ∼2.2 mm. (b-ii) PDL calculated from (b-i). In (a) and (b), W = ∼400 nm. In (a), Pin = ∼0 dBm. In (b), the input CW wavelength was ∼1550 nm.
Figure 3(b-i) shows the measured IL vs input CW power Pin. As Pin increases, the IL increases for both polarizations, with the TE polarization showing a more significant increase than the TM polarization. The increased IL was induced by the photothermal reduction of GO at high light powers, where reduced GO exhibited stronger light absorption than unreduced GO.27,39 Another interesting feature for the reduction of GO induced by photothermal effects is its reversibility within a certain power range.39,40 This originates from the instability of photothermally reduced GO, which can easily revert to the unreduced state once the input CW power is turned off. Since the hybrid waveguides exhibited stronger absorption for TE-polarized light, the reduction of GO occurred more readily for TE polarization, resulting in a more significant change in the corresponding IL in Fig. 3(b-i). We also note that the five-layer device showed more significant changes in the IL as compared to the three-layer and one-layer devices, which reflects the fact that there were more significant photothermal effects in thicker GO films. In Fig. 3(b-ii), we show the calculated PDL vs Pin, where the PDL increases for an increasing Pin. This indicates that the polarization selectivity was further improved by increasing Pin. At Pin = ∼15 dBm, the PDL for the five-layer device was ∼25 dB, representing an ∼8-dB improvement relative to the PDL at Pin = ∼0 dBm. We also characterize the measured IL vs Pin for the hybrid waveguides with different LGO, the results are shown in Note 3 of the supplementary material.
Based on the results in Figs. 2 and 3, we further analyze the properties of 2D GO films by fitting the experimental results with theoretical simulations. Figure 4(a-i) shows the propagation losses (PLs) of the hybrid waveguides vs N for both polarizations, which were extracted from the measured ILs in Figs. 2(a-i) and 2(a-ii). Figure 4(a-ii) shows the extinction coefficients (k's) of 2D GO films obtained by fitting the PL's in Fig. 4(a-i) with optical mode simulations, and the ratios of TE- to TM-polarized k values are further plotted in Fig. 4(a-iii). For all different N, the GO films exhibit much larger k values for TE polarization than TM polarization in Fig. 4(a-ii), highlighting the high anisotropy in their light absorption. For both polarizations, there is a slight increase in k as N increases. This can possibly be attributed to more significant scattering loss in thicker GO films. It is also interesting to note that in Fig. 4(a-iii), the ratio of the k values remains relatively consistent without any significant variations.
(a-i) TE- and TM-polarized waveguide propagation loss (PL) vs N for the hybrid waveguides. (a-ii) Extinction coefficients (k's) of 2D GO films obtained by fitting the results in (a-i) with optical mode simulations. (a-iii) Ratios of k values for TE and TM polarizations (kTE/kTM) extracted from (a-ii). (b-i) Measured (Exp.) and simulated (Sim.) PDL vs N. (b-ii) Fractional contributions (η's) to the overall PDL from polarization-dependent mode overlap and material loss anisotropy, which were extracted from (b-i). The simulated PDL values were obtained by using the same k value for both TE and TM polarizations. (c-i) TE- and TM-polarized PL vs input power Pin for the device with N = 5. (c-ii) k's of 2D GO films obtained by fitting the results in (c-i) with optical mode simulations. (c-iii) kTE/kTM extracted from (c-ii). In (a)–(c), LGO = ∼2.2 mm, W = ∼400 nm. In (a) and (b), Pin = ∼0 dBm.
(a-i) TE- and TM-polarized waveguide propagation loss (PL) vs N for the hybrid waveguides. (a-ii) Extinction coefficients (k's) of 2D GO films obtained by fitting the results in (a-i) with optical mode simulations. (a-iii) Ratios of k values for TE and TM polarizations (kTE/kTM) extracted from (a-ii). (b-i) Measured (Exp.) and simulated (Sim.) PDL vs N. (b-ii) Fractional contributions (η's) to the overall PDL from polarization-dependent mode overlap and material loss anisotropy, which were extracted from (b-i). The simulated PDL values were obtained by using the same k value for both TE and TM polarizations. (c-i) TE- and TM-polarized PL vs input power Pin for the device with N = 5. (c-ii) k's of 2D GO films obtained by fitting the results in (c-i) with optical mode simulations. (c-iii) kTE/kTM extracted from (c-ii). In (a)–(c), LGO = ∼2.2 mm, W = ∼400 nm. In (a) and (b), Pin = ∼0 dBm.
In Fig. 4(b-i), we compare the measured PDL values with those obtained from optical mode simulations for the devices with different N. In our simulations, we assumed that the GO films were isotropic with the same k value [i.e., kTE for N = 5 in Fig. 4(a-ii)] for both TE and TM polarizations. As a result, the simulated PDL values represent the polarization selectivity enabled by the polarization-dependent GO mode overlap, and the difference between the measured and simulated PDL values reflects the extra polarization selectivity provided by the loss anisotropy of GO films. Both the measured and simulated PDLs show positive values for all different N. This suggests that the polarization-dependent GO mode overlap contributes to the overall PDL. In Note 4 of the supplementary material, we also compare the measured and simulated PDL values for the devices with different W.
In Fig. 4(b-ii), we further calculated the fractional contributions to the overall PDL from the polarization-dependent mode overlap and the material loss anisotropy (where the sum of the two fractions equals 1). As can be seen, the contribution from the material loss anisotropy, which accounts for more than 70% for all different N, dominates the overall PDL. This further highlights the significance of the anisotropic 2D GO films in facilitating the functionality of the polarizer.
Figure 4(c-i) shows the PLs of the hybrid waveguides vs Pin for both polarizations, which were extracted from the measured ILs in Fig. 3(b). Figure 4(c-ii) shows the k values of 2D GO films obtained by fitting the PLs in Fig. 4(c-i) with optical mode simulations, and Fig. 4(c-iii) shows the ratios of the TE- to TM-polarized k values in Fig. 4(c-ii). In Fig. 4(c-ii), k increases with Pin for both polarizations, with the TE polarization showing a more significant increase than the TM polarization. In Fig. 4(c-iii), the ratio between the k values slightly increases with Pin, which is also resulting from the more significant photothermal effects in GO films induced by stronger absorption of TE-polarized light.
Except for waveguide optical polarizers, we also integrated 2D GO films onto Si MRRs to implement polarization-selective MRRs. Figures 5(a) and 5(b) show the device schematic and a microscopic image of the fabricated device with a monolayer GO film, respectively. The MRRs were fabricated together with the waveguides in Fig. 1(d) via the same processes. The rings and the bus waveguides had the same width of W = ∼400 nm, consistent with that of the waveguide polarizers in Fig. 2. The radius of the MRRs was ∼20 μm, and the length of the opened windows on the MRRs was ∼10 μm. Compared to the waveguide polarizers with GO coating lengths on the millimeter scale, the much shorter GO coating length in MRR polarizers introduces very low additional loss, allowing the resonant notches to be observed in the measured transmission spectra.
(a) Schematic illustration of a GO-coated Si microring resonator (MRR) as a polarization-selective MRR. (b) Microscopic image of a fabricated device with a monolayer GO film. (c) Measured (i) TE- and (ii) TM-polarized transmission spectra of the hybrid MRRs with N = 1, 2. The corresponding results for the uncoated MRR (N = 0) are also shown for comparison. (d) ERs for the MRRs extracted from (c). (e) PERs extracted from (d). In (c)–(e), Pin = ∼−10 dBm.
(a) Schematic illustration of a GO-coated Si microring resonator (MRR) as a polarization-selective MRR. (b) Microscopic image of a fabricated device with a monolayer GO film. (c) Measured (i) TE- and (ii) TM-polarized transmission spectra of the hybrid MRRs with N = 1, 2. The corresponding results for the uncoated MRR (N = 0) are also shown for comparison. (d) ERs for the MRRs extracted from (c). (e) PERs extracted from (d). In (c)–(e), Pin = ∼−10 dBm.
By scanning the wavelength of a CW light coupled into the bus waveguides, we measured the TE- and TM-polarized transmission spectra of the MRRs. Figure 5(c) shows the measured spectra for the hybrid MRRs with one and two layers of GO, together with those for the uncoated MRR for comparison. For all the measured spectra, the input CW power remained constant at Pin = ∼−10 dBm. Figure 5(d) shows the extinction ratios (ERs) of these MRRs extracted from Fig. 5(c). For the uncoated MRR, the ERs for TE and TM polarizations were ∼33 and ∼32 dB, respectively. In contrast, the ERs decreased to ∼16 and ∼26 dB for the one-layer device, and ∼11 and ∼19 dB for the two-layer device. The uncoated MRR is slightly under-coupled.41 Therefore, the increased round-trip loss induced by GO leads to a greater difference between the round-trip loss and the coupling strength, resulting in a decrease in the ER.
Figure 5(e) shows the PER calculated from Fig. 5(d), which is defined as the absolute difference between the ERs of the TE- and TM-polarized resonances. As can be seen, the uncoated MRR exhibited negligible polarization selectivity, with its PER being less than ∼1 dB. In contrast, the hybrid devices with one and two layers of GO exhibited PER values of ∼10 and ∼8 dB, respectively, highlighting their high polarization selectivity. Compared to the device with two layers of GO, the device with one layer of GO exhibited a higher PER. This is the opposite trend to that observed in waveguide polarizers, where a higher PDL was achieved for the two-layer device because it had increased the GO mode overlap compared to the one-layer device. For MRR polarizers, the PER is plotted on a dB scale, which results in a more significant decrease in the MRR's ER with a higher value. The one-layer device exhibited a higher PER because the TM-polarized ER decreased more significantly for the device with a thicker GO film.
We also characterize the power-dependent response for the polarization-selective MRRs, the results are provided and discussed in Note 5 of the supplementary material. It is interesting to note that although the photothermal effects at high light powers may introduce instability to the GO polarizers, they actually enhance the polarization selectivity for both waveguide and MRR polarizers [see results in Figs. 3(b-ii) and S4(b)]. By modifying the OFGs in GO, the photothermal effects can be engineered. For example, removing some temperature sensitive OFGs, such as carboxyl and hydroxyl, can increase the power threshold for photothermal effects and enhance GO's stability.42 In addition, improved stability can be achieved by fully reducing GO to remove all the OFGs, resulting in material properties similar to graphene.26
In summary, waveguide and MRR polarizers are demonstrated by integrating 2D GO films onto Si photonic devices. We fabricate devices with precise control over the thicknesses and lengths of GO films and perform detailed measurements. By optimizing the device structural parameters, up to ∼17-dB PDL and ∼10-dB PER are achieved for the waveguide and MRR polarizers, respectively. The polarizers also show a broad operation bandwidth over ∼100 nm as well as polarization selectivity improvement enabled by photothermal changes in GO. Finally, we analyze the experimental results by fitting them with theoretical simulations. Our study provides an attractive approach for implementing high-performance optical polarizers by integrating 2D GO films onto Si photonic devices.
SUPPLEMENTARY MATERIAL
See the supplementary material for the preparation and fabrication of the device. Performance comparisons of devices with different waveguide widths W and different graphene oxide coating lengths LGO are also provided. Additionally, a comparison of measured and simulated power-dependent loss, as well as the power-dependent response of polarization-selective microring resonators are elaborated.
This work was supported by the Australian Research Council Centre of Excellence Project in Optical Microcombs for Breakthrough Science (No. CE230100006), the Australian Research Council Discovery Projects Programs (Nos. P190103186 and FT210100806), the Linkage Program (Nos. LP210200345 and LP210100467), the Swinburne ECR-SUPRA program, the Industrial Transformation Training Centers scheme (No. IC180100005), the Beijing Natural Science Foundation (No. Z180007), and the Innovation Program for Quantum Science and Technology (No. 2021ZD0300703).
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Di Jin: Investigation (equal). Jiayang Wu: Investigation (equal); Supervision (equal); Writing – original draft (equal). Junkai Hu: Investigation (equal). Wenbo Liu: Investigation (equal); Supervision (equal). Yuning Zhang: Investigation (equal). Yunyi Yang: Investigation (equal). Linnan Jia: Investigation (equal). Duan Huang: Investigation (equal); Supervision (equal). Baohua Jia: Investigation (equal); Supervision (equal). David J. Moss: Supervision (equal); Writing – original draft (equal).
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding authors upon reasonable request.