1. Introduction

Polarisation control is one of the essential functions required in the realization of photonic integrated circuits (PICs) [1]. Polarisation control is also required in fibre optical communication systems that exploit polarisation diversity to increase capacity [2]. In addition, most high index-contrast waveguides currently used as platforms for PIC fabrication exhibit intrinsic birefringence, which limits the sensitivity and coherence, as well as the bandwidth, of the functional circuits required in applications such as optical sensing and optical signal processing [1, 3]. Devices may be needed that can change the polarisation state of the light, both actively and passively. Two approaches have been proposed and demonstrated in an effort to alleviate these limitations. The first is to control the birefringence of the waveguide structure, either by making it birefringence-free or by using polarisation compensation [3–5] – while the second approach is to place an integrated polariser at the input of the PIC, in order to allow only one state of polarised light to propagate in the circuit [6–8].

The principal of the integrated polariser is to make one polarisation mode lossier than the other, either by leakage or absorption loss, thereby achieving a large difference in propagation loss between the two orthogonally polarised modes. As a consequence, a large extinction ratio – defined as the ratio of the power in the desired polarisation to the power in the undesired polarisation – can be achieved. A previous approach has used the strong polarisation dependence of plasmonic modes in metal-dielectric waveguides to suppress one of the polarisations [9]. High extinction ratio between orthogonally polarised modes can be achieved using metal-cladding based integrated waveguide polarisers. The effect is wavelength dependent, requires complex structures and the incorporation of resonant buffer layers in order to produce broadband polarisation response, thus increasing the complexity of PIC fabrication.

In recent years, graphene has attracted much attention in photonics applications due to its exceptional electronic transport and related optical properties. The conductance of graphene is defined by the fine structure constant and is independent of frequency over a wide range. Bao et. al. have demonstrated a broadband fibre polariser by replacing the metal thin film in the polarising waveguide with a graphene layer [10]. Later, Kim et. al. demonstrated a planar waveguide using graphene core waveguides [11] and subsequently the polarisation-dependent coupling of plasmonic modes at the waveguide-graphene interface of polymer waveguides [12]. These results represent promising successes towards the realization of PIC devices and show that graphene photonics plays an important role in optical waveguide device technology. To date, the graphene layer/s used in graphene based waveguide polarisers have usually been deposited by the chemical vapour deposition (CVD) technique, which requires careful transfer of the graphene layer from its initial growth substrate onto the waveguide. Furthermore, achieving a uniform few-layer or single-layer graphene coating over large areas, (i.e. square-millimetre scale or larger) with a minimal amount of the defects and discontinuities that can impact critically on optical waveguiding in the graphene, is still an on-going effort. Recently, uniform coating of graphene oxide (GO) layers through drop-casting has been reported. This method provides a simple alternative for uniform coating of thin graphene-based films over a finite area [13]. By increasing the concentration of the GO solution used, it becomes possible to produce GO ‘paper’ – at micrometre-scale thicknesses [14] – using the same method.

Coincidentally, it has recently been shown that GO typically exhibits a strongly anisotropic complex dielectric function [15]. This property was demonstrated in FET-type electronic devices. At optical frequencies, such dielectric anisotropy may be expected to lead to differences in the propagation loss for different polarisation states, which can be used to provide the function of polarisation selection in an optical waveguide. It is therefore logical for these properties of GO to be explored for applications such as waveguide polarisers. In the present paper, a simple and cost-effective fabrication method has been developed. We demonstrate the deposition of GO multi-layers directly on to a polymer optical waveguide using the drop-casting method. High-concentration, chemically exfoliated, GO solution is drop-cast on to the polymer waveguide using a micropipette, in order to create a waveguide section with a micrometre-scale thickness GO-coating. We then exploit the polarisation selection capability of GO to demonstrate a TM-pass polymer waveguide-based polariser with a large extinction-ratio. A large polarising effect is observed in the GO-coated polymer waveguides. A maximum extinction ratio value of 40 dB, which is (we believe) the highest reported for graphene-based waveguide polarisers to date, was achieved in the 1550 nm wavelength band. It is worth mentioning that this extinction ratio value has been achieved with a GO coating length of only 1.3 mm along the propagation direction.

2. Device fabrication and characterization

A CR-39 polymer sheet with a thickness of 0.5 mm and a refractive index of 1.486, measured at 1550 nm with a Sairon Technology SPA-4000 Prism Coupler, was used as the substrate for the polymer waveguide. The optical characteristics of CR-39 resemble those of crown glass and it is commonly used in the manufacture of plastic lenses. SU-8 polymer with a refractive index of 1.569 measured at 1550 nm in a Sairon Technology SPA-4000 Prism Coupler was spin-coated onto the CR-39 sheet and patterned using the contact photolithography technique, forming the core of a channel waveguide structure. The measured height of the channel waveguide was 5.0 ( ± 0.1) μm and its width ranged between 10 and 15 μm over several different samples, with waveguide width variation of less then 0.2 μm over the length of a given waveguide section.

The GO was prepared by an improved version of Hummer’s method (see supplementary) [16]. 0.5 µl of GO solution containing GO flakes was carefully dispensed – using a micro-pipette – onto the waveguide channel and then allowed to dry under ambient conditions, before the next GO drop was applied in the same position. Surface mapping of the GO-coated polymer channel waveguide was carried out using a DEKTAK D150 surface profiler. It has been reported that the dielectric function of GO may be strongly modified by the presence of intercalated water (H₂O) [15, 17], implying susceptibility to variation related to ambient humidity. In the present work, the sample was covered by NOA-65 UV-sensitive resin (Norland Co. Ltd.) – which has a refractive index of 1.52 – as an immediate overcladding of the channel waveguide and the GO covered regions. This encapsulation provides both mechanical protection and, we anticipate, a major reduction in the susceptibility of the deposited GO to the possible effects of atmospheric water vapour.

The GO coated samples were then diced and polished into 1 cm long waveguide channels. The fibre butt-coupling technique was used to measure the insertion loss of both the TE- and TM-polarised modes of the GO-coated waveguide channels. Guided light polarisation in the launch fibre was controlled using a fibre polarisation controller, and the polarisation state of the GO polariser output was measured in free space using a polarimeter (Thorlabs PAX 5710) before the output was fibre-coupled, in order to measure the insertion loss. Although the SU-8 waveguide core cross-section is large enough for several characteristic modes to be available for both (quasi-)TE- and (quasi-)TM-modes, the alignment of the set-up guaranteed that, to a good approximation, only the fundamental guided mode for each polarisation was launched. In addition, extra care in the handling of the launch fibre was required, since the linear polarisation state could easily be scrambled by small movements or vibration. The polarisation-dependent loss of the uncoated polymer waveguide was measured first - and was found to be lower than 0.5 dB, limited by the performance of the measurement setup. For measurements across the broad frequency range, fibres with different cut-off wavelengths were used, in order to maintain the launching polarisation state and single waveguide-mode excitation conditions.

3. Physical properties

A schematic diagram of the proposed GO waveguide polariser is illustrated in Fig. 1(a). The GO multi-layer was deposited on to the polymer waveguide channel using the drop-casting method. As a progressively larger number of GO solution drops were applied, the GO film became thicker, as shown in Fig. 1(b), where the transparency of the GO coating decreased - while the coating diameter remained about the same. The surface morphology of the channel waveguide coated with a single drop of deposited GO solution is shown in Fig. 1(c). The average thickness of the GO film for each solution drop-casting step was ~0.5 μm. It can be seen that the GO film showed similar height coverage, both on the channel waveguide and its surroundings. This coating behaviour has implications for the overall effect of the number of GO solution drops applied during the drop-casting process and the extinction ratio of the proposed waveguide polariser, as will be discussed in the following section. Figure 1(d) shows a Scanning Electron Microscope (SEM) image of a GO-coated waveguide channel. This figure indicates that the drop-cast GO forms a continuous layer across the coating region. Wedge-shaped air gaps are also observed on either side of the waveguide channel sidewall, indicating the formation of the GO film on top of the waveguide channel during the droplet drying process before the layer formed has “collapsed” onto the waveguide channel, at the end of the drying process. This behaviour also explains the elevated ridges observed on the GO film in Fig. 1(c), which are wrinkles resulting from local folding of the GO film during “collapse”. A close-up examination of the GO film formed with 2 drops of deposited GO solutions in Fig. 1(e) shows an orderly layered GO stack, indicating the formation of GO ‘paper’ [14]. The spacing between the GO layers is calculated from the XRD result shown in Fig. 1(f) and is about 1.03 nm. This interlayer spacing indicates the presence of a large number of water molecules intercalated in the GO film, as reported by Lerf [18] and Buchsteiner [17]. No obvious gap was observed between the first and second depositions of the GO film. All these features indicate a clear distinction between the presently considered GO waveguide polariser and the graphene-based waveguide polarisers described previously [10, 12], since the number of GO layers involved is much larger than the number of graphene layers required to support plasmonic wave propagation. The GO multi-layer of the present work, in contrast, may be characterised as an anisotropically lossy dielectric overlay. It has a sufficiently large refractive index value to produce a significant redistribution of the modal fields compared with the modal fields of the uncoated polymer waveguide.

 

Fig. 1 Graphene Oxide waveguide polariser. a) Schematic diagram of the proposed GO waveguide polariser structure. b) Top view of drop-cast GO coating with different number of solution drops. The diameter of the coating remained similar while the thickness of the coating increased with the number of solution drops applied, as indicated by the increase in opacity of the coating region. c) Surface mapping of drop-cast GO coating (2 solution drop) on polymer waveguide channel, showing similar height coverage of GO coating on the waveguide channel and region surrounding the waveguide channel. Ridges across the GO coating region are the result of local folding of the GO film formed after the drying process. d) SEM image of a GO-coated waveguide channel, in which air gaps are observed between the sidewalls of the waveguide and the GO film. e) Orderly stacking of drop-cast GO layers, indicating the formation of GO ‘paper’ using the drop-casting method. f) XRD spectrum of the GO ‘paper’ indicates an interlayer gap of about 1.03 nm.

Download Full Size | PDF

4. Simulation and modelling

Based on information from the SEM images and from the literature, the field distributions of the orthogonally polarised modes were simulated using Finite Element Method (FEM) computation. Figure 2(a) illustrates the cross-section of the GO waveguide polariser used in the simulation, with air gaps between the GO film and waveguide channel sidewall being taken into consideration. The typical waveguide cross-sectional dimensions used were a rectangular cross section 5 μm high and 10 μm wide. The GO film thickness was varied in direct relation to the number of GO solution drops being applied. An air gap with a maximum width at the base of 0.5 μm was inserted at each side of the waveguide channel sidewall. A simplifying assumption made in the simulation is that the GO film thickness is uniform across the entire coating region of interest. Figure 2(b) shows the modal distributions for both the TE- and TM-polarised modes, with the GO film thickness set at 2 μm (Because of the finite width of the waveguide core, its guided modes should strictly be labelled as, respectively, quasi-TE and quasi-TM). It can be seen that a fraction of the modal field was coupled from the waveguide channel into the GO film above the waveguide. Due to the highly anisotropic complex dielectric function of the GO coating, the orthogonal modes will ‘see’ the GO film very differently. A (quasi-)TM-polarised mode will see the GO film as a simple and relatively low-loss dielectric medium - and low loss propagation is therefore possible in this layer. On the other hand, a (quasi-)TE-polarised mode will see a GO film with a relatively large (optical frequency) conductivity value, giving strong damping effects and therefore high-loss propagation in the GO film. Note that, despite the existence of the wedge-shaped air gaps between the GO film and the waveguide channel sidewall, only a small portion of the light in the waveguide is coupled into the GO multi-layer across the air gaps, as shown in the modal distributions of both the TE- and TM-polarised modes in Fig. 2(b). In this case, the (quasi-)TM-polarised mode will experience high-loss propagation while the (quasi-)TE-polarised mode will see the GO film as a relatively low-loss dielectric medium.

 

Fig. 2 Simulation of GO waveguide polariser. a) Illustration of GO-coated waveguide channel cross section. The waveguide channel cross section has a dimension of 5 μm in height and 10 μm in widths. GO coating thickness depends on the number of GO solution applied during drop casting. The maximum air gap between the waveguide channel sidewall and the GO film is 0.5 μm at the base of the channel. b) The simulated modal electric field component distributions in the waveguide channel for TE- and TM-polarised light show guiding of TM-polarised field in the GO film, while the TE-polarised light coupled into the GO film is highly damped. The GO thickness used was 2.0 μm. The wavelength used in the simulation is 1550 nm.

Download Full Size | PDF

For the simulation of the guided-light propagation in the GO-coated polymer waveguide, our choice of values for the real and imaginary parts of the anisotropic complex refractive index of the GO multi-layer is consistent with values published in the literature. For the component of the optical electric-field that is normal to the GO planes, we have assumed that the imaginary part of the refractive index may be neglected. This choice is justified by the fact that electron transport in this direction will be substantially blocked, on the atomic scale, by the barriers between nearest-neighbour graphene layers formed by layers of oxygen atoms – as well as atoms of any other element, e.g. nitrogen, that may be present. The real part of the refractive index has been set to 2, corresponding to a value of 4 for the real part of the dielectric function. This choice of n = 2 is in reasonable conformity with the values (~1.9) given by Vaupel and Stobel [19] and is an intermediate value by comparison with refractive index values that would be obtained from the low-frequency (‘quasi-dc) values for the (relative) dielectric function that are given by Loh et al [20].

For the optical electric field component that is parallel to the GO layers – and therefore to the atomic sheets of graphene – the situation is substantially different. Because of the very sub-wavelength scale of the individual GO layers, an effective (average) medium approach is appropriate. For the real part of the refractive index, we have chosen to use the same value as for the perpendicular direction, but the choice for the imaginary part reflects the fact that the graphene atomic sheets, even though they are fully (or almost fully) oxidised, are still capable of significant levels of conduction. Choice of an optical-frequency conductance value of σ = 2700 S.m⁻¹ implies a significant absorption coefficient for TE-polarised light. This value for the conductance is two orders of magnitude lower than that given by the usual Kubo expression for pure graphene (~1.8 x 10⁵ S.m⁻¹) – and it is also less than the value of 9.0 x 10⁴ S.m⁻¹ that applies for THz frequencies in reduced graphene oxide (RGO) [21]. With these choices for the refractive indices, the effective refractive index of the guided mode is obtained via simulation. By assuming a uniform coating of GO film, the optical propagation loss of the TE- and TM-polarised modes can be calculated using the Beer-Lambert law. Close agreement between simulation and experiment is obtained, as shown in Fig. 3. Our specific choices for the complex refractive indices should be taken as being indicative, rather than absolute. They are physically reasonable and are sufficiently accurate for identification of the characteristic behaviour that is found in our experimental measurements.

 

Fig. 3 Polarising effect of GO-coated waveguides. a) Polar plot of the GO waveguide polariser measured at 1550 nm. The insertion loss is lowest when the incident light is highly TM-polarised, indicating a TM-pass waveguide polariser. b) Insertion loss of TE- and TM-polarised light of waveguide polarisers coated with different number of GO solution drops corresponding to different GO film thickness. Both values are measurement averages in the 1530 to 1630 nm wavelength range. c) Top view of GO coated waveguide channel with TM- and TE- polarised light (650 nm) propagating through the waveguide. There was no observable scattering in the case of TE-polarised light, while the observed scattering in the case of TM-polarised light propagation was highly p-polarised. d) Broadband response of GO waveguide polariser, where an extinction ratio of more than 20 dB is measured from 1250 – 1640 nm.

Download Full Size | PDF

5. Optical properties and performance

The performance of the GO waveguide polariser was measured and the results are shown in Fig. 3. Figure 3(a) shows a polar plot of the measured output light from the GO waveguide polariser, as a function of the input polarisation. The output of the GO waveguide polariser is characteristically highly TM-polarised - which is similar to the observation reported by Kim et. al. [11] and indicates that the waveguide modal distribution changes substantially in the GO coated waveguide section. Figure 3(b) shows the insertion loss for the TM- and TE-polarised modes, for different GO film thicknesses determined by the number of GO solution drops applied during drop-casting. Both values are measurement averages for wavelengths in the range between 1530 and 1630 nm. The total variation of the extinction ratio over this wavelength range is 4.5 dB. The insertion loss for the TM mode increased with the number of drops of GO solution applied – and became constant after deposition of three GO drops, at approximately 6.5 dB. As the number of GO drops applied increases, there is a correspondingly rapid increase in the TE-mode propagation loss, which continues to increase steadily and reaches a maximum value for coating with five GO drops, corresponding to a GO film thickness of 2.5 μm.

The efficiency of the coupling of light into the drop-cast GO multi-layer is likely to be dependent on the thickness of the GO coating on the top channel waveguide surface. At the start of the drop-casting of the multiple GO layers, the initial GO coating thickness is small and it can only support a small part of the total propagating modal power. In this case, a large fraction of the modal power for both TE- and TM-polarised waves remains in the polymer waveguide. As additional drops of GO solution are applied, the GO coating thickness increases and our observations indicate that a point is reached where the mode coupling between the waveguide and the GO multi-layer is most efficient. This point is characterized by a sharp increase in the TE mode propagation loss (GO film thickness larger than 1.5 μm), as shown in Fig. 3(b). At the same point, the TM mode shows a partial but significant redistribution of modal power from the polymer core into the GO coating. The TM mode propagation was verified by observing the polarisation of the light leaking out from the top surface of the GO coated channel waveguide when a 650 nm wavelength laser was coupled into the channel waveguide, as shown in Fig. 3(c). Using a polarisation analyser placed between the waveguide and the microscope, the light leaking out from the top surface of the waveguide was found to be in the p-polarisation state (i.e. parallel to the plane of incidence – the plane defined by the waveguide axis and the direction of the microscope observation). This measurement indicates that TM-polarised light is coupled into and propagates partially in the GO film with relatively small propagation losses, while TE-polarised light experiences a much larger propagation loss in the GO film.

Addition of more GO solution drops did not change the size of the drop-casting area and – therefore – it should increase the number of GO layers that is observed as the film thickness increases. The addition of further drops of the GO solution beyond a certain point (typically three solution drops) does not measurably increase the propagation loss of the TM-mode, since the additional layers of GO coating are progressively further from the waveguide-GO interface and therefore interact less strongly with the light propagating in the channel waveguide.

The polarisation extinction ratio for wavelengths ranging from 650 nm to 1640 nm has been measured and is shown in Fig. 3(d). It shows a gradual decrease in value towards shorter wavelengths. The extinction ratios at wavelengths of 1590 nm, 1310 nm, 980 nm and 650 nm were found to be 40.0 dB, 25.0 dB, 16.0 dB and 8.5 dB, respectively. Though steps have been taken to ensure only fundamental mode propagation in the GO waveguide polariser, we believe that partial excitation of higher order modes of the waveguide at shorter wavelengths is quite probable and that their presence would reduce the effective strength of the interaction between the polymer waveguide and GO film, resulting in a reduced extinction ratio between the TE- and TM-polarised modes. Nevertheless, an extinction ratio of more than 20 dB has been measured over the entire wavelength range from 1250 nm to 1640 nm. The extinction ratio for wavelengths longer than 1640 nm was not measured, due to the unavailability of a suitable laser source.

The ability to introduce GO coating using the drop-casting method provides a simple and effective means for waveguide polariser fabrication. The performance of the GO waveguide polariser can be improved using the coat-and-etch method demonstrated by Kim et. al. [12], or (for example) a laser ablated waveguide [22], where the GO film will only be coated on top of the channel waveguide. In addition, this method also enables the flexibility of selective-areas drop-casting on an optical circuit intended for the polarisation function, without the need for a physical mask during deposition or mechanical transfer of the graphene layers. With the use of an automated micropipette positioning and dispensing process, accurately localized deposition of GO coatings could be achieved in volume production.

6. Conclusions

A broadband waveguide polariser with high extinction ratio has been demonstrated in a polymer waveguide coated with GO film deposited using the drop-casting method. Drop-casting of GO solution results in a substantially ordered GO layer stack, with its thickness increasing as the number of GO droplets applied increases. The polarisation effect of the GO waveguide polariser has been shown to be a result of the anisotropic complex dielectric function of GO film. The extinction ratio of the GO polariser is dependent on the GO film thickness. The average extinction ratio is 38 dB between 1530 nm and 1630 nm, with the highest extinction ratio of 40dB measured at 1590 nm and the lowest value of 35.5 dB measured at 1530 nm. The extinction ratio is achieved with only ~1.3 mm of GO coating length along the propagation direction. The short interaction length required to produce a high extinction ratio over a broad fibre-telecom wavelength band will provide a solution for the integrated waveguide polarisers required in applications that include optical Lab-on-Chip and photonic integrated circuits.