Dielectric microring and disk resonators are key building blocks of integrated optical systems. They have been extensively studied theoretically [1,2] and demonstrated later experimentally [3]. These structures can indeed be used as channel-drop filters, all-optical logic gates, optical frequency division multiplexers, and light sources [4,5]. Moreover, resonators find applications in sensing [6].

The resonator-based surface plasmon-resonance (SPR) sensors, which use surface plasmon waves to probe the interactions between molecules and sensor surfaces, have been studied rigorously in the past for optical detection of small chemical and biological entities. These include the low propagation loss channel plasmon subwavelength waveguide ring resonators [7], nanoscale rectangular resonators utilizing zero sharp bend loss, metal-insulator-metal waveguides [8], and plasmonic ring resonators with arrays of silver nanorods as waveguides [9]. Efforts are placed on the improvement of the quality factor $Q$ of the structure, which is one of the main measures of the efficiency of the ring/disk resonators. For example, dielectric-loaded plasmonic waveguide ring resonators have been introduced, where dielectric stripes deposited on a smooth metal film are used as waveguides [10]. A high-$Q$ plasmonic cavity composed of a silica core, where the light is trapped inside the thick disk, has been realized [11]. Finally, ring resonators with hybrid waveguides consisting of a dielectric nanowire on a silver layer have been studied [12]. The advantage of SPR is the strong confinement of the electric field. However, because of the high absorption losses of metals, SPR sensors suffer from a very low quality factor, which limits the sensitivity of the devices.

Similar to surface plasmon polaritons (SPPs), another well-known surface wave, called the Bloch surface wave (BSW), can be efficiently used as a probe in sensing. BSWs are surface electromagnetic modes, which can be excited in the photonic bandgap of truncated dielectric periodic multilayers [13]. In comparison to metallic layers used for SPPs, dielectric multilayers for BSWs have several advantages. Due to low losses in dielectric materials, BSWs achieve longer propagation lengths. Another advantage is the scalability. The multilayers are wavelength scalable and can be designed to sustain BSWs at a broad range of wavelengths from near-UV to IR [14]. Since the position of the surface mode within the forbidden band can be engineered by tailoring an additional layer on top of the multilayer, it is possible to tune the maximum intensity associated with the BSW on the surface. This ability to tune the localized field confinement is very appealing for sensing [15]. Therefore, BSWs have been investigated extensively for sensing applications [6,14,16,17].

In this Letter, we utilize this low loss dielectric platform for the realization of BSW-based resonators. These resonator devices take advantage of the light confinement on the surface and, hence, achieve high sensitivity for the surface environment. One advantage of BSW-based resonators is the high quality factor, in comparison to plasmonic-based resonators exhibiting quality factors on the order of ${10}^2$ [9,10,12]. This is mainly due to the low loss characteristic of BSWs.

The multilayer platform consists of a dielectric stack periodically alternating high and low refractive index materials. It is equivalent to a one-dimensional Bragg mirror. The complete structure is made up of six periods of silicon dioxide (${\mathrm{SiO}}_2$) and silicon nitride (${\mathrm{SiN}}_x$) having a refractive index of 1.45 and 1.79 at $\lambda =1550\mathrm{nm}$ and thicknesses of 472 and 283 nm, respectively. The multilayer stack fulfills the Bragg condition. The periodicity of the multilayer is terminated by a 50 nm thick top layer of ${\mathrm{SiN}}_x$. The thickness of the top layer leads the field profile of the BSW mode inside the multilayer [18]. The whole pattern, called bare multilayer (BML), is deposited on a glass wafer $(n_g=1.50)$, as shown in Fig. 1. The BML exhibits a photonic bandgap in the telecom wavelength range and can be excited with TE and/or TM polarizations. In this Letter, the multilayers are designed for TE polarization. A 60 nm additional layer of titanium dioxide (${\mathrm{TiO}}_2$) is deposited on top of the BML, which has a high refractive index of 2.23 at the wavelength of 1550 nm and is transparent in the visible and near-infrared (IR) regions of the light. The refractive index contrast $(\mathrm{\Delta }n)$ defines as the difference of the effective refractive index of the BSW with an additional layer (effective $n_{\mathrm{TiO}2}$) and without an additional layer (effective $n_{\mathrm{BML}}$) [19]. The higher value of $\mathrm{\Delta }n$ provides stronger in-plane confinement of the light inside the 2D structures. The additional layer serves as an efficient device layer allowing the lithographical fabrication of compact and low loss integrated two-dimensional (2D) photonic devices, for example, the disk resonators in this layer.

 

Fig. 1. Schematic of a 2D disk resonator designed on top of the dielectric multilayer platform. The disk resonator has been patterned into a 60 nm thick layer of ${\mathrm{TiO}}_2$. The BSW coupling zone is shown in red, which is 100 μm away from the gap area. A SNOM probe in collection mode is used for the near-field characterization of the disk resonator.

Download Full Size | PDF

To couple BSWs, a total internal reflection (TIR) configuration has been used to match the propagation constant of BSWs, the details of which are provided elsewhere [19].

Because of the evanescent nature of BSWs, near-field optical microscopy is the most suitable tool for optical characterization. Therefore, to study the characteristics of our device, we use a multi-heterodyne scanning near-field optical microscope (MH-SNOM) which collects the evanescent surface waves with a subwavelength aperture probe [20].

The BML platform is fabricated using standard plasma-enhanced chemical vapor deposition, while atomic layer deposition (ALD) was used to create the last titanium dioxide thin film. The accuracy in thickness, the conformal of the film, and the quality of the material possible with the ALD technique is ideal for nanophotonic structures [21]. As previously demonstrated, the last layer has a crucial importance in the confinement and propagation length of the BSW [18], making the precision of the deposition of this last layer a key point of the fabrication process. ${\mathrm{TiO}}_2$ has been deposited using titanium tetrachloride and water as precursors at a process temperature of 120°C. At this temperature, the deposited material is amorphous, reducing the risks of intrinsic scattering and, thus, reducing losses. The deposition rate is 0.07 nm/cycle, which explains the high accuracy of the layer thickness.

To realize the final etching of the disk resonator pattern, a chromium (Cr) mask has been used. The Cr layer was evaporated (LAB18 by Kurt. J. Lesker), and a resist (HSQ) layer was spun on the deposited multilayer. The structures were patterned to the resist by an electron beam (EBL, Vistec EBPG $5000+$ ES HR). After that, the development of the resist was done in a solution of $\mathrm{AZ}351\text{:}{\mathrm{H}}_2\mathrm{O}(1:3)$ developer. The mask was created by etching the Cr layer by an inductively coupled plasma reactive ion etching process (ICP-RIE, PlasmaLab 100 by Oxford). The ICP-RIE process involved chlorine (${\mathrm{Cl}}_2$) and oxygen (${\mathrm{O}}_2$) as process gases. Finally, the thin ${\mathrm{TiO}}_2$ layer was etched by the ICP-RIE process (PlasmaLab 80 by Oxford) with sulfur hexafluoride (${\mathrm{SF}}_6$) and argon (Ar) as etching gases. The remaining Cr was finally removed by wet etching.

A 2D micro-disk resonator is patterned into a 60 nm thick additional layer of ${\mathrm{TiO}}_2$. The thickness of the additional layer is chosen to compromise between the refractive index contrast and the propagation length with the present configuration and material. Note that the value of the refractive index contrast is not a limitation. The value of the refractive index corresponds to the difference of the effective indices of the resonator and its surrounding.

A disk resonator is introduced because it has fewer losses, compared with a ring resonator. As in disk resonators, the light interacts only with one sidewall (in comparison with ring resonators), and the travelling mode suffers fewer scattering losses from the sidewall roughness. As a result, we expect that the quality factor of a disk resonator should be higher than its ring resonator counterpart. In optical resonators with a high refractive index contrast between the cavity and the ambient (like silicon photonics), resonators with a smaller bending radius and high quality factors can be realized [22]. In this case, the ring resonators have the advantage of restraining higher-order radial modes, because the electromagnetic field propagating inside the cavity is circumscribed by two sidewalls of the ring, and higher-order radial modes are pushed to interact severely with the inner sidewall of the ring and forcibly radiate [23].

Since, in BSW-based optical resonators, the refractive index contrast is much lower than in silicon photonics, the radius of the resonator needs to be much larger. For the same reason, higher-order radial modes are strongly radiative and lossy. As a result, it is expected that a disk resonator, patterned into ${\mathrm{TiO}}_2$ layer, with a radius of 100 μm does not severely suffer from higher-order radial modes. The experimental results presented in the near-field characterization section verify this fact.

The disk resonator consists of a disk of radius $R$ coupled to two straight bus waveguides of width $w$, as shown in Fig. 1. The distance separating the disk and the waveguides is defined as the gap $g$. For the present design, a disk radius $R=100\mathrm{\mu m}$, a bus waveguide width $w=2\mathrm{\mu m}$, and a gap $g=500\mathrm{nm}$ are considered. The value of the measured effective refractive index contrast $\mathrm{\Delta }n(n_{\mathrm{TiO}2}-n_{\mathrm{BML}})$ lies between 0.1–0.2, which is sufficient to guide the BSW inside the disk of the bending radius 100 μm without significant bending loss [24]. The platform and the resonator consist of dielectric materials with low absorption in the near-IR wavelength range. However, BSWs may leak out of the multilayer because of the presence of the prism [25], which is the predominant cause of the damping of the surface mode.

The incident beam illuminates the multilayers through TIR configuration at an oblique incident angle which is higher than the critical angle, 58.26 deg. At this chosen angle, the incident beam is coupled to the BSW for a 60 nm thick of ${\mathrm{TiO}}_2$ layer. Further, the generated BSW itself is coupled to the first waveguide (input port). We keep the BSW coupling zone 100 μm away from the coupling area between the waveguide and the disk (see Fig. 1). This allows the incident light to propagate some distance to arise as a BSW mode, to increase the signal-to-noise ratio. The BSW propagates through the waveguide and couples to the disk at the coupling area, evanescently through the gap. Constructive interferences occur when the optical path of a round-trip of the disk resonator is an integer multiple of the wavelength and, hence, light builds inside the disk. This state, $2\pi R*n_{\text{eff}}=m\lambda$, is called on-resonance, which yields a dip in the signal collected at the through port. In the above equation, $\lambda$ is the BSW wavelength, $n_{\text{eff}}$ is the effective index of the mode propagating inside the cavity, and $m$ is the azimuthal mode number.

The field amplitude distribution over the disk resonator is detected with a MH-SNOM in collection mode, where the subwavelength aperture probe is connected to a single-mode fiber. To measure the transmission spectrum at the through port, we performed a scan, in wavelength and space, across the bus waveguide ($x$-direction) at a distance around 100 μm after the coupling area along the $y$-direction; see Fig. 1.

The near-field SNOM image of the field amplitude distribution over the cross section of the waveguide for different wavelengths is shown in Fig. 2(a). The range of wavelengths spans from 1551 to 1561 nm which covers five times the free spectral range (FSR) of the resonating BSW mode. As described for the through port, the field amplitude distribution at the periphery of the disk has been measured, which presents the azimuthal mode profile of the resonating wave along a 6 μm length of disk periphery; see Fig. 2(b). Measuring the field amplitude distribution is only possible with near-field measurement. The transmission spectrum of the through port and the trapped light inside the disk is obtained by plotting the cross section of the field amplitude at a particular position of the through port and periphery of the disk with respect to the wavelength; see Fig. 2(c). The amplitude is normalized with the field amplitude measured at the input port. It can be seen that the dips in the field amplitude at the resonance wavelengths at the output port appears as peaks at the same wavelengths at the periphery of the disk.

 

Fig. 2. Near-field image of a wavelength scan taken by MH-SNOM (a) at the through port and (b) at the periphery of the disk. The scanning is performed over 6 μm across the waveguide in the $x$ direction. (c) Normalized transmission spectrum at the through port and the periphery of the disk as a function of the wavelength. The field amplitude is normalized with respect to the amplitude at the input port. The scanned areas are shown in Fig. 1 (dotted rectangular area).

Download Full Size | PDF

After deducing on- and off-resonance wavelengths from the measured wavelength scan at the output port, we mapped the field distribution over the complete waveguide and disk area in space (Fig. 3). For the fine positioning of the SNOM probe, we used a piezo stage which allows the maximum lateral displacement scan of $100\times 100\mathrm{\mu m}$. Therefore, the mapping of a larger area, $250\times 250\mathrm{\mu m}$, is performed by a series of several spatially overlapping sub-maps stitched together with an error margin. To image the different sub-maps, the coarse positioning of the SNOM probe has been performed using a motorized stage. Figure 3(a), for instance, is composed of nine sub-maps. The near-field distribution over the complete disk at the on-resonance condition $(\lambda =1555.4\mathrm{nm})$ is shown in Fig. 3(a). The pixel size is $1{\mathrm{\mu m}}^2$.

 

Fig. 3. SNOM images of field amplitude distribution (a) over the complete structure of the disk resonator at on-resonance at $\lambda =1555.4\mathrm{nm}$. (b, c) Near the waveguide and disk coupling area at off-resonance $(\lambda =1554.5\mathrm{nm})$ and at on-resonance, respectively.

Download Full Size | PDF

One can observe the expected response of a disk resonator: a portion of the field is passing straight to the through port, indicating that it does not operate at critical coupling. This is the reason for the low extinction ratio. However, an important part of the field is still coupled to the disk and coupled out to the drop port, and no light can be observed in the add port. In the background, one can see a focusing effect of the part of light which is not coupled to the bus waveguide. This focusing comes from the refractive index change and the shape induced by the disk, as demonstrated in [19]. We indicate the near-field distribution around the coupling area between the waveguide and the disk for off-resonance condition at $\lambda =1554.5\mathrm{nm}$ and compare it with the corresponding scan at on-resonance; see Figs. 3(b) and 3(c), respectively. One can see the propagation of the BSW in the through port at off-resonance while it is coupled to the disk at on-resonance. However, at on-resonance, the BSW mode does not couple critically to the disk. The measured $Q$-factor is approximately $2\times {10}^3$, which is deduced from the transmission spectrum of the through port $(Q={\lambda }_0/\mathrm{\Delta }\lambda )$. We perform 2D simulations using the effective refractive index, based on the finite-difference time-domain (FDTD) method [26]. For a disk radius of 100 μm, the calculated $Q$-factor, is $2\times {10}^4$. In the experiments, the scattering losses due to fabrication imperfections and leakage losses, because of prism coupling, are the main reasons for the degradation of the $Q$-factor by an order of magnitude. Nevertheless, the $Q$-factor of our BSW resonator remains higher than a classical plasmonic-based resonator $(Q=122)$ [9]. It has been demonstrated in theory that a very high $Q$-factor is achievable for BSW-based resonators [27].

In conclusion, for the first time, to the best of our knowledge, a 2D disk resonator has been demonstrated experimentally on a BSW dielectric multilayer platform. A thin layer $(\lambda /25)$ of ${\mathrm{TiO}}_2$ has been used as a high refractive index additional layer on the top of a dielectric periodic stack to form the 2D disk resonator. To observe the near-field interaction of the BSW with the resonator, we mapped the field amplitude distribution images with a MH-SNOM. The results show that higher quality factor BSW resonators with respect to plasmonic resonators are achievable, principally because of the low loss dielectric multilayer platform. For a disk radius of 100 μm, the measured $Q$-factor was $2\times {10}^3$. The scattering and leakage are the main loss mechanisms in the present system. The quality factor can be further improved using grating couplers. With the present design, we compromise the footprint of disk resonators in terms of high $Q$-factors. Resonators with smaller dimensions can be realized by using higher refractive index materials as an additional layer, e.g., silicon. Further efforts are ongoing to improve the design to achieve higher extinction ratios and to reduce the size.