3.1. The Coupling Process from the InP Waveguide to the Microring
The 3D vertical microring coupler, presented in
Figure 1b, is structurally similar to a typical add-drop coupler configuration on a 2D plane involving one ring and two bus waveguides. Because of the self-rolled-up process of the SiN
x bilayer, the resulting 3D structure can be more precisely described as a spiral with all layers attached together. This structure can be approximated by a circular ring as long as the total thickness of the spiral structure is much less than its diameter [
14],
, where
is the winding number of the self-rolled-up tube,
is the thickness of the original bilayer
SiNx membrane, and
is the diameter of the V
μRC. This condition is satisfied for all the parameters chosen in the following simulations. This simplification allows for easier modeling and characterization of the device’s optical properties, such as its resonant modes and coupling behavior. In this add-drop coupler configuration, assuming light is injected into the device through Port A using an integrated InP laser, it may subsequently be coupled from Port C to the silicon photonics circuitry via the V
μRC.
For the study of the above model, we use the two-dimensional Wave Optics module in the commercial COMSOL Multiphysics simulation software based on the finite element method for analysis and research (
https://www.comsol.jp/). Among the indices, the effective refractive indices of
a-Si, SiN
x, SOG (Spin-on-Glass), and InP are 2.44, 1.9, 1.5, and 3.2, respectively, which are all obtained from measurements taken in previous experiments [
17]. The inner diameter of the microring is taken to be 6.5 μm in accordance with the experimental results presented in [
15,
17,
18], which can be adjusted by the differential stress within the SiN
x bilayer. The spectral position of the coupling wavelength is controlled by the structural parameters of the V
μRC. In
Figure 2a, we present the transmission spectra of three V
μRCs, whose sidewall thickness
a1 = 350, 370, and 390 nm. One may notice that when
a1 is 370 nm, the two resonant peaks of the V
μRC align precisely within the communication range of the S-band and C-Band, and the FSR of the V
μRC is determined to be 52 nm, which matches very well with our previous experimental observation [
17]. Using the same model, we calculated the transmission spectra of a series of V
μRCs with different total thicknesses of the SiN
x layer,
a1, shown in
Figure 2b. This result is consistent with what we observed experimentally [
18].
It is widely recognized that the free spectral region (FSR) of a microring cavity primarily depends on factors such as wavelength, group refractive index, and microring radius. This relationship can be expressed as follows [
25]:
where
ng is the group refractive index and
L is the length of the microcavity. In the case of a 3D inter-chip photonic coupling system using the wavelength division multiplexing (WDM) scheme, it is preferrable that the resonant coupling wavelength for each V
μRC can be easily distinguished to avoid extra coupling noise from the adjacent channel. Thus, a larger FSR of the V
μRC’s resonant peaks is desired. As shown in
Figure 2a, as the effective diameter decreases, the radiation loss increases, which causes the peaks to become broader and shallower [
18]. This relationship is validated in our simulations, as shown in
Figure 2b. It is evident that reducing the thickness of SiN
x thin films and decreasing the effective diameter leads to an increase in the 3 dB bandwidth of the coupling peak and a decrease in its peak intensity.
We have demonstrated experimentally that depositing a high-refractive-index thin silicon layer on the inner side of the silicon nitride microring results in increased axial confinement. The high-index strip within the microtube increases the effective refractive index locally. Therefore, the refractive index profile along the axial direction exhibits discontinuity similar to a step-index waveguide, which will help to confine the electromagnetic field within the high-refractive-index region and effectively form a vertical microring instead of a tube. In our previous experiment, we successfully achieved this by depositing a 21 nm
a-Si thin film inside a V
μRC [
17] and observed a greatly increased FSR. However, the thickness of this silicon layer (
a2) also affects the radius of the V
μRC, which, in turn, is related to the transmittance of the device. Therefore, in this study, we vary the thickness of the
a-Si layer around the experimentally obtained value and demonstrate the impact of
a2 on the transmission of the V
μRC at Port B when
a1 is set to 370 nm, as shown in
Figure 3. An increase in the thickness of the silicon layer results in a slight redshift of the coupling peak of the V
μRC. Based on the simulations, we identify
a2 = 25 nm as the optimal parameter for the next phase of our research. This value results in an increased effective diameter of the microring, leading to a transmission spectrum with the deepest and narrowest coupling peak. However, it is important to consider the self-rolled-up phenomenon caused by the internal stress difference in the silicon nitride bilayer. If the
a-Si layer becomes excessively thick, it can pose challenges in terms of curling. Therefore, selecting
a2 = 25 nm is a suitable choice, striking a balance between achieving the desired transmission spectrum characteristics and avoiding difficulties associated with curling caused by excessive thickness.
In the two-dimensional V
μRC model depicted in
Figure 1b, the thickness of the upper InP waveguide (
h1) plays a significant role in determining the coupling coefficient (
k1) between the InP waveguide and the microring. This, in turn, affects the transmission at Ports B and C.
Figure 4a illustrates the relationship between the transmission at Port B and the thickness of the InP waveguide. When
h1 is set to 80 nm, clear resonant dips in the transmission spectrum at Port B are observed at 1491.5 nm and 1544.4 nm with the smallest 3 dB bandwidth. To analyze the coupling coefficient
k1 between the InP waveguide and the V
μRC in this scenario, we establish a model as presented in
Figure 4c. The results of the coupling coefficient
k1 at different values of
h1 are obtained and depicted in
Figure 4b. For the optimal coefficient
h1 = 80 nm obtained in
Figure 4a,
k1 at 1491.5 nm and 1544.4 nm are 0.46 and 0.5, respectively, and the electric field distribution at the wavelength of 1544.4 nm is presented in
Figure 4c. To calculate the coupling coefficient
k1, we select a half-ring waveguide configuration to inhibit light from resonating within the ring resonator. In modeling this setup, the structure is fully enclosed by Perfectly Matched Layers (PMLs) to eliminate reflections, with both ends of the half-ring waveguide directly interfacing with the PMLs. This arrangement ensures efficient coupling from Port 1 to Port 2 across the waveguide and the circular ring while effectively preventing light reflection at the waveguide’s ports. Due to the higher polarization efficiency of TE mode for microtubes, TE-polarized light is used in the model.
Apart from the thickness of the InP waveguide, the coupling distance between the InP waveguide and the microring also plays a pivotal role in determining the transmission spectrum. To investigate this, we vary the coupling distance (
d1) between the InP waveguide and the V
μRC while keeping
a1 = 370,
a2 = 25, and
h1 = 80 nm constant. As shown in
Figure 5a,b, the transmission at Port B and the corresponding coupling coefficient
k1 of V
μRC are simulated and plotted for different
d1 values. Notably, as the coupling distance increases, the entire transmission spectrum of Port B redshifts, and the coupling coefficient
k1 exhibits a trend of first increasing and then decreasing. The maximum coupling coefficient between the InP waveguide and the microcavity is found at
d1 = 160 nm. Specifically, when
d1 = 160 nm, the coupling coefficients at the peak wavelengths of 1491.57 and 1544.4 nm was 0.46 and 0.5, respectively. Collectively, these results suggest that for the coupling process from the InP waveguide to the microring, the optimal parameters are a thickness of 370 nm for the SiN
x layer, 25 nm for the silicon layer, and 80 nm for the upper InP waveguide and a coupling distance of 160 nm. Under these conditions, we can obtain the highest coupling efficiency from the InP waveguide to the V
μRC.
3.2. The Coupling Process from the Microring to the SiNx Waveguide
In the aforementioned findings, we have successfully identified the optimal parameters for the coupling process from the InP waveguide to the V
μRC. Indeed, the thickness of the silicon nitride waveguide (
h2) also plays a crucial role in determining the transmission spectrum of Port C, as depicted in
Figure 1. By varying the thickness of
h2 within the range of 400 to 500 nm, we can observe changes in the resonance peak of the transmission spectrum.
Figure 6a displays the simulated transmission spectrum at Port C for different
h2 values. As the thickness of
h2 increases, the resonance peak of the transmission spectrum redshifts. In this case, the maximum peak is observed at
h2 = 450 nm, indicating that this thickness provides the most favorable conditions for achieving the desired transmission characteristics. This parameter differs from the parameters used in previous experiments, providing guidance for future device optimization. The impact of
h2 on the V
μRC device’s performance is further illustrated by the corresponding coupling coefficient (
k2) diagram shown in
Figure 6b. The choice of
h2 significantly affects the coupling coefficient and, consequently, the performance of the V
μRC device.
In order to facilitate the self-rolling process and the formation of the V
μRC on top of the silicon ridge waveguide, a layer of Spin-on-Glass (SOG) is applied on top of the silicon waveguide and then etched with RIE for planarization. The residual thickness of this SOG layer can be precisely controlled by the etching process. The thickness of this SOG layer, represented as
d2, assumes an important role in determining the overall coupling distance, thereby influencing the coupling coefficient between the V
μRC and the underlying silicon waveguide. To comprehensively explore the effects of varying SOG thicknesses on the transmittance of the C-port, transmission spectra are simulated for SOG thicknesses of 30, 50, and 70 nm, as illustrated in
Figure 7a. Notably, the maximum Port C transmittance is observed at a SOG thickness of 50 nm, which corresponds to coupling coefficients of 0.45 and 0.48 at wavelengths of 1491.57 nm and 1544.4 nm, respectively (see
Figure 7b). The electric field distribution at the peak wavelength of 1544.4 nm is depicted in
Figure 7c. These findings underscore the critical role of the SOG thickness in determining the performance of V
μRCs.
Moreover, the investigation delves into the impact of the refractive index of the SOG material, denoted as
nSOG, on the transmittance of Port C, as visually depicted in
Figure 8a. The variation in the refractive index of SOG exhibits negligible effects on the transmission spectrum. Subsequently, we conduct a comparative analysis between the transmission spectrum of the V
μRC and its quality factor under the following dimensional parameters:
a1 = 370,
a2 = 25,
h1 = 80,
h2 = 450,
d1 = 160, and
d2 = 50 nm. Remarkably, the coupling peak between the two components aligns notably well with the highest
Q−factor, thereby indicating a substantial correlation. The strategic application of high-index silicon layers and precision fabrication techniques significantly minimizes the impact of lateral photon leakage and surface roughness on the microcavity’s quality factor. Moreover, by examining the transmittance spectrum of the drop port in
Figure 8b, crucial information regarding the insertion losses of the V
μRC at S−band and C−band can be obtained. The insertion losses are determined to be −1.58 dB and −1.48 dB, respectively, while the corresponding 3 dB bandwidths are calculated to be 9.6 nm and 11.32 nm, respectively. Overall, these results contribute to a comprehensive understanding of the V
μRC’s performance characteristics.
Subsequently, drawing upon the comprehensive set of simulation outcomes, successfully obtain the transmission spectra of the V
μRC at Ports B, C, and D, as exemplified in
Figure 9a. Notably, these spectra are acquired under the following dimensional parameters:
a1 = 370,
a2 = 25,
h1 = 80,
h2 = 450,
d1 = 160, and
d2 = 50 nm. Regarding the parameters
a1 and
a2, which denote the thicknesses of silicon nitride (SiN
x) and amorphous silicon (
a−Si), respectively, we observe that these parameters critically influence the effective diameter of the annular cavity. Consequently, increases in
a1 and
a2 lead to the resonances becoming more narrowly defined and deeper. The dimensions of
h1,
h2,
d1, and
d2 are found to significantly impact the coupling coefficient, thereby affecting the coupling efficiency directly. Our analysis aims to identify their optimal values to enhance system performance. We also note that variations in the refractive index of the Spin−On−Glass (
nSOG) exhibit negligible impact on the transmission spectrum, aligning with our simulation results. Additionally, the electric field distributions of the coupling peaks from 1460nm to 1565 nm are depicted in
Figure 9b,c, respectively. For the V
μRC under investigation, the coupling wavelengths are determined to be 1491.57 and 1544.4 nm, respectively. As we proceed towards the subsequent stages of actual device fabrication, it is crucial to acknowledge that the V
μRC is a three-dimensional entity. The comparison between the simulation results and the optimal parameters obtained in previous experiments [
15,
17,
18] shows that the values of parameters
a1,
a2, and
d2 are consistent. As for
h2, a large waveguide thickness helps to enhance the coupling between the microring and the waveguide; improvements need to be made in future experiments.
In addition to the aforementioned parameters, the investigation of the V
μRC along the axial direction of the cylinder assumes paramount importance. Building upon our previous research results, we identify the key parameters along the axial direction, namely a waveguide width of 2 μm and an
a-Si strip width of 3 μm to correspond with it [
17]. In fact, for actual three-dimensional microtubes, without the constraint of
a−Si in the axial direction, the axial loss will be significant, and the V
μRCs may support more than one axial mode. By considering both the radial and axial dimensions of the V
μRC device, it is possible to achieve the precise control and optimization of its functionality, leading to improved performance and reliability in various photonic integrated circuit applications [
17,
18].
The device under investigation is actually a 3D structure. A simplified 2D model may expedite the simulation process, but a true 3D simulation is the best way to verify the findings obtained using the 2D model. Subsequently, we perform such a 3D simulation using the following parameters:
a1 = 370,
a2 = 25,
h1 = 80,
h2 = 450,
d1 = 160, and
d2 = 50 nm, we set the thickness of all the waveguide in the z-direction to 2 μm, and we construct a three-dimensional model. The transmission spectrum and electromagnetic field results of the three-dimensional V
μRCs’ simulation are presented in
Figure 10, which exhibit good consistency with the two-dimensional results shown in
Figure 9. Furthermore, in
Figure 10b, the electric field distribution in 3D of the resonant coupling peak at 1492 nm is elaborated. The efficacy of this structure in achieving optical coupling from the InP to SiN platform has been substantiated, thereby offering a novel approach for future heterogeneous photonics integration endeavors.