High-Performance Wavelength Division Multiplexers Enabled by Co-Optimized Inverse Design

Abstract

Wavelength division multiplexers are fundamental to the functioning and performance of integrated photonic circuits, with applications ranging from optical interconnects to sensing and quantum technologies. Current solutions are limited by trade-offs between channel spacing, crosstalk, insertion loss, and device footprint. Here, we develop a novel design approach that co-optimizes inverse-designed wavelength division multiplexers and distributed Bragg gratings to achieve ultra-low crosstalk without compromising insertion loss. We experimentally demonstrate less than -40 dB crosstalk for wavelength channel spacing of 15 nm in foundry-compatible silicon and silicon nitride devices across the telecommunications C- and L-bands. Our design process is highly adaptable, allowing for seamless scaling to a greater number of output channels, different spectral windows, and easy translation across various material platforms.

Figures (5)

• Figure 1: Approach for next generation of high-performance photonic devices with multifunctional optimization. Traditional optimization restricts the simulation region to a small design area due to constraints on the computational power of available machines. Now, through the exploitation of GPU architecture and compute speed, we have access to large-scale simulation regions. We demonstrate a design concept for the next generation of multifunctional photonics through the incorporation of existing functionality within the simulation region, resulting in highly performing devices.
• Figure 2: Design and realization of co-optimized silicon WDMs. a. Schematic representation of inverse designed Bragg WDMs. Connected to the inverse design region are output ports including Bragg filters that are designed to pass one wavelength channel and reflect the other, b. Scanning electron microscope (SEM) images of the fabricated WDM structures overlayed with the simulated electric field profiles in the structure. Two wavelength channels, 15 nm apart, are shown, c. SEM of the fabricated Bragg filters, d. SEM of one of the WDM devices, e. SEM of the inverse designed grating coupler used to efficiently couple light from free space into the integrated device, f. Simulations of 2 different WDMs designed for different wavelengths. Each of the designs are made with Bragg filters with 200 periods and all simulations are conducted with oxide cladded silicon with a thickness of 220 nm, representative of standard SOI wafers used for fabrication., g. Experimentally measured transmission spectrum of one of the WDM devices. The red and blue lines shows the output power for port 1 and port 2 of the device respectively. h. Schematic representation of a 3-channel WDM co-optimized with Bragg filters. The filters designed to pass $\lambda_1$ and $\lambda_3$ have N periods of a constant periodicity ($\Lambda_{1,3}$) whereas the middle filter has alternating periodicity between two different periods $\Lambda_{2a}$ and $\Lambda_{2b}$. in segments of $N_{ab}$ periods repeated M times., i. Simulated transmission response of the 3-channel silicon WDM. The red line shows the output of port 1, the dark blue shows the output of port 2 and the light blue shows the output of port 3.
• Figure 3: Comparison of co-optimized devices with traditional inverse design and analysis Bragg filter length dependence. a. Experimentally measured difference between devices with different number of Bragg periods at the output. The red lines show port 1 of the same device with differing lengths of Bragg filters (N=100 and N=300) and similarly for port 2 shown in blue. The shaded regions show the minimum and maximum of each port over 5 measurements and the solid lines show the moving average of the 5 measurements, b.,c. Comparison of reflection handling given co-optimized and non-co-optimized case. The device in b. was designed using the larger scale co-optimization which includes the Bragg filters during the inverse design. The gray regions signify the bands for which reflection was minimized during optimization. In c., the device was designed using a basic inverse design process with the Bragg filters added to the structure after the optimization was complete. Here we observe large peaks in the light reflected back into the input, d. Simulated comparison between WDMs created using traditional inverse design techniques and the technique of co-optimization with Bragg filters used in this work. In the case of co-optimization, there is crosstalk reduction of $>$45 dB and a narrowing of filter bandwidth while insertion loss remains comparable, e. Reflection back into input of WDM device in b., by increasing the Bragg filter length from 25 periods to 150, no significant increase in reflection can be observed, f. Comparison between the insertion loss from co-optimized devices and inverse design WDMs where Bragg filters were added after the fact, g. Comparison between crosstalk of co-optimized devices and inverse design WDMs where Bragg filters were added after the fact. All transmission plots are for Bragg period number N=150 and the shaded regions show the raw data from simulation while the solid and dashed lines show the moving average.
• Figure 4: Experimental demonstration of WDM devices. a. Schematic representation of system demonstration of the WDM device coupled to a frequency comb source for filtering of the comb lines. A frequency comb is generated from a photonic crystal microresonator in silicon nitride and this comb is sent to the WDM chip. The resulting outputs of the two ports are measured using a photodiode and power meter, b. Measured input comb spectrum from an OSA, c. Measured output spectra from each of the two WDM ports, normalized to the response of grating couplers used to couple light in and out of the WDM chip.
• Figure 5: Silicon nitride WDM design and performance. a. Schematic representation showing the resulting design for a co-optimized WDM in silicon nitride. The fabricated material stack consists of air-cladded 310 nm thick silicon nitride on SiO$_2$. The inset shows the Bragg filter dimensions., b. Scanning electron microscope (SEM) image of one of the fabricated silicon nitride structures., c. SEM of Bragg waveguide filter., d. SEM zooming in on inverse design region, showing fidelity of small fabricated features., e. SEM of 1D inverse design grating coupler used to efficiency couple light from free space into the fundamental waveguide mode., f. Simulated transmission efficiency of two silicon nitride WDMs (number of Bragg periods N=300) where less than -45 dB of crosstalk can be observed., g. Experimental measurement of WDM 2 from f.. The light red and light blue lines show the transmission spectra of devices with N=100 and the dark red and dark blue shows devices with N=300. An increase in crosstalk suppression while maintaining low insertion loss can be observed with the increase in N.