Directional couplers are fundamental components in photonic integrated circuits, enabling efficient splitting and combining of optical signals. By placing two parallel waveguides in close proximity, these passive devices exploit evanescent field coupling to control how light is distributed between channels. It is also possible to control the output by adjusting the relative phase between the inputs, enabling flexible power splitting ratios and switching behavior. Their design and compatibility with silicon-on-insulator technology have made them a popular choice for applications ranging from wavelength-division multiplexing to optical switching and signal processing.

An important consideration is how fabrication variations might affect device performance. Hence, one common approach is to sweep over different parameters and see how the quantities of interest behave.

In this notebook, we will demonstrate how to set up a simple directional coupler simulation, illustrating the common steps of the workflow:

1. Importing the geometry using gdstk. For the process of this geometry, please follow the steps presented in this example.

2. Placing the ModeSource and ModeMonitors to excite the first TE mode and analyze the transmittance and losses on both arms

3. Carry out the post-process analysis to find the 50/50 splitting frequency and efficiency.

4. Create a parametric sweep over different values of the sidewall angle and its effect on the splitting frequency and efficiency, and create a Batch object to run in parallel.

5. Use the Simulation.updated_copy method to easily create a simulation with two sources at the input ports with a $\pi$/2 phase difference.

For instructions on how to calculate the full S-matrix, please refer to our SMatrix tutorial. Additionally, if you are new to the finite-difference time-domain (FDTD) method, we highly recommend going through our FDTD101 tutorials.

First, we import the relevant structures. We will use gdstk for importing the geometry. To install it, uncomment the !pip install gdstk line.

Importing the Geometry ¶

We will use gdstk to import the geometry into a Polyslab geometry, and extrude it along the z-direction (axis=2) with a given sidewall angle. For detailed instructions, please refer to this tutorial.

Set up Structures¶

Next, we assign material properties to the imported geometries and create the td.Structures with each of those material properties.

Set up Simulation ¶

Now we set up the simulation. First, we define the wavelength of interest and set the runtime.

Now, we add the ModeSource, which excites the first order TE mode (mode_index = 0). To analyze the transmittance, we will place ModeMonitors at the end of both output arms. These monitors will carry out mode decomposition and analyze the fraction of power transmitted in each mode. Since we are only interested in the first order TE mode, the number of modes analyzed can be just 1.

We will also add a FieldMonitor to visualize the field profile in the xy-plane.

Finally, we define the grid with our AutoGrid object, which automatically creates a non-uniform mesh, and place everything together in the Simulation object.

Plot Simulation Geometry¶

It is a good practice to check the simulation and see if everything is correctly set. That can be easily achieved with the Simulation.plot method:

We can also see the 3D model.

Before running the simulation, we can check the maximum possible Flexcredit cost. If the field decays before the runtime, the simulation will auto shut-off and we will only be billed for the Flexcredits of the effective runtime.

Finally, we can run the simulation and analyze the data.

Postprocessing ¶

Now we would like to analyze the output of each arm as a function of wavelength, and find the frequency where the fields are equally distributed, and the losses.

The monitor decomposes the fields into mode components, and since the source is normalized, the transmittance of each component is simply $\text{|amplitude|}^2$.

For finding the point where the amplitudes of each arm are the same, we will just look at the signal change of the difference between both signals, which is the crossing point. For better resolution, we will interpolate the data.

Now, we can easily plot the E field with the SimulationData.plot_field method. As we can see, the input field is equally split over the two output arms.

Parametric Sweep ¶

Now, we will investigate how the sidewall angle changes the component performance. To do so, we will create a dictionary where each value is a simulation object with a different coupler sidewall.

Next, we create a Batch object to run all simulations in parallel.

Now we can easily track the crossing frequency and transmittance using our defined post-processing function.

Creating a Simulation With Two Ports ¶

We can use the updated_copy method to easily create another simulation with a second source with a $\pi$/2 phase difference.

As we can see, in this configuration, all power is directed to the upper output at the crossing frequency.