DRC violation of adjoint optimized designs

[satadrudas]

I have used the adjoint optimization method from this tutorial to design a photonic device with fabrication constraints. While the device performs well, I am facing challenges with foundry DRC violations when extracting the GDS.

Currently, I use two approaches:

Binarization & Pixel Filling:

1. I binarize the design to remove grayscale artifacts, then use GDSFactory to create rectangular waveguide elements for each pixel (Fig. 1).
2. A boolean OR operation merges these elements (Fig. 2).
3. Issue: Some foundries do not allow "staircase" edges, leading to DRC violations.

Contour Extraction & Polygon Approximation:

1. I extract contours from the bitmap (Fig. 3) and generate a GDSFactory component by connecting the midpoint of the pixel edges (Fig. 4).
2. Issue: While this reduces DRC violations, it introduces diagonal edges that may still be problematic.

Question:
What is the best way to extract a GDS that adheres to foundry DRC rules? Are there established methods to smooth the extracted contours while maintaining manufacturability? Any suggestions on alternative approaches would be highly appreciated.

Here’s the code I currently use for contour extraction:

import numpy as np
from skimage import measure
import matplotlib.pyplot as plt

# Define pixel size in microns
pixel_size = 1e-2  # Adjust this to set the length of each pixel

# Load the binary bitmap
test_gds = np.load('test.npy')
test_gds = test_gds.T  # Ensure correct orientation

# Find contours
contours = measure.find_contours(test_gds, level=0.5)


# Display the contours
fig, ax = plt.subplots()
ax.imshow(test_gds, cmap=plt.cm.gray)

for contour in contours:
    ax.plot(contour[:, 1], contour[:, 0], linewidth=2)

plt.show()

# Create a GDSFactory component
component = gf.Component("contour_polygon")

# Convert contour to GDSFactory polygon with scaling
for contour in contours:
    x = contour[:, 1] * pixel_size  # Scale x coordinates
    y = contour[:, 0] * pixel_size  # Scale y coordinates
    points = list(zip(x, y))  # Create list of (x, y) points

    # Add the scaled polygon to the component
    component.add_polygon(points, layer=(709, 727))  # Layer (layer_number, datatype)

# Save the component to a GDSII file
component.write_gds("contour_polygon_scaled.gds")

# Optional: Show the component in a viewer
component.show()

Fig 1

Fig 2

Fig 3

Fig 4