prefab 1.1.4

Artificial nanofabrication of integrated photonic circuits using deep learning

PreFab

PreFab is a virtual nanofabrication environment that leverages deep learning and computer vision to predict and correct for structural variations in integrated photonic devices during nanofabrication.

Prediction

PreFab predicts process-induced structural variations, including corner rounding, loss of small lines and islands, filling of narrow holes and channels, sidewall angle deviations, and stochastic effects. This allows designers to rapidly prototype and evaluate expected performance pre-fabrication.

Correction

PreFab corrects device designs to ensure that the fabricated outcome closely matches the intended specifications. This minimizes structural variations and reduces performance discrepancies between simulations and actual experiments.

Models

Each photonic nanofabrication process requires unique models, which are regularly updated with the latest data. The current models include (see the full list in docs/models.md):

Foundry	Process	Latest Version	Latest Dataset	Model Name
ANT	NanoSOI	ANF1 (May 6 2024)	d10 (Jun 8 2024)	ANT_NanoSOI_ANF1_d10
ANT	SiN	ANF1 (May 6 2024)	d1 (Jan 31 2024)	ANT_SiN_ANF1_d1
Generic	DUV-SOI	ANF1 (May 6 2024)	d0 (Jul 30 2024)	generic_DUV_SOI_ANF1_d0

New models are to be regularly added. Usage may change. For additional foundry and process models, feel free to contact us or raise an issue.

Installation

Install PreFab via pip:

pip install prefab

Or clone the repository and install in development mode:

git clone https://github.com/PreFab-Photonics/PreFab.git
cd PreFab
pip install -e .

Getting Started

Account setup

Before you can make PreFab requests, you will need to create an account.

To link your account, you will need an token. You can do this by running the following command in your terminal. This will open a browser window where you can log in and authenticate your token.

python3 -m prefab setup

Guides

Visit /docs/examples or our docs to get started with your first predictions.

Performance and Usage

PreFab models are hosted on a serverless cloud platform. Please keep in mind:

• 🐢 Default CPU inference may be slower.
• 🥶 The first prediction using optional GPU inference may take longer due to cold start server loading. Subsequent predictions will be faster.
• 😊 Please be considerate of usage. Start with small tasks and limit usage during the initial stages. Thank you!

License

This project is licensed under the LGPL-2.1 license. © 2024 PreFab Photonics.