Technology Computer-Aided Design (TCAD) simulations, encompassing each process and device simulations, are crucial for contemporary semiconductor manufacturing. They permit “virtual manufacturing,” allowing engineers to design, construct, and test transistors and integrated circuits digitally before committing to the costly physical fabrication process. This approach significantly reduces development time from years to months and saves billions of dollars in experimental manufacturing costs.
These simulations, nonetheless, are computationally intensive and may take so long as several weeks to finish, delaying manufacturing deadlines. AI-augmented TCAD is a key solution to deal with this challenge. That’s where NVIDIA PhysicsNeMo and NVIDIA Apollo are available in. The PhysicsNeMo framework lets developers construct high-fidelity surrogates using state-of-the-art architectures for engineering and science simulations. Apollo, announced last month at SC25, makes this easier by providing domain specific, pre-trained models.
Engineers at SK hynix, certainly one of the world’s leading memory chip manufacturers, are leveraging AI physics to develop high-fidelity surrogate models to speed up device and process simulations within the design and manufacturing of semiconductor chips. Using the NVIDIA PhysicsNeMo framework, engineers have fast-tracked the event of proprietary AI models that may unlock tools for significant innovation in device design and manufacturing.
On this blog, we’ll walk you thru the steps to start with PhysicsNeMo to develop your personal custom models and share how the TCAD Intelligence team at SK hynix used PhysicsNeMo to speed up development of its AI physics models.
Tapping into AI physics for TCAD
TCAD is a specialized field of software simulation used to model and optimize the fabrication and physics of semiconductor devices. It’s typically broken into two principal parts—process TCAD and device TCAD. Process TCAD simulations model the physical and chemical steps of chip manufacturing, resembling deposition, lithography, etching, and ion implantation. Device TCAD simulations, however, take the ultimate 3D structure predicted by the method simulation and model its electrical behavior. Engineers utilize a wide range of simulation solutions for various use cases, starting from atomic-scale density functional theory (DFT) simulations to chamber-scale computational fluid dynamics (CFD) simulations.
AI-augmented TCAD presents a fundamental disruptive opportunity for semiconductor manufacturers. As transistors shrink to the nanometer scale, the complexity of their behavior increases, making accurate simulations indispensable for designing next-generation devices but making them also orders of magnitude costlier.
AI surrogate models—which could be created with NVIDIA PhysicsNeMo—are ultra-fast, deep learning-based replicas of slow, physics-based simulations. This approach dramatically accelerates the design and optimization of semiconductor devices by reducing simulation times from hours to milliseconds, enabling engineers to explore a much wider range of possibilities.
PhysicsNeMo provides Python modules to compose scalable and optimized training and inference pipelines to develop and deploy AI surrogates. The PhysicsNeMo framework offers various AI models tuned for science and engineering and enables the mix of physics knowledge with data.
For AI physics researchers and developers exploring the usage of neural operators, GNNs, or transformers—or are fascinated by physics-informed neural networks or a hybrid approach in between—PhysicsNeMo provides an optimized stack that may enable them to coach their models at scale. The engineers use the vital constructing blocks from PhysicsNeMo to alleviate the necessity to develop from scratch. This enables them to cut back the trouble required to develop detailed AI methodologies and as an alternative give attention to using their domain expertise to develop surrogate models for specific physics problems.
Getting began with PhysicsNeMo
The best method to start with PhysicsNeMo for constructing an AI surrogate is to make use of certainly one of the reference application recipes. These examples provide you with a working template for each the training code and the info. Here is the overall step-by-step path you’d follow, using the official examples as your guide.
- Install PhysicsNeMo: First, you might want to arrange your environment.
- The easiest method is to make use of the official NVIDIA NGC container, which has all dependencies (PyTorch, CUDA, etc.) pre-installed. Next, clone the PhysicsNeMo GitHub repository to get the relevant reference application recipes.
- If you’ve got an existing dev environment setup for PyTorch, you possibly can pip install from source following the steps outlined here.
- Let’s assume you’re fascinated by developing a GNN-based surrogate model for TCAD CFD simulations, you’d start with the vortex shedding recipe. After replicating the sample, you possibly can begin to customize the training pipeline to your personal custom data.
- You can even evaluate other model architectures like DoMINO or Transolver in your custom data.
- The built-in distributed functionality in PhysicsNeMo recipes lets you scale any of the above architectures to full 3D chip scale simulations.
Let’s take a have a look at how SK hynix engineers used PhysicsNeMo for certainly one of the numerous TCAD use cases.
How SK hynix uses AI physics for TCAD
South Korea-based SK hynix is a world leader in producing high-bandwidth memory (HBM), an important component in advanced AI accelerators and GPUs. Its products are vital for a wide selection of electronics, from data center servers and PCs to smartphones and next-generation AI systems.
The corporate’s engineers are pioneering the usage of AI physics by developing high-fidelity surrogate models to speed up device and process simulations. Utilizing the NVIDIA PhysicsNeMo framework, they’ve rapidly advanced their proprietary AI models. An example is the SK hynix TCAD intelligence team’s work on AI surrogate models for etching, an increasingly critical process in semiconductor front-end manufacturing, particularly for advanced memory technologies. By employing predictive modeling to guide the etching process, SK hynix goals to expedite the event of next-generation memory devices.
Accurate prediction of time-varying structures within the etching process is crucial for SK hynix. While neural operators are helpful, they often require large datasets and struggle with data scarcity. To handle this, SK hynix adopted the Graph Network-based Simulator (GNS) architectures grounded in Graph Neural Networks (GNNs), which mix numerical time-stepping methods to effectively model geometry changes over time. GNS captures local interactions, representing critical physical properties with minimal training data. Nevertheless, the present GNS models were insufficient for effectively emulating the etching process, necessitating the event of additional AI models to reinforce the accuracy and efficiency of the emulations.
| Methodologies | Improvement |
| MeshGraphNet(MGN) | Memory requirement decreased |
| Chamfer Loss used for velocity calculation | Training loss reduced |
| Re-meshing each Iteration steps | Inference accuracy improved |
| Feature selection | Inference accuracy improved |
| Multi-scale message passing | Training loss reduced |
| Material feature update each iteration steps | Inference accuracy improved |
The TCAD Intelligence team at SK hynix believes that AI-augmented TCAD will develop into a key enabler of research productivity within the semiconductor industry. By leveraging AI-accelerated TCAD predictions, engineers will give you the option to realistically evaluate tens of 1000’s of process cases generated from dozens of recipe combos. This advancement allows TCAD to evolve beyond qualitative guidance and function a quantitative optimization framework for semiconductor R&D.
A wide selection of AI models that were developed using the PhysicsNeMo framework and GPU-accelerated libraries play an important role in enabling these capabilities efficiently.
Tips on how to start with NVIDIA PhysicsNeMo
If you happen to are a TCAD application developer or an AI physics researcher, PhysicsNeMo is a robust tool in your arsenal to speed up your AI model development. As an alternative of constructing every part from scratch, you possibly can leverage PhysicsNeMo modules and model architectures to construct enterprise scale Physics AI solutions with unprecedented speed and ease.
TCAD engineers at SK hynix used this approach to focus their domain expertise and efforts on modeling their problems effectively and constructing skillful models as an alternative of writing training pipelines using low-level libraries.
You may learn more by utilizing these resources:
Yiyi Wang and Alexey Kamenev contributed to the project featured on this blog.
