PatternPaint: Generating Layout Patterns Using Generative AI and Inpainting Techniques

Guanglei Zhou,Bhargav Korrapati,Gaurav Rajavendra Reddy,Jiang Hu,Yiran Chen,Dipto G. Thakurta

2024-10-26

Abstract:Generation of diverse VLSI layout patterns is crucial for various downstream tasks in design for manufacturing (DFM) studies. However, the lengthy design cycles often hinder the creation of a comprehensive layout pattern library, and new detrimental patterns may be discovered late in the product development process. Existing training-based ML pattern generation approaches struggle to produce legal layout patterns in the early stages of technology node development due to the limited availability of training <a class="link-external link-http" href="http://samples.To" rel="external noopener nofollow">this http URL</a> address this challenge, we propose PatternPaint, a training-free framework capable of generating legal patterns with limited DRC Clean training samples. PatternPaint simplifies complex layout pattern generation into a series of inpainting processes with a template-based denoising scheme. Our framework enables even a general pre-trained image foundation model (stable-diffusion), to generate valuable pattern variations, thereby enhancing the library. Notably, PatternPaint can operate with any input size. Furthermore, we explore fine-tuning a pre-trained model with VLSI layout images, resulting in a 2x generation efficiency compared to the base model. Our results show that the proposed model can generate legal patterns in complex 2D metal interconnect design rule settings and achieves a high diversity score. The designed system, with its flexible settings, supports pattern generation with localized changes and design rule violation correction. Validated on a sub-3nm technology node (Intel 18A), PatternPaint is the first framework to generate a complex 2D layout pattern library using only 20 design rule clean layout patterns as input.

Computer Vision and Pattern Recognition,Computational Engineering, Finance, and Science,Machine Learning

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

This paper aims to solve the crucial problem of generating diverse VLSI (Very - Large - Scale Integration) layout patterns in semiconductor manufacturing. Specifically, the paper focuses on how to efficiently generate legal and diverse layout patterns under strict design rules (Design Rules, DRs), especially in the early stage of technology node development when the available DR Clean samples are very limited.

#### Background and Challenges

1. **Long Design Cycle**: Traditional layout pattern generation methods rely on manual design or rule - based methods, which require a great deal of time and engineering effort to understand and implement thousands of design rules.
2. **Insufficient Samples**: In the early stage of new technology node development, since the design rules are not yet fully determined, the available DR Clean samples are very limited, which makes it difficult for training - based machine - learning methods to work effectively.
3. **Limitations of Existing Methods**:
- **Rule - Based Methods**: It is necessary to convert a large number of design rules into algorithmic constraints, and these methods are usually tightly coupled with specific technology nodes and difficult to adapt to new technology nodes.
- **Training - Based ML Methods**: Although they can reduce engineering effort and increase pattern diversity, in the early stage, due to the lack of sufficient training samples, these methods are difficult to generate legal layout patterns.

#### Solutions

To solve the above problems, the paper proposes **PatternPaint**, a training - free framework that can generate legal layout patterns under the condition of limited DR Clean samples. Its main features include:

1. **Unsupervised Generation**: PatternPaint does not require a large number of training samples, and only a small number of DR Clean samples are needed to generate diverse legal layout patterns.
2. **Repair - Based Generation Method**: Through a series of inpainting processes and a template - based denoising scheme, the complex layout pattern generation task is simplified.
3. **Flexibility**: It supports any input size and can further improve the generation efficiency by fine - tuning the pre - trained model.
4. **Industrial Verification**: It has been verified on Intel 18A (sub - 3nm technology node) and can generate more than 4,000 legal layout patterns using only 20 DR Clean samples.

### Main Contributions

1. **Proposing PatternPaint**: The first training - free automated layout generation framework, which accelerates future semiconductor manufacturing yield improvement and reduces engineering effort.
2. **Efficient Denoising Scheme**: Through the template - based denoising method, the legality of the generated patterns is significantly improved.
3. **Iterative Generation Strategy**: Combining PCA to select representative layouts and gradually build a diverse pattern library.
4. **Industrial Verification**: For the first time, the effectiveness of the ML method has been verified on an industrial PDK (Process Design Kit) and passed the sign - off design rule check.

In conclusion, PatternPaint provides an innovative, training - free solution that can efficiently generate diverse legal layout patterns in a complex design rule environment, especially suitable for the early stage of new technology node development.