TLDR: Researchers developed an automated method for Directed Self-Assembly (DSA) lithography, crucial for sub-7nm chip manufacturing. They introduced a simple Gaussian descriptor for template design and proposed using AB/AB binary block copolymer blends for better material adaptability. Bayesian optimization co-optimizes both the template shape and blend composition, ensuring high-precision patterns and improved template manufacturability by constraining curvature variation. This approach significantly advances DSA technology for complex multi-hole patterns.
The relentless pursuit of smaller and more powerful microchips continues to drive innovation in semiconductor manufacturing. A key technology enabling this miniaturization is lithography, and within this field, Directed Self-Assembly (DSA) of block copolymers (BCPs) offers a highly promising route for fabricating ultra-small features, such as contact holes or vertical interconnect access (VIA) at advanced sub-7nm technology nodes.
Historically, creating these precisely sized and positioned circular holes has relied on guiding the self-assembly of BCPs with meticulously designed templates. However, efficiently optimizing the shape of these templates has presented a significant challenge. Furthermore, for practical applications, the optimized templates must also be easily manufacturable.
A recent research paper, titled “Fully automated inverse co-optimization of templates and block copolymer blending recipes for DSA lithography,” introduces a groundbreaking approach to address these critical issues. The authors, Yuhao Zhou, Huangyan Shen, Qingliang Song, Qingshu Dong, Jianfeng Li, and Weihua Li, propose a novel Gaussian descriptor to characterize template shapes using only two parameters. This elegant simplification drastically reduces the complexity involved in template design.
Beyond template design, the study also explores an innovative material strategy. Instead of using pure diblock copolymers, the researchers suggest employing AB/AB binary blends. These blends, composed of two different diblock copolymers, demonstrate enhanced adaptability to the template shape. This improved material flexibility allows for better conformity to the guiding patterns, leading to more precise and controlled self-assembly.
The core of their optimization strategy is Bayesian optimization (BO), a powerful computational technique. This method is applied to simultaneously co-optimize both the binary blend composition and the template shape. The results are compelling, showing that BO, when combined with the Gaussian descriptor, can efficiently generate optimal templates for a diverse range of multi-hole patterns, all yielding highly matched self-assembled morphologies.
A crucial aspect addressed in this research is the manufacturability of the templates. By imposing constraints on the variation of curvature of the template during the optimization process, the team ensures that the resulting templates are not only effective in guiding self-assembly but also practical to fabricate. This represents a significant step forward for the real-world application of DSA technology.
The study also highlights that the key parameters of the blend exhibit a relatively wide tunable window, allowing for high precision while maintaining experimental flexibility. This work provides invaluable insights for advancing DSA technology, potentially propelling its practical applications in the semiconductor industry. For a deeper dive into the methodology and results, the full research paper is available at arXiv:2510.02715.
The researchers illustrate their findings with examples such as double-hole patterns. They demonstrate that for pure diblock copolymers, increasing the distance between holes can lead to templates with sharp, difficult-to-manufacture “peanut-like” shapes. However, by incorporating the AB/AB binary blend, the curvature variation of the template is considerably reduced, enhancing manufacturability without compromising precision. This improvement is attributed to the longer B-blocks in the blend, which can effectively fill spaces further away from the holes, thereby increasing the material’s adaptability to complex template geometries.
The paper further details the multi-objective function used for optimization, which integrates three key metrics: circularity deviation, center position accuracy, and redundancy distribution. These metrics rigorously quantify the discrepancies between the self-assembled patterns and the target designs, ensuring high fidelity. The application of Bayesian optimization, a gradient-free black-box optimization method, is instrumental in efficiently navigating the vast parameter space involved in the co-optimization of both material and template.
Also Read:
- AI Agents Accelerate New Alloy Development for 3D Printing
- Smart Manufacturing: Automating Job Scheduling with AI and Specialized Languages
In conclusion, this research presents a comprehensive and automated approach to designing templates and material blends for DSA lithography. By meticulously balancing precision with manufacturability, it paves the way for more efficient and cost-effective fabrication of next-generation semiconductor devices.
