TL;DR
This paper introduces a novel Inverse Lithography Technique (ILT) for double exposure lithography, enabling finer feature printing for next-generation semiconductor manufacturing. It presents a method for decomposing aerial images rather than target patterns for conventional resists (k₁ ≥ 0.25) and explores the use of novel materials like Contrast Enhancement Layers (CEL) for pushing the limits to k₁ < 0.25.
What Problem Does This Paper Address?
As semiconductor manufacturing approaches the physical limits of traditional lithography, the industry faces the challenge of continuing to shrink feature sizes to meet Moore's Law. This requires achieving smaller patterning resolutions, often characterized by a low k₁ factor (k₁ < 0.30). Double exposure lithography is a promising technique to overcome these limitations, but it traditionally involves complex processing and can be constrained by the linear superposition of light intensities. This paper addresses the need for more efficient and effective double exposure techniques, particularly for achieving very low k₁ values, by leveraging ILT and exploring new material properties.
Methods
The paper proposes two main approaches:
- Double Exposure ILT (DEL-ILT) for k₁ ≥ 0.25: This method adapts a single-exposure ILT framework to a double exposure scenario. Instead of decomposing the target pattern, it decomposes the aerial image into two parts, each corresponding to a separate mask. This approach automatically resolves phase conflicts and aims to produce a combined aerial image with superior contrast. The ILT algorithm synthesizes the masks by minimizing the difference between the desired pattern and the combined aerial image.
- Contrast Enhancement Layer (CEL) for k₁ < 0.25: For even finer features, the paper investigates the use of novel materials that exhibit non-linear responses to light. CEL is a photo-bleachable film that, when coated on conventional resist, enhances contrast. The authors propose a mathematical model for CEL, defining its properties and deriving bounds on image log-slope improvement. They then integrate this CEL model into a DEL framework, combined with ILT, to synthesize masks for 2D patterns at k₁ = 0.185. They also briefly discuss the potential of two-photon absorption resists.
Datasets
Not discussed in the paper. The paper focuses on simulation results and theoretical modeling, using synthesized patterns for evaluation.
Key Insights and Ideas
- Aerial Image Decomposition: Decomposing the aerial image, rather than the target pattern, for double exposure lithography offers a more robust way to handle phase conflicts and achieve higher contrast in the final printed image.
- Non-linear Materials are Key for Ultra-Low k₁: To push lithography beyond k₁ = 0.25, materials with non-linear responses to light, such as CEL and two-photon absorption resists, are essential. These materials allow for a non-linear combination of exposures, overcoming the limitations of linear superposition.
- CEL as a Viable Enabler: The proposed mathematical model for CEL and its integration with DEL and ILT demonstrate its potential for printing features at k₁ < 0.25, specifically showing promise for 80nm gratings and 2D patterns at k₁ = 0.185.
- Reversibility of CEL is Crucial: For DEL with CEL to work effectively, the CEL material must possess reversible bleaching properties, meaning it returns to its original state after the first exposure, allowing for a clean second exposure.
- ILT Adapts to Material Properties: The ILT framework can be extended to account for the non-linear behavior of novel materials like CEL, enabling the synthesis of masks that are optimized for these specific material responses.
Key Results
- DEL-ILT for k₁ ≥ 0.25: Simulations showed that the DEL-ILT approach can decompose patterns and generate masks that result in combined aerial images with good contrast, enabling patterning of 60nm features at k₁ = 0.289 and 80nm gratings at k₁ = 0.35.
- CEL for k₁ < 0.25:
- The CEL model demonstrated an improvement in image log-slope (ILS) for 1D gratings.
- Simulations showed that DEL with CEL can enable printing 80nm gratings using dry lithography (kpitch = 0.34).
- Combining ILT, CEL, and DEL allowed for the synthesis of 2D patterns with k₁ = 0.185, printing 40nm features.
- Two-Photon Absorption Resists: While theoretically suitable for DEL due to their squared intensity response, the practical challenge lies in the unavailability of materials with sufficiently high two-photon absorption cross-sections at relevant wavelengths, requiring impractically high laser power.
Limitations and Future Work
- CEL Model Simplifications: The proposed CEL model assumes ideal 100% opacity and 100% transparency for unbleached and fully bleached states, respectively, which is impractical. Future work should refine the model to incorporate realistic transmittance values.
- ILT Convergence for k₁ < 0.25: The DEL-ILT algorithm with the integrated CEL model did not converge to an acceptable local minimum for k₁ < 0.25. The authors suspect this is due to the local gradient-based search technique used for optimization and suggest that better initialization points might help. This remains a direction for future work.
- Mask Alignment Robustness: A critical requirement for DEL is robustness to mask alignment errors. This could be incorporated as a regularization function in the ILT framework to obtain masks tolerant to misalignment.
- Experimental Validation: The paper primarily presents simulation results. Future work should include carrying out physical experiments using CEL to validate the proposed models and techniques.
- Two-Photon Absorption Materials: The major challenge for two-photon absorption resists is the lack of materials with high enough cross-sections, making them economically unsuitable for high-volume manufacturing. Future research could focus on developing such materials.
Summary updated