The goal of this paper is to explore machine learning solutions to improve the run-time of model-based retargeting in the mask synthesis flow. The purpose of retargeting is to re-size non-lithography friendly designs so that the design geometries are shifted to a more lithography-robust design space. However, current model-based approaches can take significant run-time. As a result, this step is rarely done in production settings. Different machine learning solutions for resolution enhancement techniques (RETs) have been previously proposed. For instance, to model optical proximity correction (OPC) or inverse lithography (ILT). In this paper, we compare and expand some of these solutions. In the end, we will discuss the experimental results that can achieve a nearly 360x run-time improvement while maintaining similar accuracy to traditional retargeting techniques.
We provide background on differences between traditional and machine learning modeling. We then discuss how these differences impact the different validation needs of traditional and machine learning OPC compact models. We then provide multiple diverse examples of how machine learning OPC compact validation modeling can be appropriately validated both for modeling-specific production requirements such as model signal/contour accuracy, predictiveness, coverage and stability; and also general OPC mask synthesis requirements such as OPC/ILT stability, convergence, etc. Finally we conclude with thoughts on how machine learning modeling methods and their required validation methods are likely to evolve for future technology nodes.
As feature resolution and process variations continue to shrink for new nodes of both DUV and EUV lithography, the density and number of devices on advanced semiconductor masks continue to increase rapidly. These advances cause significantly increased pressure on the accuracy and efficiency of OPC and assist feature (AF) optimization methods for each subsequent process technology. To meet manufacturing yield requirements, significant wafer retargeting from the original design target is often performed before OPC to account for both lithographic limitations and etch effects. As retargeting becomes more complex and important, rule-table based approaches become ineffective. Alternatively, modelbased optimization approaches using advanced solvers, e.g., inverse lithography technology (ILT), have demonstrated process window improvement over rule-based approaches. However, model-based target optimization is computationally expensive which typically limits its use to smaller areas like hotspot repairs. In this paper, we present results of a method that uses machine-learning (ML) to predict optimal retargeting for line-space layers. In this method, we run ILT co-optimization of the wafer target and process window to generate the training data used to train a machine learning model to predict the optimum wafer target. We explore methods to avoid ML model overfitting and show the ML infrastructure used to integrate ML solution into a manufacturable OPC flow. Both lithographic quality and runtime performance are evaluated for an ML enabled retargeting flow, an ILT flow and a simple rule table flow at advanced node test cases.
As feature resolution and process variations continue to shrink for new nodes of both DUV and EUV lithography, the density and number of devices on advanced semiconductor masks continue to increase rapidly. These advances cause significantly increased pressure on the accuracy and efficiency of OPC and assist feature (AF) optimization methods for each subsequent process technology. Several publications and industry presentations have discussed the use of neural networks or other machine learning techniques to provide improvements in efficiency for OPC main feature optimization or AF placement.
In this paper, we present results of a method for using machine-learning to predict OPC mask segment displacements. We will review several detailed examples showing the accuracy and overall OPC TAT benefits of our method for advanced node manufacturing test cases. We will also discuss the experiments testing the amount and diversity of training data required to achieve true production level OPC stability.
As feature resolution and process variations continue to shrink for new nodes of both DUV and EUV lithography, the density and number of devices on advanced semiconductor masks continue to increase rapidly. These advances cause significantly increased pressure on the accuracy and efficiency of OPC and assist feature (AF) optimization methods for each subsequent process technology. Several publications and industry presentations have discussed the use of neural networks or other machine learning (or even deep learning) to provide improvements in efficiency for OPC main feature optimization or AF placement. However, these two mask synthesis steps are not independent. OPC affects AF optimum position and size; and AF position and size both affect the final optimum OPC main feature correction. A challenging example of these interactions is the need for OPC and AF methods to be aware of potential AF wafer printing. AF printing on the wafer can lead to catastrophic device failure. If an AF is at risk of printing in photoresist, both the OPC and the size (and potentially the position) of the AF need to be modified accurately and efficiently. Recent advancements in lithography utilizing negative tone develop (NTD) photoresists (resists) with strong physical shrink effects also further increase the difficulty of accurately modeling AF printing. In this paper, we present results of our work to explore the requirements, the issues and the overall potential for developing robust, accurate and fast integrated machine learning methods to optimize OPC and AFs.
Below the 28nm node the difficulty of using subresolution assist features (SrAFs) in OPC/RET schemes increases substantially with each new device node. This increase in difficulty is due to the need for tighter process window control for smaller target patterns, the increased risk of SrAF printing , and also the increased difficulty of SrAF mask manufacture and inspection. Therefore, there is a substantially increased risk of SrAFs which violate one or more manufacturability limits.
In this paper, we present results of our work to evaluate methods to pre-characterize designs which are likely to become problematic for SrAF placement. We do this by evaluating different machine learning methods, inputs and functions.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.