Test pattern selection plays a vital role in the model calibration in the optical proximity correction process. Traditional OPC resist models mainly use the image parameters such as the minimum intensity, the maximum intensity, the slope of intensity along the cut lines crossing the gauge points as their input parameters to calculate the resist contour position. To guarantee the accuracy of the resist model over the whole design layout, it is important that the image parameter space of the test patterns used to calibrate the OPC model covers the image parameter space of the original design layout. We present a method to generate test patterns based on the provided image parameters. The method is based on the adversarial neural network. With this method, we can prepare the test patterns with the desired image parameter coverage.
As chip feature sizes have continued to shrink, resolution enhancement techniques such as Optical Proximity Correction (OPC) have been utilized in advanced technology nodes. In recent years, Inverse Lithography Technology (ILT), a new OPC technique, has been widely applied in advanced Logic and Memory applications to improve imaging performance. Compared to the conventional OPC, ILT enables better process windows (PW) with low edge placement error (EPE) and high wafer critical dimension uniformity (CDU), etc. However, the nonrectilinear mask shapes in ILT make mask writing extremely complex and slow, which can potentially cause more mask manufacturing errors. Therefore, it’s important to quantitatively study the MEEF in ILT masks. In this work, we studied the MEEFs of 2D patterns corrected by ILT and conventional OPC and the differences between these two techniques. The results show that the MEEF at different positions (local MEEF) on an ILT mask has a bigger mean of ~3.14 and a smaller σ of ~0.09 relative to the mean of ~2.14 and σ of ~0.67 from a conventional OPC mask. The MEEF budget is analyzed based on the separated main features (MF) and subresolution assist features (SRAF). With SRAFs being inserted into the entire layout of the ILT mask, it contributes to all individual patterns with ~ 45% (1.49) of the total MEEF. Meanwhile, a conventional OPC mask only has SRAFs on the edges. Thus, SRAFs only contribute MEEF to the patterns located in the edge region (within the proximity effect range). Thus, the main center region of the OPC Mask has a lower MEEF contribution (~1.7). These results suggest that in the ILT recipe tuning process, MEEF should also be included in the cost function as a nonlinear factor so that the inversion can minimize MEEF while optimizing PW and EPE. Furthermore, the manhattanization of the ILT Mask can effectively reduce MEEF.
The contour data extracted from SEM wafer images after the lithography are widely used in the critical dimension (CD), edge placement error (EPE) measurement. It is important to obtain the contours fast and accurate before the analysis of lithographic process and calibration of the lithographic models. Without the accurate contour data, the complete CDU, PVband analysis and inverse lithography technique are hard to realize. With the continuous shrink of the technology nodes, the demand for the accurate contour extraction increases. However, fast and accurate contour extraction from SEM images with defects and noises is challenging. We apply the U-Net to the semantic segmentation of SEM images. The contour extraction and evaluation can be done better after the image segmentation. Our experimental results show that satisfactory contour data of various types of lithographic patterns can be obtained with noisy SEM images.
The technology node shrinks years after years. To guarantee the functionality and yield of IC production, the resolution enhancement technology becomes more and more important. Both optical proximity correction and inverse lithography technique need a precisely calibrated lithographic model. A mask of test patterns needs to be prepared and the lithographic experiment has to be done with it to obtain the CD SEM data for the model fitting. It is beneficial to select the test pattern efficiently. Fewer number of test patterns should be selected without compromising their coverage capability and the accuracy of the lithographic model. We present a machine learning method based on the convolutional autoencoder and core set selection method to achieve above goal. We optimize the existing test pattern mask by selecting parts of gauges out. The OPC models calibrated with the selected data are compared with the models calibrated with original test patterns to evaluate our method.
OPC is a key step to improve design fidelity when people transfer patterns from the photomask to the wafer. However, to complete a traditional OPC job in advanced technology node, a huge number of CPU cores and above several days are required. In this article, we proposed a pixel based OPC and deep learning OPC hybrid optimization framework. First, the pixOPC is done with the raw training clips. The pairs of the raw training clip and post OPC clip form the training data set. The training clip pairs are fed into GAN (Generative Adversarial Network) OPC architecture and the GAN network is trained. The GAN OPC generator is then validated to ensure that it has enough accuracy and does not overfit the data. The validated GAN OPC generator is then applied to generate OPC masks for the new design clips and the generated masks are refined with traditional OPC to exclude some unexpected outliers generated by the GAN method. We design the reversed high discretion pix2pix GAN to generate OPC masks. Its runtime and performance are compared with the model based OPC, pixOPC and U-Net. The generated OPC masks, simulated lithographic contours, EPE, PVBAND and NILS are compared. We find the GAN generative models have better performance compared with the traditional OPC, and the runtime are also much shorter.
Inverse lithography technology (ILT) can optimize the mask to gain the best process window and image quality when the design dimension shrinks. However, as a pixel level correction method, ILT is very time-consuming. In order to make the ILT method useful in real mask fabrication, the runtime of ILT-based optical proximity correction mask must evidently decrease while keeping the good lithographic metric performance. Our study proposes a framework to obtain the curvilinear ILT mask with generative adversarial network (GAN). It is subsequently refined with the traditional ILT to exclude unexpected outliers generated by the GAN method. We design conditional GAN, reverse GAN (RGAN), and high discretion GAN (HDGAN) to generate curvilinear ILT mask. Their runtime and the performance are compared. Compared with the CILT method, the speed of GAN type methods with the afterward refinement is increased by an order of magnitude. The RGAN has a better performance in edge placement error and process variation band evaluation, and HDGAN has a better performance in the mask error enhancement factor evaluation. The designed RGAN and HDGAN are promising in actual application to generate the curvilinear mask. They can evidently decrease the runtime and have better lithographic metric performance.
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.