spie proceedings [spie photomask technology - monterey, ca (monday 6 october 2008)] photomask...

9
Enhancing OPC Model Stability and Predictability Using SEM Image Contours Mohamed Serag El-Din Habib Mentor Graphics Corporation, Wilsonville, OR 97070 [email protected] ABSTRACT The process model is a major factor affecting the quality of the Model Based Optical Proximity Correction (OPC). A better process model directly leads to better OPC, hence better yield and more profit. The traditional way in calibrating these process models is using CD measurements at sample locations in the test chip. However, the use of Scanning Electron Microscope (SEM) image contours for process model calibration and optimization has been recently introduced in an attempt to build more predictable models. In this study, we characterize the traditional flow models versus the contour calibrated models and study the effect of using different combinations and weighting schemes on the quality of the resulting process models, its stability and its ability to correctly predict the process. INTRODUCTION As the Nano-fabrication technologies advances to catch up with the famous Moore's law, the need for strong resolution enhancement techniques (RET) grow larger as well. One famous RET technique that is widely used nowadays to enhance the printed resist image is the Model-Based Optical Proximity Correction (MB OPC). A detailed exposition of MB OPC can be found in N. Cobb PhD thesis [6]. As the name suggests, the MB OPC utilizes statistical or compact models to predict the behavior of the lithography process. These process models are used to calculate a virtual or simulated image of the designed patterns after the lithography process, henceforth, calculate the Edge Placement Error (EPE) needed for each fragment. The EPE is the difference between the printed – or simulated – edge and the edge of the original layout while a Fragment is a segment of an edge in the designed pattern [6]. The fragment is then moved perpendicular to its orientation according to biases calculated as function of the EPE of the fragment. Therefore, the quality of the MB OPC is strongly dependent on the predictability of the used process models. Thus, the quality of the process models is important for a good MB OPC. The traditional way of calibrating the process models is the CD based model calibration. In this method, we collect sample measurements from structures across the test chip. These test structure are usually symmetrical and usually are selected from some famous structures such as isolated lines, well patterned pitches and line ends. A test structure often yields one sample measurement; hence hundreds and even thousands of test structures have to be prepared and printed. This traditional way has been subjected to several characterizations and it has been found that it has some limitations. One limitation is the need to carefully select measurement location to ensure symmetry. Also due to the dominant 1D nature of the CD based measurements, it's considered inefficient in representing the 2D effects in the printing process which may lead to a false prediction of the resist image. The difficulty of capturing the Photomask Technology 2008, edited by Hiroichi Kawahira, Larry S. Zurbrick, Proc. of SPIE Vol. 7122, 712244 · © 2008 SPIE · CCC code: 0277-786X/08/$18 · doi: 10.1117/12.801513 Proc. of SPIE Vol. 7122 712244-1 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Upload: larry-s

Post on 03-Dec-2016

216 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

Enhancing OPC Model Stability and Predictability Using SEM Image Contours

Mohamed Serag El-Din Habib

Mentor Graphics Corporation, Wilsonville, OR 97070 [email protected]

ABSTRACT

The process model is a major factor affecting the quality of the Model Based Optical Proximity Correction (OPC). A better process model directly leads to better OPC, hence better yield and more profit. The traditional way in calibrating these process models is using CD measurements at sample locations in the test chip. However, the use of Scanning Electron Microscope (SEM) image contours for process model calibration and optimization has been recently introduced in an attempt to build more predictable models. In this study, we characterize the traditional flow models versus the contour calibrated models and study the effect of using different combinations and weighting schemes on the quality of the resulting process models, its stability and its ability to correctly predict the process.

INTRODUCTION

As the Nano-fabrication technologies advances to catch up with the famous Moore's law, the need for strong resolution enhancement techniques (RET) grow larger as well. One famous RET technique that is widely used nowadays to enhance the printed resist image is the Model-Based Optical Proximity Correction (MB OPC). A detailed exposition of MB OPC can be found in N. Cobb PhD thesis [6]. As the name suggests, the MB OPC utilizes statistical or compact models to predict the behavior of the lithography process. These process models are used to calculate a virtual or simulated image of the designed patterns after the lithography process, henceforth, calculate the Edge Placement Error (EPE) needed for each fragment. The EPE is the difference between the printed – or simulated – edge and the edge of the original layout while a Fragment is a segment of an edge in the designed pattern [6]. The fragment is then moved perpendicular to its orientation according to biases calculated as function of the EPE of the fragment. Therefore, the quality of the MB OPC is strongly dependent on the predictability of the used process models. Thus, the quality of the process models is important for a good MB OPC. The traditional way of calibrating the process models is the CD based model calibration. In this method, we collect sample measurements from structures across the test chip. These test structure are usually symmetrical and usually are selected from some famous structures such as isolated lines, well patterned pitches and line ends. A test structure often yields one sample measurement; hence hundreds and even thousands of test structures have to be prepared and printed. This traditional way has been subjected to several characterizations and it has been found that it has some limitations. One limitation is the need to carefully select measurement location to ensure symmetry. Also due to the dominant 1D nature of the CD based measurements, it's considered inefficient in representing the 2D effects in the printing process which may lead to a false prediction of the resist image. The difficulty of capturing the

Photomask Technology 2008, edited by Hiroichi Kawahira, Larry S. Zurbrick, Proc. of SPIE Vol. 7122, 712244 · © 2008 SPIE · CCC code: 0277-786X/08/$18 · doi: 10.1117/12.801513

Proc. of SPIE Vol. 7122 712244-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Page 2: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

information of the pinching and the bridging can also be considered another limitation of this CD based approach [1, 3]. Recently, the concept of using Scanning Electron Microscope (SEM) image contours to calibrate the process model has been introduced [2]. The algorithms of the calibration via contour data try to minimize the differences between the simulated and the contours extracted from the SEM images. Figure 1 shows an example of a SEM image used and its equivalent extracted contours [3].

Figure 1: SEM image of a test structure (on the left) and an image of the corresponding contour (on the right) [3].

The use the SEM image contours for process model calibration opens the ability to capture effects that couldn't be captured by the traditional calibration approach. One of its advantages is that the test structures don't have to be symmetrical and complex structures can be used. Further more; the test structures can be extracted form the designed layout itself to enrich the coverage of the model over the designed patterns. Also the 2D effect as well as pinching and bridging are captured effectively by the contours.

Proc. of SPIE Vol. 7122 712244-2

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Page 3: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

Target Layer

Contour Layer

MANIPULATING CONTOURS

It’s a common trend in the process modeling to give less weight or even exclude the parts of the measurements data that are less important to the design. For example, if the design is mostly consisting of 1D structures, the 2D measurement data is then given the least weight. This also can be helpful when calibrating optical models by giving less weight to the 2D parts. This contour weighting should help in preventing the optics from over fitting itself to capture the fine details in 2Ds.

Figure 2: an example of weighting layer (in green) overlaid on the target contour (in blue).

However due to its nature, weighting contours is not as simple as weighting the CD sample measurements. For the sake of weighting the SEM image contours, we have to create weighting layers. These weighting layers are overlaid over the parts of the contour that will receive a different weight than the original contour. Figure 2 shows an example of a weighting layer overlaid on the test pattern and its corresponding SEM image contours.

Proc. of SPIE Vol. 7122 712244-3

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Page 4: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

We developed two approaches for weighting the contours. The first approach is based on the geometrical shape of the designed polygons. It considers the corners point as the center of the 2D region. This algorithm is fast and fairly accurate however its complexity had to be raised to account for the OPC'ed edges of the test patterns in order to distinguish between the actual corners and jogs made by the OPC. The other approach is based on the image – simulated – properties of the designed patterns. It considers the 2D regions as the regions with image slope larger than a certain values. This algorithm is slower than the shape based algorithm but more accurate especially when the test pattern shapes involve OPC'ed patterns or jogs.

MODEL CALIBRATION

In order to characterize the impact of using the SEM image contours, we calibrated several optical and resist models. We used combinations of SEM contours – with and without weighting – and CD measurements to calibrate these models. Also a Reference model was calibrated using the traditional CD measurements model calibration flow to judge the quality of the resulting models.

Optical Model CM1 Resist Model Contours Contours

Weighted Contours (1:2, 1:5, 1:10, 1:100) Contour Weighted Contours (1:0) Weighted Contours (1:2, 1:5, 1:10, 1:100)

Contours + CD measurements Contours + CD measurements Weighted Contours (1:2, 1:5, 1:10, 1:100) + CDs

Contours + CD measurements Weighted Contours (1:0) + CD measurements Weighted Contours (1:2, 1:5, 1:10, 1:100) + CDs

Table 1: the different data set used to calibrate the process models used in the characterization study. The ratio beside each "weighted contour" indicates the ratio between the 1D areas to the 2D areas (1D: 2D).

Table 1 shows the different data sets used to calibrate the process models. The original datasets is of 1000 CD sample measurements and 100 contour blocks. The contour blocks were divided into two groups, 80 blocks used for the calibration of the models, and the rest is for verification. The usage of the weighting layers obtained from the above section of this paper is indicated in the above table by the word "Weighted Contours". It was used to clip the 2D areas in case of optical models or give them extra weight in case of resist models. We used Calibre® ContourCal® for the process of calibrating the Optical and CM1 models. CM1 is a resist model designed for use in OPC and OPC verification introduced in [7]. The calibrated optical models showed no difference in results in case of using the CD measurements along with the contours than in case of using contours only. This can be a result of the huge number of points forming the contours (4.6 million points) when compared to the number CD measurements (1000 readings). However, in case of CM1 models, using the CDs in conjunction with the contours caused slight changes in the model parameters. Yet, the overall error RMS is the same as in case of contours only.

Proc. of SPIE Vol. 7122 712244-4

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Page 5: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

Model verification

Table 2 shows the RMS error value for the above calibrated models. It is shown in the table that the traditionally calibrated model is having the least RMS error value. However, the RMS error value may be deceiving since the number of points that was used to calibrate the contour models was huge when compared to the case of the CD measurements used for the Reference model. Therefore, the best way to judge a model is through verification.

# Optical Model CM1 Resist Model errRMS R CDs CDs 1.70nm 1 Contours 6.25nm 2 Weighted Contours (1:2) 6.14nm 3 Weighted Contours (1:5) 5.89nm 4

Contours

Weighted Contours ( 1:10) 5.26nm 5 Contours 6.10nm 6 Weighted Contours (1:2) 6.18nm 7 Weighted Contours (1:5) 5.96nm 8

Weighted Contours (1:0)

Weighted Contours (1:10) 5.31nm

Table 2: showing the RMS error values for the calibrated models. The resulting RMS error values when using the Contours and CDs together is exactly the same as the above results.

In this study, we used the division of the contour blocks that was excluded in the model calibration phase for verifying the resulting models. Figure 3 shows screen shots of the overlaying of the printed image over the verification contours. Figure 3a shows good compliance with the SEM contour. However, in Figure 3b, the printed image predicted a false bridging in the reference model case, while the SEM image contour shows no such bridging. This is due to the lake of the 2D details captured by the CD measurements. Figure 4 shows the results of models calibrated via the contours (Process models # 5 in Table 2). The screen shots show a great compliance with the SEM image contours. Figure 4b shows that the model calibrated via the SEM contours predicted the image accurately and avoided the false bridging condition as in the reference model case (figure 3b). Figure 5: shows the results from model #1. The printed image is following the trend of the SEM contour; however, it has fairly contracted form the contour. This can be the result of keeping the 2D areas while calibrating the optics, causing the optical model to be over fit trying to catch up with the rapid changes in the SEM contours. This figure illustrates the importance of using the contour weighting layers in the optical calibration phase. Increasing the weight of the 2D areas (models # 2, 3, 4, 6, 7 and 8) produced nearly the same result as the model #5. This result is clearly shown in figure 6, as the printed image of the models 5 and 8 smoothly fit over each other. This can be considered as an indication on the stability of the model.

Proc. of SPIE Vol. 7122 712244-5

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Page 6: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

a-.

Figure 3a: screen shot sowing the printed image of the reference model (model #R – Table 2) hashed in violet and the

verification SEM image contours in Green.

Figure 3b: screen shot showing the printed image by the reference model (model #R – Table 2) hashed in violet and the SEM

image contours in Green. false bridging is shown in the printed image.

Proc. of SPIE Vol. 7122 712244-6

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Page 7: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

Figure 4a: screen shot showing the printed image by model # 5 (Table 2) hashed in orange and the verification SEM image

contours in Green.

Figure 4b: screen shot showing the printed image by model # 5 (Table 2) hashed in orange and the verification SEM image

contours in Green.

Proc. of SPIE Vol. 7122 712244-7

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Page 8: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

IC__

Figure 5: screen shot showing the printed image by model #1 (Table 2) in dotted red and the Verification SEM image contours

in green.

Figure 6: screen shot for printed image by model #8 (Table 2) in blue and printed image of model #5 hashed in orange, both

overlaid on the verification SEM image contours (in green).

Proc. of SPIE Vol. 7122 712244-8

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms

Page 9: SPIE Proceedings [SPIE Photomask Technology - Monterey, CA (Monday 6 October 2008)] Photomask Technology 2008 - Enhancing OPC model stability and predictability using SEM image contours

CONCLUSION

The process model is essential for performing MB OPC. The accuracy and stability of the process models is a major factor affecting the quality of the resulting OPC. Recently, the concept of the process model calibration via the SEM image contours has been introduced. This opens the door for taking in account the effects that couldn't be captured by the traditional CD sample measurements flow. In this study, we characterized the effect of calibrating the process models using the recently introduced SEM image contours. We also characterized the effect of utilizing contour weighting layers while calibrating both the optical and the resist models.

REFERENCES

1. Y. Granik, I. Kusnadi, "Challenges of OPC Model Calibration from SEM Contours", Metrology, Inspection, and Process Control for Microlithography XXII, Proc. of SPIE Vol. 6922, 2008.

2. K. N. Taravade, E. H. Croffie, A. Jost, “Two-dimensional image-based model calibration for OPC applications", Proceedings of SPIE 5377, 2004.

3. Y. Granik, "Calibration of Compact OPC Models Using SEM Contours", 25th Annual BACUS Symposium on Photomask Technology, Proceedings of SPIE, vol. 5992, 2005.

4. G. E. Bailey, T. Do, Y. Granik, I. Kusnadi, A. Estroff, "Intensive 2D SEM model calibration for 45nm and beyond", Optical Microlithography XIX, Proceedings of SPIE, vol. 6143, 2006.

5. A. N. Drozdov, M. L. Kempsell, Y. Granik, "Fitness and runtime correlation of compact model complexity", Optical Microlithography XXI, Proceedings of SPIE Vol. 6924, 2008.

6. Nick Cobb, “Fast Optical and Process Proximity Correction Algorithms for Integrated Circuit Manufacturing,” PhD Thesis, University of California, Berkeley, 1998.

7. Y. Granik, D. Medvedev, N. Cobb, "Toward standard process models for OPC", Optical Microlithography XX, Proc. of SPIE vol. 6520, 2007.

Proc. of SPIE Vol. 7122 712244-9

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/18/2013 Terms of Use: http://spiedl.org/terms