Optical proximity correction operation method and optical proximity correction operation device

Abstract

An optical proximity correction operation method and an optical proximity correction operation device are disclosed. The optical proximity correction operation method includes the following steps. An optical mask layout is obtained. If the optical mask layout has at least one defect hot spot, at least one local area pattern is extracted from the optical mask layout according to the defect hot spot. The local area pattern is analyzed by a machine learning model to obtain at least one optical proximity correction strategy. The optical proximity correction strategy is executed to correct the optical mask layout.

Description

Technical Field

[0001] This invention relates to an operating method and an operating apparatus, and more particularly to an optical proximity correction operating method and an optical proximity correction operating apparatus. Background Technology

[0002] Photolithography is a crucial step in semiconductor manufacturing. In photolithography, a photomask pattern defines the exposed and non-exposed areas, allowing a predetermined photoresist pattern to be developed on a photoresist layer. Then, in the etching process, circuitry is etched using this photoresist pattern. However, the etched circuitry often contains defects such as broken lines and short circuits, necessitating optical proximity correction (OPC) to correct the photomask pattern. The goal of OPC is to ensure that the etched circuitry closely matches the predetermined pattern and is free of defects.

[0003] However, different defects and different patterns require different optical proximity correction strategies. Traditionally, these have all relied on operator judgment, which is not only prone to errors but also makes it difficult to accumulate experience. Summary of the Invention

[0004] This invention relates to an optical proximity correction operation method and an optical proximity correction operation device. When faced with various photomask patterns, it analyzes one or more local region patterns through a machine learning model to obtain one or more corresponding optical proximity correction strategies. This eliminates the need for operators to make judgments, resulting in more accurate judgments and continuous accumulation of experience.

[0005] According to a first aspect of the present invention, an optical proximity correction (OPC) operation method is proposed. The OPC operation method includes the following steps: obtaining a photomask layout pattern; if the photomask layout pattern has at least one defect hotspot, extracting at least one local region pattern from the photomask layout pattern based on the defect hotspot; analyzing the local region pattern using a machine learning model to obtain at least one optical proximity correction strategy; and executing the optical proximity correction strategy to correct the photomask layout pattern.

[0006] According to a second aspect of the present invention, an optical proximity correction (OPC) operation apparatus is provided. The OPC operation apparatus includes an input unit, a region extraction unit, a machine learning model, and a correction unit. The input unit is used to obtain a photomask layout pattern. If the photomask layout pattern has at least one defect hotspot, the region extraction unit extracts at least one local region pattern based on the defect hotspot. The machine learning model is used to analyze the local region pattern to obtain at least one optical proximity correction strategy. The correction unit is used to execute the optical proximity correction strategy to correct the photomask layout pattern.

[0007] To provide a better understanding of the above and other aspects of the present invention, specific embodiments are described below in conjunction with the accompanying drawings: Attached Figure Description

[0008] Figure 1 This is a schematic diagram of an embodiment of optical proximity correction (OPC);

[0009] Figure 2 This is a schematic diagram of an optical proximity correction operation device according to an embodiment;

[0010] Figure 3 A flowchart of an embodiment of an optical proximity correction operation method;

[0011] Figures 4-9 for Figure 3 A schematic diagram of each step;

[0012] Figure 10 A flowchart illustrating a retraining method for a machine learning model, as shown in one embodiment.

[0013] Figure 11 for Figure 10 A diagram illustrating each step. Detailed Implementation

[0014] Please refer to Figure 1 The diagram illustrates an optical proximity correction (OPC) according to one embodiment. Optical proximity correction is primarily used to correct photomask patterns. The goal of optical proximity correction is to ensure that the etched lines conform as closely as possible to the predetermined line pattern and are free of defects. Figure 1 As shown, after exposure / development / etching using the photomask pattern MK11, the circuit pattern PT11 can be obtained. The bridging defect hotspot DF11 can be clearly observed from the circuit pattern PT11.

[0015] To avoid the occurrence of the defect hotspot DF11, researchers can use an optical proximity correction strategy to modify the photomask pattern MK11 into the photomask pattern MK12. For example... Figure 1 As shown, after exposure / development / etching using the photomask pattern MK12, the circuit pattern PT12 can be obtained. From the circuit pattern PT12, it can be seen that the defect hotspot DF11 has disappeared. However, the states of defect hotspots are quite diverse, and trace density and curvature also affect the circuit pattern. Therefore, facing various different situations, it is often difficult to correctly determine the required optical proximity correction strategy.

[0016] Please refer to Figure 2 The diagram illustrates an optical proximity correction (OPC) operating device 100 according to an embodiment. The OPC operating device 100 includes an input unit 110, a defect analysis unit 120, a region extraction unit 130, a machine learning model 140, a correction unit 150, and a retraining unit 160. The functions of each component are summarized below. The input unit 110 is used to input data. The defect analysis unit 120 is used to perform defect analysis. The region extraction unit 130 is used to segment out regions of the image. The machine learning model 140 is used to perform image analysis / recognition procedures. The correction unit 150 is used to perform optical proximity correction. The retraining unit 160 is used to retrain the machine learning model 140. The input unit 110 is, for example, a database, a transmission line, a network cable, or a wireless transmission module. The defect analysis unit 120, region extraction unit 130, machine learning model 140, correction unit 150, and / or retraining unit 160 are, for example, a circuit, a chip, a circuit board, program code, or a storage device storing program code. During the operation of the optical proximity correction operating device 100, the machine learning model 140 can analyze at least one local region pattern to obtain at least one optical proximity correction strategy. The operation of each component is further explained in detail below with flowcharts.

[0017] Please refer to Figures 3-9 , Figure 3 A flowchart illustrating an optical proximity correction operation method according to one embodiment is shown. Figures 4-9 Example description Figure 3 Each step. Figure 3 The optical proximity correction operation method includes steps S110 to S150. In step S110, as... Figure 4 As shown, input unit 110 obtains a photomask layout pattern MK21. The photomask layout pattern MK21 is, for example, a GDS file. Before correction, the edges of the photomask layout pattern MK21 are quite smooth. However, according to the above... Figure 1As explained, the uncorrected photomask layout MK21 may contain defect hotspots. These defect hotspots include, for example, short circuits, open circuits, and bridging. Therefore, the correction unit 150 can comprehensively correct the photomask layout MK21 using a basic optical proximity correction strategy to obtain... Figure 5 Photomask layout diagram MK22.

[0018] Next, in step S120, as Figure 5 As shown, the defect analysis unit 120 determines whether the photomask layout pattern MK22 has any defect hotspots. If the photomask layout pattern MK22 does not have any defect hotspots, the process ends; if the photomask layout pattern MK22 has any defect hotspots, the process proceeds to step S130. The number of defect hotspots in the photomask layout pattern MK22 may be one or more. Figure 5 As shown, the photomask layout MK22 has defective hotspots DF221, DF222, and DF223. Defective hotspot DF221 may be a broken wire, for example, while defective hotspots DF222 and DF223 may be bridging, for example.

[0019] In step S130, as Figure 6 As shown, the region extraction unit 130 extracts local region patterns LM221, LM222, and LM223 based on the defect hotspots DF221, DF222, and DF223. The local region patterns LM221, LM222, and LM223 are not the entire photomask layout MK22, but only the adjacent regions of the defect hotspots DF221, DF222, and DF223. The sizes of the local region patterns LM221, LM222, and LM223 can be the same or not exactly the same. The sizes of the local region patterns LM221, LM222, and LM223 can be 5 to 20 times the linewidth.

[0020] Next, in step S140, as Figure 6 As shown, machine learning model 140 analyzes local region patterns LM221, LM222, and LM223 to obtain optical proximity correction strategies ST221, ST222, and ST223. Optical proximity correction strategy ST221 is used to correct local region pattern LM221 to avoid line breaks. Optical proximity correction strategies ST222 and ST223 are used to correct local region patterns LM222 and LM223 to avoid bridging. Optical proximity correction strategies ST221, ST222, and ST223 can be completely different, partially the same, or completely identical. Furthermore, for the same local region pattern, machine learning model 140 may analyze multiple optical proximity correction strategies, and cost analysis can be used to select the most appropriate optical proximity correction strategy.

[0021] Then, in step S150, as Figure 6 As shown, the correction unit 150 executes optical proximity correction strategies ST221, ST222, and ST223 to correct the photomask layout pattern MK22. In this step, the correction unit 150 executes optical proximity correction strategies ST221, ST222, and ST223 for local region patterns LM221, LM222, and LM223 to obtain corrected local region patterns LM221', LM222', and LM223'. The correction unit 150 then fuses the corrected local region patterns LM221', LM222', and LM223' into the photomask layout pattern MK22 to obtain... Figure 7 The photomask pattern LM23. In this step, the optical proximity correction strategies ST221, ST222, and ST223 can be executed simultaneously or sequentially.

[0022] In another embodiment, optical proximity correction strategies ST221, ST222, and ST223 can be executed sequentially on the entire photomask layout MK22 without the need for a fusion procedure.

[0023] Next, the process returns to step S120. In step S120, as follows... Figure 7 As shown, the defect analysis unit 120 determines whether the photomask layout pattern MK23 has any defect hotspots. If the photomask layout pattern MK23 does not have any defect hotspots, the process ends; if the photomask layout pattern MK23 has any defect hotspots, the process proceeds to step S130. Figure 7 As shown, the photomask layout MK23 has a defective hotspot DF231. The defective hotspot DF231 corresponds to the corrected local area pattern LM222'. In other words, the photomask layout MK23 still needs further correction.

[0024] In step S130, as Figure 8 As shown, the region extraction unit 130 extracts the local region pattern LM231 based on the defect hotspot DF231.

[0025] Next, in step S140, as Figure 8 As shown, machine learning model 140 analyzes the local region pattern LM231 to obtain the optical proximity correction strategy ST231.

[0026] Then, in step S150, as Figure 8As shown, the correction unit 150 executes an optical proximity correction strategy ST231 to correct the photomask layout pattern MK23. In this step, the correction unit 150 executes the optical proximity correction strategy ST231 on the local region pattern LM231 to obtain a corrected local region pattern LM231'. The correction unit 150 then fuses the corrected local region pattern LM231' into the photomask layout pattern LM23 to obtain... Figure 9 Photomask layout diagram MK24.

[0027] Next, the process returns to step S120. In step S120, as follows... Figure 9 As shown, the defect analysis unit 120 determines whether the photomask layout pattern MK24 has any defect hotspots. If the photomask layout pattern MK24 does not have any defect hotspots, the process ends; if the photomask layout pattern MK24 has any defect hotspots, the process proceeds to step S130. Figure 9 As shown, the photomask layout diagram MK24 does not have any defect hotspots, so this process ends.

[0028] Through the above embodiments, when faced with various photomask patterns, the machine learning model 140 can analyze one or more local region patterns to obtain one or more corresponding optical proximity correction strategies, eliminating the need for operator judgment. This not only makes the judgment more accurate but also allows for continuous accumulation of experience. The following further explains how to accumulate experience to continuously retrain the machine learning model 140.

[0029] Please refer to Figures 10-11 , Figure 10 A flowchart illustrating a retraining method for a machine learning model 140 according to one embodiment is shown. Figure 11 Example description Figure 10 The steps are as follows. In step S161, the retraining unit 160 retrains the machine learning model 140 based on the defect hotspot analysis result RS of the local region patterns LM221, LM222, LM223, the optical proximity correction strategies ST221, ST222, ST223, and the corrected local region patterns LM221', LM222', LM223' (for example, local region patterns LM221' and LM222' no longer have defect hotspots, while local region pattern LM223' has defect hotspot DF231). The defect hotspot analysis result RS is the ground truth of the training data. Through this data, the machine learning model 140 can be retrained to accumulate experience, which can significantly improve the accuracy of the analysis.

[0030] In summary, although the present invention has been disclosed in conjunction with the above embodiments, it is not intended to limit the invention. Those skilled in the art to which this invention pertains can make various modifications and refinements without departing from the spirit and scope of the invention. Therefore, the scope of protection of this invention should be defined by the appended claims.

Claims

1. An optical proximity correction operation method, comprising: Obtain the photomask layout diagram; If the photomask layout has at least one defect hot spot, then based on the defect hot spot, extract the pattern of the area adjacent to the defect hot spot from the photomask layout. The machine learning model is used to analyze the pattern of the area adjacent to the at least one defect hotspot to obtain more than one optical proximity correction strategy, and the most appropriate optical proximity correction strategy is selected by cost analysis. as well as The most appropriate optical proximity correction strategy is executed to correct the photomask layout.

2. The optical proximity correction operation method as claimed in claim 1, wherein the number of the at least one defect hotspot proximity region pattern extracted from the photomask layout is greater than one.

3. The optical proximity correction operation method as described in claim 1, wherein the optical proximity correction strategies are not entirely the same.

4. The optical proximity correction operation method as described in claim 1, wherein some of the optical proximity correction strategies are the same.

5. The optical proximity correction operation method as claimed in claim 1, wherein the optical proximity correction strategies are performed simultaneously to correct the photomask layout.

6. The optical proximity correction operation method as claimed in claim 1, wherein the step of performing the most appropriate optical proximity correction strategy to correct the photomask layout includes: The most appropriate optical proximity correction strategy is applied to the pattern of the area adjacent to the defect hotspot to correct the pattern of the area adjacent to at least one defect hotspot. as well as The modified pattern of the adjacent area of ​​the at least one defect hotspot is incorporated into the photomask layout to correct the photomask layout.

7. The optical proximity correction operation method as claimed in claim 1, wherein the step of performing the most appropriate optical proximity correction strategy to correct the photomask layout includes: The most appropriate optical proximity correction strategy is applied to the photomask layout to correct the photomask layout.

8. The optical proximity correction operation method as described in claim 1, further comprising: Collect the defect hotspot analysis results of the at least one defect hotspot's neighborhood pattern, the corresponding most appropriate optical proximity correction strategy, and the corresponding corrected photomask layout; and The machine learning model is then trained based on the defect hotspot analysis results of the at least one defect hotspot neighborhood pattern, the corresponding most appropriate optical proximity correction strategy, and the corresponding corrected photomask layout.

9. An optical proximity correction operating device, comprising: Input unit, used to obtain photomask layout diagram; If the photomask layout has at least one defect hot spot, the region extraction unit extracts the pattern of the region adjacent to the defect hot spot based on the defect hot spot. A machine learning model is used to analyze the pattern of the area adjacent to at least one defect hotspot to obtain more than one optical proximity correction strategy, and the most appropriate optical proximity correction strategy is selected by cost analysis. as well as The correction unit is used to execute the most appropriate optical proximity correction strategy to correct the photomask layout.

10. The optical proximity correction operation apparatus of claim 9, wherein the number of the at least one defect hotspot proximity region patterns extracted by the region extraction unit from the photomask layout is greater than one.

11. The optical proximity correction operating apparatus of claim 9, wherein the optical proximity correction strategies are not entirely the same.

12. The optical proximity correction operating apparatus of claim 9, wherein some of the optical proximity correction strategies are the same.

13. The optical proximity correction operation apparatus of claim 9, wherein the optical proximity correction strategies are performed simultaneously to correct the photomask layout.

14. The optical proximity correction operation apparatus of claim 9, wherein the correction unit performs the most appropriate optical proximity correction strategy for the defect hotspot proximity region pattern to correct the at least one defect hotspot proximity region pattern, and the correction unit fuses the corrected at least one defect hotspot proximity region pattern into the photomask layout to correct the photomask layout.

15. The optical proximity correction operation apparatus of claim 9, wherein the correction unit performs the most appropriate optical proximity correction strategy for the photomask layout to correct the photomask layout.

16. The optical proximity correction operating apparatus of claim 9, further comprising: The retraining unit is used to retrain the machine learning model based on the defect hotspot analysis results of the at least one defect hotspot neighborhood pattern, the most appropriate optical proximity correction strategy, and the modified photomask layout.

Images