A method and device for accelerating lithography hot spot detection

Abstract

The present invention discloses a method and device for accelerating lithography hot spot detection, including: obtaining product layout data; performing lithography process hot spot inspection on the product layout data, locating hot spots, and setting the radius of the hot spot graphic area; converting the hot spot area graphic into a feature vector and mapping it to the nodes of a graph, and using a greedy algorithm to realize the formation and update of a graph-based dynamic hot spot library; localizing the product layout data and screening high-risk hot spot layout areas; performing lithography simulation and outputting lithography hot spot graphic detection results. Through the hot spot detection of layout localization, the present invention can transform the lithography process hot spot simulation from global to local processing, greatly reducing the lithography hot spot inspection time; the improved hot spot library hot spot matching method, including the hot spot graphic size selection method and the graph-based dynamic hot spot library establishment algorithm combined with the greedy algorithm, makes the hot spot library graphics easier to match and speeds up the training speed.

Description

Technical Field

[0001] The present invention relates to the field of lithography processes, and particularly to a method and device for accelerating lithography hot spot detection. Background Art

[0002] The existing technologies mainly include a hot spot matching mode based on known hot spot patterns and a hot spot detection method based on machine learning. First, it is the hot spot matching based on known hot spot patterns. Generally, manufacturers will establish a hot spot pattern library for known process hot spot patterns according to the risk level. When new design data is filed, the design layout is matched one by one with the hot spot patterns in the process hot spot library to detect whether there are known lithography process hot spots in the design layout. It can be seen from the process that the detection mode of known process hot spots includes two parts: the analysis stage and the application stage. The analysis stage mainly involves the establishment of the hot spot library. During the long-term manufacturing process, a large number of wafer-level defect patterns will be accumulated. These patterns are mapped to the designer's layout patterns, and then the process hot spot pattern library is extracted according to the danger level of the hot spot patterns. During application, the process hot spots are matched one by one with the input layout for detection. However, as the complexity of the layout increases, the complexity of the hot spot patterns will also increase exponentially, the amount of accumulation in the hot spot library will become larger and larger, the retrieval of the hot spot library will become more and more complex and time-consuming, and the efficiency will become lower and lower.

[0003] The algorithm process of the hot spot detection technology based on machine learning includes: the graphic training stage, in which relevant layout feature vectors need to be extracted first, and then machine learning methods are used to train hot spots and finally establish a recognition model. During this process, optical proximity effect correction and lithography simulation will be performed on the layout training set, and the obtained simulation results will be used as the calibration basis for machine learning. In the final detection stage, the hot spot positions will be found. This process mainly relies on the trained hot spot recognition model to check the new layout. Currently, there are many machine learning methods used for training, such as ANN (Artificial Neural Networks,) and SVM (Support Vector Machines, artificial neural networks), etc. Of course, the feature extraction method will have a greater impact on the training efficiency and model accuracy to a large extent.

[0004] In short, the hot spot detection technology based on machine learning has a fast execution speed and can also predict some hot spots that have not appeared in the training. However, due to the fact that the training set cannot cover many layout structures, the limitation of predicting new hot spots is very obvious. Moreover, training the model is very time-consuming, there are also problems with model convergence, and problems such as omission and misjudgment will occur during the actual application of the trained model.

[0005] Graph-based hotspot detection is a key area for rapid lithography hotspot search. The dynamic hotspot library built on a sparse hierarchical graph and the greedy algorithm for hotspot detection can improve search efficiency by dividing the hotspot matching and search process into multiple stages, ranking each stage in ascending order of search complexity. Furthermore, since the dynamic graph-based hotspot library detection coupled with the greedy algorithm utilizes graph node matching instead of model training, new hotspots can be mapped as nodes and embedded into the hotspot library node graph, improving the effectiveness of new hotspot detection. Summary of the Invention

[0006] The present invention aims to address the deficiencies of the prior art and to provide a method and device for accelerating photolithography hotspot detection.

[0007] The object of the present invention is achieved through the following technical solution: a method for accelerating lithography hotspot detection, the method comprising:

[0008] S1. Obtain product layout data;

[0009] S2. Check the hot spots of the lithography process on the product layout data, locate the hot spots, and set the radius of the hot spot graphic area;

[0010] S3. Convert the hotspot area graph into feature vectors and map them to graph nodes. Use a greedy algorithm to implement the formation and update of a dynamic hotspot library based on the graph.

[0011] S4. Localization of product map data and screening of high-risk hotspot map areas;

[0012] S5. Output of lithography simulation and lithography hotspot pattern detection results.

[0013] Furthermore, the radius of the hot spot pattern area is greater than 3 times the minimum design rule CD and less than 15 times the minimum design rule CD.

[0014] Furthermore, the conversion of hotspot area graphics into feature vectors and mapping them into nodes of the graph includes: converting each hotspot area graphic into a feature vector by calling a code library function, using the feature vector as the identifier of the hotspot and mapping the identifier into a node of the graph, and the nodes are connected to each other at different levels.

[0015] Furthermore, the use of the greedy algorithm to form a graph-based dynamic hotspot library is specifically as follows:

[0016] Preset the number of layers of the layout based on the size of the hotspot area graph, select the topmost layer as the traversal layer, select node J, and establish the nearest neighbor list, waiting list, and visited list of this node in the traversal layer. Set an upper limit k on the nearest neighbor list and initialize the nearest neighbor list to any point in the traversal layer except the previously selected node J.

[0017] Determine the distance relationship between each node in the waiting list and node J, and add the nodes with a distance less than the maximum distance in the nearest list to the nearest neighbor list, otherwise ignore them, until all nodes in the waiting list in this layer are traversed; if the number of nearest neighbor lists in this layer exceeds the upper limit k, select the k points closest to point J, if the number of nodes in this layer is less than k, then all nodes in this layer are added to the nearest neighbor list; all traversed nodes are added to the visited list;

[0018] After all nodes in this layer are traversed, enter the next layer to establish a new nearest neighbor list, waiting list and visited list for traversal, and loop back to the bottom layer. The nodes in the last bottom layer's nearest neighbor list whose distance to point J is less than or equal to the threshold Dth are regarded as the same type of hotspots. The selection of Dth is determined by the size of the hotspot graph; the same type of nodes maintain the original node graph, if they are not the same type of nodes, the nodes are inserted and the node graph is updated.

[0019] Furthermore, the number of layers of the layout preset according to the size of the hotspot area graphic is specifically:

[0020]

[0021] Among them, r is the radius of the intercepted figure, a represents the obtained size standard, l is the fixed length value of the segmented edge, V i Represents the number of graphic vertices, d is the size of the polygonal graphic within the layout; the radius of the hotspot graphic setting is set between 2.5a and a, and the corresponding Ly value is set to 10 to 20 layers.

[0022] Furthermore, the updating process of the hotspot database is as follows: a random number method is used to specify a layer number L for the node P that needs to be matched, and the P point will be embedded in all layers from the L layer to the bottom layer.

[0023] Then set the top layer as M, take all the nodes in the top layer as the starting point, and search the nearest neighbor list layer by layer according to the greedy algorithm. In each layer, the nearest n points in the nearest neighbor list are selected as the starting point for the next layer, until the Lth layer and the bottom layer, and the nearest neighbor list is searched layer by layer with the starting point of the Lth layer. Select the nearest j nodes in the nearest neighbor list of each layer to connect with P, and then remove the edges connecting those nodes in the Lth to the 2nd layer that are not included in the j nodes connected to P in the bottom layer and point P, and update the node graph.

[0024] Furthermore, the product layout data localization and high-risk hotspot layout area screening are specifically as follows: the principle of screening processing is to select areas with a higher probability of hotspot occurrence and exclude areas with a lower probability of hotspot occurrence.

[0025] Furthermore, in the lithography simulation, the simulation area is divided according to the results of product layout data localization and high-risk hotspot layout area screening, including

[0026] After removing the simulated IP library from the low-risk area, the intersection with the high-risk area is taken to obtain the area R that needs to be simulated. The edge area of R is extended outward by a distance d to form an annular reference area D, where the value of d is used to expand all edges in the hotspot to 0.1 to 0.5 times the minimum critical dimension CD value of the layout design; the lithography simulation area is the union of the lithography hotspot simulation area R and the annular reference area D.

[0027] According to another aspect of the specification, an accelerated lithography hotspot detection device is provided, comprising a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the executable code, the accelerated lithography hotspot detection method is implemented.

[0028] According to another aspect of the specification, a computer-readable storage medium is provided, on which a program is stored. When the program is executed by a processor, the method for accelerating lithography hotspot detection is implemented.

[0029] Beneficial effects of the present invention:

[0030] 1) Localized hotspot detection on the layout can transform the lithography process hotspot simulation from global to local processing, greatly reducing the lithography hotspot inspection time;

[0031] 2) Improved hotspot library matching methods, including hotspot graph size selection methods and a graph-based dynamic hotspot library creation algorithm combined with a greedy algorithm, effectively solve the following problems:

[0032] 1. The hotspot size is too large, which makes the hotspot layout information complex and difficult to match the hotspot library graphics, which is not conducive to the simplification and compression of hotspot graphics;

[0033] 2. The hotspot size is too small, which greatly increases the hotspot detection and matching time;

[0034] 3. Imperfect machine learning hotspot classification and matching results in a large number of graphics of the same type being divided into different hotspot graphics, increasing the complexity of the hotspot library and its matching; or a large number of graphics of different types being divided into the same hotspot graphics, leading to misjudgment;

[0035] 4. The training speed of the machine learning hotspot detection model is slow. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A flow chart of a method for accelerating lithography hotspot detection provided by an embodiment of the present invention;

[0037] Figure 2 Schematic diagram of hot spot pattern size selection provided by an embodiment of the present invention;

[0038] Figure 3 Graph of the relationship between hot spot map size selection and the number of vertices included provided by an embodiment of the present invention;

[0039] Figure 4 Flowchart of local lithography hot spot detection for the layout provided by an embodiment of the present invention;

[0040] Figure 5 Flowchart of establishing a graph-based dynamic hot spot library combined with a greedy algorithm provided by an embodiment of the present invention;

[0041] Figure 6 Flowchart of inserting new patterns into the graph-based hot spot library and updating the hot spot library provided by an embodiment of the present invention;

[0042] Figure 7 Schematic diagram of matching search for the hot spot pattern node mapping diagram provided by an embodiment of the present invention;

[0043] Figure 8 Schematic diagram of localizing the lithography hot spot area provided by an embodiment of the present invention;

[0044] Figure 9 Schematic diagram of simulation area division provided by an embodiment of the present invention;

[0045] Figure 10 Schematic diagram of a device for accelerating lithography hot spot detection provided by an embodiment of the present invention. Detailed implementation manners

[0046] The following further elaborates on the detailed implementation manners of the present invention in conjunction with the accompanying drawings.

[0047] The present invention focuses on the simulation of lithography hot spots based on hot spot library matching and layout localization. On the basis of detecting hot spots by traditional hot spot library matching, it is equivalent to appropriately processing the original design layout, and performing targeted lithography hot spot simulation will greatly shorten the time of lithography simulation.

[0048] As Figure 1 shown, a method for accelerating lithography hot spot detection provided by the present invention includes:

[0049] S1. Obtain data from the product layout;

[0050] S2. Then perform hot spot inspection on the layout (the hot spots are determined for reference in establishing the hot spot library), select an appropriate hot spot size (radius of the hot spot pattern area) according to the specific hot spot situation, and collect hot spot pattern data;

[0051] S3. Convert the hotspot area graph into a feature vector and map it to a node of the graph. Then, use a graph-based algorithm combined with a greedy algorithm to build a dynamic hotspot database and store the collected hotspot data into the hotspot database.

[0052] S4. Localize the layout data of the new product that needs to be checked for hotspots, that is, focus on the key areas of the layout and ignore the unimportant parts at the edge of the layout. Then match the hotspot graphic areas from the hotspot library. The obtained hotspot graphic areas are screened as high-risk hotspot layout areas.

[0053] S5. Perform lithography simulation on high-risk hotspot layout areas, and finally output the lithography hotspot pattern detection results.

[0054] Specifically, for Figure 1 Step S2 in the process is to locate the hotspot and select the hotspot radius. The hotspot radius selection is related to how to set the appropriate hotspot graphic size. Setting the appropriate hotspot graphic size can achieve excellent hotspot search results. Specifically, the radius value generally needs to be set to be greater than 3 times the minimum design rule CD, otherwise the established hotspot library graphic will have no matching meaning. However, a radius that is too large will include more areas for lithography process hotspot inspection, so the upper limit of the radius value of the present invention is set to 15 times the minimum design rule CD.

[0055] The size of hotspot graphics in the hotspot library directly affects their ability to match design layout graphics. Larger hotspot library graphics, with richer layout information, naturally reduce the number of matching graphics, hindering subsequent graphic compression. Conversely, smaller hotspot library graphics allow for more matching graphics, significantly increasing hotspot checking time. Further research is needed to determine the appropriate hotspot library graphic size.

[0056] Figure 2 The left half of Figure 2 shows an SEM image of a defect pattern found during the tape-out process, where a pinch-off is clearly visible in the center. Figure 2 The right half of the image shows the development simulation image of the hotspot pattern's design layout generated by the lithography process inspection simulation software package. The defect pattern selection method is to expand the process defect in four directions by a length r, ultimately forming a square area of 2r*2r. This area is ultimately stored in the hotspot library for hotspot pattern matching.

[0057] In the actual layout production process, as long as the electrical performance meets the design requirements, the layout line width will always be arranged according to the minimum design rule size. This method can obtain more effective integrated circuit devices within a certain area. Based on the area A of a single polygon in this layout i and the polygonal figure size d and perimeter C iThe relationship can be approximated by Equation 1-1,

[0058] (1-1)

[0059] Due to the requirements of the etching process and chemical mechanical polishing process for pattern uniformity, dummy patterns are added during the actual process production to keep the area density of the layout patterns basically consistent, and the segmentation of the edges is also ensured to be as consistent as possible. Let the radius of the intercepted pattern be represented by r, a represent a common constant, E i represent the number of edges in the intercepted pattern, and l be the fixed length value of the segmented edge, that is, it conforms to the following Equations 1-2 and 1-3,

[0060] (1-2)

[0061] (1-3)

[0062] In a planar figure, all figures satisfy Euler's formula, where V i represents the number of vertices of the figure, and F i represents the number of faces of the polygon figure. Since the integrated circuit design layout Figure 1 is generally a simple single-sided figure and does not involve complex topological figures. In this case, F i is always equal to l, and E i represents the number of sides of the polygon, thus obtaining Equation 1-4,

[0063] (1-4)

[0064] In summary, under the above-mentioned restrictive conditions, after simple processing, Equation 1-5 can be obtained,

[0065] (1-5)

[0066] At this point, it can be found that when the radius r of the intercepted pattern is relatively large, both the number of vertices of the design layout and the number of edges of the design layout will become larger and larger, and the complexity of the figure will become higher and higher.

[0067] Figure 3 The following shows the verification results. For the Figure 2 hotspot patterns, the radii are set to 5a, 2.5a, 1.5a, a, and 0.7a respectively to generate the corresponding number of vertices V. It can be clearly found that the number of figures and the number of vertices decrease in turn, and the complexity of the figure decreases. According to experience and experimental results, the radius is usually set between 2.5a and a, and it is judged according to the specific situation.

[0068] Figure 4The figure shows the flowchart of lithography hot spot detection for layout localization, which shows the basic process of lithography process hot spot inspection for regionalization (localization). The lithography process hot spot library is compared with the layout data of the new product, and the matching parts are positively screened out as high-risk areas. At the same time, the simulation IP library with existing simulation experience is compared with the layout data of the new product. The data in the simulation IP library are low-risk lithography layout patterns without hot spots, and the matching parts are reversely filtered out as low-risk areas. The processed areas are then subjected to lithography process hot spot simulation by a lithography process friendliness inspection software package, and finally the lithography process hot spots are obtained.

[0069] Figure 5 The figure shows the specific process of forming a graph-based dynamic hot spot library combined with the greedy algorithm and matching with high-risk areas (hot spot areas), corresponding to Figure 1 the part of forming the dynamic hot spot library in S3 and the matching of the hot spot area (high-risk area) in S4 in it. The greedy algorithm matches the high- and low-risk area matching corresponding to Figure 4 in it. The formation of the existing simulation IP library and the hot spot library adopt the same algorithm process, except that the hot spot pattern is replaced with a determined pattern of a hot spot-free (low-risk) area.

[0070] For the process of establishing and matching the hot spot library, first, the hot spot area graph with a selected radius size is converted into feature nodes. The conversion method is to call the code library function to convert each hot spot area graph into a feature vector μ h H = [y 1， y 2， y 3，.......， y n , where h represents a single hot spot pattern, and H represents the set of hot spot patterns of the entire layout. The feature vector is used as a unique identifier for the hot spot, and this identifier is mapped to the nodes of the graph. The nodes are connected to each other at different levels. The points existing at the higher level will be embedded in all lower levels, thus forming a hot spot library node graph that is sparse from the higher level to the lower level and dense. The traversal of the graph must go from one node through the connection to another node.

[0071] Specifically, the graph has a Ly layer. Ly is a predefined variable, and Ly is selected according to the radius size. For a radius size ranging from 2.5a to a, the corresponding Ly value is set to 10 to 20 layers. First, select the top layer (the Ly layer is the top layer and the first layer is the bottom layer) as the P layer. Select a node J, and establish a list C of the nearest neighbors of point J, a waiting list W, and an accessed list V in the P layer. There is no specific limit on the capacities of lists W and V, while the upper limit of the capacity of list C is k, and the value of k ranges from 50 to 2000 according to different layout situations. C is initialized to any point in the P layer other than J, and W is initialized to all points in the top layer other than point J. Judge the distance relationship between the nodes in W and point J and the distance from the nodes in C to point J. The nodes in W that are less than the maximum distance in C are put into C, otherwise they are ignored. Repeat this process until all points in C in the P layer are selected.

[0072] The meaning of "selected" is to traverse all points in W in this layer. If the number of points in this layer of the graph is greater than or equal to k, the k points with the closest distance to point J are all selected. Or since the number of nodes in this layer is less than k, all points in this layer are put into list C. The browsed nodes in W and C will be put into list V, and the nodes existing in list V will not be traversed again to prevent repeated access and improve efficiency. Specifically, the distance between nodes is defined by the Euclidean distance d: , where J = [j 1， j 2，.......， j n , Q = [q 1， q 2，.......， q n represent the feature vectors of the corresponding hot spot patterns of two nodes. Then select the next layer as the P layer, and establish a list C of the nearest neighbors of point J, a waiting list W, and an accessed list V in the new P layer. The V list of the new P layer will save and continue the data of the V list of the upper layer. The new W list is updated to all points in the current P layer except point J and the nodes in list C. The C list remains the same as the upper layer, and then repeat the node distance judgment step of the upper layer. Repeat this process until the bottom layer is traversed. Finally, the nodes in the C list of the bottom layer with a distance less than or equal to a certain threshold Dth from point J are regarded as the same type of hot spots. The selection of Dth is determined according to the size of the hot spot pattern, and the range of Dth is [1e-6, 1e-4]. The larger the pattern size, the larger Dth, and vice versa. If there are the same type of hot spots, the original node graph remains unchanged. Otherwise, execute the Figure 6 insertion algorithm process shown, and update the node graph. If all the nodes corresponding to the patterns on the product layout are traversed, the hot spot library is established.

[0073] Figure 6The figure shows a schematic diagram of the new graph insertion and hotspot library update process based on a graph. Through an algorithm, a layer number L is assigned to a specified node P that needs to be searched and matched using a random number method. Node P will be embedded in all layers from layer L to the bottom layer. This mechanism ensures that the bottom layer graph is more complex than the upper layer graphs, achieving a hierarchical node graph structure from sparse to dense, and all traversed hotspot graph corresponding nodes will be embedded in the bottom layer graph to ensure that the matching results appear in the bottom layer. Specifically, the top layer is set as layer M, and an initial starting point O is selected. Point O is selected as all nodes in the top layer because the number of top layer nodes is small. Then, in the next layer, i.e., layer M + 1, the nearest neighbor list C of O is searched. The greedy algorithm in Figure 5 is used for searching and sorting to select the n points closest to node P in C as the starting point O for the next layer search until layer L is reached. The value of n is selected from 3 to 5 according to experience based on different hotspot radius sizes. From layer L to the bottom layer, the nearest neighbor list C of P is searched using the starting point O of layer L. The j nodes closest to P in each layer C are selected to be connected to P. Then, the edges connecting the nodes that are in layers from L to 2 (the layer above the bottom layer) and not included in the j nodes connected to P in the bottom layer and connected to P are removed, and then the node graph is updated. It should be additionally noted that for building a dynamic node graph hotspot library from scratch, the initial nodes will be directly embedded in the graph

[0074] Figure 7 The figure shows a schematic diagram of the matching search of the hotspot graph node mapping graph. 701 corresponds to the graph to be searched, 705 corresponds to the matching graph, 702 corresponds to the top layer of the hotspot library node graph, 703 corresponds to several intermediate layers, and 704 represents the bottom layer. First, the required graph is mapped to nodes, and the nodes search for neighbor points layer by layer from the upper layer to the lower layer in the multi-layer graph. The neighbor points in the bottom layer whose distance from the node to be searched is less than the specified threshold are the successfully matched hotspot graphs.

[0075] Figure 8 The figure shows a schematic diagram of the layout data localization, Figure 8 、 Figure 9 corresponding to Figure 1 the layout data localization in step S4. The screening principle is to select, as much as possible, the regions with a higher probability of hotspot occurrence and exclude the regions with a lower probability of hotspot occurrence. Assume that the lithography test layout contains a global region A, and a local region B is screened out as a high-risk hotspot region (left figure). In such a case, the hotspot simulation region is reduced, and the hotspot inspection time is reduced. Similarly, if region B is found to be a lower-risk hotspot region in the global region A (right figure), then region B does not need to be inspected for lithography process hotspots, and the purpose of reducing the lithography process hotspot simulation time can also be achieved.

[0076] Figure 9The specific process rules for the simulation area division are shown as follows: Assume that the original layout data is A, the selected high-risk hotspots are B, and the IP libraries that have been simulated are C. After processing the data area, the area to be simulated is R, where R represents the graphical part in A that matches the IP libraries C that have been simulated. The logic is shown in Equation 2-1 as follows,

[0077] (2-1)

[0078] During the lithography process simulation of the final area R, it should be noted that the entire R area cannot be directly used as the lithography hot spot simulation. The optical proximity correction effect at the area edge should also be considered to prevent the area edge from affecting the lithography simulation of the lithography hot spot simulation area R. The specific method is to extend the edge area of R outward by a distance d to form an annular reference area D. During the optical proximity correction process, the logical operation of the lithography simulation area T is shown in Equation 2-2 as follows,

[0079] (2-2)

[0080] The final process simulation checks that the lithography hot spot area is still the R area. The expansion distance d of the hot spot library is mainly to expand all the edges in the hot spot by 0.1 to 0.5 times the minimum critical dimension CD value of the layout design, that is, the minimum spacing of the layout pattern profile.

[0081] Corresponding to the foregoing embodiment of a method for accelerating lithography hot spot detection, the present invention also provides an embodiment of an apparatus for accelerating lithography hot spot detection.

[0082] Refer to Figure 10 , an apparatus for accelerating lithography hot spot detection provided by an embodiment of the present invention includes a memory and one or more processors. An executable code is stored in the memory. When the processor executes the executable code, it is used to implement a method for accelerating lithography hot spot detection in the foregoing embodiment.

[0083] An embodiment of the apparatus for accelerating lithography hot spot detection provided by the present invention can be applied to any device with data processing capabilities. The any device with data processing capabilities can be a device or apparatus such as a computer. The apparatus embodiment can be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a logically meaningful apparatus, it is formed by the processor of any device with data processing capabilities where it is located reading the corresponding computer program instructions in the non-volatile memory into the memory for running. From the hardware level, as Figure 10 shown, it is a hardware structure diagram of any device with data processing capabilities where the apparatus for accelerating lithography hot spot detection provided by the present invention is located. In addition to Figure 10 In addition to the processor, memory, network interface, and non-volatile memory shown, any device with data processing capabilities where the device in the embodiment is located may generally include other hardware according to the actual functions of the device with data processing capabilities, which will not be elaborated here.

[0084] The implementation processes of the functions and roles of each unit in the above device are specifically described in the implementation processes of the corresponding steps in the above method, which will not be elaborated here.

[0085] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment. The device embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present invention. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0086] The embodiment of the present invention also provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements an accelerated lithography hot spot detection method in the above embodiment.

[0087] The computer-readable storage medium may be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium may also be an external storage device of any device with data processing capabilities, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc. equipped on the device. Further, the computer-readable storage medium may also include both an internal storage unit and an external storage device of any device with data processing capabilities. The computer-readable storage medium is used to store the computer program and other programs and data required by the device with data processing capabilities, and can also be used to temporarily store the data that has been output or will be output.

[0088] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the accelerated lithography hot spot detection method described above.

[0089] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the contents disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered merely as exemplary, and the true scope and spirit of the present application are indicated by the claims.

[0090] It should be understood that the above general description and the detailed description that follows are exemplary and explanatory only and do not limit the present application. The present application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes may be made without departing from the scope of the present application. The scope of the present application is limited only by the appended claims.

Claims

1. A method for accelerating lithography hot spot detection, characterized in that, The method includes: S1. Obtain product layout data; S2. Conduct lithography process hot spot inspection on the product layout data, locate hot spots, and set the radius of the hot spot pattern area; S3. Convert the hot spot area pattern into a feature vector and map it to a node of a graph, and use the greedy algorithm to achieve the formation and update of a dynamic hot spot library based on the graph; specifically, the use of the greedy algorithm to achieve the formation of a dynamic hot spot library based on the graph is as follows: Preset the number of layout layers according to the size of the hot spot area pattern, select the top layer as the traversal layer, select node J, and establish a list of its nearest neighbors, a waiting list, and a visited list in the traversal layer. Set a capacity limit k for the list of nearest neighbors, and initialize the list of nearest neighbors as any point in this traversal layer except the previously selected node J; Judge the distance relationship between each node in the waiting list and node J, and supplement the nodes with a distance less than the maximum distance in the nearest list into the list of nearest neighbors, otherwise ignore them until all nodes in the waiting list in this layer are traversed; if the number of nodes in the nearest neighbor list in this layer exceeds the upper limit k, then select the k points closest to point J, if the number of nodes in this layer is less than k, then put all the nodes in this layer into the list of nearest neighbors; the traversed nodes are all put into the visited list; After all nodes in this layer are traversed, enter the next layer to establish a new list of nearest neighbors, a waiting list, and a visited list for traversal, and so on until the bottom layer. The nodes in the bottom layer's list of nearest neighbors with a distance less than or equal to the threshold Dth from point J are regarded as the same type of hot spots, and Dth is selected according to the hot spot pattern size; the same type of nodes keep the original node graph, and if they are not the same type of nodes, then insert nodes and update the node graph; S4. Localize the product layout data of the new hot spots to be inspected, that is, focus on the key areas of the layout and ignore the unimportant parts at the edge of the layout, and then match the hot spot pattern area from the hot spot library, and the obtained hot spot pattern area is screened as a high-risk hot spot layout area; S5. Conduct lithography simulation on the high-risk hot spot layout area, and finally output the lithography hot spot pattern detection result.

2. The method for accelerating lithography hot spot detection according to claim 1, wherein The radius of the hot spot pattern area is greater than 3 times the minimum design rule CD and less than 15 times the minimum design rule CD.

3. The method for accelerating lithography hot spot detection according to claim 1, wherein The conversion of the hot spot area pattern into a feature vector and mapping it to a node of a graph includes: calling the code library function to convert each hot spot area pattern into a feature vector, using the feature vector as the identifier of the hot spot and mapping the identifier to a node of the graph, and the nodes are connected to each other according to different levels.

4. A method for accelerating lithography hot spot detection according to claim 1, characterized in that, Specifically, presetting the number of layout layers according to the size of the hot spot area pattern is as follows: Among them, r is the dimension radius of the intercepted figure, a represents the obtained dimension standard quantity, l is the fixed length value of the segmented side, V i represents the number of vertices of the figure, and d is the dimension of the polygon figure within the layout; the radius of the hot spot figure is set between 2.5a and a, and the corresponding Ly value is set to 10 to 20 layers.

5. A method for accelerating lithography hot spot detection according to claim 1, characterized in that, The update process of the hot spot library is specifically: use the random number method to specify a layer number L for the node P to be searched and matched, and point P will be embedded in all layers from layer L to the bottom layer. Then set the topmost layer as layer M, and use all the nodes in the top layer as the starting points. According to the greedy algorithm, search for the nearest neighbor list layer by layer. Among them, select the n nearest points in the nearest neighbor list of each layer as the starting points for searching in the next layer until reaching the Lth layer and then until the bottom layer. For each layer from the Lth layer to the bottom layer, use the starting points of the Lth layer to search for the nearest neighbor list layer by layer. Select the j nearest nodes in the nearest neighbor list of each layer to connect to P, and then remove the edges connecting the nodes from the Lth layer to the second layer that are not included in the j nodes connected to P in the bottom layer and are connected to P. Update the node graph.

6. A method for accelerating lithography hot spot detection according to claim 1, characterized in that, The localization of the product layout data and the screening of high-risk hot spot layout areas are specifically as follows: The principle of the screening process is to select the areas with a relatively high probability of hot spots and exclude the areas with a relatively low probability of hot spots.

7. A method for accelerating lithography hot spot detection according to claim 1, characterized in that, In the lithography simulation, according to the results of the localization of the product layout data and the screening of high-risk hot spot layout areas, the simulation area is divided. After removing the already simulated IP libraries from the low-risk areas and taking the intersection with the high-risk areas, the area R that needs to be simulated is obtained. Extend the edge area of R outward by a distance d to form an annular reference area D, where the value of d is used to expand all the edges in the hot spots to 0.1 times to 0.5 times the minimum critical dimension CD value of the layout design; the lithography simulation area is the union of the lithography hot spot simulation area R and the annular reference area D.

8. An apparatus for accelerating lithography hot spot detection, comprising a memory and one or more processors, wherein executable code is stored in the memory, characterized in that, When the processor executes the executable code, it implements an accelerated lithography hot spot detection method as described in any one of claims 1-7.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements an accelerated lithography hot spot detection method as described in any one of claims 1-7.

Images