Photoetching hot spot detection method and system based on graph neural network

Abstract

The invention belongs to the technical field of semiconductors, and particularly discloses a photoetching hot spot detection method and system based on a graph neural network. The method comprises the following steps: decomposing each polygonal metal in a layout into rectangles by scanning in the horizontal direction, and taking the rectangles as nodes of the layout; adding an internal edge and an external edge between the rectangles according to a specific rule; the diagonal vertex coordinates of the rectangles are used as node features, the minimum distance between the rectangles represents edge features, and graph representation of the layout is completed; inputting the graph representation of the layout into a message passing layer of a GNN model to perform graph feature extraction, and performing global maximum pooling and global average pooling on each dimension of all node features to obtain global graph representation; and classifying the global graph representation into hotspots and non-hotspots. According to the invention, the accuracy and calculation efficiency of layout hot spot detection can be improved.

Description

Technical Field

[0001] The present application belongs to the field of semiconductor technology, and more specifically, to a lithography hotspot detection method and system based on graph neural network. Background Art

[0002] The prior art proposes various lithography hotspot detection methods, including: methods based on lithography simulation, methods based on pattern matching, methods based on machine learning, and methods based on deep learning. The lithography simulation-based method uses physical and mathematical models of the manufacturing process to simulate manufacturing results and locate hotspots. It has high computing resource requirements and the slowest detection speed. The pattern matching-based method establishes a hotspot library based on known hotspots, then scans the layout to be detected, and compares each area in the layout with the known hotspot patterns in the library. If a certain area in the layout is found to be similar to a certain topological structure in the hotspot library, the area is considered to be a hotspot. The pattern matching method is faster than the lithography simulation detection speed, but the generalization is cross-over. When the detected layout has never appeared in the hotspot library, the detection accuracy is low and it takes time to build the hotspot library. The machine learning-based method applies a machine learning model to perform hotspot detection on the artificially designed layout features. Compared with the pattern matching method, the machine learning method further improves the detection accuracy and generalization, but it requires a large amount of data for training and requires manual extraction of layout features. The deep learning-based method applies a convolutional neural network (CNN). The method uses convolutional neural networks (CNNs) to automatically extract features, where the layout is directly represented as an image and used as the input of a convolutional neural network for feature extraction and binary classification, providing the highest hotspot detection accuracy compared to previous pattern matching and machine learning-based methods and faster detection speed than lithography simulation.

[0003] However, the current CNN-based lithography hotspot detection methods still have the following shortcomings: (1) The model is complex, computationally demanding, and time-consuming to train. Complex CNN models require powerful computing resources, including high-performance GPUs and large amounts of memory. They also take a lot of time to train, which is particularly evident when processing large-scale integrated circuit datasets. (2) Although the model complexity is increasing, the improvement in detection accuracy is limited. As CNN models become more and more complex, their training and inference time and model size increase significantly. However, the corresponding improvement in detection accuracy is not proportional, but only slightly improved. Summary of the invention

[0004] In view of the defects of the prior art, the purpose of this application is to provide a lithography based on graph neural network, aiming to solve the problems of low computational efficiency and insufficient measurement accuracy when the existing CNN model is used for layout hotspot detection.

[0005] To achieve the above objectives, in a first aspect, the present application provides a lithography hotspot detection method based on a graph neural network, comprising the following steps: Step 1: Based on the layout constructed by polygonal metal, decompose each polygonal metal into several independent rectangles by horizontal scanning, and take each rectangle as a node; Step 2: When adjacent rectangles form a corner of the polygonal metal, an internal edge is added between the adjacent rectangles; when the projections of adjacent rectangles overlap and the distance is less than a preset threshold, an external edge is added between the adjacent rectangles; Step 3: Use the coordinates of the diagonal vertices of the rectangle as node features and the minimum distance between rectangles as edge features to complete the graph representation of the layout; Step 4: Input the graph representation of the layout into the message passing layer of the GNN model to extract graph features by updating edge features and node features, perform global maximum pooling and global average pooling on each dimension of all node features, and obtain the global graph representation; Step 5: Input the global graph representation into the classification layer of the GNN model and classify the global graph representation into hot spots and non-hot spots.

[0006] Further preferably, the training method of the GNN model comprises the following steps: Based on the design layout, photolithography simulation is used to generate a simulated manufacturing layout, and defects are detected on the simulated manufacturing layout, and defect locations are marked; Use the effective area window for detection and calculate the overlap ratio between the defect mark and the effective area window; If the overlapping area exceeds 30%, the layout classification label is set as a hotspot, otherwise it is set as a non-hotspot; Divide the simulated manufacturing layout with classification labels into a training set and a test set; The simulated manufacturing layout in the training set is sequentially processed through steps 1, 2, and 3 to obtain the corresponding graphical representation of each polygonal metal, and input into the GNN model. The unique hot code of the classification label and the cross entropy loss of the model prediction probability are used as the cost function, and the GNN model is iteratively updated by back propagation to obtain the trained GNN model. The simulated manufacturing layout in the test set is sequentially obtained through steps 1, 2, and 3 to obtain a graphical representation, which is input into the trained GNN model for forward calculation, and the category with the highest predicted probability corresponding to the simulated manufacturing layout is output; The class with the highest predicted probability corresponding to the simulated manufacturing layout is compared with the classification label to evaluate the trained GNN model.

[0007] Further preferably, the preset threshold in step 2 is 1.2-1.5 times the minimum spacing in the layout.

[0008] Further preferably, the graph feature extraction specifically includes the following steps: The graph representation of the map is input into the message passing layer of the GNN model, and the nodes are used to aggregate the information of their neighboring nodes; after the kth message passing layer, each node will contain the information of its neighbors up to k hops away; Based on the edge features and node features between nodes in the kth message passing layer, combined with the single-layer perceptron, the edge features between nodes in the k+1th message passing layer are obtained; Based on the node features in the k message passing layers and the edge features between the nodes in the k+1th message passing layer, the node features in the k+1th message passing layer are updated using the attention coefficient and the edge-related node feature update function of the two-layer perceptron.

[0009] Further preferably, the edge feature is updated as follows:

[0010]

[0011] in, and Respectively represent nodes v and w The node characteristics of and is a single-layer perceptron; r represents The edge type of Represents a node v and nodes w The edge between them; when r=0, it represents an internal edge; when r=1, it represents an external edge; is the node in the k+1th message passing layer v and w The edge features between For cascade.

[0012] Further preferably, the node feature is updated as follows:

[0013]

[0014] in, Is with the node v A set of neighbor nodes connected by internal or external edges; It is an edge-correlated node feature update function using a two-layer perceptron; is the attention coefficient, control node v The attention it gets from its neighbors.

[0015] Further preferably, the global graph is represented as:

[0016] in, Represents the node in the last message passing layer features; GMP and GAP represent the global maximum pooling and global average pooling of each dimension of all node features respectively; V Represents a collection of nodes.

[0017] In a second aspect, the present application provides a lithography hotspot detection system based on a graph neural network, comprising: The polygonal metal decomposition module is used for the layout built based on polygonal metals. Each polygonal metal is decomposed into several independent rectangles by horizontal scanning, and each rectangle is regarded as a node. An edge adding module is used to add internal edges between adjacent rectangles when the adjacent rectangles form a corner of the polygonal metal; when the projections of adjacent rectangles overlap and the distance is less than a preset threshold, add external edges between adjacent rectangles; The feature determination module is used to use the coordinates of the diagonal vertices of the rectangle as the node features and the minimum distance between the rectangles as the edge features to complete the graph representation of the layout; The global graph representation acquisition module is used to input the graph representation of the layout into the message passing layer of the GNN model to extract graph features by updating edge features and node features, and perform global maximum pooling and global average pooling on each dimension of all node features to obtain the global graph representation; The classification and discrimination module is used to input the global graph representation into the classification layer of the GNN model and classify the global graph representation into hot spots and non-hot spots.

[0018] Further preferably, the lithography hotspot detection system further includes a GNN model training module, and the GNN model training module includes: A defect marking unit, used to generate a simulated manufacturing layout based on the design layout by using photolithography simulation, and to perform defect detection on the simulated manufacturing layout and mark defect positions; The classification unit is used to detect using the effective area window and calculate the overlap ratio between the defect mark and the effective area window. If the overlap area exceeds 30%, the layout classification label is set as a hotspot, otherwise it is set as a non-hotspot. A layout division unit, used for dividing the simulated manufacturing layout with classification labels into a training set and a test set; The model training unit is used to drive the polygon metal decomposition module, the edge addition module and the feature determination module to obtain the graph representation corresponding to each polygon metal with the simulated manufacturing layout in the training set as the object, and input it into the GNN model, use the unique hot code of the classification label and the cross entropy loss of the model prediction probability as the cost function, and iteratively update the GNN model by back propagation to obtain the trained GNN model; The model evaluation unit is used to drive the polygonal metal decomposition module, the edge addition module and the feature determination module to obtain the graph representation corresponding to each polygonal metal with the simulated manufacturing layout in the test set as the object, and input it into the trained GNN model for forward calculation, and output the category with the largest predicted probability corresponding to the simulated manufacturing layout; compare the category with the largest predicted probability corresponding to the simulated manufacturing layout with the classification label to evaluate the trained GNN model.

[0019] Further preferably, the global graph representation acquisition module includes a graph feature extraction unit, and the graph feature extraction unit includes: The node aggregation subunit is used to input the graph representation of the layout into the message passing layer of the GNN model, and use the nodes to aggregate the information of its neighboring nodes; after the kth message passing layer, each node will contain the information of its neighbors up to k hops away; An edge feature updating subunit, for obtaining edge features between nodes in the k+1th message passing layer based on edge features and node features between nodes in the kth message passing layer in combination with a single-layer perceptron; The node feature updating subunit is used to update the node features in the k+1th message passing layer based on the node features in the k message passing layers and the edge features between the nodes in the k+1th message passing layer, using the attention coefficient and the edge-related node feature update function of the two-layer perceptron.

[0020] Further preferably, the edge feature updating subunit in the edge feature updating subunit is as follows:

[0021]

[0022] in, and Respectively represent nodes v and w The node characteristics of and is a single-layer perceptron; r represents The edge type of Represents a node v and nodes w The edge between them; when r=0, it represents an internal edge; when r=1, it represents an external edge; is the node in the k+1th message passing layer v and w The edge features between For cascade.

[0023] Further preferably, the node feature update in the node feature update subunit is as follows:

[0024]

[0025] in, Is with the node v A set of neighbor nodes connected by internal or external edges; It is an edge-correlated node feature update function using a two-layer perceptron; is the attention coefficient, control node v The attention it gets from its neighbors.

[0026] Further preferably, the global graph representation in the global graph representation acquisition module is:

[0027] in, Represents the node in the last message passing layer features; GMP and GAP represent the global maximum pooling and global average pooling of each dimension of all node features respectively; V Represents a collection of nodes.

[0028] In a third aspect, the present application provides an electronic device comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method described in the first aspect or any possible implementation of the first aspect.

[0029] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0030] In a fifth aspect, the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0031] It can be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0032] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the prior art: This application provides a lithography hotspot detection method based on graph neural network, which represents the layout as a graph, that is, decomposing the polygons in the layout into rectangles and treating the rectangles as nodes of the graph, adding different types of edges according to the positional relationship between the rectangles, and finally using the graph representation of the layout as the input of GNN for hotspot detection. Compared with the existing CNN-based hotspot detection technology that requires the entire layout image as input, it greatly reduces the input size and model size, and improves the computational efficiency.

[0033] The present application provides a lithography hotspot detection method based on a graph neural network, which designs node features and edge features in combination with key geometric information related to the generation of hotspots, that is, the absolute position (coordinates) of rectangles are used as node features, and the relative position (distance) between rectangles are used as edge features; and a GNN architecture that is consistent with the graph representation and efficient is designed, that is, a graph attention mechanism is combined and different update functions are used according to the type of edge to perform deeper feature extraction on the graph representation, thereby improving detection accuracy and overcoming the problem of low detection accuracy caused by the existing hotspot detection model based on deep learning that uses features unrelated to hotspots for classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is the construction and feature representation of the layout diagram provided by the embodiments of the present application; FIG. 2 ( a ) is an edge diagram of a construction layout provided in an embodiment of the present application; FIG2( b ) is a schematic diagram of the internal edge of a layout provided in an embodiment of the present application; FIG2( c ) is a schematic diagram of the outer edge of a layout provided in an embodiment of the present application; Figure 3 is a schematic diagram of geometric information of two adjacent rectangles provided in an embodiment of the present application; Figure 4 It is a GNN-based hotspot detector architecture provided in an embodiment of the present application; Figure 5 The hotspot layout slices and non-hotspot layout slices provided in the embodiment of the present application and their output after lithography simulation; wherein, the polygons in the dotted area are too close to each other, which will cause a short circuit in the area after lithography, i.e., a lithography hotspot is generated; Figure 6 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0036] The term "and / or" in this article is a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The symbol " / " in this article indicates that the associated objects are in an or relationship, for example, A / B means A or B.

[0037] The terms “first”, “second” and the like in the specification and claims herein are used to distinguish different objects rather than to describe a specific order of the objects.

[0038] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0039] In the description of the embodiments of the present application, unless otherwise specified, “plurality” means two or more than two.

[0040] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application.

[0041] The lithography hotspot detection method based on Graph Neural Network (GNN) includes two steps: the first step is to represent the layout with a graph; the second step is to use the graph representation of the layout as the input of the graph neural network for classification; like Figure 1 As shown, the construction and characterization of the layout diagram representation specifically includes the following steps: Step 1.1: polygon decomposition; The layout is composed of complex polygons, and the interactions between them will produce lithography hotspots; however, treating each polygon as a single graph node will lose key information for hotspot detection, because the interactions also exist inside the metal polygons; for example, the corner information of the polygon involves the horizontal and vertical parts of the same metal, containing rich information about local printability; in addition, the shapes of metal polygons vary greatly, and it is difficult to describe the position and shape of the polygons in a simple and unified format; therefore, the present application decomposes each polygon into several independent rectangles using horizontal scan lines, and regards each rectangle as a node in the layout; wherein the polygons can be decomposed into regular rectangles; the vertical direction refers to the x-axis direction in the Cartesian coordinate system; Step 1.2: Graph construction; After decomposing the metal polygons to obtain the nodes represented by the layout graph, the connections between the interacting rectangles are added as edges in Figure 2 (a); the present application defines two types of edges between the decomposed rectangles, namely internal edges and external edges; rectangles from the same polygon are naturally connected due to their internal interactions to form corners; as shown in Figure 2 (b), the edges between adjacent rectangles are added as internal edges; when the projections of adjacent polygons overlap (such as Figure 3 When the distance is less than a predefined threshold r, an external edge is added between adjacent polygons. For example, in FIG2(c), an edge is added between rectangles with distances d1 and d2 (less than r), while there is no edge connection between rectangles with distance d3 (greater than r); the external edge represents the interaction between rectangles of adjacent polygons; wherein r is set to 80 nm, which is approximately equal to 1.2 times the minimum spacing in the design rules of the layout data set used in this application; Step 1.3: Extract node and edge features represented by the layout graph; With the nodes and edges represented by the layout graph, the features of the nodes and edges can be defined. The features can represent the key information used for hotspot detection in the layout slice, including the position and shape information of the decomposed rectangles and the interaction between them; the vertex coordinates of each rectangle in the layout plane are used as node features to accurately reflect the position and shape information of the rectangle; the smaller the distance between rectangles with external edges, the more likely it is to form a hotspot, so the distance between rectangles is used as the edge feature; the distance of the internal edges is regarded as 0; specifically, the feature representation of nodes and edges is defined as follows: Node feature representation: coordinates; the coordinates of the upper left corner vertex and the lower right corner vertex of the decomposed rectangle directly describe their absolute positions and contain their shape information, such as Figure 3 The node features corresponding to the middle rectangle R1 are (x1, y1) and (x2, y2); Edge features represent: Distance; the distance between two rectangles represents the minimum gap between their closest parallel boundaries, e.g. Figure 3d in; After obtaining the layout representation, it is used as the input of the graph neural network and classified. The graph neural network architecture for layout hotspot detection is designed as follows: The hotspot detection architecture based on the graph neural network includes graph feature extraction, reading and classification, such as Figure 4 As shown, |V| is the number of nodes and |E| is the number of edges; The graph neural network performs a hotspot classification method based on graph representation, which specifically includes the following steps: Step 2.1: Graph feature extraction; Similar to the convolutional layers in CNN that extract image features in computer vision tasks, the message passing layers in GNN ( Figure 4 Message-Passing in also extracts graph features; in each message passing layer, each node collects and aggregates information about its neighboring nodes (i.e., nodes with edges to the node) to capture the local graph structure and update its feature representation; by stacking multiple message passing layers, information is propagated between nodes in the graph; after the kth message passing layer, each node will contain information about its neighbors that are up to k hops away; among them, node 1 is connected to node 2; node 2 is connected to node 3; node 1 is not connected to node 3; after one message passing layer, node 1 contains the features of node 2, and then proceeds to the next message passing layer. Since node 2 is connected to node 3, node 1 also contains the features of node 3; the value of k is set manually, specifically, k is an integer greater than or equal to 1; More specifically, graph feature extraction includes the following steps: Step 2.1.1: Edge feature update method in graph feature extraction, specifically: In order to extract more effective features of the interactions between graph nodes, i.e., edges, different update functions are applied to internal and external edges. Specifically, in the kth message passing layer, the edge features are updated , as shown in formula (1) and formula (2); (1) (2) in, and Respectively represent nodes v and w The node characteristics of and is a single-layer perceptron; r represents The edge type of v and w For different nodes; Represents a node v and w The edge between them; when r=0, it represents an internal edge; when r=1, it represents an external edge; is the node in the k+1th message passing layer v and w The edge features between For cascade; Step 2.1.2: Node feature update method in graph feature extraction, specifically: Similarly, the node features are updated as shown in formula (3) and formula (4); (3) (4) in, It is the set of neighbor nodes connected to node v by internal or external edges; It is an edge-correlated node feature update function using a two-layer perceptron; In practical applications, edge features are updated first, followed by node features; Inspired by Grape Attention Networks (GAT), a graph attention mechanism is introduced into the architecture to learn the different importance of adjacent nodes in contributing to node feature updates; attention coefficient The calculation of is shown in formula (5): (5) in, a is a learnable vector that controls the node v the attention it gets from its neighbors; Step 2.2: Reading and classification; When the features represented by the layout are k Layer Message Passing After layer extraction, the aggregate feature representation of all polygonal metals can be obtained by formula (6); (6) in, Represents the node in the last message passing layer The features of , GMP and GAP represent the global maximum pooling and global average pooling of each dimension of all node features respectively; by connecting the global maximum value and global average information, the global graph representation y is obtained; the two-layer perceptron can further y Classified into hot spots and non-hot spots.

[0042] Example 1 Perform lithography simulation on the collected design layout to generate a simulated manufacturing layout; perform defect detection on the simulated manufacturing layout and mark the defect location; use an effective area window, i.e. Figure 5The predefined region of interest (ROI) shown in the figure is inspected and the overlap ratio of the equivalent defect mark and the effective area window is calculated; if the overlap area exceeds 30%, the layout classification label is set to "hotspot", otherwise it is set to "non-hotspot"; the data set is divided into a training set and a test set; The layout is as follows Figure 1 The steps to construct a graph representation are as follows: Figure 4 The GNN model shown in the figure is constructed, and the constructed layout graph representation is used as the input of the GNN model. In the training phase, the one-hot code of the "classification label" and the cross entropy loss of the model prediction probability are used as the cost function, and the model is iteratively updated through back propagation. In the testing phase, the above steps are also followed to use the layout graph representation as the input of the GNN model for forward calculation, and the category with the largest prediction probability is output.

[0043] Based on the GNN model obtained through the above training, this application provides the following specific example of testing the GNN model to illustrate the advantages of using the GNN model to classify hotspots and non-hotspots in the layout: More specifically, the following steps are included: Step a. Build a dataset The proposed GNN-based hotspot detector is evaluated on two commonly used datasets, one is the smaller ICCAD2012 dataset used by many previous methods; the other is an enhanced dataset in VTS 2018, which contains a large number of layout slices and is later used by other methods; the ICCAD 20212 dataset consists of 4 subsets (28 nm designs) and 1 subset (32 nm designs); this application merged and used all 28 nm designs in the experiment; The VTS2018 dataset is made based on the freely available 45nm FreePDK and Mentor Calibre verified lithography technology; the following table lists the detailed information of the two datasets (the size of the hotspot HS and non-hotspot NHS slices in different datasets ( ) and quantity); Convert the layout slices in ICCAD 2012 to a resolution of 1200 1200 images and convert the layout slices in VTS 2018 to a resolution of 1110 1110; specifically, the metal area of ​​the layout slice is represented by an image pixel with a normalized value of 1, and the layout background is represented by a value of 0;

[0044] Step b. Detection accuracy evaluation index Hotspot Accuracy (HA): the proportion of hotspots detected in real hotspot slices; False Alarm (FA): the number of non-hotspot slices that are mistakenly detected as hotspot slices; This application proposes a GNN-based hotspot detector and compares it with three state-of-the-art deep learning-based hotspot detectors; among them, the model architecture of TCAD'19 and TCAD'22 is end-to-end CNN, and DATE'22 uses GNN for hotspot detection; TCAD'19 proposed using the Discrete Cosine Transform (DCT) coefficients of layout slices as input features of CNN; TCAD'22 applies attention modules and deep metric learning to more unique hotspot and non-hotspot feature extraction, which stacks five Inception modules with CBAM attention and three fully connected layers for classification and metric learning; DATE'22 uses five node features and four edge features to represent the layout slice as a graph and further processes it using a GNN model for binary classification; The following table shows the detection accuracy and inference time of each layout slice and model size of various deep learning-based hotspot detectors on two datasets, where “Time ( )” represents the reasoning time; Detection accuracy: The GNN model achieves the highest HA and the lowest FA on the ICCAD 2012 and VTS 2018 datasets; specifically, on the larger and more difficult VTS 2018 dataset, GNN has an advantage over other models in detection accuracy; Inference time: Compared to the slowest TCAD'19, GNNs require less inference time than CNNs, up to 31.8x and 29.2x faster on both datasets; in general, GNNs are an order of magnitude faster than CNNs; GNNs take slightly longer to infer than DATE'22, which also uses GNNs; Model size: Compared with all existing technologies, GNN has the smallest model size, which is only about 1 / 2.5 and 1 / 64 of TCAD'19 and TCAD'22 respectively;

[0045] Example 2 The present application provides a lithography hotspot detection system based on a graph neural network, comprising: The polygon metal decomposition module is used for the layout built based on polygons. Each polygon metal is decomposed into several independent rectangles by horizontal scanning, and each rectangle is regarded as a node. An edge adding module is used to add internal edges between adjacent rectangles when the adjacent rectangles form a corner of the polygonal metal; when the projections of adjacent rectangles overlap and the distance is less than a preset threshold, add external edges between adjacent rectangles; The feature determination module is used to use the coordinates of the diagonal vertices of the rectangle as the node features and the minimum distance between the rectangles as the edge features to complete the graph representation of the layout; The global graph representation acquisition module is used to input the graph representation of the layout into the message passing layer of the GNN model to extract graph features by updating edge features and node features, and perform global maximum pooling and global average pooling on each dimension of all node features to obtain the global graph representation; The classification and discrimination module is used to input the global graph representation into the classification layer of the GNN model and classify it into hot spots and non-hot spots.

[0046] Further preferably, the lithography hotspot detection system further includes a GNN model training module, and the GNN model training module includes: A defect marking unit, used to generate a simulated manufacturing layout based on the design layout by using photolithography simulation, and to perform defect detection on the simulated manufacturing layout and mark defect positions; The classification unit is used to detect using the effective area window and calculate the overlap ratio between the defect mark and the effective area window. If the overlap area exceeds 30%, the layout classification label is set as a hotspot, otherwise it is set as a non-hotspot. A layout division unit, used for dividing the simulated manufacturing layout with classification labels into a training set and a test set; The model training unit is used to drive the polygon metal decomposition module, the edge addition module and the feature determination module to obtain the graph representation corresponding to each polygon metal with the simulated manufacturing layout in the training set as the object, and input it into the GNN model, use the unique hot code of the classification label and the cross entropy loss of the model prediction probability as the cost function, and iteratively update the GNN model by back propagation to obtain the trained GNN model; The model evaluation unit is used to drive the polygonal metal decomposition module, the edge addition module and the feature determination module to obtain the graph representation corresponding to each polygonal metal with the simulated manufacturing layout in the test set as the object, and input it into the trained GNN model for forward calculation, and output the category with the largest predicted probability corresponding to the simulated manufacturing layout; compare the category with the largest predicted probability corresponding to the simulated manufacturing layout with the classification label to evaluate the trained GNN model.

[0047] Further preferably, the global graph representation acquisition module includes a graph feature extraction unit, and the graph feature extraction unit includes: The node aggregation subunit is used to input the graph representation of the layout into the message passing layer of the GNN model, and use the nodes to aggregate the information of its neighboring nodes; after the kth message passing layer, each node will contain the information of its neighbors up to k hops away; An edge feature updating subunit, for obtaining edge features between nodes in the k+1th message passing layer based on edge features and node features between nodes in the kth message passing layer in combination with a single-layer perceptron; The node feature updating subunit is used to update the node features in the k+1th message passing layer based on the node features in the k message passing layers and the edge features between the nodes in the k+1th message passing layer, using the attention coefficient and the edge-related node feature update function of the two-layer perceptron.

[0048] Further preferably, the edge feature updating subunit in the edge feature updating subunit is as follows:

[0049]

[0050] in, and Respectively represent nodes v and w The node characteristics of and is a single-layer perceptron; r represents The edge type of Represents a node v and nodes w The edge between them; when r=0, it represents an internal edge; when r=1, it represents an external edge; is the node in the k+1th message passing layer v and w The edge features between For cascade.

[0051] Further preferably, the node feature update in the node feature update subunit is as follows:

[0052]

[0053] in, Is with the node v A set of neighbor nodes connected by internal or external edges; It is an edge-correlated node feature update function using a two-layer perceptron; is the attention coefficient, controlling the node v The attention it gets from its neighbors.

[0054] Further preferably, the global graph representation in the global graph representation acquisition module is:

[0055] in, Represents the node in the last message passing layer features; GMP and GAP represent the global maximum pooling and global average pooling of each dimension of all node features respectively; V Represents a collection of nodes.

[0056] In summary, compared with the prior art, the present application has the following advantages: This application provides a lithography hotspot detection method based on graph neural network, which represents the layout as a graph, that is, decomposing the polygons in the layout into rectangles and treating the rectangles as nodes of the graph, adding different types of edges according to the positional relationship between the rectangles, and finally using the graph representation of the layout as the input of GNN for hotspot detection. Compared with the existing CNN-based hotspot detection technology that requires the entire layout image as input, it greatly reduces the input size and model size, and improves the computational efficiency.

[0057] The present application provides a lithography hotspot detection method based on a graph neural network, which designs node features and edge features in combination with key geometric information related to the generation of hotspots, that is, the absolute position (coordinates) of rectangles are used as node features, and the relative position (distance) between rectangles are used as edge features; and a GNN architecture that is consistent with the graph representation and efficient is designed, that is, a graph attention mechanism is combined and different update functions are used according to the type of edge to perform deeper feature extraction on the graph representation, thereby improving detection accuracy and overcoming the problem of low detection accuracy caused by the existing hotspot detection model based on deep learning that uses features unrelated to hotspots for classification.

[0058] It should be understood that the above-mentioned system is used to execute the methods in the above-mentioned embodiments. The implementation principles and technical effects of the corresponding program modules in the system are similar to those described in the above-mentioned methods. The working process of the device can refer to the corresponding process in the above-mentioned method and will not be repeated here.

[0059] Based on the method in the above embodiment, Figure 6 As shown, an embodiment of the present application provides an electronic device, which may include: a processor (Processor) 810, a communication interface (Communications Interface) 820, a memory (Memory) 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 can call the logic instructions in the memory 830 to execute the method in the above embodiment.

[0060] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application.

[0061] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0062] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0063] It is understandable that the processor in the embodiment of the present application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0064] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0065] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0066] It should be understood that the various numerical numbers involved in the embodiments of the present application are only used for the convenience of description and are not used to limit the scope of the embodiments of the present application.

[0067] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A lithography hotspot detection method based on graph neural network, characterized in that: The following steps are involved: Step 1: Based on the layout constructed by polygonal metal, decompose each polygonal metal into several independent rectangles by horizontal scanning, and take each rectangle as a node; Step 2: When adjacent rectangles form a corner of the polygonal metal, an internal edge is added between the adjacent rectangles; when the projections of adjacent rectangles overlap and the distance is less than a preset threshold, an external edge is added between the adjacent rectangles; Step 3: Use the coordinates of the diagonal vertices of the rectangle as node features and the minimum distance between rectangles as edge features to complete the graph representation of the layout; Step 4: Input the graph representation of the layout into the message passing layer of the GNN model to extract graph features by updating edge features and node features, perform global maximum pooling and global average pooling on each dimension of all node features, and obtain the global graph representation; Step 5: Input the global graph representation into the classification layer of the GNN model and classify the global graph representation into hot spots and non-hot spots.

2. The photolithography hotspot detection method according to claim 1, characterized in that: The training method of the GNN model specifically includes the following steps: Based on the design layout, photolithography simulation is used to generate a simulated manufacturing layout, and defects are detected on the simulated manufacturing layout, and defect locations are marked; Use the effective area window for detection and calculate the overlap ratio between the defect mark and the effective area window; If the overlapping area exceeds 30%, the layout classification label is set as a hotspot, otherwise it is set as a non-hotspot; Divide the simulated manufacturing layout with classification labels into a training set and a test set; The simulated manufacturing layout in the training set is sequentially input into the GNN model through steps 1, 2 and 3 to obtain the graph representation, and the cross entropy loss of the unique hot code of the classification label and the model prediction probability is used as the cost function, and the GNN model is iteratively updated by back propagation to obtain the trained GNN model; The simulated manufacturing layout in the test set is sequentially processed through steps 1, 2, and 3 to obtain the graph representation, which is input into the trained GNN model for forward calculation, and the output graph representation corresponds to the category with the highest prediction probability; The class with the highest predicted probability corresponding to the simulated manufacturing layout is compared with the classification label to evaluate the trained GNN model.

3. The photolithography hotspot detection method according to claim 1, characterized in that: The preset threshold in step 2 is 1.2-1.5 times the minimum spacing in the layout.

4. The photolithography hotspot detection method according to claim 1 or 2, characterized in that: The image feature extraction in step 4 specifically includes the following steps: The graph representation of the map is input into the message passing layer of the GNN model, and nodes are used to aggregate the information of their neighboring nodes; after the kth message passing layer, each node will contain the information of its neighbors up to k hops away; Based on the edge features and node features between nodes in the kth message passing layer, combined with the single-layer perceptron, the edge features between nodes in the k+1th message passing layer are obtained; Based on the node features in the k message passing layers and the edge features between the nodes in the k+1th message passing layer, the node features in the k+1th message passing layer are updated using the attention coefficient and the edge-related node feature update function of the two-layer perceptron.

5. The photolithography hotspot detection method according to claim 4, characterized in that: The edge feature is updated as: in, and Respectively represent nodes v and w The node characteristics of and is a single-layer perceptron; r represents The edge type of Represents a node v and nodes w The edge between them; when r=0, it represents an internal edge; when r=1, it represents an external edge; is the node in the k+1th message passing layer v and w The edge features between For cascade; The node features are updated as follows: in, Is with the node v A set of neighbor nodes connected by internal or external edges; It is an edge-correlated node feature update function using a two-layer perceptron; is the attention coefficient, controlling the node v The attention it gets from its neighbors.

6. The photolithography hotspot detection method according to claim 5, characterized in that: The global graph is represented as: in, Represents the node in the last message passing layer features; GMP and GAP represent the global maximum pooling and global average pooling of each dimension of all node features respectively; V Represents a collection of nodes.

7. A lithography hotspot detection system based on graph neural network, characterized in that: include: The polygonal metal decomposition module is used for the layout built based on polygonal metals. Each polygonal metal is decomposed into several independent rectangles by horizontal scanning, and each rectangle is regarded as a node. An edge adding module is used to add internal edges between adjacent rectangles when the adjacent rectangles form a corner of the polygonal metal; when the projections of adjacent rectangles overlap and the distance is less than a preset threshold, add external edges between adjacent rectangles; The feature determination module is used to use the coordinates of the diagonal vertices of the rectangle as the node features and the minimum distance between the rectangles as the edge features to complete the graph representation of the layout; The global graph representation acquisition module is used to input the graph representation of the layout into the message passing layer of the GNN model to extract graph features by updating edge features and node features, and perform global maximum pooling and global average pooling on each dimension of all node features to obtain the global graph representation; The classification and discrimination module is used to input the global graph representation into the classification layer of the GNN model and classify the global graph representation into hot spots and non-hot spots.

8. The photolithography hotspot detection system according to claim 7, characterized in that: It also includes a GNN model training module, which includes: A defect marking unit, used to generate a simulated manufacturing layout based on the design layout by using photolithography simulation, and to perform defect detection on the simulated manufacturing layout and mark defect positions; The classification unit is used to detect using the effective area window and calculate the overlap ratio between the defect mark and the effective area window. If the overlap area exceeds 30%, the layout classification label is set as a hotspot, otherwise it is set as a non-hotspot. A layout division unit, used for dividing the simulated manufacturing layout with classification labels into a training set and a test set; The model training unit is used to drive the polygon metal decomposition module, the edge addition module and the feature determination module to obtain the graph representation corresponding to each polygon metal with the simulated manufacturing layout in the training set as the object, and input it into the GNN model, use the unique hot code of the classification label and the cross entropy loss of the model prediction probability as the cost function, and iteratively update the GNN model by back propagation to obtain the trained GNN model; The model evaluation unit is used to drive the polygonal metal decomposition module, the edge addition module and the feature determination module to obtain the graph representation corresponding to each polygonal metal with the simulated manufacturing layout in the test set as the object, and input it into the trained GNN model for forward calculation, and output the category with the largest predicted probability corresponding to the simulated manufacturing layout; compare the category with the largest predicted probability corresponding to the simulated manufacturing layout with the classification label to evaluate the trained GNN model.

9. The photolithography hotspot detection system according to claim 7, characterized in that: The global graph representation acquisition module includes a graph feature extraction unit, which includes: The node aggregation subunit is used to input the graph representation of the layout into the message passing layer of the GNN model, and use the nodes to aggregate the information of its neighboring nodes; after the kth message passing layer, each node will contain the information of its neighbors up to k hops away; An edge feature updating subunit, for obtaining edge features between nodes in the k+1th message passing layer based on edge features and node features between nodes in the kth message passing layer in combination with a single-layer perceptron; The node feature updating subunit is used to update the node features in the k+1th message passing layer based on the node features in the k message passing layers and the edge features between the nodes in the k+1th message passing layer, using the attention coefficient and the edge-related node feature update function of the two-layer perceptron.

10. The photolithography hotspot detection system according to claim 9, characterized in that: The edge feature update subunit is as follows: in, and Respectively represent nodes v and w The node characteristics of and is a single-layer perceptron; r represents The edge type of Represents a node v and nodes w The edge between them; when r=0, it represents an internal edge; when r=1, it represents an external edge; is the node in the k+1th message passing layer v and w The edge features between For cascade; The node features in the node feature update subunit are updated as follows: in, Is with the node v A set of neighbor nodes connected by internal or external edges; It is an edge-correlated node feature update function using a two-layer perceptron; is the attention coefficient, controlling the node v the attention it gets from its neighbors; The global graph representation in the acquisition module is represented as: in, Represents the node in the last message passing layer features; GMP and GAP represent the global maximum pooling and global average pooling of each dimension of all node features respectively; V Represents a collection of nodes.

Images