Model division method of semiconductor device and application thereof

Abstract

The invention provides a model division method of a semiconductor device and application of the model division method, and relates to the technical field of semiconductor devices.The method comprises the steps that a three-dimensional structure model of the semiconductor device is read; performing initial division on the three-dimensional structure model to obtain a plurality of initial grids; constructing feature points based on the index relationship of the initial grids; wherein in different initial grids, the feature points corresponding to the same basic feature coincide; the feature points comprise one or more of body feature points, surface feature points and edge feature points; dividing the initial grid into a plurality of final grids by combining the original nodes and the feature points corresponding to the initial grid; and after all the initial grids are divided, all the final grids are summarized to obtain a target structure model, and the division of the three-dimensional structure model is completed, so that the calculation amount is reduced and the prediction rate is improved when the performance expression data for optimizing the performance design is generated.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor devices, and in particular, to a method and system for model partitioning and application of a semiconductor device, an electronic device, a computer storage medium, and a computer program product. Background Art

[0002] A semiconductor device is an electronic device that has electrical conductivity between that of a good conductor and an insulator and uses the special electrical properties of semiconductor materials to perform specific functions.

[0003] In order to improve the performance design of semiconductor devices in terms of photoelectric conversion efficiency, response speed, energy consumption control, and device stability, it is necessary to evaluate the performance of semiconductor devices. Among them, the generation process of photo-generated carriers is an important factor in the performance of semiconductor devices. Therefore, accurately predicting the generation rate of photo-generated carriers in semiconductor devices is of great significance for optimizing the design of semiconductor devices and improving device performance.

[0004] However, existing prediction methods mainly rely on a variety of numerical simulation methods, which generally face problems such as a large amount of calculation, long calculation time, and insufficient accuracy.

[0005] A method and system for model partitioning and application of a semiconductor device, an electronic device, a computer storage medium, and a computer program product are hereby proposed. Summary of the Invention

[0006] This specification provides a method and system for model partitioning and application of a semiconductor device, an electronic device, a computer storage medium, and a computer program product. By subdividing the three-dimensional structure model into final grids, non-uniform and complex geometric structures can be flexibly processed, which helps to overcome the limitations of the transfer matrix method when dealing with such structures; when predicting performance data in combination with ray tracing technology, the amount of calculation is reduced by reducing unnecessary acceleration structure construction and complex intersection detection; by directly performing intersections on the subdivided final grids, the prediction efficiency and accuracy are improved.

[0007] A method for model partitioning of a semiconductor device provided in this application adopts the following technical solutions, including: Read the three-dimensional structure model of the semiconductor device; Perform an initial partition on the three-dimensional structure model to obtain a number of initial grids; Construct feature points based on the index relationship of the initial grids; among them, in different initial grids, the feature points corresponding to the same basic feature coincide; the feature points include one or more of volume feature points, surface feature points, and edge feature points; Combine the original nodes and feature points corresponding to the initial grids to divide the initial grids into a number of final grids; After all the initial meshes are divided, all the final meshes are aggregated to obtain the target structure model, completing the division of the three-dimensional structure model.

[0008] Optionally, the three-dimensional structure model is initially divided to obtain a plurality of initial grids, including: Obtain the partition attribute value input by the user as a configuration parameter; Finite element meshing is performed on the three-dimensional structure model based on the configuration parameters to obtain a plurality of initial meshes.

[0009] Optionally, the volume feature point is the centroid of the initial grid; and / or, The surface feature point is the centroid of the original surface of the initial mesh; and / or, The edge feature point is the midpoint of the original edge of the initial grid.

[0010] Optionally, combining the original nodes and feature points corresponding to the initial grid to divide the initial grid into a plurality of final grids includes: Determine the current original node; Determine whether the node index number of the current original node exceeds the maximum node index number; If the node index number does not exceed the maximum node index number, searching for a target feature point associated with the current original node; The final grid is constructed by combining the current original node, the corresponding target feature point set and the preset topological relationship.

[0011] Optionally, after combining the current original node, the corresponding target feature point set and the preset topological relationship to construct the final grid, the method further includes: Determine whether the current original node is the last original node in the original node list; If yes, determine whether the current initial grid is the last initial grid in the initial grid list. If yes, it indicates that all initial grids have been divided. Otherwise, take the next initial grid of the current initial grid as the new current initial grid. Otherwise, the next original node of the current original node is used as the new current original node.

[0012] The present application provides an application method of model division, which predicts the performance of semiconductor devices based on the target structure model obtained by the aforementioned model division method, and adopts the following technical solutions, including: Extracting light information of the irradiated light and a target structure model of the semiconductor device; the target structure model includes a plurality of final grids; Determine a first intersection point based on the intersection of the light information and the final grid; Update the light information according to the reflection result of the irradiation light at the first intersection point; Find the first final grid corresponding to the first intersection point, simulate the light trace of the irradiation light in the first final grid in combination with the light information, and determine the second intersection point and the performance data of the second intersection point; If the second basic surface where the second intersection point is located is of the target type, obtain the performance data of the semiconductor device in combination with the simulation result and / or performance data of the light tracing.

[0013] A model partitioning system for a semiconductor device provided by this application adopts the following technical solutions, including: A reading module, configured to read a three-dimensional structure model of a semiconductor device; An initial partitioning module, configured to perform an initial partition on the three-dimensional structure model to obtain a plurality of initial grids; A feature point construction module, configured to construct feature points based on the index relationship of the initial grids; wherein, in different initial grids, the feature points corresponding to the same basic feature coincide; the feature points include one or more of volume feature points, surface feature points, and edge feature points; A final partitioning module, configured to partition the initial grids into a plurality of final grids in combination with the original nodes and feature points corresponding to the initial grids; A summarizing module, configured to summarize all the final grids to obtain a target structure model after all the initial grids are partitioned, and complete the partitioning of the three-dimensional structure model.

[0014] Optionally, the initial partitioning module includes: A configuration sub-module, configured to obtain the partitioning attribute value input by the user as a configuration parameter; An initial partitioning sub-module, configured to perform a finite element grid partition on the three-dimensional structure model based on the configuration parameter to obtain a plurality of the initial grids.

[0015] Optionally, the volume feature point is the centroid of the initial grid; Optionally, the surface feature point is the centroid of the original surface of the initial grid; Optionally, the edge feature point is the midpoint of the original edge of the initial grid.

[0016] Optionally, the final partitioning module includes: A node determination sub-module, configured to determine the current original node; A first judgment sub-module, configured to judge whether the node index number of the current original node exceeds the node index maximum value; If the node index number does not exceed the maximum node index value, search for the target feature points associated with the current original node; Construct the final mesh by combining the current original node, the corresponding set of target feature points, and the preset topological relationship.

[0017] Optionally, the final partitioning module further includes: A second judgment sub-module for judging whether the current original node is the last original node in the list of original nodes; If so, judge whether the current initial mesh is the last initial mesh in the list of initial meshes. If so, it indicates that all the initial meshes have been partitioned; otherwise, use the next initial mesh of the current initial mesh as the new current initial mesh; Otherwise, use the next original node of the current original node as the new current original node.

[0018] An application system for model partitioning provided by the present application predicts the performance of a semiconductor device based on the target structure model obtained by the foregoing model partitioning method, and adopts the following technical solutions, including: An extraction module for extracting the light information of the irradiated light and the target structure model of the semiconductor device; the target structure model includes a number of final meshes; An intersection processing module for determining the first intersection point based on the intersection situation between the light information and the final mesh; A reflection processing module for updating the light information according to the reflection result of the irradiated light at the first intersection point; A tracking prediction module for searching for the first final mesh corresponding to the first intersection point, simulating the light trace of the irradiated light in the first final mesh in combination with the light information, and determining the second intersection point and the performance data of the second intersection point; An acquisition module for, if the second basic plane where the second intersection point is located is of the target type, obtaining the performance data of the semiconductor device in combination with the simulation result and / or performance data of the light tracing.

[0019] This specification also provides an electronic device, where the electronic device includes: A processor; and, A memory storing computer-executable instructions, and the executable instructions, when executed, cause the processor to execute any one of the above methods.

[0020] This specification also provides a computer-readable storage medium, where the computer-readable storage medium stores one or more programs, and the one or more programs, when executed by a processor, implement any one of the above methods.

[0021] This specification also provides a computer program product, where the computer program product includes: a computer program / instructions, and when the computer program / instructions are executed by a processor, the above-mentioned method is implemented.

[0022] In this application, by reading the three-dimensional structure model of a semiconductor device; performing an initial division on the three-dimensional structure model to obtain a number of initial meshes; constructing feature points based on the index relationship of the initial meshes; where in different initial meshes, the feature points corresponding to the same basic feature coincide, so as to avoid the geometric feature misalignment caused by the local mesh independence in the traditional mesh division; the feature points include one or more of volume feature points, surface feature points, and edge feature points; combining the original nodes and feature points corresponding to the initial meshes, dividing the initial meshes into a number of final meshes; after all the initial meshes are divided, summarizing all the final meshes to obtain a target structure model, completing the division of the three-dimensional structure model, so as to reduce the calculation amount when generating performance data for optimizing performance design; and improving the prediction rate during the process of simulating the propagation process of light in the semiconductor device and predicting the generation rate of carriers based on the ray tracing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic diagram of the principle of a method for dividing a model of a semiconductor device provided by an embodiment of this specification; Figure 2 It is a related schematic diagram of dividing a tetrahedron into a hexahedron; Figure 3 It is a flowchart of steps S3 - S4 of a method for dividing a model of a semiconductor device provided by an embodiment of this specification Figure 1 ; Figure 4 It is a schematic diagram of the structure of the second division of a method for dividing a model of a semiconductor device provided by an embodiment of this specification; Figure 5 It is a schematic diagram of the structure of the final mesh of a method for dividing a model of a semiconductor device provided by an embodiment of this specification; Figure 6 It is a flowchart of steps S3 - S4 of a method for dividing a model of a semiconductor device provided by an embodiment of this specification Figure 2 ; Figure 7 It is a schematic diagram of the principle of an application method for model division provided by an embodiment of this specification; Figure 8 It is a schematic diagram of the structure of a system for dividing a model of a semiconductor device provided by an embodiment of this specification; Figure 9 It is a schematic diagram of the structure of an application system for model division provided by an embodiment of this specification; Figure 10 A schematic structural diagram of an electronic device provided by an embodiment of this specification; Figure 11 A schematic structural diagram of a computer-readable storage medium provided by an embodiment of this specification. Detailed implementation manners

[0024] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles defined in the following description of the present invention can be applied to other implementation manners, variations, improvements, equivalent manners, and other technical solutions without departing from the spirit and scope of the present invention.

[0025] Now, the exemplary embodiments of the present invention will be described more comprehensively with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be understood that the present invention is limited to the embodiments described herein. On the contrary, providing these exemplary embodiments can make the present invention more comprehensive and complete, and more convenient to fully convey the inventive concept to those skilled in the art. The same reference numerals in the figures represent the same or similar elements, components, or parts, and thus their repeated description will be omitted.

[0026] On the premise of conforming to the technical concept of the present invention, the features, structures, characteristics, or other details described in a specific embodiment may not be excluded from being combined in a suitable manner in one or more other embodiments.

[0027] In the description of specific embodiments, the features, structures, characteristics, or other details described in the present invention are for those skilled in the art to fully understand the embodiments. However, it does not exclude that those skilled in the art can practice the technical solutions of the present invention without one or more of the specific features, structures, characteristics, or other details.

[0028] The flowcharts shown in the accompanying drawings are only exemplary illustrations, and do not necessarily include all the contents and operations / steps, nor do they necessarily need to be executed in the described order. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may be changed according to the actual situation.

[0029] The block diagrams shown in the accompanying drawings are only functional entities, and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0030] The term "and / or" or "and / or" includes all combinations of any one or more of the associated listed items.

[0031] If the technical solution of this application involves personal information, before the product applying the technical solution of this application processes personal information, it has clearly informed the personal information processing rules and obtained the independent consent of the individual. If the technical solution of this application involves sensitive personal information, before the product applying the technical solution of this application processes sensitive personal information, it has obtained the individual's separate consent and at the same time meets the requirements of "express consent".

[0032] Figure 1 It is a schematic diagram of the principle of a method for model partitioning of a semiconductor device provided for the embodiments of this specification. The method includes: S1 Read the three-dimensional structure model of the semiconductor device; S2 Perform an initial partition on the three-dimensional structure model to obtain a number of initial meshes; S3 Construct feature points based on the index relationship of the initial meshes; wherein, in different initial meshes, the feature points corresponding to the same basic feature coincide; the feature points include one or more of volume feature points, surface feature points, and edge feature points; S4 Combine the original nodes and feature points corresponding to the initial meshes, and partition the initial meshes into a number of final meshes according to a preset topological relationship; S5 After all the initial meshes are partitioned, summarize all the final meshes to obtain the target structure model, and complete the partitioning of the three-dimensional structure model.

[0033] Due to their unique electrical and optical properties, semiconductor materials have become indispensable basic materials in modern electronics, optoelectronics, energy conversion, communication and other fields, and are widely used in semiconductor devices. For example, common semiconductor materials such as silicon (Si), gallium arsenide (GaAs), cadmium telluride (CdTe), etc. are widely used in various semiconductor devices such as integrated circuits, solar cells, photodetectors, lasers, light-emitting diodes (LEDs), and photoelectric sensors.

[0034] With the continuous progress of technology, it is necessary to improve the performance requirements of semiconductor devices in aspects such as photoelectric conversion efficiency, response speed, energy consumption control, and device stability. Among them, in semiconductor optoelectronic devices, their performance is decisively related to the generation, migration, and recombination processes of photo-generated carriers (including photo-generated electrons and photo-generated holes).

[0035] That is to say, the generation process of photo-generated carriers is an important factor in the performance of semiconductor devices. Therefore, accurately quantifying the generation rate of photo-generated carriers has become one of the core tasks for optimizing device performance.

[0036] Currently, the prediction of the generation rate of photo-generated carriers mainly relies on various numerical simulation methods such as the transfer matrix method and the finite-difference time-domain method. However, the related numerical simulation methods mainly have the following problems: (1) The transfer matrix method calculates the propagation and distribution of light in a semiconductor device by constructing the optical transfer matrices of each layer in a multi-layer structure to describe the reflection and transmission behaviors of light between different layers. This method is applicable to periodic and layered structures, with high computational efficiency and easy to implement. For non-uniform, complex geometric structures and multi-dimensional light propagation problems, the applicability of the transfer matrix method is limited.

[0037] (2) The finite-difference time-domain method (FDTD) discretizes Maxwell's equations in the time domain and the spatial domain to simulate the propagation, reflection, refraction, and scattering processes of electromagnetic waves in a medium. This method has high time-domain and spatial-domain resolutions, can simulate complex electromagnetic phenomena, and is applicable to various optical materials and structures. When dealing with large-scale devices and high-resolution requirements, the FDTD method consumes a huge amount of computing resources and has low computational efficiency. In addition, this method faces challenges in accurately describing the carrier generation and recombination mechanisms and is difficult to comprehensively capture the microscopic carrier behaviors.

[0038] Therefore, the existing prediction methods for the generation rate of photo-generated carriers generally face problems such as huge computational amounts, long calculation times, and insufficient accuracy when dealing with complex geometric structures, high-frequency dynamic processes, and multi-physical field couplings. There is an urgent need to develop more efficient and accurate calculation methods to overcome the existing limitations.

[0039] Ray tracing technology shows significant advantages in optical simulations. Therefore, the present invention considers applying ray tracing technology to the prediction / calculation of the rate of semiconductor photo-generated carriers. However, it still faces many challenges, including: being unable to handle complex boundary conditions; being unable to effectively achieve multi-physical field couplings, etc.

[0040] Based on this, in order to effectively integrate ray tracing technology into the prediction method of the generation rate of photo-generated carriers in semiconductor devices to reduce the computational complexity and improve the prediction effect, the present invention proposes a method for partitioning the model of a semiconductor device, which specifically includes: S1 Read the three-dimensional structure model of the semiconductor device; Obtain the physical structure data of the semiconductor device; construct the three-dimensional structure model of the semiconductor device according to the physical structure data.

[0041] Among them, the physical structure data includes but is not limited to: the size of the semiconductor device, the material of the semiconductor device.

[0042] Based on the fusion modeling of multi-dimensional physical characteristics (size, material), it provides a basis for the simulation of complex optical effects (such as refraction, absorption) and ensures the reliability of subsequent analyses.

[0043] S2 performs an initial division on the three-dimensional structure model to obtain a number of initial meshes; Through the initial division, the complex three-dimensional structure model is simplified into multiple initial meshes that are easier to process. By means of the division parameters configured by the user, the division method of the meshes can be flexibly adjusted to adapt to different analysis requirements, thereby improving the processing ability for complex geometric shapes.

[0044] S21 Obtains the division attribute value input by the user as a configuration parameter; S211 Pre-constructs a number of division parameters; the division parameters include but are not limited to: element type, mesh density, and interval.

[0045] Among them, the element type is used to determine what type of geometric elements the three-dimensional structure model is divided into. The mesh density is used to control the density of the meshes after division. The interval is used to determine the gap between elements or a specific division spacing.

[0046] S212 Based on the first division instruction, presents a configuration interface to the user. The configuration interface displays the division parameters and the corresponding parameter selection areas. The parameter selection areas are used for the user to configure (input / select) the division parameter values corresponding to the division parameters.

[0047] In an embodiment of the present specification, the parameter selection area corresponding to the element type is a drop-down menu. The drop-down menu displays all the element type options. Among them, in order to improve the calculation efficiency, the element type options include: the first element type. The first element type is used to represent that the three-dimensional structure model is divided into hexahedron meshes.

[0048] In order to flexibly fill irregular areas and improve the processing ability for complex geometric shapes, the element type options also include: the second element type. The second element type is used to represent that the three-dimensional structure model is divided into tetrahedron meshes. The specific implementation manner of the present invention mainly takes tetrahedron meshes (initial meshes) as an example.

[0049] The parameter selection area corresponding to the mesh density is a numerical input box, allowing the user to input a number or select a preset density level.

[0050] The parameter selection area corresponding to the interval is a numerical input box, allowing the user to input a number or select a preset interval.

[0051] S213 After the user finishes inputting / selecting in the parameter selection area, based on the user's parameter submission instruction, obtains the division parameter values configured by the user as configuration parameters; S22 Performs a finite element mesh division on the three-dimensional structure model based on the configuration parameters to obtain a number of the initial meshes.

[0052] Input the configuration parameters into the finite element analysis tool, and automatically perform the first mesh generation on the three-dimensional structure model by the finite element analysis tool, dividing it into several hexahedron meshes / tetrahedron meshes.

[0053] Among them, the finite element analysis tool includes but is not limited to: ANSYS, ABAQUS, COMSOL, etc. The finite element analysis tool belongs to the prior art, and its specific analysis process will not be elaborated here.

[0054] In an embodiment of this specification, when the element type selected by the user is the first element type, perform finite element mesh generation on the three-dimensional structure model to construct several hexahedron meshes. At this time, the initial mesh is a hexahedron mesh.

[0055] In another embodiment of this specification, when the element type selected by the user is the second element type, perform finite element mesh generation on the three-dimensional structure model to construct several tetrahedron meshes. At this time, the initial mesh is a tetrahedron mesh.

[0056] S23 Construct an initial mesh list; S231 Based on the coordinate system of the three-dimensional structure model, determine the coordinates of each vertex; obtain the coordinates of all vertices and the number of initial meshes; S232 Traverse each initial mesh and establish the corresponding relationship between the initial mesh and the original nodes, original edges, and original faces: S232-1 Determine the basic characteristics of the initial mesh; Among them, take the vertices of the initial mesh as the original nodes; take the edges of the initial mesh as the original edges; take the faces of the initial mesh as the original faces. The basic characteristics include: all the original nodes corresponding to the initial mesh, all the original edges corresponding to the initial mesh, and all the original faces corresponding to the initial mesh.

[0057] S232-2 Establish the first corresponding relationship between the initial mesh and the original nodes, and the first corresponding relationship is used to represent each original node corresponding to each initial mesh; For example, when the initial mesh is a tetrahedron mesh, the original nodes and their node coordinates of the current initial mesh include: ( , , ), ( , , ), ( , , ), ( , , ); S232-3 Establish a second correspondence between the initial meshes and the original edges, where the second correspondence is used to represent each original edge corresponding to each initial mesh; S232-4 Establish a third correspondence between the initial meshes and the original faces, where the third correspondence is used to represent each original face corresponding to each initial mesh; S232-5 Aggregate the first correspondence, the second correspondence, and the third correspondence to construct an index relationship for the initial meshes; Based on the index relationship, the respective basic features corresponding to each initial mesh and the material information of each original face can be found.

[0058] Subsequent operations will be performed on each initial mesh in the present invention. Therefore, for the convenience of subsequent processing, all index relationships are aggregated to construct a list of initial meshes. The list of initial meshes is used to represent the mapping relationship between the initial meshes and the index relationships.

[0059] Preferably, a mesh index number is assigned to each initial mesh. In an embodiment of the present specification, the mesh index number is a natural number starting from 1 and increasing incrementally, and the maximum value of the mesh index is the total number of initial meshes. The initial meshes in the list of initial meshes are sorted in ascending order according to the mesh index numbers. Of course, the initial meshes can also be sorted based on the adjacent relationship of each initial mesh to generate a list of initial meshes.

[0060] For the convenience of coupled simulation, in an embodiment of the present invention, based on the material to which each original face belongs, the material information is determined; an association is established between the original faces and the material information of the original faces; the material information includes: the material refractive index and the material absorption coefficient of the material to which it belongs; The partitioning method of the present invention supports the precise association of material information (material refractive index, material absorption coefficient), laying a foundation for subsequent coupled simulation of multiple physical fields (such as electric field, thermal field). For example, the refractive index fluctuation caused by temperature change can be achieved by dynamically updating the material information (material refractive index, material absorption coefficient).

[0061] S3 Construct feature points based on the index relationship of the initial meshes; When performing the second partitioning, the prior art usually divides tetrahedral meshes into hexahedral meshes. The process mainly includes: Select the centroid (usually the center of gravity or the circumcenter) of an initial mesh , and draw perpendicular lines from the centroid to the four faces. As shown in Figure 2 (a), in the initial mesh , , , is the foot of the perpendicular. If a hexahedron needs to be divided, a point needs to be found on the edge such that , , , , are coplanar.

[0062] As shown in Figure 2 (b), an adjacent initial grid of the initial grid , and the common original face of the two initial grids is . In the initial grid , if a hexahedron needs to be divided, similarly, a point needs to be found on the edge such that the four points are coplanar.

[0063] Generally, the of the initial grid may not coincide with . The centroid of the initial grid on the plane is the foot of the perpendicular and also may not coincide with .

[0064] In this way, for the plane , it will be divided into six sub - planes, as shown in Figure 2 (c), that is: the planes divided on the initial grid , , (red line division). The planes divided on the initial grid , , (blue line division).

[0065] The sub - plane division result of the initial grid may not coincide with the sub - plane division result of the initial grid .

[0066] Light passes through the initial grid and through the initial grid . Assuming the light passes through the plane , after that, it is impossible to determine which plane ( , , , ) on the initial grid it enters. Extra calculations are needed to determine which plane it passes through.

[0067] Therefore, although the properties of hexahedral meshes are relatively good, however, if the initial mesh is divided into finer hexahedral meshes, the ray tracing process is relatively complex. Therefore, as Figure 3 shown, the present invention will perform a second division on each initial mesh based on other methods to reduce the computational complexity.

[0068] S31 Initialize the fundamental face list and the final mesh list; Among them, the fundamental face list is used to represent the mapping relationship between each fundamental face and the face information, and the face information includes but is not limited to: the material information of the fundamental face; that is, according to the fundamental face list, the corresponding face information can be found based on the fundamental face. The final mesh list is used to represent the mapping relationship between the final meshes and the fundamental faces, that is, according to the final mesh list, each final mesh can be found, as well as all the fundamental faces corresponding to each final mesh.

[0069] By initializing the fundamental face list and the final mesh list, subsequent traversal of the initial mesh and completion of the division of the final mesh can be facilitated.

[0070] S32 Traverse each initial mesh and find the feature points of each initial mesh; S321 Take the first initial mesh in the initial mesh list as the current initial mesh; then execute step S322.

[0071] S322 Determine whether the mesh index number of the current initial mesh exceeds the maximum mesh index value; if so, return the fundamental face list and the final mesh list; otherwise, execute step S323.

[0072] If the index information of the current initial mesh exceeds the maximum mesh index value, it indicates that subsequent operations cannot be performed on the current initial mesh.

[0073] S323 Based on the basic features of the current initial mesh, construct the feature points of the current initial mesh; According to the index relationship of the current initial mesh, find the basic features of the current initial mesh; combined with the basic features, create feature points. Among them, the feature points include one or more of volume feature points, surface feature points, and edge feature points.

[0074] S323-1 Construct the volume feature points of the current initial mesh; For the sake of facilitating the division of space, the volume feature points ( 、 、 ) are located inside the initial mesh. Preferably, the volume feature point is the centroid of the current initial mesh.

[0075] S323-2 Find all the original faces of the current initial mesh; determine the surface feature points of each original face; As described above, if the surface feature point is the orthocenter, the surface feature points of multiple adjacent initial meshes on the common original surface generally do not coincide. It is also necessary to additionally calculate which original surface of the next initial mesh the light will reach after passing through an initial mesh, which significantly increases the subsequent workload.

[0076] Therefore, in order to eliminate the discontinuity of the common surface of adjacent meshes and reduce the surface matching complexity during the light path penetration, the present invention makes the surface feature points of adjacent initial meshes coincide on the common original surface, so that the division results of different initial meshes on the common original surface are consistent.

[0077] Preferably, the surface feature point is the centroid of the original surface; of course, the surface feature point can also be the incenter of the original surface.

[0078] In an embodiment of the present specification, as Figure 4 shown, when the current initial mesh is a tetrahedral mesh, the current initial mesh includes four original surfaces: surface , surface , surface , surface . Among them, the surface feature point of surface is the centroid of surface ; the surface feature point of surface is the centroid of surface ; the surface feature point of surface is the centroid of surface ; the surface feature point of surface is 's centroid .

[0079] S323-3 Search for all the original edges of the current initial mesh; determine the edge feature points of each original edge; Preferably, the edge feature point is the midpoint of the original edge.

[0080] In an embodiment of the present specification, when the current initial mesh is a tetrahedral mesh, the current initial mesh includes six original edges: edge , edge , edge , edge , edge , edge .

[0081] Among them, the edge feature point of edge is the midpoint of edge ; the edge feature point of edge is the midpoint of edge ​ ; The edge The edge feature point of is the midpoint of edge ; The edge The edge feature point of is the midpoint of edge ; The edge The edge feature point of is the midpoint of edge ; The edge The edge feature point of is the midpoint of edge .

[0082] In the present invention, in different initial meshes, the feature points corresponding to the same basic feature coincide, so that the division of the common basic surfaces of the final mesh completely coincides; furthermore, based on the intersection points on the basic surface of the current final mesh, the basic surface of the next final mesh into which the light penetrates is directly determined, reducing the complexity of the calculation.

[0083] S4 Combine the original nodes and feature points corresponding to the initial mesh, and divide the initial mesh into a plurality of final meshes; Traverse all the original nodes of the current initial mesh, and divide the current initial mesh based on each original node. Specifically: S41 Based on the original nodes of the current initial mesh, construct a list of original nodes; then execute step S42.

[0084] Assign a node index number to each original node, and establish an index relationship for each original node of the current initial mesh based on the association relationship between the original node and the node index number.

[0085] In an embodiment of the present specification, the node index number is a natural number, and the node index numbers increase in sequence; the largest node index number (the total number of original nodes) is used as the node index maximum value. The specific allocation method of the numbers is not limited herein.

[0086] Subsequent operations in the present invention will be performed based on each original node respectively. Therefore, in order to facilitate subsequent processing, a list of original nodes is constructed. The list of original nodes is used to represent the mapping relationship between the original nodes and the node index numbers. Preferably, the original nodes in the list of original nodes are arranged in ascending order according to the node index numbers.

[0087] S42 Take the first original node in the list of original nodes as the current original node; S43 Determine whether the node index number of the current original node exceeds the node index maximum value; If so, execute step S32; otherwise, execute step S44.

[0088] S44 Find the target feature points corresponding to the current original node; If the node index number of the current original node does not exceed the maximum node index, find the target feature points associated with the current original node; specifically: S441 According to the current initial grid where the current original node is located, find the target body feature points; S442 Retrieve the basic features of the current initial grid, and find the target feature points according to the association relationship between the current original node and the basic features; S442-1 Find the original faces and / or original edges related to the current original node from the basic features of the current initial grid; S442-2 Based on the original faces related to the current original node, find the corresponding face feature points and use them as the target face feature points; S442-3 Based on the original edges related to the current original node, find the corresponding edge feature points and use them as the target edge feature points; S442-4 Aggregate all the target body feature points, target face feature points and target edge feature points to construct a target feature point set.

[0089] In an embodiment of the present specification, the current original node is ; Based on the current initial grid where the current original node is located 、Find the body feature points corresponding to the current initial grid and use them as the target body feature points.

[0090] The original faces related to the current original node include: original face 、original face 、 .

[0091] The original edges related to the current original node include: original edge 、original edge 、original edge .

[0092] Among them, the target face feature point of face is face feature point ; the target face feature point of face is face feature point ; the target face feature point of face is face feature point .

[0093] The target edge feature point of edge is edge feature point ; the target edge feature point of edge is edge feature point ; The target edge feature points of the edge are edge feature points .

[0094] Therefore, the set of target feature points corresponding to the current node includes: , , , , , , .

[0095] S45 constructs the final mesh by combining the current original node, the corresponding set of target feature points, and the preset topological relationship.

[0096] In one embodiment of this specification, the target volume feature points are respectively connected to each target surface feature point to construct basic edges; the target surface feature points are respectively connected to each target edge feature point to construct basic edges; the current original node is respectively connected to each target edge feature point to construct basic edges; The target volume feature points are respectively connected to each target edge feature point to construct basic edges; after determining all the basic edges, according to the closed loops formed by the basic edges, the basic surfaces are obtained, and then the final mesh is obtained. Through the above feature point selection method, it is found that the divided final mesh is a nonahedron.

[0097] In another embodiment of this specification, the target volume feature points are respectively connected to each target surface feature point to construct basic edges; the target surface feature points are respectively connected to each target edge feature point to construct basic edges; the current original node is respectively connected to each target edge feature point to construct basic edges; The target volume feature points are respectively connected to each target edge feature point to construct implicit edges; As Figure 5 shown, after determining all the basic edges and implicit edges (orange dashed lines), the closed loops formed by the basic edges and / or implicit edges are obtained, and the basic surfaces are obtained. Search for two basic surfaces that contain implicit edges; Perform a coplanarity detection on the basic surfaces. Specifically, determine whether the two basic surfaces containing the same implicit edge are coplanar; if so, fuse the two basic surfaces into one basic surface; remove the implicit edge; otherwise, use the implicit edge as a basic edge.

[0098] The shape of the final mesh may vary depending on the coplanarity of the feature points, including hexahedrons, heptahedrons, octahedrons, nonahedrons, etc. , , , are all located on the surface , so , , , are coplanar. , , , are all located on the plane , so , , , are coplanar. , , , are all located on the plane , so , , , are coplanar.

[0099] , , , may (not) be coplanar; , , , may (not) be coplanar; , , , may (not) be coplanar. Among them: When the above three groups are all coplanar, the final mesh is a hexahedron. When two of them are coplanar, the final mesh is a heptahedron. When one of them is coplanar, the final mesh is an octahedron.

[0100] When none of them is coplanar, the final mesh is a nonahedron. For example, when , , , are not coplanar, these four points form two triangles, namely triangle and triangle . When , , , are not coplanar, these four points form two triangles, namely triangle and triangle . When , , , are not coplanar, these four points form two triangles, namely triangle and triangle . The current node The corresponding final mesh includes nine faces, namely 3 quadrilaterals and 6 triangles. At this time, the final mesh is a nonahedron.

[0101] Based on the above partitioning method, all other basic faces except those at the boundary can correspond to two final meshes. By looking up each other between the basic faces and the final meshes, it is possible to know which final mesh the light will enter in the next stage after exiting one final mesh.

[0102] The present invention can flexibly process non-uniform and complex geometric structures by constructing a subdivided nonahedron, which helps to overcome the limitations of the transfer matrix method when dealing with such structures.

[0103] S46 Add the final mesh to the final mesh list; add all basic faces of the final mesh to the basic face list; S461 Update the final mesh list according to the newly partitioned final mesh; Obtain all basic faces of the final mesh; Establish a body-face correspondence relationship between the final mesh and the basic face; the body-face correspondence relationship is used to represent each basic face corresponding to each final mesh; Add the body-face correspondence relationship to the final mesh list, and the final mesh list is used to represent the mapping relationship between each final mesh and the body-face correspondence relationship.

[0104] S462 Update the basic face list according to the newly partitioned final mesh; Judge whether each basic face exists in the basic face list; If the basic face does not exist in the basic face list, determine the original node corresponding to the basic face, and use it as the node information; determine the material information of the basic face; construct the face information of the basic face by combining the material information and the node information; add the basic face and the corresponding face information to the basic face list. In an embodiment of the present specification, it further includes: marking the basic face as the first type; the first type is used to represent that the basic face is the boundary face of the three-dimensional structure model; that is, the basic face corresponds to only one final mesh.

[0105] If the basic face exists in the basic face list, determine the original node corresponding to the basic face, and use it as the node information; add the node information to the face information of the basic face. In an embodiment of the present specification, it further includes: marking the basic face as the second type; at this time, there is no need to add the basic face repeatedly. The second type is used to represent that the basic face is the common face of the three-dimensional structure model; that is, the basic face corresponds to two final meshes.

[0106] S47 Judge whether the current original node is the last original node in the original node list; If so, execute step S48; otherwise, use the next original node of the current original node as the new current original node, and re-execute step S43.

[0107] In one embodiment of this specification, if the current original node is the last original node in the original node list, it indicates that the initial mesh of a tetrahedron has been divided into four parts and assigned to four nodes, that is, in the initial mesh each original node corresponds to a final mesh.

[0108] S48 determines whether the current initial mesh is the last initial mesh in the initial mesh list; if so, it indicates that the division of all initial meshes is completed, and step S4 is ended; otherwise, use the next initial mesh of the current initial mesh as the new current initial mesh, and re-execute step S32.

[0109] Based on the above operations, each basic face can correspond to one (when the basic face belongs to the boundary face) or two final meshes (when the basic face belongs to the common face) in the present invention.

[0110] The following briefly describes steps S3 - S4 in conjunction with Figure 6 : ① Initialize the final mesh list, and the mesh index number i = 0; ② Determine whether the mesh index number i of the current initial mesh exceeds the maximum mesh index value (the total number of initial meshes); if so, execute ③; otherwise, return the basic face list and the final mesh list.

[0111] ③ Calculate the centroid of the current initial mesh, the centroids of all original faces, and the midpoints of all original edges.

[0112] ④ Initialize the basic face list, and the node index number j = 0; ⑤ Determine whether the node index number j of the current original node exceeds the maximum node number (the total number of original nodes); if so, execute ⑥; otherwise, execute ⑧.

[0113] ⑥ Obtain the original nodes and characteristic points required for the final mesh corresponding to this original node; construct a final mesh, add all face information to this final mesh, and add the final mesh to the final mesh list. Determine whether the constructed basic face exists in the basic face list; if so, add node information to the basic face; if not, construct the basic face; add node information to the basic face, and add the basic face to the basic face list.

[0114] ⑦ j = j + 1, and execute ⑤.

[0115] ⑧ i = i + 1, and execute ②.

[0116] After all the initial meshes are divided, all the final meshes are aggregated to obtain the target structure model, completing the division of the three-dimensional structure model.

[0117] The present invention divides the three-dimensional structure model of a semiconductor device into initial meshes, and each initial mesh includes original nodes. Then each initial mesh is further divided into final meshes. Based on the above two divisions, the three-dimensional structure model of the semiconductor device is refined into a target structure model including final meshes. By accurately dividing the meshes and optimizing the model structure, the accuracy and reliability of the simulation results are improved, providing strong support for the design and optimization of semiconductor devices.

[0118] Of course, the model division method of the present invention is not limited to semiconductor devices (ray tracing therein), and can also be applied to other field scenarios with internal discrete absorption (not necessarily light).

[0119] Figure 7 FIG. is a schematic diagram of the principle of an application method for model division provided by an embodiment of this specification. The method includes: S6 Extract the ray information of the incident ray and the target structure model of the semiconductor device; the target structure model includes a number of final meshes; S7 Based on the intersection situation between the ray information and the final meshes, determine the first intersection point; S8 Update the ray information according to the reflection result of the incident ray at the first intersection point; S9 Locate the first final mesh corresponding to the first intersection point, and combine the ray information to simulate the ray trace of the incident ray in the first final mesh, and determine the second intersection point and the performance data of the second intersection point; S10 If the second basic surface where the second intersection point is located is of the target type, obtain the performance data of the semiconductor device by combining the simulation results and / or performance data of the ray tracing.

[0120] The high efficiency and flexibility of ray tracing in simulating optical phenomena such as the propagation path, reflection, refraction, and scattering of light. By tracking the paths of a large number of rays, ray tracing can accurately simulate the distribution of complex light fields and the interaction between light and matter with high precision. With its flexibility and efficiency, it can better meet the design and optimization requirements of superstructures and metasurfaces, efficiently handle these complex sub-wavelength geometries and multi-layer structures, accurately predict their impact on light propagation, and promote the development of new optical devices.

[0121] Although ray tracing technology has shown significant advantages in optical simulations, its application to predicting / calculating the rate of photo-generated carriers in semiconductors still faces many challenges, such as: (1) Unable to handle complex boundary conditions: The surface of semiconductor devices may have complex shapes (e.g., uneven, micro-structured, etc.), and these boundary conditions will affect the path of light propagation, resulting in multiple refractions and reflections of light between different surfaces.

[0122] (2) Unable to effectively implement multi-physics field coupling: During the operation of semiconductor devices, the electric field and thermodynamic field in the devices will change with the working conditions, which will affect the refractive index of the material, etc. Therefore, in ray tracing simulations, multi-physics field coupling needs to be considered to accurately simulate the dynamic changes of light characteristics.

[0123] The present invention takes into account that the scene construction of ray tracing includes four basic elements: light source, material, optical properties, and geometry.

[0124] Among them, the light source determines the starting position, direction, wavelength, and energy of the light.

[0125] Materials are used to characterize the properties of different objects, such as: material refractive index and material absorption coefficient.

[0126] Optical properties refer to the properties exhibited when materials or objects interact with light. These properties determine how light is reflected, refracted, absorbed, and scattered, thereby affecting optical performance. In this scenario, we mainly use the Fresnel formula to determine whether light is refracted or reflected.

[0127] Geometry refers to dividing a three-dimensional structure model or region into discrete grids (tetrahedral grids, hexahedral grids, etc.) and the discrete nodes that make up these grids based on the finite element mesh algorithm. Based on this grid information, the three-dimensional structure model is divided into tiny geometries.

[0128] In the foregoing content, the present invention reshapes the basic element of "geometry" to construct a target structure model. Next, the present invention will also combine other basic elements to accurately simulate the dynamic changes of light characteristics.

[0129] Based on the limitations of existing methods, the present invention proposes an application method for model division, which is applicable to efficient and accurate calculation of the rate of photo-generated carriers based on ray tracing, specifically including: S6 Extract the ray information of the incident light and the target structure model of the semiconductor device; The target structure model includes a number of final grids; S61 When the incident light irradiates the semiconductor device, collect the ray information of the incident light; When there is incident light irradiating on the semiconductor device, light information is constructed according to the incident situation of the incident light. The light information includes but is not limited to: power, incident position, incident direction, wavelength.

[0130] S62 Retrieve the target structure model of the semiconductor device; If the three-dimensional structure model of the semiconductor device includes initial meshes, original nodes, after dividing them into tetrahedral meshes, there will be 4 final meshes, assigned to original nodes.

[0131] According to the traditional ray tracing modeling process, it mainly is: First, construct 4 final meshes (N-sided polyhedra), then there will be 4 basic surfaces. Then, based on the bounding boxes of these basic surfaces, perform spatial partitioning, such as acceleration structures like quadtree, octree, KD-tree, BVH, etc. When the incident light intersects with the three-dimensional structure model in the scene, first intersects with the geometric bounding box in the acceleration structure, after confirming the candidate set of possible intersection points, then for each geometric body in the candidate set, perform precise intersection point calculation and return the final intersection point.

[0132] However, the above method mainly has the following problems: (1) Constructing the acceleration structure is very time-consuming; (2) Although the cost of a single intersection of the light with the geometric bounding box of the acceleration structure is very low, due to the density of the scene, the number of intersections of the light with the bounding box is extremely large, resulting in extremely time-consuming ray tracing.

[0133] Therefore, in order to improve the calculation efficiency, the present invention performs model partitioning on the three-dimensional structure model of the semiconductor device based on the foregoing steps S1 - S5. Through a two-time partitioning strategy, the three-dimensional structure model is converted into several dense and continuous final meshes (nonagons), and subsequent operations are performed according to the target structure model composed of the final meshes.

[0134] S7 Determine the first intersection point based on the intersection situation between the light information and the final meshes; A light source emits an incident light. According to the light information of the incident light, perform intersection point detection on the boundary of the target structure model to determine the intersection point of the incident light and the target structure model, and use it as the first intersection point. In an embodiment of this specification, use the ray - triangle / voxel intersection algorithm to calculate the first intersection point of the incident light and all boundary surfaces.

[0135] S8 Update the light information according to the reflection result of the incident light at the first intersection point; Find the basic surface where the first intersection point is located as the first basic surface; Determine the incident angle of the irradiation light on the first basic plane in combination with the relative position of the light information and the first basic plane ; According to the material information of the first basic plane, use the Fresnel formula to judge the reflection ratio; if , it indicates total reflection, and end the ray tracing; otherwise, update the light information; S9 Search for the first final grid corresponding to the first intersection point, simulate the ray trace of the irradiation light in the first final grid in combination with the light information, and determine the second intersection point and the performance data of the second intersection point; S91 Search for the final grid where the first basic plane is located according to the final grid list as the first final grid; S92 Based on the pre-constructed final grid list, obtain the adjacent basic planes according to the mapping relationship between the first final grid and the basic plane. The adjacent basic planes are other basic planes except the first basic plane in the first final grid; The present invention reduces the computational amount by reducing unnecessary acceleration structure construction and complex intersection detection; by directly performing intersection on the subdivided final grid (nonahedron), the computational efficiency and computational accuracy are improved.

[0136] S93 Search for the second intersection point and the second basic plane in combination with the relative position of the updated light information and the adjacent basic planes.

[0137] The present invention does not apply traditional acceleration structures (such as BVH and KD tree), and uses the method of constructing adjacent indexes (establishing index relationships for adjacent nonahedrons) to improve the tracing efficiency.

[0138] S94 Determine the performance data at the second intersection point; Among them, the performance data is preferably the generation rate of photo-generated carriers.

[0139] S941 Search for the second basic plane where the second intersection point is located; determine the material absorption coefficient of the second intersection point based on the material information of the second basic plane ; S942 Determine the ray trace route of the irradiation light in the first final grid according to the coordinates of the first intersection point and the second intersection point, and determine the path length of the irradiation light in the first final grid ; S943 Combine the material absorption coefficient of the second intersection point , the path length and the power in the light information to determine the power of the second intersection point ; Take the power in the light information as the power of the irradiation light at the first intersection point .

[0140] Power of the second intersection point 。

[0141] S944 Construct the volume of the first final grid according to the coordinates of the first final grid ; S945 Calculate the generation rate of photo-generated carriers 。

[0142] When the irradiated light reaches the second intersection point, the generation rate of photo-generated carriers 。

[0143] Wherein: is the quantum efficiency (constant), which describes the efficiency of generating an electron-hole pair after absorbing a photon; Planck's constant; speed of light (constant); volume of the first final grid; wavelength.

[0144] The present invention uses ray tracing technology to accurately simulate the propagation process of light in semiconductor materials, combines the generation rate formula of photo-generated carriers, accurately calculates the generation distribution of carriers, and improves the authenticity and reliability of calculation results.

[0145] In another embodiment of this specification, the performance data further includes: light absorption amount (distribution of light absorption in the device), that is - 。

[0146] S10 If the second fundamental plane where the second intersection point is located is of the target type, obtain the performance data of the semiconductor device by combining the simulation results of the ray tracing and / or the performance data.

[0147] Obtain the performance data of the semiconductor device according to the type of the second fundamental plane where the second intersection point is located.

[0148] S101 If the second fundamental plane is of the first type (boundary plane), then determine that the second fundamental plane where the second intersection point is located is of the target type, and obtain the performance data of the semiconductor device by combining all the ray traces and the performance data corresponding to the second intersection point.

[0149] The performance data can be the simulation results of ray tracing. In an embodiment of this specification, summarize each ray trace route in sequence to obtain the simulation results of ray tracing.

[0150] Of course, the performance data can be all the performance data and / or the latest performance data.

[0151] Based on performance data, assist users in optimizing the performance design of semiconductor devices.

[0152] S92 If the second fundamental plane is of the second type (common plane), use the second intersection point as the new first intersection point, update the ray information; execute step S9 to continue simulating the ray trace.

[0153] The method for calculating the rate of optically generated carriers in a semiconductor based on ray tracing proposed by the present invention realizes efficient and accurate simulation of the behavior of optically generated carriers through innovative algorithm design and optimization, providing strong theoretical support and technical guarantee for the high-performance design of semiconductor devices. This method can not only improve the calculation accuracy and shorten the calculation time, but also better meet the simulation requirements of complex semiconductor devices.

[0154] When the final mesh is an enneahedron, the following briefly describes the specific execution process of this application method: ① A light source generates a ray, and the ray information includes: position, direction, wavelength, power.

[0155] ② The ray intersects with the surface of the physical structure to find the corresponding (basic) plane of the intersection point.

[0156] ③ Obtain the material refractive index and absorption coefficient of the node corresponding to the plane, and apply the Fresnel formula to judge the next direction of the ray.

[0157] ④ Judge whether the ray is reflected. If so, stop the trace; otherwise, execute ⑤.

[0158] ⑤ Update the ray information, find the enneahedron according to the intersection plane, and let the ray intersect with the other eight planes to find the nearest intersection point.

[0159] ⑥ Update the ray power and the rate of optically generated carriers of the node according to the intersection information.

[0160] ⑦ Confirm the next enneahedron according to the plane where the intersection point is located.

[0161] ⑧ Judge whether there is a next enneahedron. If so, execute ⑤; otherwise, stop the trace.

[0162] In an application scenario of the present invention, the method of the present invention is applied to a simulation scenario. According to the calculated distribution of optically generated carriers, the surface optical structure (anti-reflection, crosstalk prevention, etc.) and the optical design of the semiconductor thin film structure (optical absorption film thickness, multi-junction photocurrent matching, etc.) are designed and optimized.

[0163] In an application scenario of the present invention, the above method is applied to a light-absorbing semiconductor device, and the main functions of its performance data include: (1) By simulating the behaviors of light reflection, refraction, and scattering in semiconductor devices, the parts and causes that may lead to light loss are discovered, and then the device structure is improved and optimized. For example, in the design of solar concentrators, ray tracing technology can be used to adjust the shape, position, and angle of mirrors or lenses so that more light can be absorbed by solar cells, improving the light collection efficiency of the entire system.

[0164] (2) Considering the complex lighting conditions and environmental factors in practical applications, such as different solar altitude angles, cloud cover, shadows, etc., accurately evaluate the performance of light-absorbing semiconductor devices in various environments, providing a reference for their applications in different scenarios. For example, in the design of distributed solar power generation systems, ray tracing simulation can be used to predict the power generation of solar panels at different installation positions and angles at different times of the day and in different seasons.

[0165] In an application scenario of the present invention, applying the above method to a light-detecting semiconductor device, the functions of its performance data mainly include: (1) It can track the propagation and interaction process of incident light in a photodetector, accurately calculate the generation, separation, and collection efficiency of photo-generated carriers, thereby deeply understanding the light response characteristics (responsivity, quantum efficiency, response time, etc.) of the detector. For example, in the design of high-speed photodetectors, ray tracing can be used to optimize the device structure and materials to reduce the recombination and transport time of carriers and improve the response speed of the detector.

[0166] (2) By simulating the propagation path of light in the detector and the distribution of photo-generated carriers, defects or unreasonable parts in the structure are discovered, and then the structure of the detector is optimized. For example, in the design of avalanche photodiodes, ray tracing technology can be used to adjust the electric field distribution and structural parameters of the multiplication region, improve the avalanche multiplication efficiency and stability, and reduce noise.

[0167] (3) Combining the generation and transport process of photo-generated carriers, analyze the noise sources in the detector, such as shot noise, thermal noise, etc., and study the variation law of noise under different structures and working conditions through ray tracing simulation, providing a theoretical basis for reducing noise and improving the signal-to-noise ratio of the detector. For example, in the design of low-noise photodetectors, ray tracing can be used to optimize the package structure and materials of the detector, reduce the interference of the external environment on the detector, and lower the noise level.

[0168] In an application scenario of the present invention, applying the above method to a light-sensing semiconductor device, the functions of its performance data mainly include: (1) By tracking the propagation of light in the optical sensor device and its interaction process with semiconductor materials, deeply understand the physical mechanism of optical sensing, such as the generation of photo-generated carriers and the change of electrical properties of semiconductor materials, etc., providing a theoretical basis for optimizing the performance and design of the sensor. For example, when studying semiconductor gas sensors, ray tracing can help researchers understand the influence of gas molecules adsorbed on the semiconductor surface on light absorption and scattering, so as to optimize the sensitivity and selectivity of the sensor.

[0169] (2) Simulate the propagation path and distribution of light in the sensor, discover the key factors affecting the sensor performance, such as the incident angle of light, the optical and electrical properties of materials, etc., and improve the sensor performance by optimizing the structure and materials of the sensor. For example, in the design of charge-coupled devices, ray tracing technology can be used to optimize the structure and layout of photosensitive units to improve the resolution and sensitivity of images.

[0170] (3) Consider the influence of environmental factors in practical applications, such as temperature, humidity, gas concentration, etc. on the performance of optical sensing semiconductor devices. Through ray tracing simulation, study the change of the response characteristics of the sensor under different environmental conditions, providing a basis for the calibration and compensation of the sensor. For example, when studying semiconductor humidity sensors, ray tracing can simulate the propagation and scattering of light in the sensor under different humidity environments, as well as the influence on the electrical properties of the sensor, so as to achieve accurate measurement and compensation of humidity.

[0171] Figure 8 The structural schematic diagram of a model partitioning system for a semiconductor device provided by an embodiment of this specification. The system includes: A reading module 810, configured to read the three-dimensional structure model of the semiconductor device; An initial partitioning module 820, configured to perform an initial partitioning on the three-dimensional structure model to obtain a number of initial meshes; A feature point construction module 830, configured to construct feature points based on the index relationship of the initial meshes; wherein, in different initial meshes, the feature points corresponding to the same basic feature coincide; the feature points include one or more of volume feature points, surface feature points, and edge feature points; A final partitioning module 840, configured to combine the original nodes and feature points corresponding to the initial meshes to partition the initial meshes into a number of final meshes; A summarizing module 850, configured to, after all the initial meshes are partitioned, summarize all the final meshes to obtain a target structure model, completing the partitioning of the three-dimensional structure model.

[0172] Optionally, the initial partitioning module 820 includes: A configuration sub-module, configured to obtain the partitioning attribute value input by the user as a configuration parameter; An initial division sub-module, configured to perform finite element mesh division on the three-dimensional structure model based on the configuration parameters to obtain a plurality of the initial meshes.

[0173] Optionally, the body feature point is the centroid of the initial mesh; Optionally, the face feature point is the centroid of the original face of the initial mesh; Optionally, the edge feature point is the midpoint of the original edge of the initial mesh.

[0174] Optionally, the final division module 840 includes: A node determination sub-module, configured to determine a current original node; A first judgment sub-module, configured to judge whether the node index number of the current original node exceeds the maximum node index value; If the node index number does not exceed the maximum node index value, search for a target feature point associated with the current original node; Construct the final mesh by combining the current original node, the corresponding target feature point set, and the preset topological relationship.

[0175] Optionally, the final division module 840 further includes: A second judgment sub-module, configured to judge whether the current original node is the last original node in the original node list; If so, judge whether the current initial mesh is the last initial mesh in the initial mesh list. If so, it indicates that all the initial meshes have been divided; otherwise, use the next initial mesh of the current initial mesh as the new current initial mesh; Otherwise, use the next original node of the current original node as the new current original node.

[0176] Figure 9 It is a schematic structural diagram of an application system for model division provided by an embodiment of this specification. Based on the target structure model obtained by the foregoing model division method, the performance of a semiconductor device is predicted. The system includes: An extraction module 910, configured to extract light information of the irradiated light and the target structure model of the semiconductor device; the target structure model includes a plurality of final meshes; An intersection processing module 920, configured to determine a first intersection point based on the intersection situation between the light information and the final mesh; A reflection processing module 930, configured to update the light information according to the reflection result of the irradiated light at the first intersection point; The tracking and prediction module 940 is configured to find the first final grid corresponding to the first intersection point, simulate the light trace of the illumination light in the first final grid in combination with the light information, and determine the second intersection point and the performance data of the second intersection point; The acquisition module 950 is configured to, if the second basic plane where the second intersection point is located is of a target type, obtain the performance data of the semiconductor device in combination with the simulation result and / or performance data of the ray tracing.

[0177] The functions of the system according to the embodiments of the present invention have been described in the above method embodiments. Therefore, for the details not described in this embodiment, reference may be made to the relevant descriptions in the foregoing embodiments, and details will not be repeated here.

[0178] Figure 10 FIG. is a schematic structural diagram of an electronic device provided by an embodiment of this specification. The electronic device includes: a memory 1001 and a processor 1002. The memory 1001 is configured to store computer-executable instructions. When the computer-executable instructions are executed by the processor 1002, the steps of the above method embodiments can be implemented.

[0179] Figure 11 FIG. is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of this specification. The computer-readable storage medium 1100 stores one or more computer programs. When the one or more computer programs are executed by a processor, the steps of the above method embodiments can be implemented.

[0180] An embodiment of this specification further provides a computer program product, including a computer program / computer-executable instructions. When the computer program / computer-executable instructions are executed by a processor, the steps of the above method embodiments can be implemented.

[0181] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. When the computer program is executed, it may include the processes of the above method embodiments.

[0182] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A semiconductor device model partitioning method, characterized in that: include: Read the three-dimensional structure model of semiconductor devices; Performing initial division on the three-dimensional structure model to obtain a plurality of initial grids; Constructing feature points based on the index relationship of the initial grid; wherein, in different initial grids, feature points corresponding to the same basic feature overlap; the feature points include one or more of volume feature points, surface feature points, and edge feature points; Combining the original nodes and feature points corresponding to the initial grid, dividing the initial grid into a plurality of final grids; After all the initial meshes are divided, all the final meshes are aggregated to obtain the target structure model, completing the division of the three-dimensional structure model.

2. The method according to claim 1, characterized in that The three-dimensional structure model is initially divided to obtain a plurality of initial grids, including: Obtain the partition attribute value input by the user as a configuration parameter; Finite element meshing is performed on the three-dimensional structure model based on the configuration parameters to obtain a plurality of initial meshes.

3. The method according to claim 1, characterized in that The volume feature point is the centroid of the initial grid; and / or, The surface feature point is the centroid of the original surface of the initial mesh; and / or, The edge feature point is the midpoint of the original edge of the initial grid.

4. The method according to claim 1, characterized in that The combining of the original nodes and feature points corresponding to the initial grid to divide the initial grid into a plurality of final grids includes: Determine the current original node; Determine whether the node index number of the current original node exceeds the maximum node index number; If the node index number does not exceed the maximum node index number, searching for a target feature point associated with the current original node; The final grid is constructed by combining the current original node, the corresponding target feature point set and the preset topological relationship.

5. The method according to claim 4, characterized in that After combining the current original node, the corresponding target feature point set and the preset topological relationship to construct the final grid, the method further includes: Determine whether the current original node is the last original node in the original node list; If yes, determine whether the current initial grid is the last initial grid in the initial grid list. If yes, it indicates that all initial grids have been divided. Otherwise, take the next initial grid of the current initial grid as the new current initial grid. Otherwise, the next original node of the current original node is used as the new current original node.

6. An application method of model partitioning, characterized in that: Based on the target structure model obtained by the model partitioning method according to any one of claims 1 to 5, the performance prediction of the semiconductor device is performed, specifically comprising: Extracting light information of the irradiated light and a target structure model of the semiconductor device; the target structure model includes a plurality of final grids; Determining a first intersection point based on the intersection of the light information and the final grid; updating light information according to a reflection result of the irradiation light at the first intersection point; Finding a first final grid corresponding to the first intersection, simulating a light trace of the irradiation light in the first final grid in combination with the light information, and determining performance data of the second intersection and the second intersection; If the second base surface where the second intersection point is located is of a target type, the performance data of the semiconductor device is obtained by combining the simulation result and / or performance data of the ray tracing.

7. A semiconductor device model partitioning system, characterized in that: include: A reading module, used for reading a three-dimensional structure model of a semiconductor device; An initial partitioning module, used to perform initial partitioning on the three-dimensional structure model to obtain a plurality of initial grids; A feature point construction module, used to construct feature points based on the index relationship of the initial grid; wherein, in different initial grids, feature points corresponding to the same basic feature overlap; the feature points include one or more of volume feature points, surface feature points, and edge feature points; A final division module, used for combining the original nodes and feature points corresponding to the initial grid to divide the initial grid into a plurality of final grids; The summary module is used to summarize all the final grids to obtain the target structure model after all the initial grids are divided, thereby completing the division of the three-dimensional structure model.

8. A model partitioning application system, characterized in that: Based on the target structure model obtained by the model partitioning method described in claims 1-5, the performance prediction of the semiconductor device is performed, specifically including: An extraction module, used to extract light information of the irradiated light and a target structure model of the semiconductor device; the target structure model includes a plurality of final grids; An intersection processing module, used for determining a first intersection point based on the intersection of the light information and the final grid; A reflection processing module, used for updating light information according to a reflection result of the irradiation light at the first intersection point; a tracing prediction module, configured to find a first final grid corresponding to the first intersection, simulate a light trace of the irradiating light in the first final grid in combination with the light information, and determine performance data of the second intersection and the second intersection; An acquisition module is used to obtain the performance data of the semiconductor device in combination with the simulation result and / or performance data of the ray tracing if the second base surface where the second intersection point is located is of the target type.

9. An electronic device, wherein: The electronic device includes: processor; and, A memory storing computer executable instructions which, when executed, cause the processor to perform the method according to any one of claims 1-6.

10. A computer-readable storage medium, wherein: The computer-readable storage medium stores one or more computer programs, and when the one or more computer programs are executed by a processor, the method of any one of claims 1 to 6 is implemented.

Images