Integrated Circuit 3D Visualization Simulation Analysis Method and System

By generating a three-dimensional process disturbance evolution model, the propagation path of accumulated disturbances and deviation amplification within the integrated circuit is traced, solving the problem of failure location under multiple process steps, and realizing accurate root cause location and process optimization in three-dimensional visualization simulation analysis of integrated circuits.

CN123116984APending Publication Date: 2026-11-17JIANGSU DONN TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611217878.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-12
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing integrated circuit simulation and analysis techniques are insufficient for unified modeling of cumulative disturbances under continuous action of multiple process steps, and lack the ability to reverse-engineer the disturbance propagation and amplification paths, resulting in difficulties in failure localization.

Method used

The system collects the three-dimensional structure of the integrated circuit, process step parameters, online metrology results, and local deformation detection information to generate a three-dimensional process disturbance evolution model. It tracks the cumulative disturbance propagation path and deviation amplification path, and performs correlation matching analysis to identify key failure factors.

Benefits of technology

It realizes the end-to-end causal relationship from the origin of process deviation to final failure, accurately identifies the dominant failure process segment, structural layer and deviation region, and improves the root cause localization accuracy and process improvement efficiency of three-dimensional visualization simulation analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116984A_ABST
    Figure CN123116984A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of simulation analysis technology and discloses a method and system for three-dimensional visualization simulation analysis of integrated circuits. It includes: collecting three-dimensional structural information, process step parameter information, online metrology result information, local deformation detection information, and final-state failure marker information of the integrated circuit; mapping the process step parameter information, online metrology result information, and local deformation detection information to generate a three-dimensional process disturbance evolution body, and tracking the cumulative disturbance propagation path and deviation amplification path of different process steps to form a process disturbance backtracking result; performing correlation matching analysis on the process disturbance backtracking result and the final-state failure marker information to form a failure root cause locking result, and generating a three-dimensional process disturbance backtracking view, a key area highlight distribution map, and process compensation suggestion results. This invention can comprehensively improve the root cause localization accuracy and process improvement efficiency of three-dimensional visualization simulation analysis of integrated circuits in multi-step continuous manufacturing scenarios.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of simulation analysis technology, and more specifically, to a method and system for three-dimensional visualization simulation analysis of integrated circuits. Background Technology

[0002] With the continuous development of advanced packaging, 3D integration, and high-density interconnect technologies, the internal structure of integrated circuits is evolving towards greater complexity, including multi-layer stacking, heterogeneous material coupling, and microscale interconnects. Against this backdrop, issues such as localized deformation, interface mismatch, residual stress accumulation, and amplified process deviations during manufacturing have become key factors affecting the electrical performance stability and structural reliability of integrated circuits. Especially under the continuous superposition of multiple manufacturing processes, minute structural deviations introduced by a single process step often do not immediately manifest as detectable defects, but rather accumulate, diffuse, and couple amplify in subsequent steps, ultimately manifesting as warping, cracks, interface peeling, conduction failure, or electrical anomalies. Therefore, accurately modeling and visually analyzing the disturbance formation mechanism of integrated circuits in complex process chains has become an important technical direction for improving yield and enhancing manufacturability.

[0003] However, most existing integrated circuit simulation and analysis technologies focus on displaying local simulation results under a single process step, or only observe static defects in the final structure. They typically lack the ability to uniformly model the propagation process of accumulated disturbances under the continuous action of multiple process steps. At the same time, traditional analysis methods often treat online metrology results, local deformation detection information, and final failure results in isolation, making it difficult to establish a continuous correlation link from process step parameters to final failure. As a result, when integrated circuits fail, engineers can usually only rely on experience to repeatedly check multiple possible processes, making it difficult to accurately locate the key process segment, key structural layer, and core deviation area that truly dominates the failure. In addition, existing 3D visualization methods are usually more inclined to display results and lack the ability to reverse-engineer and dynamically trace the propagation and amplification paths of manufacturing disturbances within integrated circuits. Therefore, there is an urgent need for an integrated circuit 3D visualization simulation and analysis system that can perform 3D modeling, back-engineering, and root cause identification of accumulated disturbances within integrated circuits.

[0004] In view of this, the present invention proposes a three-dimensional visualization simulation analysis method and system for integrated circuits to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and achieve the above objectives, the present invention provides the following technical solution: a three-dimensional visualization simulation analysis method for integrated circuits, comprising: Collect three-dimensional structural information, process step parameter information, online metrology result information, local deformation detection information and final state failure marker information of integrated circuits to form a basic dataset of process disturbances; Based on the basic dataset of process disturbances, process step parameter information, online measurement result information and local deformation detection information are mapped to the three-dimensional spatial mesh corresponding to the three-dimensional structural information to generate a three-dimensional process disturbance evolution body that characterizes the cumulative physical influence of each process step. Based on the three-dimensional process disturbance evolution model, the cumulative disturbance propagation path and deviation amplification path formed by different process steps inside the integrated circuit are traced, and the process disturbance backtracking results are generated. By performing correlation and matching analysis on the process disturbance backtracking results and the final state failure marking information, the key process segments, key structural layers and key deviation areas that dominate the failure are identified, and the failure root cause locking results are formed. Based on the failure root cause identification results, a three-dimensional backtracking view of process disturbances, a highlighted distribution map of key areas, and process compensation suggestions are generated.

[0006] Furthermore, methods for generating three-dimensional process perturbation evolution bodies include: Based on the three-dimensional structural information, the three-dimensional spatial range of the integrated circuit is determined; the three-dimensional spatial range of the integrated circuit is divided into multiple voxel units; process step parameter information, online metrology result information, and local deformation detection information are mapped to each voxel unit to obtain the corresponding process history record, metrology deviation record, and deformation state record for each voxel unit; material property mapping is performed on each voxel unit to obtain material property fields and material physical parameter fields; the material property fields, material physical parameter fields, process history record, metrology deviation record, and deformation state record corresponding to each voxel unit are summarized and integrated to form a three-dimensional process disturbance evolution body.

[0007] Furthermore, methods for tracing the propagation path of accumulated disturbances include: For all voxel units, deviation origin voxel identification is performed; for each deviation origin voxel, downstream process steps are determined from all corresponding process steps; starting from each deviation origin voxel, the spatial propagation behavior of the deviation is tracked along the step execution order in the process step parameter information to each corresponding downstream process step to obtain an initial propagation relationship; for each initial propagation relationship, the rationality of the propagation direction is checked, the deviation propagation receiving voxel is identified, and a deviation propagation chain is constructed; for each deviation propagation receiving voxel, the spatial propagation behavior of the deviation is continued to be tracked to each corresponding downstream process step, and a deviation propagation chain is constructed; multiple deviation propagation chains originating from the same deviation origin voxel are arranged according to the step execution order to form a cumulative disturbance propagation path.

[0008] Furthermore, methods for tracing the path of deviation amplification include: For each deviation propagation link, calculate the corresponding deviation increment value; based on the deviation increment value, identify the amplification links in all deviation propagation links; for each amplification link, perform amplification mechanism type labeling; according to the step execution order, extract all amplification links on the same cumulative disturbance propagation path to form a deviation amplification path, and sequentially determine the dense amplification segment record of each deviation amplification path.

[0009] Furthermore, methods for correlation and matching analysis between process disturbance backtracking results and final state failure marker information include: Based on the final state failure marker information, identify the failed voxels from all voxel units and determine the failure neighborhood voxel set for each failed voxel; based on the failure neighborhood voxel set and the cumulative disturbance propagation path, determine the failure association propagation path set for each failed voxel; for each failure association propagation path in each failure association propagation path set, perform causal compatibility verification, and mark the failure association propagation path that passes the causal compatibility verification as a causal confirmation propagation path; summarize all causal confirmation propagation paths to form a causal confirmation propagation path set.

[0010] Furthermore, methods for identifying key process segments, key structural layers, and key deviation regions that dominate failure include: Based on the set of causal confirmation propagation paths, identify causal amplification paths from all deviation amplification paths; determine the corresponding process segments based on the dense amplification segment records of each causal amplification path; calculate the segment occurrence frequency of each process segment, select the process segment with the highest segment occurrence frequency as the key process segment, and perform dominant amplification mechanism type identification on the key process segment. Based on the causal confirmation propagation path set, the total number of associated voxels in each structural layer is counted, and the structural layer with the largest total number of associated voxels is taken as the key structural layer; the interface amplification ratio of the key structural layer is calculated, and based on the interface amplification ratio, it is determined whether to add an interface mismatch dominant marker in the key structural layer. Extract all voxel units corresponding to each amplification link on all causal confirmation propagation paths and label them as causal voxels; perform spatial clustering analysis on all causal voxels to obtain clustered core voxels; merge all clustered core voxels with all causal voxels within the corresponding preset spatial clustering neighborhood radius to form a critical deviation region, and determine the dominant stress gradient direction of the critical deviation region.

[0011] Furthermore, methods for generating three-dimensional retrospective views of process disturbances include: For each causal confirmation propagation path, a corresponding directed 3D path segment sequence is constructed. In a pre-constructed 3D rendering space, using a 3D spatial mesh as the background skeleton, each structural layer is rendered in a semi-transparent layered form, and each directed 3D path segment sequence is superimposed on the 3D spatial mesh with colored arrow lines. For amplified links, color-separated highlight rendering based on the amplification mechanism type is performed. A temporal animation frame sequence is introduced, and a keyframe annotation mechanism is introduced on the basis of the temporal animation frame sequence. The rendering results of the 3D path segment sequence, the color-separated highlight rendering results of the amplified links, the temporal animation frame sequence, and the keyframe annotation results are integrated to form a 3D backtracking view of process disturbance.

[0012] Furthermore, methods for generating key region highlight distribution maps include: From the final failure marker information, the failure location coordinates and failure type are extracted. In the 3D rendering space, all voxel units, key structural layers and failure location coordinates in the critical deviation area are rendered. The dominant stress gradient direction, dominant amplification mechanism type and failure type are marked with arrow symbols and text respectively. The key process segments are displayed in the form of bar time sequence annotations. Explicit correlation lines are drawn between the rendering results of the critical deviation area and the bar time sequence annotations, and between the rendering results of the key structural layer and the bar time sequence annotations, to form a key area highlight distribution map.

[0013] Furthermore, methods for generating process compensation recommendations include: Each process step within the critical process segment is matched against a pre-defined process compensation rule library to obtain corresponding process parameter adjustment suggestions. The first exceeding step of each causal confirmation propagation path is obtained, and the number of times each process step within the critical process segment is the first exceeding step is counted to obtain the initial introduction frequency of each process step. The initial introduction frequencies of each process step are sorted from high to low, and the corresponding step compensation priorities are marked accordingly. The process parameter adjustment suggestions corresponding to each process step within the critical process segment are integrated with the step compensation priorities to form the process compensation suggestion results.

[0014] An integrated circuit 3D visualization simulation analysis system, implementing the method, includes: The disturbance acquisition module is used to acquire three-dimensional structural information of integrated circuits, process step parameter information, online metrology result information, local deformation detection information and final state failure marker information to form a basic dataset of process disturbances. The disturbance modeling module is used to map process step parameter information, online measurement result information and local deformation detection information to the three-dimensional spatial mesh corresponding to the three-dimensional structural information based on the process disturbance basic dataset, and generate a three-dimensional process disturbance evolution body that characterizes the cumulative physical effects of each process step. The backtracking and deduction module is used to trace the cumulative disturbance propagation path and deviation amplification path formed inside the integrated circuit by different process steps based on the three-dimensional process disturbance evolution, and to generate process disturbance backtracking results; The root cause locking module is used to perform correlation and matching analysis between process disturbance backtracking results and final state failure marking information, identify key process segments, key structural layers and key deviation areas that dominate the failure, and form failure root cause locking results. The visual output module is used to generate a 3D backtracking view of process disturbances, a highlighted distribution map of key areas, and process compensation suggestions based on the failure root cause identification results.

[0015] The technical effects and advantages of the integrated circuit three-dimensional visualization simulation analysis method and system of this invention are as follows: By constructing a fundamental dataset of process disturbances encompassing integrated circuit 3D structural information, process step parameter information, online metrology results, local deformation detection information, and final failure marker information, this approach overcomes the limitations of existing simulation analysis techniques that treat online metrology, deformation detection, and final failure results in isolation, achieving unified and correlated acquisition of data across the entire manufacturing chain. Furthermore, by mapping process step parameters, online metrology deviations, and local deformation information onto a voxel-by-voxel spatial mesh, and introducing material physical parameter fields, deviation monotonicity annotations, and maximum stress gradient direction annotations at the voxel level, a 3D model is generated with voxel units as spatial resolution units and the execution order of steps as the temporal index dimension. This process disturbance evolution model overcomes the limitation of only being able to perform local simulation of a single process step and not uniformly modeling multi-step continuous cumulative disturbances. It constructs a cumulative disturbance propagation path by tracing the propagation and receiving voxels of the deviation propagation sequentially along the process steps, starting from the origin voxel. During the tracing process, it introduces a mechanism for verifying the rationality of the propagation direction and screening material interface associations. By introducing three amplification mechanism types—process type switching amplification, material interface mismatch amplification, and process condition cumulative amplification—on the deviation amplification path, it achieves a leap from simply quantifying the deviation increment to identifying the amplification physical mechanism. Finally, it performs spatial neighborhood matching between the terminal voxel of the cumulative disturbance propagation path and the final failure location coordinates. The system is configured and performs causal compatibility verification between deviation types and failure types via a causal compatibility mapping table, establishing an end-to-end causal link from the origin of process deviation to final failure. This solves the core problem in traditional analysis methods where engineers can only rely on experience to repeatedly check multiple processes. By performing step frequency statistics and voxel space aggregation analysis on the causal confirmation propagation path set, the system accurately identifies the key process segments (including the dominant amplification mechanism type), key structural layers (including the dominant interface mismatch markers), and key deviation regions (including the dominant stress gradient direction) that dominate failure. The system generates data including directed propagation path rendering, amplification link color highlighting, time-series animation frame sequences, and side... The sidebar information panel and keyframe annotation mechanism provide a 3D backtracking view of process disturbances, along with a key area highlight distribution map containing multi-element overlay rendering and explicit associated line segments. This allows engineers to intuitively perceive the 3D propagation and amplification trajectory of deviations within integrated circuits, their physical mechanism attribution, and spatial-temporal coupling relationships. By generating process compensation suggestions based on process compensation rule base matching and step compensation priority ranking, the system provides engineers with a comprehensive process optimization solution that combines process parameter direction guidance with deviation contribution priority ranking. This comprehensively improves the root cause localization accuracy and process improvement efficiency of integrated circuit 3D visualization simulation analysis in multi-step continuous manufacturing scenarios. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the integrated circuit three-dimensional visualization simulation analysis system according to Embodiment 1 of the present invention; Figure 2This is a flowchart of the integrated circuit three-dimensional visualization simulation analysis method according to Embodiment 2 of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: Please refer to Figure 1 As shown, the integrated circuit three-dimensional visualization simulation analysis system described in this embodiment includes a disturbance acquisition module, a disturbance modeling module, a backtracking and deduction module, a root cause locking module, and a visual output module; each module is connected via wired and / or wireless means to realize data transmission between modules.

[0019] The disturbance acquisition module is used to collect three-dimensional structural information of integrated circuits, process step parameter information, online measurement result information, local deformation detection information and final state failure marker information to form a basic dataset of process disturbances.

[0020] Methods for acquiring three-dimensional structural information include: The three-dimensional structural information of the integrated circuit is obtained from the layout design system (i.e., the chip design platform used to store and manage the physical layout design data of the integrated circuit). This three-dimensional structural information includes a set of structural layer codes, the material type of each structural layer, the interlayer sequence of each structural layer, the design thickness of each structural layer, the set of design geometric dimensions of the internal interconnect features of each structural layer, and the layout boundary dimensions. The structural layer code set includes the structural layer code corresponding to each structural layer, which is a predefined unique identifier for each structural layer. The material type is used to identify the type of main material used in the corresponding structural layer, including but not limited to silicon substrate, oxide, etc. Silicon dielectric layer, low dielectric constant dielectric layer, copper interconnect layer, barrier metal, passivation layer material, etc.; interlayer order relationship is used to identify the relative position order of each structural layer in the vertical stacking direction, represented by integer numbers increasing sequentially from the bottom layer to the top layer; design thickness is the nominal thickness value marked in the layout design of the corresponding structural layer; design geometry set includes the design width, design spacing and design position coordinates of each interconnect feature (such as via, metal line, pad, etc.) inside the corresponding structural layer, and the design position coordinates are used to identify the center position coordinates of the corresponding interconnect feature; layout boundary dimensions are used to identify the overall geometric range of the integrated circuit in the horizontal plane.

[0021] Methods for collecting process step parameter information include: The process step parameters of each process step that the integrated circuit undergoes in the manufacturing process are obtained from the process execution system (i.e., the manufacturing execution platform used to record and manage the operating parameters of each process step in the integrated circuit manufacturing process). These process step parameters include step code, step type, step execution order, step association structure layer code, and step process condition set. The step code is a predefined unique identifier for each process step. The step type identifies the process category of the corresponding process step, including but not limited to deposition, etching, chemical mechanical polishing, ion implantation, annealing, photolithography, and cleaning. The step execution order identifies the sequence of the corresponding process steps in the entire manufacturing process chain. The execution position is represented by an integer number, incrementing sequentially from the first step to the last. The step-related structure layer code is used to identify the target structure layer directly affected by the corresponding process step. The step process condition set contains the key control parameters of the corresponding process step during execution. The specific content varies depending on the step type. For example, the process condition set for the deposition step includes deposition temperature, deposition pressure, and deposition rate; the process condition set for the etching step includes etching gas composition, etching power, and etching time; and the process condition set for the chemical mechanical polishing step includes polishing pressure, polishing fluid flow rate, and polishing time. The process step parameters of each process step are summarized and integrated to form process step parameter information.

[0022] Methods for collecting online measurement results include: The online metrology results parameters for the corresponding structural layer after each process step are obtained from the online metrology system (i.e., an online inspection platform used to measure key dimensions, film thickness, overlay accuracy, and other parameters of each structural layer in real time during integrated circuit manufacturing). These parameters include the step code, structural layer code, measured film thickness value, set of measured key dimension values, and measured overlay deviation value. The measured film thickness value is the actual measured thickness of the corresponding structural layer after the corresponding process step. The set of measured key dimension values ​​contains the measured key dimension values ​​of each interconnect feature within the corresponding structural layer; these values ​​represent the actual measured dimensions of the corresponding interconnect features. For line-type interconnect features, the measured key dimension value is the actual measured width; for via-type interconnect features, the measured key dimension value is the actual measured aperture. The measured overlay deviation value is the actual alignment offset of the corresponding structural layer relative to the underlying structure. The online metrology results parameters for each process step are integrated to form online metrology result information.

[0023] Methods for collecting local deformation detection information include: The local deformation detection parameters are obtained from the deformation detection system (i.e., the deformation monitoring platform used to detect the warpage, residual stress distribution, and local deformation of the surface of each structural layer during the manufacturing process of integrated circuits). The local deformation detection parameters include step codes, structural layer codes, a set of detection position coordinates, warpage values ​​at each detection position, and residual stress values ​​at each detection position. The set of detection position coordinates contains the three-dimensional spatial coordinates of multiple preset detection points on the surface of the corresponding structural layer, which are preset by those skilled in the art based on factors such as the geometric range of the corresponding structural layer, material interface distribution, dense interconnect feature areas, and estimated stress concentration areas. The warpage value is the offset of the surface at the corresponding detection position relative to the ideal plane. The residual stress value is the accumulated internal stress level at the corresponding detection position. The local deformation detection parameters corresponding to each process step are integrated to form local deformation detection information.

[0024] Methods for collecting final-state failure marker information include: The final-state failure marking parameters of the integrated circuit are obtained from the final-state testing system (i.e., the final-state testing platform used to perform electrical performance testing and physical failure detection on the finished integrated circuit after all manufacturing processes are completed). The final-state failure marking parameters include failure type, failure location coordinates, and failure associated structure layer code. The failure type is used to identify the physical characterization category of the corresponding failure, such as conduction failure, short circuit failure, leakage failure, warpage failure, etc. The failure location coordinates are the coordinates of the corresponding failure location in the three-dimensional space of the integrated circuit. The failure associated structure layer code is the code of the structure layer involved in the corresponding failure. All final-state failure marking parameters are integrated to form final-state failure marking information.

[0025] Methods for generating a basic dataset of process perturbations include: Using step codes and structural layer codes as the association primary keys, three-dimensional structural information, process step parameter information, online measurement result information, local deformation detection information, and final state failure labeling information are associated and integrated to form a basic dataset of process disturbances; among them, the final state failure labeling information is associated and integrated with the corresponding failure association structural layer code as the association primary key.

[0026] The disturbance modeling module is used to map process step parameter information, online measurement result information and local deformation detection information to a three-dimensional spatial mesh corresponding to the three-dimensional structural information based on the process disturbance basic dataset, and generate a three-dimensional process disturbance evolution body that characterizes the cumulative physical effects of each process step.

[0027] Methods for constructing three-dimensional spatial meshes corresponding to three-dimensional structural information include: Three-dimensional structural information is extracted from the basic dataset of process disturbances; based on the layout boundary dimensions in the three-dimensional structural information, the horizontal dimensions of the integrated circuit are determined. Direction and Horizontal Spatial extent in the direction; based on the interlayer sequence and design thickness of each structural layer in the three-dimensional structural information, determine the spatial extent of each structural layer in the vertical direction. The positional range in the direction; in the horizontal direction, the design width and design spacing of each interconnecting feature are extracted from the set of design geometric dimensions, and all extracted design widths and design spacings are compared, with the minimum value selected as the minimum feature scale in the horizontal direction; a preset mesh refinement granularity coefficient is set by those skilled in the art according to the required analysis accuracy and computational resource balance requirements, used to control the division density of the spatial mesh in the horizontal and vertical directions; the horizontal... Direction and Horizontal Spatial range in direction and vertical The positional ranges of each structural layer in the direction are integrated to form the three-dimensional spatial range of the integrated circuit; The horizontal mesh step size is obtained by multiplying the minimum feature scale by the mesh refinement granularity coefficient. For each structural layer, the vertical mesh step size is obtained by multiplying the corresponding design thickness by the mesh refinement granularity coefficient. Based on the horizontal mesh step size, the three-dimensional spatial range of the integrated circuit is divided along the horizontal direction. Direction and Horizontal The grid is divided into equal-spaced sections along the direction; based on the grid step size in the vertical direction corresponding to each structural layer, the three-dimensional spatial range of the integrated circuit is divided along the vertical direction. The direction is divided into non-uniform grids according to the structural layers; after the division is completed, each smallest grid unit is defined as a voxel unit; each voxel unit is assigned a globally unique voxel code in sequence, and the corresponding structural layer code and three-dimensional center coordinates are recorded; all voxel units are summarized to form a three-dimensional spatial grid; Furthermore, to ensure that the spatial mesh reflects the differences in material properties among different structural layers, material property mapping is performed on each voxel unit. Specifically, the material type of each structural layer is extracted from the 3D structural information. For each voxel unit, the corresponding material type is obtained according to the structural layer code and written into the material property field of the corresponding voxel unit. A preset material physical parameter table is used, which records physical parameters such as elastic modulus, coefficient of thermal expansion, and Poisson's ratio corresponding to each material type. These parameters are pre-set by those skilled in the art based on standard physical data sheets for various materials in the field of integrated circuit manufacturing. Based on the material type of each voxel unit, the corresponding elastic modulus, coefficient of thermal expansion, and Poisson's ratio are obtained from the material physical parameter table and written into the material physical parameter field of the corresponding voxel unit. The material physical parameter field is used to determine whether deviations will be amplified or refracted due to material interface mismatch when propagating between voxels with different material properties.

[0028] Methods for mapping process step parameter information to spatial grids include: From the basic dataset of process disturbances, process step parameter information is extracted. For each process step, the corresponding process step parameters are extracted from the process step parameter information. Based on the step association structure layer code in the process step parameters, all voxel units whose structure layer codes match the step association structure layer codes are retrieved in the 3D space grid to form the target voxel set for the corresponding process step. For each voxel unit in the target voxel set, the step code, step type, step execution order, and step process condition set of the corresponding process step are written into the process history of the corresponding voxel unit. The process history is an ordered data list attached to each voxel unit, which stores all process step information experienced by the corresponding voxel unit in the order of step execution from first to last. This is used to track the complete process action sequence of the corresponding voxel unit in the entire manufacturing process chain. For each process step in the process history corresponding to each voxel unit, a standardized process type code is assigned. Specifically, a process type code mapping table is pre-set, which records the standardized process type codes corresponding to each step type. These codes are pre-defined by those skilled in the art based on the physical mechanisms of various process steps in the field of integrated circuit manufacturing. For example, deposition steps and epitaxial growth steps belonging to additive manufacturing processes correspond to the same process type code; etching steps and chemical mechanical polishing steps belonging to subtractive manufacturing processes correspond to the same process type code; and ion implantation steps and annealing steps belonging to modification processes correspond to the same process type code. Based on the step type in the process history corresponding to each voxel unit, the standardized process type code for each voxel unit is obtained from the process type code mapping table and written into the corresponding voxel unit's process history. The standardized process type code is used to subsequently determine whether there is a sudden change in interface stress or a reversal in the direction of deviation propagation caused by a process type switch between adjacent process steps.

[0029] Methods for mapping online measurement results to spatial grids include: From the basic dataset of process disturbances, online metrology results are extracted. For each process step, corresponding online metrology parameters are extracted from the online metrology results. Based on the structural layer code in the online metrology results parameters, all voxel units of the corresponding structural layer are retrieved in the 3D spatial grid. The absolute value of the difference between the measured film thickness and the design thickness for each structural layer is calculated to obtain the film thickness deviation value for each structural layer. Each film thickness deviation value is written into the metrology deviation record of all voxel units corresponding to the corresponding structural layer. For each interconnect feature in each structural layer, the absolute value of the difference between the corresponding measured critical dimension value and the design width is calculated to obtain the corresponding critical dimension deviation value. Based on the design position coordinates and design width of each interconnect feature, the coverage of each interconnect feature is determined. Based on the 3D center coordinates of each voxel unit, the interconnect feature to which each voxel unit belongs is determined. Specifically, with the design position coordinates as the center and half the design width as the radius, the horizontal... Direction and Horizontal The coverage area in the direction; if the three-dimensional center coordinates of the voxel unit fall within the coverage area of ​​the corresponding interconnect feature, then the corresponding voxel unit is determined to belong to the corresponding interconnect feature; for each critical dimension deviation value, it is written into the measurement deviation record of all voxel units contained in the corresponding interconnect feature; for each measured overlay deviation value, it is written into the measurement deviation record of all voxel units corresponding to the corresponding structural layer; among them, the measurement deviation record is a data structure attached to each voxel unit, used to store the measured deviation information of the corresponding voxel unit after each process step is completed, and is coded and indexed by step for subsequent step-by-step analysis of the deviation accumulation trajectory; For the measurement deviation record of each voxel unit, deviation monotonicity annotation is performed. Specifically, for the measurement deviation record of each voxel unit, the film thickness deviation value, critical dimension deviation value, and measured overlay deviation value are sorted in order of execution, from first to last, to obtain film thickness deviation value sequence, critical dimension deviation value sequence, and measured overlay deviation value sequence. For each deviation value sequence, every two adjacent deviation values ​​are considered as a deviation combination. The number of deviation combinations in each deviation value sequence is counted to obtain the total number of combinations. Finally, it is determined that in each deviation value sequence, the deviation value of the later process step is greater than the deviation value of the previous process step. The number of deviation combinations is used to obtain the total monotonicity. For each deviation value sequence, the ratio of the corresponding total monotonicity to the total combination is calculated to obtain the monotonicity ratio. The monotonicity ratio of each voxel unit is compared with a preset monotonicity increasing judgment ratio threshold, which is preset by those skilled in the art based on the sensitivity requirements for determining the continuous deterioration of deviation. If the monotonicity ratio is greater than the monotonicity increasing judgment ratio threshold, the deviation monotonicity of the corresponding voxel unit is marked as a continuously deteriorating trend. If the monotonicity ratio is less than or equal to the preset monotonicity increasing judgment ratio threshold, the deviation monotonicity of the corresponding voxel unit is not marked as a continuously deteriorating trend.

[0030] Methods for mapping local deformation detection information to spatial meshes include: Local deformation detection information is extracted from the basic dataset of process disturbances. For each process step, corresponding local deformation detection parameters are extracted from the local deformation detection information, and all voxel elements of the corresponding structural layer are retrieved in the 3D spatial mesh based on the structural layer encoding in the local deformation detection parameters. Spatial interpolation mapping is performed on the warpage value and residual stress value in the local deformation detection parameters respectively. Specifically, the horizontal coordinates of each detection point in the detection location coordinate set are used as the basis for spatial interpolation mapping. Coordinates and horizontal The coordinates are known sampling points. Using the warpage or residual stress value of the corresponding detection point as known sampling values, inverse distance weighted interpolation is used to perform interpolation calculations on the three-dimensional center coordinates of each voxel unit in the corresponding structural layer, obtaining the interpolated warpage or residual stress value for each voxel unit. The inverse distance weighted interpolation method is a conventional technique in this field, and its specific process will not be elaborated upon here. The interpolated warpage and residual stress values ​​at each voxel unit are written into the deformation state record of the corresponding voxel unit. The deformation state record is a data structure attached to each voxel unit, indexed by step, storing the local warpage and stress state of the corresponding voxel unit after each process step, for subsequent analysis of the gradual evolution and accumulation trend of deformation in the process chain. For each voxel element's deformation state record, the maximum stress gradient direction is labeled; specifically, for each voxel element, the direction along the positive and negative stress gradients is obtained in the 3D mesh. Direction, positive and negative Direction and positive / negative Directly adjacent voxel elements are labeled as adjacent voxels. The absolute value of the difference between the interpolated residual stress value of each voxel element and the interpolated residual stress value of each corresponding adjacent voxel is calculated to obtain the stress gradient component. For each voxel element, the direction with the largest stress gradient component is selected as the direction of the maximum stress gradient of the corresponding voxel element. The direction of the maximum stress gradient is used to determine which spatial direction the deviation is most likely to propagate to the adjacent voxels at the corresponding voxel element, providing a quantitative physical basis for inferring the direction of the deviation propagation path.

[0031] Methods for generating three-dimensional process perturbation evolution bodies include: The material property fields, material physical parameter fields, process history records (including standardized process type codes), measurement deviation records (including deviation monotonicity annotations), and deformation state records (including maximum stress gradient direction annotations) corresponding to each voxel unit in the 3D spatial mesh are summarized and integrated to form a 3D process perturbation evolution body. Among them, the 3D process perturbation evolution body is the core intermediate data structure. It uses voxel units as spatial resolution units and the step execution sequence as the temporal index dimension. It fully records the physical field information such as the sequence of process conditions, measured deviation evolution trajectory, deformation state change history, material properties, and stress gradient direction accumulated in each spatial micro-region after the integrated circuit has undergone all manufacturing process steps. The 3D process perturbation evolution body enables subsequent tracing back along the step execution sequence at any voxel position. Combined with material interface mismatch judgment, deviation monotonicity trend identification, and stress gradient direction inference, it can identify where the deviation starts to accumulate, at which step it begins to amplify, and through what path it propagates to adjacent regions.

[0032] The backtracking and deduction module is used to trace the cumulative disturbance propagation path and deviation amplification path formed inside the integrated circuit by different process steps based on the three-dimensional process disturbance evolution body, and generate process disturbance backtracking results.

[0033] Methods for tracing the propagation path of cumulative disturbances include: In the multi-step continuous manufacturing process of integrated circuits, physical deviations (such as film thickness deviations, critical dimension deviations, or warpage) introduced by a certain process step in a specific spatial region will not only remain at the voxel position where the deviation initially occurs, but will propagate to adjacent voxels along the vertical direction (interlayer propagation) and the horizontal direction (intralayer diffusion). Especially under the superposition of subsequent process steps, the original deviation may be further amplified or shifted to a new spatial region by the new process conditions. Therefore, the goal of tracing the propagation path of accumulated disturbances is to start from each voxel unit with significant deviations in the three-dimensional process disturbance evolution body, and trace backward along the step execution sequence how the corresponding deviation is propagated to adjacent voxels in subsequent process steps, whether it is attenuated or amplified during the propagation process, so as to construct a continuous propagation link in three-dimensional space from the voxel of origin of the deviation to the voxel of the final state of the deviation. For all voxel units, deviation origin voxel identification is performed to obtain a deviation origin voxel set. Specifically, from the 3D process perturbation evolution volume, the metrological deviation records of all voxel units are traversed, and the film thickness deviation value, critical dimension deviation value, and measured overlay deviation value after each process step are extracted. A deviation origin identification threshold set is preset, which includes the film thickness deviation origin threshold, critical dimension deviation origin threshold, and overlay deviation origin threshold. The deviation origin identification threshold set is preset by those skilled in the art based on the typical minimum deviation level that each type of deviation in the field of integrated circuit manufacturing begins to have a detectable impact on subsequent process steps. For each voxel... The unit's measurement deviation record identifies the process step in the execution sequence where the deviation value first exceeds the corresponding deviation origin threshold and marks it as the first deviation step. If a first deviation step exists in all corresponding process steps, the corresponding voxel unit is marked as the deviation origin voxel, and the corresponding first exceeding step code (i.e., the step code corresponding to the first deviation step) and the first exceeding deviation type (i.e., the type of deviation that exceeds the corresponding deviation origin threshold in the first deviation step, specifically including film thickness deviation, critical dimension deviation, or overlay deviation) are recorded. All deviation origin voxels are summarized with their corresponding first exceeding step codes and first exceeding deviation types to form a deviation origin voxel set. For each voxel element, a material interface association filter is performed, and voxel elements located near the material interface are marked with a material interface tag. Specifically, for each voxel element, the material type in the corresponding material attribute field is obtained, and its vertical orientation in the 3D mesh is retrieved. The material type of two directly adjacent voxel units is considered. If the material type of a voxel unit is different from that of the adjacent voxel unit above or below, the corresponding voxel unit is determined to be near the material interface, and a material interface marker is added to the corresponding voxel unit. If the material type of a voxel unit is the same as that of both the adjacent voxel unit above and below, no material interface marker is added to the corresponding voxel unit. The material interface marker is used to prioritize the analysis of propagation behavior across the material interface direction in subsequent deviation propagation tracking, because deviation propagation at the interface usually has an amplification effect. Starting from each deviation origin voxel, the propagation behavior of the deviation in space is tracked along the execution sequence to each corresponding downstream process step. Specifically, the position of the first exceeding step corresponding to each deviation origin voxel in the execution sequence is obtained and marked as the first exceeding position. For each deviation origin voxel, all process steps located after the corresponding first exceeding position are extracted from all process steps in the corresponding process history in the execution sequence and marked as downstream process steps. For each adjacent voxel corresponding to each deviation origin voxel, the corresponding measurement deviation record is obtained. For each downstream process step corresponding to each deviation origin voxel, deviation propagation determination is performed. Specifically, in all downstream process steps corresponding to the deviation origin voxel, if there is an adjacent voxel whose measurement deviation record after the completion of the corresponding downstream process step shows a deviation value that did not exist before (i.e., the corresponding adjacent voxel did not exceed the corresponding deviation origin threshold in the previous process step, but exceeded the corresponding deviation origin threshold for the first time in the current downstream process step), it is preliminarily determined that the corresponding deviation has propagated from the deviation origin voxel to the corresponding adjacent voxel in space and formed an initial propagation relationship. For each initially defined propagation relationship, a propagation direction rationality check is performed to identify the deviation propagation receiver voxels. Specifically, the corresponding maximum stress gradient direction is extracted from the deformation state record of the deviation origin voxel. If the adjacent voxel in the initially defined propagation relationship is located on the maximum stress gradient direction of the corresponding deviation origin voxel, the propagation direction is determined to be reasonable, and the corresponding adjacent voxel is marked as a deviation propagation receiver voxel. If the adjacent voxel in the initially defined propagation relationship is not located on the maximum stress gradient direction of the corresponding deviation origin voxel, the deviation monotonicity of the corresponding adjacent voxel is further checked to see if it is marked as a continuously deteriorating trend. If the deviation monotonicity is marked as a continuously deteriorating trend, it indicates that although the corresponding adjacent voxel is not on the maximum stress gradient direction, there is indeed an independent continuous deviation accumulation, and it is still marked as a deviation propagation receiver voxel. If the deviation monotonicity is not marked as a continuously deteriorating trend, the propagation direction is determined to be unreasonable, and the corresponding adjacent voxel is not marked as a deviation propagation receiver voxel. The voxel codes of the originating voxel, the voxel codes of the receiving voxels of the deviation propagation, the step codes of the corresponding downstream process steps, and the deviation types involved in the propagation (i.e., the deviation type to which the adjacent voxel first exceeds the corresponding deviation origin threshold within the downstream process step corresponding to the initial propagation relationship) are recorded to form a deviation propagation chain. For each receiving voxel of the deviation propagation, starting from the step code in the corresponding deviation propagation chain, the same logic is used to trace the further propagation to the corresponding adjacent voxel in subsequent downstream process steps until no new receiving voxels of the deviation propagation appear in all downstream process steps. All voxel codes and step codes in all propagation branches formed from the same originating voxel through multiple deviation propagation chains are arranged according to the execution order of the steps in which the propagation occurs to form the cumulative perturbation propagation path of the corresponding originating voxel. The cumulative perturbation propagation paths of all originating voxels are summarized to form a set of cumulative perturbation propagation paths.

[0034] Methods for tracing the path of deviation amplification include: The cumulative disturbance propagation path describes where the deviation travels from in space, but it does not quantify whether the deviation increases or decreases during propagation. In actual manufacturing, some process steps can further amplify existing deviations (e.g., chemical mechanical polishing can cause more severe dish-shaped indentations in areas with existing uneven film thickness), while other process steps may partially reduce the deviation (e.g., annealing can release residual stress to some extent). The goal of deviation amplification path tracing is to quantify the changing trend of the deviation amplitude step by step along the identified cumulative disturbance propagation path, thereby identifying the key process steps and key spatial locations that cause the deviation to rapidly amplify from a small level to a severe level. For each deviation propagation link in each cumulative disturbance propagation path, the difference between the deviation value of the corresponding deviation propagation receiver voxel (i.e., the value corresponding to the deviation type obtained from the measurement deviation record) and the deviation value of the corresponding deviation propagation receiver voxel in the previous deviation propagation link is calculated to obtain the deviation increment value of the corresponding deviation propagation link. Specifically, for the first deviation propagation link in the cumulative disturbance propagation path, the deviation value of the corresponding deviation origin voxel after the completion of the first exceedance step is used as the deviation value of the previous deviation propagation link. If the deviation increment value is greater than zero, it indicates that the deviation has been amplified in the corresponding deviation propagation link; if the deviation increment value is less than zero, it indicates that the deviation has been amplified in the corresponding deviation propagation link. If the value is less than or equal to zero, it indicates that the deviation has not been amplified or has been reduced in the corresponding deviation propagation link. A significant deviation amplification threshold is preset, which is pre-set by those skilled in the art based on the typical incremental level of deviation amplification leading to a significant increase in reliability risk in the field of integrated circuit manufacturing. The deviation increment value of each deviation propagation link is compared with the significant deviation amplification threshold. If the deviation increment value is greater than the significant deviation amplification threshold, the corresponding deviation propagation link is marked as an amplified link. If the deviation increment value is less than or equal to the significant deviation amplification threshold, the corresponding deviation propagation link is not marked as an amplified link. For each amplification link, amplification mechanism type labeling is performed. Specifically, the standardized process type code for the corresponding process step is extracted from the process history of the deviation propagation receiver voxel corresponding to the amplification link; the standardized process type code for the previous process step is extracted from the process history of the deviation propagation receiver voxel corresponding to the previous deviation propagation link. If the two extracted standardized process type codes are different, the amplification mechanism type of the corresponding amplification link is labeled as process type switching amplification, indicating that deviation amplification is related to the interface stress mutation caused by the process type conversion between adjacent process steps (such as switching from additive manufacturing to subtractive manufacturing). If the two extracted standardized process type codes are the same, it is further determined whether the deviation propagation receiver voxel in the corresponding amplification link has an attached material interface mark. If a material interface mark is attached, the amplification mechanism type of the corresponding amplification link is labeled as material interface mismatch amplification, indicating that deviation amplification is related to the mismatch of elastic modulus or thermal expansion coefficient between different material layers. If no material interface mark is attached, the amplification mechanism type of the corresponding amplification link is labeled as process condition cumulative amplification. The large number of amplification steps indicates that the deviation amplification is mainly caused by the conditional cumulative effect of the same process type in consecutive steps. Following the step execution order, all amplification links on the same cumulative disturbance propagation path are extracted sequentially to form the deviation amplification path corresponding to the cumulative disturbance propagation path. For each deviation amplification path, consecutive amplification links in the process steps are extracted to form an amplification link subsequence. Multiple amplification link subsequences corresponding to the same deviation amplification path are compared, and the amplification link subsequence containing the most amplification links is taken as the dense amplification segment of the corresponding deviation amplification path. For each dense amplification segment, the step code corresponding to the first amplification link is marked as the amplification start step code, and the step code corresponding to the last amplification link is marked as the amplification end step code. For each deviation amplification path, the corresponding amplification start step code, amplification end step code, all voxel codes contained in the dense amplification segment, and the amplification mechanism type of each amplification link are summarized and integrated to form a dense amplification segment record for the corresponding deviation amplification path. All deviation amplification paths and their corresponding dense amplification segment records are summarized and integrated to form a deviation amplification path set.

[0035] Methods for generating process disturbance backtracking results include: The cumulative disturbance propagation path set and the deviation amplification path set are correlated and integrated to form the process disturbance backtracking result. The process disturbance backtracking result takes the deviation origin voxel as the starting point and completely records the step-by-step propagation link and step-by-step amplification trajectory of the deviation in three-dimensional space, providing a complete spatial, temporal and physical mechanism backtracking basis for establishing the causal relationship between process steps and final failure.

[0036] The root cause identification module is used to perform correlation and matching analysis between process disturbance backtracking results and final state failure marking information, identify key process segments, key structural layers and key deviation areas that dominate the failure, and form failure root cause identification results.

[0037] Methods for correlation matching analysis between process disturbance backtracking results and final state failure marker information include: From the basic dataset of process disturbances, final-state failure marker information is extracted. From the final-state failure marker information, the failure type, failure location coordinates, and failure-associated structure layer code are extracted from each set of final-state failure marker parameters. For each failure location coordinate, the Euclidean distance between it and the corresponding three-dimensional center coordinates of each voxel unit in the three-dimensional spatial grid is calculated and marked as the failure location distance. All failure location distances corresponding to the same failure location coordinate are compared, and the voxel unit with the smallest failure location distance is selected as the failure voxel. The failure spatial neighborhood radius is preset, which is determined by those skilled in the art based on the typical spatial expansion range of the physical characterization of failures inside integrated circuits. The scope is pre-defined; with the failed voxel as the center, all voxel units in the three-dimensional spatial grid whose voxel spacing with the corresponding failed voxel is less than or equal to the radius of the failed spatial neighborhood are obtained, forming the set of failed neighborhood voxels for the corresponding failed voxel; where the voxel spacing refers to the maximum value of the absolute value of the difference between each dimension between the corresponding three-dimensional center coordinates of two voxel units; for each failed voxel, from the set of cumulative disturbance propagation paths in the process disturbance backtracking results, the cumulative disturbance propagation paths of the terminal voxel (i.e., the last deviation propagation receiving voxel in the cumulative disturbance propagation path) belonging to the set of failed neighborhood voxels are selected, forming the set of failure association propagation paths for the corresponding failed voxel; For each failure propagation path in each set of failure propagation paths, a causal compatibility check is performed between the deviation type and the failure type. Specifically, a causal compatibility mapping table is pre-defined, recording whether there is a known physical causal relationship between different deviation types and different failure types. This table is pre-defined by those skilled in the art based on engineering knowledge in the field of integrated circuit failure analysis. For each cumulative disturbance propagation path in each set of failure propagation paths, the first out-of-range deviation type of the corresponding deviation origin voxel and the deviation type of each amplification link are obtained, and both the obtained first out-of-range deviation type and deviation type are marked as deviation association types. Each deviation association type is then further classified. The failure type of the failure voxel corresponding to the type and the corresponding failure propagation path is queried in the causal compatibility mapping table. If there is a deviation correlation type and the corresponding failure type marked as having a causal relationship in the causal compatibility mapping table, the corresponding failure propagation path is determined to have passed the causal compatibility check and is marked as a causal confirmation propagation path. If there is no deviation correlation type and the corresponding failure type marked as having a causal relationship in the causal compatibility mapping table, the corresponding failure propagation path is determined to have failed the causal compatibility check and is not marked. All causal confirmation propagation paths are summarized to form a causal confirmation propagation path set.

[0038] Methods for identifying key process segments that dominate failure include: From the set of deviation amplification paths, extract the deviation amplification paths corresponding to each causal confirmation propagation path in the set of causal confirmation propagation paths, and mark them as causal amplification paths. For each causal amplification path, obtain the amplification start step code and amplification end step code from the corresponding dense amplification segment record. Based on the amplification start step code and amplification end step code, determine the corresponding process segment. Here, a process segment refers to the interval of continuous process steps from the amplification start step code to the amplification end step code (including the process steps corresponding to the amplification start step code and amplification end step code). For all process segments, count the number of times each process segment appears in the set of causal confirmation propagation paths to obtain the segment occurrence frequency corresponding to each process segment. Here, the segment occurrence frequency is used to reflect the corresponding process segment's role in leading to final state failure. The degree to which a process segment is repeatedly covered in the entire causal confirmation propagation path is considered. The higher the frequency of a segment's occurrence, the more likely the corresponding process segment is to be a core amplification segment that multiple failure propagation paths jointly depend on. The frequency of occurrence of each process segment is compared, and the process segment with the highest frequency is selected as the critical process segment. If multiple process segments have the same frequency of occurrence and are all at the highest value, then all of these process segments are marked as candidate process segments. For each candidate process segment, the sum of the deviation increment values ​​of each amplification link is calculated to obtain the corresponding total deviation increment. The total deviation increment of each candidate process segment is compared, and the candidate process segment with the largest total deviation increment is selected as the critical process segment. Here, the total deviation increment is used to reflect the overall contribution of the corresponding process segment to deviation amplification in the entire causal confirmation propagation path. For critical process segments, a dominant amplification mechanism type identification is performed. Specifically, from all dense amplification segment records corresponding to the critical process segment, the amplification mechanism type of each amplification link is extracted, and the frequency of occurrence of different amplification mechanism types is statistically analyzed. The amplification mechanism type with the most frequent occurrence is taken as the dominant amplification mechanism type of the critical process segment. The dominant amplification mechanism type is used to characterize the most important physical mechanism of deviation amplification within the corresponding critical process segment. The critical process segment is associated and integrated with the corresponding dominant amplification mechanism type to form the critical process segment identification result of dominant failure. The critical process segment identification result not only identifies the most critical process step interval leading to final state failure, but also identifies the core physical mechanism of deviation amplification within the corresponding interval.

[0039] Methods for identifying key structural layers that dominate failure include: From the set of causal confirmation propagation paths, extract the structural layer codes of each voxel unit on all causal confirmation propagation paths and mark them as causal codes; count the number of structural layer codes in all causal codes to obtain the total number of associated voxels in each structural layer; compare the total number of associated voxels in each structural layer to obtain the structural layer with the largest number of associated voxels and designate it as the critical structural layer; the critical structural layer refers to the structural layer that carries the most deviation propagation links in all causal confirmation propagation paths leading to final state failure; extract all amplification links located in the critical structural layer from the set of causal confirmation propagation paths and mark them as critical amplification links; count the number of voxel units with additional material interface markings in all critical amplification links to obtain the number of interfaces; The number of voxel units in all critical amplification links is counted to obtain the voxel count; the ratio of the number of interfaces to the number of voxels is calculated to obtain the interface amplification ratio of the critical structural layer; a pre-set interface amplification dominance threshold is set, and the interface amplification ratio is compared with the interface amplification dominance threshold; wherein, the interface amplification dominance threshold is pre-set by those skilled in the art based on the typical contribution ratio of deviation amplification caused by material interface mismatch in the overall amplification effect; if the interface amplification ratio is greater than the interface amplification dominance threshold, an interface mismatch dominance marker is added to the critical structural layer, indicating that the deviation amplification in the critical structural layer is mainly driven by the material interface mismatch between the critical structural layer and the adjacent structural layer, and subsequent process compensation recommendations should prioritize adjustments to the interface treatment process of the critical structural layer.

[0040] Methods for identifying key deviation regions that dominate failure include: From the set of causal confirmation propagation paths, all voxel units corresponding to each amplification link on all causal confirmation propagation paths are extracted and marked as causal voxels. Spatial clustering analysis is performed on the distribution of all causal voxels in a three-dimensional spatial grid. Specifically, a spatial clustering neighborhood radius is preset, which is pre-set by those skilled in the art based on the spatial scale of typical defect-affected regions within integrated circuits. For each causal voxel, the number of voxel units that are also causal voxels within the spatial clustering neighborhood radius is counted to obtain the local clustering density of the corresponding causal voxel. A significant clustering density threshold is preset, and the local clustering density of each causal voxel is compared with the significant clustering density threshold. The significant clustering density threshold is pre-set by those skilled in the art based on the typical spatial concentration of the deviation amplification region. If the local clustering density is greater than the significant clustering density threshold, the corresponding causal voxel is marked as a cluster core voxel. If the local clustering density is less than or equal to the significant clustering density threshold, the corresponding causal voxel is not marked as a cluster core voxel. All cluster core voxels are merged with all causal voxels within the corresponding spatial clustering neighborhood radius to form a critical deviation region. For each voxel element within the critical deviation region, the corresponding maximum stress gradient direction is extracted from the corresponding deformation state record. The frequency of each maximum stress gradient direction within the critical deviation region is statistically analyzed to obtain the frequency of occurrence of multiple directions. The frequency of occurrence of each direction corresponding to the maximum stress gradient direction is compared, and the maximum stress gradient direction with the highest frequency of occurrence is selected as the dominant stress gradient direction in the critical deviation region. The dominant stress gradient direction is used to indicate to engineers which spatial direction the deviation within the critical deviation region is most likely to continue to expand, providing a physical basis for adjusting the directional parameters in subsequent process compensation suggestions.

[0041] Methods for generating root cause determination results include: By associating and integrating key process segments (including dominant amplification mechanism types), key structural layers (including dominant interface mismatch markers), key deviation regions (including dominant stress gradient directions), and the set of causal confirmation propagation paths, a failure root cause locking result is formed.

[0042] The visual output module is used to generate a 3D backtracking view of process disturbances, a highlighted distribution map of key areas, and process compensation suggestions based on the failure root cause identification results.

[0043] Methods for generating three-dimensional retrospective views of process disturbances include: The propagation and amplification of deviations within integrated circuits is a complex evolutionary process that unfolds simultaneously in two dimensions: three-dimensional space and process timing. Traditional two-dimensional cross-sectional diagrams or static three-dimensional renderings at a single moment cannot fully express the dynamic evolutionary characteristics of deviations, which gradually expand from a tiny origin and undergo significant amplification at specific process steps. Therefore, the three-dimensional backtracking view of process disturbances, based on three-dimensional structure rendering, further introduces a timing animation mechanism based on the execution order of steps and a color-separated highlighting mechanism based on the amplification mechanism type. This allows engineers to simultaneously perceive the spatial trajectory and temporal evolution of deviations and intuitively identify the distribution patterns of amplification links driven by different physical mechanisms in three-dimensional space. Specifically, from the failure root cause locking results, each causal confirmation propagation path within the causal confirmation propagation path set is extracted; each voxel unit on the same causal confirmation propagation path is connected sequentially according to the execution order of the deviation propagation chain links, forming a directed three-dimensional path segment sequence from the deviation origin voxel to the end voxel; a three-dimensional rendering space is pre-constructed, in which each structural layer is rendered in a semi-transparent layered form with a three-dimensional spatial mesh as the background skeleton, making the interlayer sequence relationship of each structural layer visually clear; the three-dimensional rendering space is a visualization display environment with the three-dimensional spatial range of the integrated circuit as the boundary and the three-dimensional center coordinates of the voxel unit as the geometric reference, used to present the voxel units in the three-dimensional process disturbance evolution and the causal confirmation propagation path in the failure root cause locking results in the form of three-dimensional graphics; the directed three-dimensional path segment sequence corresponding to each causal confirmation propagation path is superimposed and rendered on the three-dimensional spatial mesh with colored arrow lines; the deviation origin voxel is rendered with spherical nodes and the end voxel is rendered with diamond nodes, making the start and end points of deviation propagation clear at a glance in the view. For amplification links, color-separated highlight rendering based on amplification mechanism type is performed. Specifically, from the deviation amplification path corresponding to each causal confirmation propagation path, the step code, voxel code, and amplification mechanism type of all amplification links are extracted. In the 3D rendering space, the arrow lines corresponding to the amplification links are rendered with a highlight color that distinguishes them from ordinary propagation links (i.e., deviation propagation links not marked as amplification links). Specifically, a preset amplification mechanism highlight color mapping table is used, which records the rendering color corresponding to each amplification mechanism type. This color is then determined by those skilled in the art based on the visualization contrast requirements. Pre-defined; wherein, the rendering colors corresponding to each amplification mechanism type are different from each other, to ensure that the amplification links driven by different physical mechanisms can be clearly distinguished visually in the 3D rendering space; according to the amplification mechanism highlight color mapping table, each amplification link is highlighted with the corresponding rendering color according to the corresponding amplification mechanism type; for example, the amplification link with the amplification mechanism type of process type switching amplification is highlighted in red, the amplification link with the amplification mechanism type of material interface mismatch amplification is highlighted in orange, and the amplification link with the amplification mechanism type of process condition accumulation amplification is highlighted in yellow; A temporal animation mechanism is introduced, and a keyframe annotation mechanism is added on top of this. Specifically, the execution order of the steps is used as the timeline of the temporal animation frame sequence. In each frame of the temporal animation frame sequence, only the deviation propagation chain and amplification chain formed up to the corresponding process step are displayed, while deviation propagation chains that have not yet appeared in subsequent process steps are hidden. In each frame of the temporal animation frame sequence, the step code, step type, and set of process conditions for the corresponding process step are also displayed synchronously in the form of a sidebar information panel, as well as the deviation increment value and amplification mechanism type of the newly appearing amplification chain in the corresponding frame. By playing the animation frames one by one, engineers can dynamically observe the temporal evolution process of deviation expansion and amplification step by step in the manufacturing process. The system simultaneously understands the contribution of each process step to the process conditions of deviation propagation and amplification; extracts the amplification start step code and amplification end step code corresponding to the key process segment from the failure root cause locking results; marks all frames between the frame containing the corresponding amplification start step code and the frame containing the corresponding amplification end step code in the time-series animation frame sequence as key frames, and summarizes all key frames to obtain the key frame interval; when playing the key frame interval, a prominent highlighted border is used to mark the current key frame interval at the view border, and key process segment prompt information is simultaneously added to the sidebar information panel; the key frame annotation mechanism enables engineers to automatically identify the core process step interval where deviation continues to amplify during continuous animation playback, avoiding manual frame-by-frame comparison in all steps; The rendering results of the 3D path segment sequence, the color-separated highlight rendering results of the magnified chain section, the rendering results of the deviation origin voxel and end voxel, the time-series animation frame sequence, the sidebar information panel and key frame annotation results are integrated to form a 3D backtracking view of process disturbance.

[0044] Methods for generating key region highlight distribution maps include: In the root cause analysis results, the key process segments, key structural layers, and key deviation regions represent the core process steps, core physical carrier structural layers, and core deviation spatial locations that dominate the final state failure, respectively. However, these three types of information are physically interconnected, and displaying any one type in isolation cannot present the spatial coupling relationship between them. Therefore, the key region highlight distribution map adopts a multi-element overlay rendering strategy, simultaneously presenting the voxel highlight of the key deviation region, the hierarchical labeling of the key structural layer, the temporal labeling of the key process segments, and the location marker of the final state failure in the same 3D view. By introducing associated line segments, the spatial distribution and temporal location are visually and explicitly linked, enabling engineers to simultaneously perceive the complete distribution of the root cause in both the spatial and temporal domains. Specifically, from the root cause of failure locking results, the voxel codes and 3D center coordinates of all voxel units in the critical deviation region are extracted; in the 3D rendering space, all voxel units in the critical deviation region are rendered with a highlight fill color (such as orange voxel blocks); from the root cause of failure locking results, the dominant stress gradient direction of the critical deviation region is extracted; in the highlighted rendering results of the critical deviation region, the dominant stress gradient direction is marked with an arrow symbol to indicate which spatial direction the deviation is most likely to continue to expand in the corresponding critical deviation region; From the failure root cause locking results, extract the structural layer code and interface mismatch dominant marker of the key structural layer; in the 3D rendering space, render the entire space corresponding to the key structural layer with a semi-transparent highlighted border that is different from other structural layers; if the key structural layer has an interface mismatch dominant marker, then mark it with a dashed highlighted boundary at the interface between the key structural layer and the adjacent structural layer to indicate that the main deviation amplification of the corresponding key structural layer occurs at the material interface. From the failure root cause identification results, extract the amplification start step code, amplification end step code, and dominant amplification mechanism type corresponding to the key process segment; based on the amplification start step code and amplification end step code, display the position range of the key process segment in the entire process chain in the form of bar time sequence labels in the sidebar, and label the dominant amplification mechanism type with text next to the bar labels; Explicit linking lines are drawn between the highlighted rendering results of the critical deviation area and the bar timeline annotations of the critical process segments in the sidebar. These explicit linking lines are visible connecting lines between the main screen of the 3D rendering space and the sidebar, used to visually indicate the correspondence between spatial elements and timeline annotations, enabling engineers to intuitively perceive the correspondence between the spatial position of the critical deviation area and the temporal position of the critical process segment. Explicit linking lines are also drawn between the highlighted border of the critical structural layer and the bar timeline annotations of the critical process segments in the sidebar. The introduction of explicit linking lines enables a clear visual binding of information between the 3D spatial domain and the process timeline domain in a single view, overcoming the limitation of traditional 3D views that only present spatial distribution and lack temporal linking annotations. Extract final-state failure marker information from the basic dataset of process disturbances; extract failure location coordinates and failure type from the final-state failure marker information; render the corresponding failure location coordinates with red crosses in the 3D rendering space, and add text labels for the failure type; draw dashed line segments between the failure location markers and the critical deviation area, so that engineers can intuitively perceive the spatial distance and interlayer relationship between the critical deviation area and the final-state failure location. The key deviation area highlight rendering results, the arrow symbol annotation results of the dominant stress gradient direction, the highlight border rendering results of the key structural layer, the dashed highlight boundary annotation results of the interface mismatch dominant mark, the bar time sequence annotation results of the key process segment, the text annotation results of the dominant amplification mechanism type, the explicit associated line segment, the rendering results of the failure location coordinates, and the text annotation results of the failure type are integrated to form a key area highlight distribution map.

[0045] Methods for generating process compensation recommendations include: From the failure root cause identification results, the amplification start step code, amplification end step code, and dominant amplification mechanism type corresponding to the key process segments are extracted; from the process step parameter information in the process disturbance basic dataset, the step code, step type, and step process condition set of each process step in the key process segments are extracted; a pre-set process compensation rule library is established, which records process parameter adjustment suggestions corresponding to different step types, different deviation types, different amplification mechanism types, and different interface mismatch dominant marker combinations; the process compensation rule library is pre-established by those skilled in the art based on engineering experience and physical mechanism knowledge in the field of integrated circuit manufacturing processes; For example, if the dominant amplification mechanism of a critical process segment is process type switching amplification and the corresponding step type is switched from deposition to chemical mechanical polishing, the recommended process parameter adjustment is to insert a buffer annealing step between the two corresponding process steps to release interfacial stress concentration; if the dominant amplification mechanism is material interface mismatch amplification and the critical structural layer has an interface mismatch dominant marker, the recommended process parameter adjustment is to adjust the surface pretreatment process parameters before deposition of the corresponding structural layer to improve interfacial bonding quality; if the dominant amplification mechanism is process condition cumulative amplification and the corresponding step type is chemical mechanical polishing, the recommended process parameter adjustment is to reduce polishing pressure or shorten polishing time to mitigate the dish-shaped depression effect. For each process step within a critical process segment, the deviation type, deviation increment value, and amplification mechanism type of the corresponding amplification chain segment are extracted from the corresponding deviation amplification path. If the critical structure layer has an interface mismatch dominant marker, the combination of step type, deviation type, amplification mechanism type, and interface mismatch dominant marker is used as a matching condition and matched with the process compensation rule base to obtain corresponding process parameter adjustment suggestions. If the critical structure layer does not have an interface mismatch dominant marker, the combination of step type, deviation type, and amplification mechanism type is used as a matching condition and matched with the process compensation rule base to obtain corresponding process parameter adjustment suggestions. From the process disturbance backtracking results, the first out-of-specification step codes corresponding to the deviation origin voxels of each causal confirmation propagation path are obtained, and the process steps corresponding to each first out-of-specification step code are marked as first out-of-specification steps. The number of times each process step is considered as a first out-of-specification step in all causal confirmation propagation paths within the critical process segment is counted, obtaining the first introduction frequency for each process step. The first introduction frequency reflects the degree to which the corresponding process step appears as the initial deviation introduction node in all causal confirmation propagation paths leading to final state failure. A higher first introduction frequency indicates that the corresponding process step is more likely to be a deviation. The original triggering process for the accumulation of deviations; sorting the frequency of the first introduction of each process step from high to low to form a step compensation priority ranking; according to the ascending order of the step compensation priority ranking, each process step is marked with its corresponding step compensation priority; that is, the process step with the highest first introduction frequency has the highest step compensation priority, and the process step with the lowest first introduction frequency has the lowest step compensation priority; the introduction of step compensation priority enables engineers to implement process adjustments step by step according to the contribution of deviation introduction in order of high to low, under limited resources, rather than treating all process steps in the critical process segment equally; The process compensation suggestion results are formed by integrating the process parameter adjustment suggestions and step compensation priority for each process step within the key process segment. The process compensation suggestion results take the key process segment as the core framework and the step compensation priority ranking as the execution order guide, providing manufacturing engineers with a comprehensive process optimization scheme that combines information on both process parameter direction and step priority.

[0046] This embodiment overcomes the limitation of existing simulation analysis techniques that treat online metrology, deformation detection, and final failure results as separate entities by constructing a basic dataset of process disturbances that encompasses integrated circuit three-dimensional structural information, process step parameter information, online metrology results, local deformation detection information, and final failure marker information. This enables unified and correlated acquisition of data across the entire manufacturing chain. By mapping process step parameters, online metrology deviations, and local deformation information onto a voxel-by-voxel spatial mesh, and introducing material physical parameter fields, deviation monotonicity annotations, and maximum stress gradient direction annotations at the voxel level, a dataset is generated that uses voxel units as spatial resolution units and the execution order of steps as the temporal index dimension. This paper presents a three-dimensional process disturbance evolution model, overcoming the limitation of only being able to perform local simulation of a single process step and failing to uniformly model multi-step continuous cumulative disturbances. It constructs a cumulative disturbance propagation path by tracing the deviation propagation receiving voxel sequentially along the process steps, starting from the deviation origin voxel. During the tracing process, a mechanism for verifying the rationality of the propagation direction and screening material interface associations is introduced. By introducing three amplification mechanism types—process type switching amplification, material interface mismatch amplification, and process condition cumulative amplification—on the deviation amplification path, it achieves a leap from simply quantifying the deviation increment to identifying the amplification physical mechanism. Finally, it establishes spatial adjacency between the terminal voxel of the cumulative disturbance propagation path and the final failure location coordinates. Domain matching is performed, and causal compatibility verification between deviation types and failure types is conducted via a causal compatibility mapping table. This establishes an end-to-end causal link from the origin of process deviation to final failure, solving the core problem in traditional analysis methods where engineers can only rely on experience to repeatedly check multiple processes. By performing step frequency statistics and voxel space aggregation analysis on the causal confirmation propagation path set, the key process segments (including the dominant amplification mechanism type), key structural layers (including the dominant interface mismatch markers), and key deviation regions (including the dominant stress gradient direction) that dominate failure are accurately identified. Furthermore, by generating a sequence of directed propagation paths, highlighting amplified links in color, and creating a time-series animation frame sequence, the system achieves the desired results. The sidebar information panel and keyframe annotation mechanism provide a 3D backtracking view of process disturbances, along with a key area highlight distribution map that includes multi-element overlay rendering and explicit associated line segments. This allows engineers to intuitively perceive the 3D propagation and amplification trajectory of deviations within integrated circuits, their physical mechanism attribution, and spatial-temporal coupling relationships. By generating process compensation suggestions based on process compensation rule base matching and step compensation priority ranking, the system provides engineers with a comprehensive process optimization solution that combines process parameter direction guidance with deviation contribution priority ranking. This comprehensively improves the root cause localization accuracy and process improvement efficiency of integrated circuit 3D visualization simulation analysis in multi-step continuous manufacturing scenarios.

[0047] Example 2: Please refer to Figure 2 As shown, parts not described in detail in this embodiment are described in Embodiment 1. A method for three-dimensional visualization simulation analysis of integrated circuits is provided, the method including: Collect three-dimensional structural information, process step parameter information, online metrology result information, local deformation detection information and final state failure marker information of integrated circuits to form a basic dataset of process disturbances; Based on the basic dataset of process disturbances, process step parameter information, online measurement result information and local deformation detection information are mapped to the three-dimensional spatial mesh corresponding to the three-dimensional structural information to generate a three-dimensional process disturbance evolution body that characterizes the cumulative physical influence of each process step. Based on the three-dimensional process disturbance evolution model, the cumulative disturbance propagation path and deviation amplification path formed by different process steps inside the integrated circuit are traced, and the process disturbance backtracking results are generated. By performing correlation and matching analysis on the process disturbance backtracking results and the final state failure marking information, the key process segments, key structural layers and key deviation areas that dominate the failure are identified, and the failure root cause locking results are formed. Based on the failure root cause identification results, a three-dimensional backtracking view of process disturbances, a highlighted distribution map of key areas, and process compensation suggestions are generated.

[0048] Example 3: This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable code, which, when executed by the one or more processors, can perform the integrated circuit 3D visualization simulation analysis method described above.

[0049] The method or system according to the embodiments of this application can also be implemented using the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as ROM or hard disk, may store the integrated circuit three-dimensional visualization simulation analysis method provided in this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is merely exemplary; when implementing different devices, one or more components in the electronic device shown in this application may be omitted according to actual needs.

[0050] Example 4: One embodiment of this application discloses a computer-readable storage medium. The computer-readable storage medium stores computer-readable instructions. When the computer-readable instructions are executed by a processor, the integrated circuit three-dimensional visualization simulation analysis method according to an embodiment of this application, as described with reference to the above figures, can be performed. The storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0051] Furthermore, according to embodiments of this application, the processes described in the above-referenced flowcharts can be implemented as computer software programs. For example, this application provides a non-transitory machine-readable storage medium storing machine-readable instructions that can be executed by a processor to perform instructions corresponding to the method steps provided in this application, such as an integrated circuit three-dimensional visualization simulation analysis method. When this computer program is executed by a central processing unit (CPU), it performs the functions defined in the method of this application.

[0052] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0053] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0054] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A three-dimensional visualization simulation analysis method for integrated circuits, characterized in that, include: Collect three-dimensional structural information, process step parameter information, online metrology result information, local deformation detection information and final state failure marker information of integrated circuits to form a basic dataset of process disturbances; Based on the basic dataset of process disturbances, process step parameter information, online measurement result information and local deformation detection information are mapped to the three-dimensional spatial mesh corresponding to the three-dimensional structural information to generate a three-dimensional process disturbance evolution body that characterizes the cumulative physical influence of each process step. Based on the three-dimensional process disturbance evolution model, the cumulative disturbance propagation path and deviation amplification path formed by different process steps inside the integrated circuit are traced, and the process disturbance backtracking results are generated. By performing correlation and matching analysis on the process disturbance backtracking results and the final state failure marking information, the key process segments, key structural layers and key deviation areas that dominate the failure are identified, and the failure root cause locking results are formed. Based on the failure root cause identification results, a three-dimensional backtracking view of process disturbances, a highlighted distribution map of key areas, and process compensation suggestions are generated.

2. The method according to claim 1, characterized in that, Methods for generating three-dimensional process perturbation evolution bodies include: Based on the three-dimensional structural information, the three-dimensional spatial range of the integrated circuit is determined; the three-dimensional spatial range of the integrated circuit is divided into multiple voxel units; process step parameter information, online metrology result information, and local deformation detection information are mapped to each voxel unit to obtain the corresponding process history record, metrology deviation record, and deformation state record for each voxel unit; material property mapping is performed on each voxel unit to obtain material property fields and material physical parameter fields; the material property fields, material physical parameter fields, process history record, metrology deviation record, and deformation state record corresponding to each voxel unit are summarized and integrated to form a three-dimensional process disturbance evolution body.

3. The method according to claim 2, characterized in that, Methods for tracing the propagation path of cumulative disturbances include: For all voxel units, deviation origin voxel identification is performed; for each deviation origin voxel, downstream process steps are determined from all corresponding process steps; starting from each deviation origin voxel, the spatial propagation behavior of the deviation is tracked along the step execution order in the process step parameter information to each corresponding downstream process step to obtain an initial propagation relationship; for each initial propagation relationship, the rationality of the propagation direction is checked, the deviation propagation receiving voxel is identified, and a deviation propagation chain is constructed; for each deviation propagation receiving voxel, the spatial propagation behavior of the deviation is continued to be tracked to each corresponding downstream process step, and a deviation propagation chain is constructed; multiple deviation propagation chains originating from the same deviation origin voxel are arranged according to the step execution order to form a cumulative disturbance propagation path.

4. The method according to claim 3, characterized in that, Methods for tracing the path of deviation amplification include: For each deviation propagation link, calculate the corresponding deviation increment value; based on the deviation increment value, identify the amplification links in all deviation propagation links; for each amplification link, perform amplification mechanism type labeling; according to the step execution order, extract all amplification links on the same cumulative disturbance propagation path to form a deviation amplification path, and sequentially determine the dense amplification segment record of each deviation amplification path.

5. The method according to claim 4, characterized in that, Methods for correlation matching analysis between process disturbance backtracking results and final state failure marker information include: Based on the final state failure marker information, identify the failed voxels from all voxel units and determine the failure neighborhood voxel set for each failed voxel; based on the failure neighborhood voxel set and the cumulative disturbance propagation path, determine the failure association propagation path set for each failed voxel; for each failure association propagation path in each failure association propagation path set, perform causal compatibility verification, and mark the failure association propagation path that passes the causal compatibility verification as a causal confirmation propagation path; summarize all causal confirmation propagation paths to form a causal confirmation propagation path set.

6. The method according to claim 5, characterized in that, Methods for identifying key process segments, key structural layers, and key deviation regions that dominate failure include: Based on the set of causal confirmation propagation paths, identify causal amplification paths from all deviation amplification paths; determine the corresponding process segments based on the dense amplification segment records of each causal amplification path; calculate the segment occurrence frequency of each process segment, select the process segment with the highest segment occurrence frequency as the key process segment, and perform dominant amplification mechanism type identification on the key process segment. Based on the causal confirmation propagation path set, the total number of associated voxels in each structural layer is counted, and the structural layer with the largest total number of associated voxels is taken as the key structural layer; the interface amplification ratio of the key structural layer is calculated, and based on the interface amplification ratio, it is determined whether to add an interface mismatch dominant marker in the key structural layer. Extract all voxel units corresponding to each amplification link on all causal confirmation propagation paths and label them as causal voxels; perform spatial clustering analysis on all causal voxels to obtain clustered core voxels; merge all clustered core voxels with all causal voxels within the corresponding preset spatial clustering neighborhood radius to form a critical deviation region, and determine the dominant stress gradient direction of the critical deviation region.

7. The method according to claim 6, characterized in that, Methods for generating three-dimensional retrospective views of process disturbances include: For each causal confirmation propagation path, a corresponding directed 3D path segment sequence is constructed. In a pre-constructed 3D rendering space, using a 3D spatial mesh as the background skeleton, each structural layer is rendered in a semi-transparent layered form, and each directed 3D path segment sequence is superimposed on the 3D spatial mesh with colored arrow lines. For amplified links, color-separated highlight rendering based on the amplification mechanism type is performed. A temporal animation frame sequence is introduced, and a keyframe annotation mechanism is introduced on the basis of the temporal animation frame sequence. The rendering results of the 3D path segment sequence, the color-separated highlight rendering results of the amplified links, the temporal animation frame sequence, and the keyframe annotation results are integrated to form a 3D backtracking view of process disturbance.

8. The method according to claim 7, characterized in that, Methods for generating key region highlight distribution maps include: From the final failure marker information, the failure location coordinates and failure type are extracted. In the 3D rendering space, all voxel units, key structural layers and failure location coordinates in the critical deviation area are rendered. The dominant stress gradient direction, dominant amplification mechanism type and failure type are marked with arrow symbols and text respectively. The key process segments are displayed in the form of bar time sequence annotations. Explicit correlation lines are drawn between the rendering results of the critical deviation area and the bar time sequence annotations, and between the rendering results of the key structural layer and the bar time sequence annotations, to form a key area highlight distribution map.

9. The method according to claim 8, characterized in that, Methods for generating process compensation recommendations include: Each process step within the critical process segment is matched against a pre-defined process compensation rule library to obtain corresponding process parameter adjustment suggestions. The first exceeding step of each causal confirmation propagation path is obtained, and the number of times each process step within the critical process segment is the first exceeding step is counted to obtain the initial introduction frequency of each process step. The initial introduction frequencies of each process step are sorted from high to low, and the corresponding step compensation priorities are marked accordingly. The process parameter adjustment suggestions corresponding to each process step within the critical process segment are integrated with the step compensation priorities to form the process compensation suggestion results.

10. An integrated circuit three-dimensional visualization simulation analysis system, implementing the method according to any one of claims 1-9, characterized in that, include: The disturbance acquisition module is used to acquire three-dimensional structural information of integrated circuits, process step parameter information, online metrology result information, local deformation detection information and final state failure marker information to form a basic dataset of process disturbances. The disturbance modeling module is used to map process step parameter information, online measurement result information and local deformation detection information to the three-dimensional spatial mesh corresponding to the three-dimensional structural information based on the process disturbance basic dataset, and generate a three-dimensional process disturbance evolution body that characterizes the cumulative physical effects of each process step. The backtracking and deduction module is used to trace the cumulative disturbance propagation path and deviation amplification path formed inside the integrated circuit by different process steps based on the three-dimensional process disturbance evolution, and to generate process disturbance backtracking results; The root cause locking module is used to perform correlation and matching analysis between process disturbance backtracking results and final state failure marking information, identify key process segments, key structural layers and key deviation areas that dominate the failure, and form failure root cause locking results. The visual output module is used to generate a 3D backtracking view of process disturbances, a highlighted distribution map of key areas, and process compensation suggestions based on the failure root cause identification results.