CN-121980912-A - Semiconductor device reliability evaluation method, device, equipment and storage medium

CN121980912ACN 121980912 ACN121980912 ACN 121980912ACN-121980912-A

Abstract

The application discloses a method, a device, equipment and a storage medium for evaluating the reliability of a semiconductor device, wherein the method comprises the steps of obtaining geometric structure data and material data of a target semiconductor device; the method comprises the steps of establishing a finite element model based on geometric structure data and material data, applying a target load and boundary conditions to the finite element model, wherein the target load at least comprises a thermal load determined based on the working power consumption of a transistor of a target semiconductor device, carrying out mechanical simulation on the finite element model applied with the target load and the boundary conditions to obtain mechanical response data of the target semiconductor device, inputting the mechanical response data into a pre-trained target machine learning model, and outputting a reliability evaluation result of the target semiconductor device by the target machine learning model.

Inventors

ZHANG BAO
HUANG QISHENG
DANG WEN

Assignees

西安电子科技大学

Dates

Publication Date: 20260505
Application Date: 20251217

Claims (10)

1. A method for evaluating reliability of a semiconductor device, comprising: acquiring geometric structure data and material data of a target semiconductor device; Establishing a finite element model based on the geometry data and the material data; Applying a target load and boundary conditions to the finite element model, the target load including at least a thermal load determined based on transistor operating power consumption of the target semiconductor device; performing mechanical simulation on the finite element model applied with the target load and the boundary condition to obtain mechanical response data of the target semiconductor device; Inputting the mechanical response data into a pre-trained target machine learning model, and outputting a reliability evaluation result of the target semiconductor device by the target machine learning model.
2. The method of evaluating reliability of a semiconductor device according to claim 1, wherein the establishing a finite element model based on the geometry data and the material data comprises: Performing three-dimensional entity reconstruction based on the geometric structure data to generate an initial three-dimensional geometric model of the target semiconductor device; Defining corresponding material attribute parameters for each material region in the initial three-dimensional geometric model based on the material data to obtain an intermediate three-dimensional geometric model; and performing grid discrete operation on the intermediate three-dimensional geometric model to obtain the finite element model.
3. The semiconductor device reliability evaluation method according to claim 2, characterized in that: And in the grid discrete operation process, carrying out local grid encryption operation on the corners of metal interconnection lines of the intermediate three-dimensional geometric model and/or interface areas among different material layers of the intermediate three-dimensional geometric model.
4. The method for evaluating reliability of a semiconductor device according to claim 1, wherein the thermal load is obtained by: Obtaining the instantaneous power loss of the transistor in a dynamic working state through electrothermal coupling simulation; applying the instantaneous power loss as a bulk heat source to a corresponding transistor structure region in the finite element model; based on the bulk heat source, calculating to obtain non-uniform transient temperature field distribution of the target semiconductor device in a dynamic working state by solving a heat conduction equation; And obtaining the thermal load according to the non-uniform transient temperature field distribution.
5. The method according to claim 1, wherein the mechanically simulating the finite element model to which the target load and the boundary condition are applied to obtain mechanical response data of the target semiconductor device, comprises: And solving the finite element model applied with the target load and the boundary condition based on a transient dynamics analysis method, and outputting stress distribution data and strain distribution data of all areas of the finite element model to obtain the mechanical response data.
6. The method for evaluating reliability of a semiconductor device according to claim 1, wherein the target machine learning model is obtained by: acquiring finite element simulation mechanical response historical data of a plurality of semiconductor devices and corresponding experimental reliability test historical results; And taking the mechanical response historical data as input, taking the experimental reliability test historical result as a training target, and training an initial machine learning model to obtain the target machine learning model.
7. The method for evaluating the reliability of a semiconductor device according to claim 1, further comprising: acquiring actual reliability test data of the target semiconductor device; Comparing the actual reliability test data with a reliability evaluation result output by the target machine learning model to obtain a comparison error; and if the comparison error exceeds a preset tolerance, performing incremental learning on the target machine learning model by using the actual reliability test data.
8. A mechanical behavior prediction apparatus of a semiconductor device, comprising: The acquisition module is used for acquiring geometric structure data and material data of the target semiconductor device; A building module for building a finite element model based on the geometry data and the material data; an application module for applying a target load and boundary conditions to the finite element model, the target load including at least a thermal load determined based on transistor operating power consumption of the target semiconductor device; the simulation module is used for carrying out mechanical simulation on the finite element model applied with the target load and the boundary condition to obtain mechanical response data of the target semiconductor device; And the output module is used for inputting the mechanical response data into a pre-trained target machine learning model, and outputting a reliability evaluation result of the target semiconductor device by the target machine learning model.
9. An electronic device, comprising: At least one processor; At least one memory for storing at least one program; The semiconductor device reliability evaluation method according to any one of claims 1 to 7, when at least one of the programs is executed by at least one of the processors.
10. A computer-readable storage medium, in which a processor-executable program is stored, which when executed by a processor is for implementing the semiconductor device reliability evaluation method according to any one of claims 1 to 7.

Description

Semiconductor device reliability evaluation method, device, equipment and storage medium Technical Field The present application relates to the field of reliability analysis technologies for semiconductor devices, and in particular, to a method, an apparatus, a device, and a storage medium for evaluating reliability of a semiconductor device. Background With the continued progress in semiconductor manufacturing processes, device feature sizes continue to shrink and integration density increases significantly. Technologies such as three-dimensional heterogeneous integrated packaging and microelectromechanical systems are being widely used because of their advantages in terms of improving performance, reducing size, and implementing heterogeneous integration. Because of the highly complex physical configurations of multi-layer chip stacking, heterogeneous material interfaces, movable microstructures and the like, the mechanical problems of thermal mechanical stress, mechanical fatigue under dynamic load and the like caused by mismatch of thermal expansion coefficients of materials become extremely prominent. These problems have become critical factors affecting the long-term reliability of semiconductor devices, which are prone to metal interconnect line breakage, dielectric layer cracking, through-silicon via interface delamination, and the like. In the related art, the evaluation of the mechanical behavior and the reliability mainly depends on the traditional method of finite element simulation combined with experimental verification. However, the method needs to rely on a large amount of time-consuming and expensive physical experiments for calibration and verification, and is difficult to predict the reliability of the semiconductor device efficiently and accurately. Disclosure of Invention The application aims to provide a method, a device, equipment and a storage medium for evaluating the reliability of a semiconductor device, which can efficiently and accurately predict the reliability of the semiconductor device. In a first aspect, an embodiment of the present application provides a method for evaluating reliability of a semiconductor device, including: acquiring geometric structure data and material data of a target semiconductor device; Establishing a finite element model based on the geometry data and the material data; Applying a target load and boundary conditions to the finite element model, the target load including at least a thermal load determined based on transistor operating power consumption of the target semiconductor device; performing mechanical simulation on the finite element model applied with the target load and the boundary condition to obtain mechanical response data of the target semiconductor device; Inputting the mechanical response data into a pre-trained target machine learning model, and outputting a reliability evaluation result of the target semiconductor device by the target machine learning model. According to some embodiments of the application, the building a finite element model based on the geometry data and the material data comprises: Performing three-dimensional entity reconstruction based on the geometric structure data to generate an initial three-dimensional geometric model of the target semiconductor device; Defining corresponding material attribute parameters for each material region in the initial three-dimensional geometric model based on the material data to obtain an intermediate three-dimensional geometric model; and performing grid discrete operation on the intermediate three-dimensional geometric model to obtain the finite element model. According to some embodiments of the application, during the grid discretization operation, a local grid encryption operation is performed at metal interconnect line corners of the intermediate three-dimensional geometric model and/or at interface regions between different material layers of the intermediate three-dimensional geometric model. According to some embodiments of the application, the thermal load is obtained by: Obtaining the instantaneous power loss of the transistor in a dynamic working state through electrothermal coupling simulation; applying the instantaneous power loss as a bulk heat source to a corresponding transistor structure region in the finite element model; based on the bulk heat source, calculating to obtain non-uniform transient temperature field distribution of the target semiconductor device in a dynamic working state by solving a heat conduction equation; And obtaining the thermal load according to the non-uniform transient temperature field distribution. According to some embodiments of the application, the mechanical simulation of the finite element model applied with the target load and the boundary condition, to obtain mechanical response data of the target semiconductor device, includes: And solving the finite element model applied with the target load and the boundary condition bas