Multi-physics field coupling silicon-based optical interconnection chip degradation model construction method and application

Abstract

The invention relates to the technical field of semiconductor photoelectronics, and discloses a multi-physics field coupling silicon-based optical interconnection chip degradation model construction method and application, and the method comprises the steps: firstly constructing a multi-physics field coupling analysis model, and quantitatively representing the coupling relation of thermodynamics, optics, electricity and material aging through an inter-field coupling equation and a material aging dynamic model; simulating field distribution of a chip key area under multi-environment stress through finite element analysis, and extracting data; then historical degradation data is collected by combining an accelerated aging test and long-term operation monitoring, and a degradation parameter prediction model is established after data processing based on machine learning; and finally, inputting a finite element result and a prediction result into a material aging dynamic model, and finally forming a multi-physics field coupling degradation model. The problems that the prior art depends on a single physical field, degradation mechanism analysis is incomplete, and service life evaluation is inaccurate are solved, and support can be provided for reliability evaluation, structure optimization and service life prediction of the silicon-based optical interconnection chip.

Description

Technical Field

[0001] This application relates to the field of semiconductor optoelectronics technology, specifically a method for constructing a degradation model of a multi-physics coupled silicon-based optical interconnect chip and its application. Background Technology

[0002] With the rapid development of 5G communication, data centers, and artificial intelligence, silicon-based optical interconnect chips have become core devices supporting high-speed information transmission due to their advantages of high bandwidth, low power consumption, and compatibility with CMOS processes. However, in actual service, silicon-based optical interconnect chips are exposed to complex environmental stresses such as high temperature, high humidity, and electromagnetic interference for extended periods. Their degradation process is not the result of a single physical field (such as thermodynamics or electricity), but rather a comprehensive effect of thermodynamics, optics, electricity, and material aging characteristics coupled together. The deterioration of key chip performance parameters (such as optical loss and wavelength drift) directly determines their lifespan and system reliability. Therefore, constructing an accurate degradation model is crucial for chip lifespan assessment and application optimization.

[0003] Currently, existing degradation analysis and lifetime assessment techniques for silicon-based optical interconnect chips still have limitations. On the one hand, most methods focus only on the influence of a single physical field, such as analyzing the effect of temperature on chip performance through thermal simulation or assessing the interference of electric fields on device stability through electrical simulation, failing to consider the coupling mechanisms between multiple physical fields. In actual operating conditions, increased temperature leads to changes in the dielectric constant of chip materials, which in turn affects the electric field distribution. Anomalies in the electric field can exacerbate fluctuations in the light absorption coefficient. Such cross-field coupling effects make it difficult for single-physical-field analysis to accurately reflect the true degradation law of the chip, easily leading to significant deviations between lifetime assessment results and actual conditions. On the other hand, existing degradation models do not adequately integrate the aging characteristics of materials. They typically only use simplified empirical aging formulas, failing to incorporate the aging dynamics of silicon-based optical interconnect chip-specific material systems (such as single-crystal silicon waveguides, metal electrodes, and packaging materials) under multiple stresses. This makes it impossible to quantify the differentiated impact of different combinations of environmental stresses (such as the synergy of high temperature and high humidity, and the superposition of electromagnetic interference and temperature) on chip lifetime.

[0004] Therefore, existing technologies are insufficient to effectively address the degradation modeling requirements of silicon-based optical interconnect chips under complex operating conditions. There is an urgent need for a degradation model construction method that can comprehensively consider the coupling effects of multiple physical fields, material aging characteristics, and multi-source data fusion. Summary of the Invention

[0005] The purpose of this application is to provide a method for constructing a degradation model of a multi-physics coupled silicon-based optical interconnect chip and its application, so as to solve the technical problems mentioned in the background art.

[0006] To achieve the above objectives, this application discloses the following technical solutions:

[0007] In a first aspect, this application discloses a method for constructing a degradation model of a multi-physics coupled silicon-based optical interconnect chip, the method comprising the following steps:

[0008] A multiphysics coupling analysis model is constructed. The coupling analysis model quantitatively characterizes the coupling relationship between thermodynamics, optics, electricity and material aging characteristics through inter-field coupling equations. The material aging characteristics are mathematically described by a material aging kinetic model. The inter-field coupling equations include the correlation between temperature and dielectric constant and the correction equation between electric field and light absorption coefficient.

[0009] The stress distribution and evolution of the chip under multiple environmental stress conditions were numerically simulated using the finite element method. Spatial distribution data of temperature field, electric field and optical field in key areas of the chip were extracted. The multiple environmental stress conditions include high temperature, high humidity and electromagnetic interference. The key areas include optical waveguide, electrode and packaging interface.

[0010] Based on accelerated aging tests and long-term operation monitoring, historical degradation data of the chip under different stress conditions were collected. The historical degradation data includes optical loss, wavelength drift and changes in material microstructure.

[0011] Based on machine learning algorithms, feature extraction and normalization are performed on the historical degradation data, and a degradation parameter prediction model is established through machine learning. The degradation parameter prediction model outputs the evolution trend curve of chip parameters with working time through time series analysis.

[0012] The finite element analysis results and the prediction results of the degradation parameter prediction model are input into the material aging kinetic model to quantify the chip lifetime decay law under different stress conditions and form a multi-physics coupled degradation model.

[0013] Preferably, the correlation between temperature and dielectric constant is expressed as a quadratic function to characterize the linear and nonlinear effects of temperature on dielectric constant.

[0014] Preferably, the modified equation for the electric field and the light absorption coefficient is characterized by a linear function to represent the effect of the electric field strength on the gain of the light absorption coefficient.

[0015] Preferably, the method of numerically simulating the stress distribution and evolution of the chip under multiple environmental stress conditions using finite element analysis, and extracting spatial distribution data of temperature field, electric field, and optical field in key areas of the chip, includes the following steps:

[0016] Based on the aforementioned inter-field coupling equation, the correlation between temperature and dielectric constant, as well as the corrected equation between electric field and light absorption coefficient, are embedded into the physical field control equation of finite element analysis as the basis for numerical simulation.

[0017] The core region of the chip's optical waveguide is divided into micrometer-level grids, and the chip's packaging structure region is divided into millimeter-level grids;

[0018] Automatically refine the mesh in regions with drastic changes in stress gradient to optimize the calculation accuracy of stress distribution in such regions;

[0019] Multiple environmental stress conditions are applied to the chip, numerical simulations are performed and iterative solutions are obtained, and finally the spatial distribution data of temperature field, electric field and light field in key areas of the chip are extracted.

[0020] Preferably, the method of collecting historical degradation data of the chip under different stress conditions based on accelerated aging tests and long-term operation monitoring includes the following steps:

[0021] Based on the extracted stress distribution data, the multi-environment stress conditions for accelerated aging tests were determined.

[0022] An accelerated aging test platform was built according to the determined environmental stress conditions. The chip sample was placed on the test platform to perform accelerated aging tests. Degradation parameters were continuously monitored and recorded to obtain short-term accelerated degradation data.

[0023] Select chip samples of the same specifications and conduct long-term monitoring under normal conditions to obtain natural aging data;

[0024] By integrating short-term accelerated degradation data with natural aging data, historical degradation data of the chip under different stress conditions can be obtained.

[0025] Preferably, the feature extraction and normalization processing of the historical degradation data includes:

[0026] Feature extraction: In the historical degradation dataset, multidimensional features are extracted, including: the hourly average change in light loss during the continuous monitoring period, the daily cumulative change in wavelength drift during the continuous monitoring period, the difference in light loss under high temperature conditions and normal temperature conditions, the ratio of wavelength drift change rate under high humidity conditions and normal humidity conditions, and the ratio of light loss or wavelength drift change per 100 hours in the accelerated aging test to the corresponding data in the long-term operation monitoring.

[0027] Normalization: The extracted multidimensional features are mapped to the [0,1] interval using the Min-Max normalization method. The optical loss feature has a lower limit of 0dB / cm and an upper limit of the maximum allowable optical loss specified in the chip design document. The wavelength drift feature has a lower limit of 0pm and an upper limit of the wavelength drift failure judgment value specified in the industry standard for silicon-based optical interconnect chips.

[0028] Preferably, the method of establishing a degradation parameter prediction model through machine learning includes the following steps:

[0029] Using the extracted multidimensional features as input, a long short-term memory network based on the attention mechanism is selected as the basic model. The attention mechanism assigns 3-5 times the weight of the features corresponding to sudden increases in light loss and wavelength jumps to the features of the normal time period.

[0030] The training and validation sets are divided in a 7:3 ratio. An adaptive moment estimator optimizer is used for iterative training. The root mean square error of the validation set does not exceed 5% as the convergence condition. This yields a degradation parameter prediction model that can output predicted values ​​of optical loss and wavelength drift at future moments.

[0031] Preferably, the method of outputting the evolution trend curve of chip parameters with operating time through time series analysis includes:

[0032] Based on the degradation parameter prediction model, time series data are generated using a rolling prediction method.

[0033] Trend fitting was performed on the prediction results to obtain the evolution trend curves of optical loss and wavelength drift over working time.

[0034] Preferably, the formation of the multiphysics coupling degradation model includes:

[0035] The finite element analysis results and the output of the degradation parameter prediction model are used as input parameters to input the material aging kinetics model.

[0036] By using the multiphysics coupling coefficients preset in the material aging kinetics model, the environmental stress and chip performance degradation parameters are correlated, the decay rate of chip lifetime over time under different stress conditions is calculated, and finally integrated into a multiphysics coupling degradation model that can directly output the relationship between stress conditions, performance degradation and remaining lifetime.

[0037] Secondly, this application discloses the application of the multiphysics field coupled silicon-based optical interconnect chip degradation model construction method described above in the reliability assessment, structural optimization or lifetime prediction of silicon-based optical interconnect chips.

[0038] Beneficial effects: This application constructs a multi-physics field coupling analysis model to quantitatively characterize the coupling relationship between thermodynamics, optics, electricity and material aging characteristics, breaking through the limitations of single-physics field analysis; it combines finite element analysis to obtain field distribution data of key areas of the chip, relies on comprehensive historical degradation data from accelerated aging and long-term monitoring, and establishes an accurate degradation parameter prediction model through machine learning. Finally, it integrates to form a multi-physics field coupling degradation model, which can effectively quantify the chip lifetime decay law under different stresses, solve the problems of incomplete degradation mechanism analysis and inaccurate lifetime assessment in existing technologies, and provide accurate support for reliability assessment, structural optimization and lifetime prediction of silicon-based optical interconnect chips. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart illustrating the method for constructing a degradation model of a multiphysics-coupled silicon-based optical interconnect chip provided in this application embodiment. Detailed Implementation

[0041] The technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0042] In this document, the term "comprising" is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0043] This embodiment provides, in a first aspect, a method such as Figure 1 The proposed method for constructing a degradation model for multiphysics-coupled silicon-based optical interconnect chips aims to address the current technical problems of inaccurate lifetime assessment and incomplete degradation mechanism analysis. The method includes the following steps:

[0044] A multiphysics coupling analysis model is constructed. This model quantitatively characterizes the coupling relationship between thermodynamics, optics, electricity, and material aging properties through inter-field coupling equations. The material aging properties are mathematically described by a material aging kinetic model. The inter-field coupling equations include the correlation between temperature and dielectric constant, as well as the corrected equations between electric field and light absorption coefficient. The material aging kinetic model is a multi-parameter aging rate equation based on the Arrhenius equation or the Alling model, modified according to the material properties of silicon-based optical interconnect chips (including single-crystal silicon optical waveguides, electrode metal layers, and packaging materials). It is used to describe the relationship between the aging reaction rate and time of materials under multiple stress couplings of temperature, electric field, and humidity.

[0045] The stress distribution and evolution of the chip under multiple environmental stress conditions were numerically simulated using the finite element method. Spatial distribution data of temperature field, electric field and optical field in key areas of the chip were extracted. The multiple environmental stress conditions include high temperature, high humidity and electromagnetic interference. The key areas include optical waveguide, electrode and packaging interface.

[0046] Based on accelerated aging tests and long-term operation monitoring, historical degradation data of the chip under different stress conditions were collected. The historical degradation data includes optical loss, wavelength drift and changes in material microstructure.

[0047] Based on machine learning algorithms, feature extraction and normalization are performed on historical degradation data, and a degradation parameter prediction model is established through machine learning. The degradation parameter prediction model outputs the evolution trend curve of chip parameters with working time through time series analysis.

[0048] The finite element analysis results and the prediction results of the degradation parameter prediction model are input into the material aging kinetic model to quantify the chip lifetime decay law under different stress conditions and form a multi-physics coupled degradation model.

[0049] Based on the above, the multi-physics coupled silicon-based optical interconnect chip degradation model construction method provided in this embodiment overcomes the limitations of existing technologies that rely solely on single-physics field analysis by constructing a multi-physics coupled analysis model that encompasses the correlation between temperature and dielectric constant, the correction equation for electric field and light absorption coefficient, and the material aging kinetics model. This method can quantitatively characterize the coupling relationship between thermodynamics, optics, electricity, and material aging characteristics, accurately reflecting the complex degradation mechanism under actual chip operating conditions. Finite element analysis is used to numerically simulate and extract data on the temperature, electric, and light field distributions of key chip regions under multiple environmental stress conditions, providing realistic physical field foundation data for subsequent degradation modeling. Historical degradation data is collected through a combination of accelerated aging tests and long-term operational monitoring, ensuring data coverage of both short-term accelerated and long-term natural degradation under different stress conditions. The degradation scenario improves the comprehensiveness and reliability of the data. Based on machine learning algorithms, feature extraction and normalization are performed on historical degradation data to establish a degradation parameter prediction model. Furthermore, time series analysis is used to output the evolution trend curve of chip parameters over working time, enabling accurate prediction of the degradation trend of key chip performance. Finally, the finite element analysis results and prediction results are input into the material aging kinetics model to quantify the chip lifetime decay law under different stress conditions. The resulting multi-physics coupled degradation model can systematically integrate multi-dimensional technical means, effectively solving the problems of incomplete chip degradation mechanism analysis and inaccurate lifetime assessment in existing technologies. It provides scientific and accurate technical support for the reliability assessment, structural optimization, and lifetime prediction of silicon-based optical interconnect chips, improving the model's ability to characterize chip degradation laws under complex working conditions and its practical application value.

[0050] In this embodiment, the relationship between temperature and dielectric constant is characterized by a quadratic function, representing both the linear and nonlinear effects of temperature on the dielectric constant. The specific expression is:

[0051] ε(T)=ε0×[1+γ1×(T-T0)+γ2×(T-T0) 2 ]

[0052] Where ε(T) is the dielectric constant of the chip material at temperature T; ε0 is the dielectric constant of the chip material at reference temperature T0; T is the actual operating temperature of the chip; T0 is the reference temperature, which is 25℃; γ1 is the linear influence coefficient of temperature on dielectric constant, obtained through experimental fitting; γ2 is the nonlinear influence coefficient of temperature on dielectric constant, obtained through experimental fitting.

[0053] Based on the above, the linear and nonlinear effects of temperature on dielectric constant are accurately characterized by a quadratic function-form temperature-dielectric constant correlation, overcoming the simplification defects of traditional linear models and improving the quantitative accuracy of thermodynamic and electrical interactions in multiphysics coupling analysis.

[0054] In this embodiment, the corrected equation for the electric field and light absorption coefficient is characterized by a linear function to represent the effect of the electric field intensity on the gain of the light absorption coefficient. The specific expression is:

[0055] α(E)=α0×[1+δ×E]

[0056] Where α(E) is the light absorption coefficient of the chip optical waveguide material when the electric field strength is E; α0 is the reference light absorption coefficient of the chip optical waveguide material when there is no external electric field (E=0); δ is the influence coefficient of the electric field on the light absorption coefficient, which is obtained through experimental fitting; and E is the actual electric field strength inside the chip.

[0057] Based on the above, the gain effect of electric field intensity on light absorption coefficient is clarified by using the electric field-light absorption coefficient correction equation in the form of a linear function. This fills the gap in the quantization of optical and electrical field coupling and makes the multiphysics model more consistent with the physical nature of light-electric interaction in actual chip operation.

[0058] In this embodiment, the stress distribution and evolution of the chip under multiple environmental stress conditions are numerically simulated using the finite element method, and the spatial distribution data of temperature field, electric field and light field in key areas of the chip are extracted, including the following steps:

[0059] Based on the field coupling equation, the correlation between temperature and dielectric constant and the modified equation between electric field and light absorption coefficient are embedded into the physical field control equation of finite element analysis as the basis for numerical simulation.

[0060] The core region of the chip's optical waveguide is divided into micrometer-level grids, and the chip's packaging structure region is divided into millimeter-level grids;

[0061] The mesh is automatically refined in regions with drastic changes in stress gradient to optimize the calculation accuracy of stress distribution in such regions, including the interface between the electrode and the optical waveguide.

[0062] Multiple environmental stress conditions are applied to the chip, numerical simulations are performed and iterative solutions are obtained, and finally the spatial distribution data of temperature field, electric field and light field in key areas of the chip are extracted.

[0063] Based on the above, by using multi-scale meshing and adaptive densification algorithms, the computational efficiency of the encapsulation structure region is optimized while ensuring the computational accuracy of core regions such as optical waveguides. This solves the problem of balancing accuracy and efficiency in finite element analysis and provides high-quality field distribution data for subsequent degradation modeling.

[0064] Furthermore, based on accelerated aging tests and long-term operational monitoring, historical degradation data of the chip under different stress conditions were collected, including the following steps:

[0065] Based on the extracted stress distribution data, the multi-environment stress conditions for accelerated aging tests were determined.

[0066] An accelerated aging test platform was built according to the determined environmental stress conditions. The chip sample was placed on the test platform to perform accelerated aging tests. Degradation parameters were continuously monitored and recorded to obtain short-term accelerated degradation data.

[0067] Select chip samples of the same specifications and conduct long-term monitoring under normal conditions for no less than 10,000 hours. Record degradation parameters such as optical loss and wavelength drift every 24 hours to obtain natural aging data.

[0068] By integrating short-term accelerated degradation data with natural aging data, historical degradation data of the chip under different stress conditions can be obtained.

[0069] Based on the above, accelerated test conditions are determined using finite element stress data. Short-term accelerated and long-term natural aging data are integrated to ensure that the historical degradation dataset covers different stress scenarios and time scales, thus solving the problem of insufficient model generalization ability caused by a single data source and improving the comprehensiveness of data support.

[0070] In this embodiment, feature extraction and normalization processing are performed on historical degradation data, including:

[0071] Feature extraction: Multidimensional features are extracted from the historical degradation dataset. These features include: the hourly average change in light loss during the continuous monitoring period, the daily cumulative change in wavelength drift during the continuous monitoring period, the difference in light loss under high temperature conditions and normal temperature conditions, the ratio of wavelength drift change rate under high humidity conditions and normal humidity conditions, and the ratio of light loss or wavelength drift change per 100 hours in the accelerated aging test to the corresponding data in the long-term operation monitoring.

[0072] Normalization: The extracted multidimensional features are mapped to the [0,1] interval using the Min-Max normalization method. The optical loss feature has a lower limit of 0dB / cm and an upper limit of the maximum allowable optical loss specified in the chip design document. The wavelength drift feature has a lower limit of 0pm and an upper limit of the wavelength drift failure judgment value specified in the industry standard for silicon-based optical interconnect chips.

[0073] Based on the above, by extracting multi-dimensional features and normalizing them based on design / standard thresholds, the dimensional differences of different parameters are eliminated, highlighting the differential impact of stress conditions on degradation, providing high-quality input for machine learning models, and avoiding the decrease in prediction accuracy caused by feature redundancy or bias.

[0074] Secondly, a degradation parameter prediction model is established through machine learning, including the following steps:

[0075] Using the extracted multidimensional features as input, a long short-term memory network based on the attention mechanism is selected as the basic model. The attention mechanism assigns 3-5 times the weight of the features corresponding to sudden increases in light loss and wavelength jumps to the features of the normal time period.

[0076] The training and validation sets are divided in a 7:3 ratio. An adaptive moment estimator optimizer is used for iterative training. The root mean square error of the validation set does not exceed 5% as the convergence condition. This yields a degradation parameter prediction model that can output predicted values ​​of optical loss and wavelength drift at future moments.

[0077] Based on the above, an attention mechanism is introduced to strengthen the feature weights of key degradation nodes, and an adaptive moment estimation optimizer is combined to improve the model training efficiency and convergence stability, thereby solving the problem that traditional machine learning is insufficient in capturing nonlinear degradation trends and improving the prediction accuracy of optical loss and wavelength drift.

[0078] In addition, time series analysis is used to output the evolution trend curves of chip parameters over operating time, including:

[0079] Based on the degradation parameter prediction model, time series data are generated by the rolling prediction method (predicting the parameter values ​​for the next 100 hours each time and feeding back the measured values ​​to the model for updates);

[0080] Trend fitting was performed on the predicted results to obtain the evolution trend curves of optical loss and wavelength drift over operating time (0-10000 hours). The evolution trend curve contains a high probability range, which was obtained through multiple simulation calculations. This means that during actual chip operation, there is a 95% probability that the actual values ​​of optical loss and wavelength drift will fall within this range. Through such curves, we can intuitively see the approximate range of parameter degradation, as well as the key time points when the parameters will change significantly (such as a sudden increase in optical loss or a sudden increase in wavelength drift).

[0081] Based on the above, an evolution curve containing a 95% confidence interval is generated by rolling prediction and trend fitting, which intuitively presents the high probability range and key inflection points of parameter degradation, solves the problem of insufficient visualization of degradation trends, and provides a clear reference for chip performance monitoring.

[0082] In this embodiment, forming a multiphysics coupling degradation model includes:

[0083] The finite element analysis results (temperature field, electric field, and optical field distribution data of key areas of the chip) and the output of the degradation parameter prediction model (optical loss and wavelength drift degradation data over time) are used as input parameters to input the material aging kinetics model.

[0084] By using the multiphysics coupling coefficients preset in the material aging kinetics model, the environmental stress (temperature, humidity, electromagnetic interference) and chip performance degradation parameters are correlated, the decay rate of chip lifetime over time under different stress conditions is calculated, and finally integrated into a multiphysics coupling degradation model that can directly output the relationship between stress conditions, performance degradation and remaining lifetime.

[0085] Based on the above, by integrating finite element field distribution data and machine learning prediction results, and using a material aging dynamics model to quantify the relationship between stress, performance, and lifespan, the resulting multiphysics coupled degradation model can directly output the lifespan decay law under different stress conditions, solving the problem that existing models are difficult to directly support engineering applications.

[0086] This embodiment provides, in its second aspect, the application of the multiphysics-coupled silicon-based optical interconnect chip degradation model construction method described above in the reliability assessment, structural optimization, or lifetime prediction of silicon-based optical interconnect chips.

[0087] In detail

[0088] In reliability assessment, the multi-physics coupling degradation model constructed in this invention is applied to the reliability verification process of silicon-based optical interconnect chips. The model is input with stress parameters of the chip's actual service environment (e.g., the typical temperature control range of commercial data centers is 18-27℃, relative humidity is 40%-60%, and the typical electromagnetic interference frequency of surrounding circuits is 500MHz). The model outputs the time (e.g., 7500 hours) for the chip's optical loss to rise from the initial value (0.12dB / cm for a single-crystal silicon optical waveguide) to the failure threshold (2dB / cm for industry-standard optical loss failure) under this stress condition, and the time (e.g., 7200 hours) for the wavelength drift to increase from 0pm to the failure threshold (100pm). The minimum of the two values ​​is taken as the chip's reliable operating life. Compared with the traditional assessment results that only consider thermal stress (e.g., 9000 hours), this model can accurately reflect the impact of multi-physics coupling on reliability and provide a scientific basis for the redundancy design of data center optical interconnect systems (e.g., the number of spare chips deployed).

[0089] In structural optimization, the model of this invention is used to analyze the influence of different structural parameters on degradation patterns in the high-stress region at the interface between the chip's optical waveguide and electrodes. The conventional width range of silicon-based optical waveguides (increasing from 1μm to 2μm) and electrode spacing (increasing from 3μm to 5μm) are selected as optimization variables. The model calculates the uniformity of electric field distribution (e.g., 15% electric field non-uniformity at 1μm width, decreasing to 8% at 2μm width) and changes in optical absorption coefficient under different parameter combinations, outputting the annual growth rate of optical loss for the corresponding structure (0.25dB / cm / year before optimization, 0.12dB / cm / year after optimization). Based on this, the parameter combination of "2μm optical waveguide width + 5μm electrode spacing" is determined, which avoids the high electric field concentration problem in narrow waveguides and prevents mode diffusion loss in wide waveguides, thus achieving reliability optimization of the chip structure.

[0090] In the lifespan prediction, the model is applied to the entire lifecycle management of the chip: For a chip that has been running online for 3000 hours, real-time optical loss (e.g., 0.5dB / cm) and wavelength drift (e.g., 25pm) data are collected by the online monitoring sensors integrated into the chip. After being input into the model of this invention, the model combines historical degradation trends (e.g., the average increase in optical loss of 0.127dB / cm / thousand-hours in the first 3000 hours) with the current stress conditions to output the remaining lifespan prediction value (e.g., 4200 hours) and a 95% confidence interval (3800-4600 hours). This result can be directly used to formulate a chip replacement plan (e.g., to start replacement before the remaining lifespan is 1000 hours), reducing the risk of data transmission interruption caused by sudden failures in the optical interconnect system.

[0091] The above applications are all based on the multi-physics coupling degradation model constructed in this invention. By quantifying the correlation between stress, performance and lifetime, it enables scientific management of silicon-based optical interconnect chips from design to service, thereby improving their reliability and economy in high-speed communication systems.

[0092] In the embodiments provided in this application, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor may be implemented in one or more of the following: application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments may be performed by a computer program instructing the associated hardware. During implementation, the program may be stored in a computer-readable storage medium or transmitted as one or more instructions or code on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media may be any available medium accessible to a computer. Computer-readable storage media may include, but are not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer.

[0093] Finally, it should be noted that the above description is only a preferred embodiment of this application and is not intended to limit this application. Although this application 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 this application should be included within the protection scope of this application.

Claims

1. A method for constructing a degradation model of a multiphysics-coupled silicon-based optical interconnect chip, characterized in that, The method includes the following steps: A multiphysics coupling analysis model is constructed. The coupling analysis model quantitatively characterizes the coupling relationship between thermodynamics, optics, electricity and material aging characteristics through inter-field coupling equations. The material aging characteristics are mathematically described by a material aging kinetic model. The inter-field coupling equations include the correlation between temperature and dielectric constant and the correction equation between electric field and light absorption coefficient. The stress distribution and evolution of the chip under multiple environmental stress conditions are numerically simulated using the finite element method. Spatial distribution data of temperature field, electric field and optical field in key areas of the chip are extracted. The multiple environmental stress conditions include high temperature, high humidity and electromagnetic interference. The key areas include optical waveguide, electrodes and packaging interface. Based on accelerated aging tests and long-term operation monitoring, historical degradation data of the chip under different stress conditions were collected. The historical degradation data includes optical loss, wavelength drift and changes in material microstructure. Based on machine learning algorithms, feature extraction and normalization are performed on the historical degradation data, and a degradation parameter prediction model is established through machine learning. The degradation parameter prediction model outputs the evolution trend curve of chip parameters with working time through time series analysis. The finite element analysis results and the prediction results of the degradation parameter prediction model are input into the material aging kinetic model to quantify the chip lifetime decay law under different stress conditions and form a multi-physics coupled degradation model.

2. The method for constructing a degradation model for a multiphysics-coupled silicon-based optical interconnect chip according to claim 1, characterized in that, The correlation between temperature and dielectric constant is characterized by a quadratic function to represent the linear and nonlinear effects of temperature on dielectric constant.

3. The method for constructing a degradation model for a multiphysics-coupled silicon-based optical interconnect chip according to claim 1, characterized in that, The modified equation for the electric field and light absorption coefficient characterizes the effect of electric field strength on the gain of light absorption coefficient through a linear function.

4. The method for constructing a degradation model for a multiphysics-coupled silicon-based optical interconnect chip according to claim 1, characterized in that, The method described above uses finite element analysis to numerically simulate the stress distribution and evolution of a chip under multiple environmental stress conditions, and extracts spatial distribution data of temperature, electric, and optical fields in key areas of the chip. This includes the following steps: Based on the aforementioned inter-field coupling equation, the correlation between temperature and dielectric constant, as well as the corrected equation between electric field and light absorption coefficient, are embedded into the physical field control equation of finite element analysis as the basis for numerical simulation. The core region of the chip's optical waveguide is divided into micrometer-level grids, and the chip's packaging structure region is divided into millimeter-level grids; Automatically refine the mesh in regions with drastic changes in stress gradient to optimize the calculation accuracy of stress distribution in such regions; Multiple environmental stress conditions are applied to the chip, numerical simulations are performed and iterative solutions are obtained, and finally the spatial distribution data of temperature field, electric field and light field in key areas of the chip are extracted.

5. The method for constructing a degradation model for a multiphysics-coupled silicon-based optical interconnect chip according to claim 4, characterized in that, The method of collecting historical degradation data of the chip under different stress conditions based on accelerated aging tests and long-term operation monitoring includes the following steps: Based on the extracted stress distribution data, the multi-environment stress conditions for accelerated aging tests were determined. An accelerated aging test platform was built according to the determined environmental stress conditions. The chip sample was placed on the test platform to perform accelerated aging tests. Degradation parameters were continuously monitored and recorded to obtain short-term accelerated degradation data. Select chip samples of the same specifications and conduct long-term monitoring under normal conditions to obtain natural aging data; By integrating short-term accelerated degradation data with natural aging data, historical degradation data of the chip under different stress conditions can be obtained.

6. The method for constructing a degradation model of a multiphysics-coupled silicon-based optical interconnect chip according to claim 1, characterized in that, The aforementioned feature extraction and normalization processing of the historical degradation data includes: Feature extraction: In the historical degradation dataset, multidimensional features are extracted, including: the hourly average change in light loss during the continuous monitoring period, the daily cumulative change in wavelength drift during the continuous monitoring period, the difference in light loss under high temperature conditions and normal temperature conditions, the ratio of wavelength drift change rate under high humidity conditions and normal humidity conditions, and the ratio of light loss or wavelength drift change per 100 hours in the accelerated aging test to the corresponding data in the long-term operation monitoring. Normalization: The extracted multidimensional features are mapped to the [0,1] interval using the Min-Max normalization method. The optical loss feature has a lower limit of 0dB / cm and an upper limit of the maximum allowable optical loss specified in the chip design document. The wavelength drift feature has a lower limit of 0pm and an upper limit of the wavelength drift failure judgment value specified in the industry standard for silicon-based optical interconnect chips.

7. The method for constructing a degradation model for a multiphysics-coupled silicon-based optical interconnect chip according to claim 6, characterized in that, The aforementioned method of establishing a degradation parameter prediction model through machine learning includes the following steps: Using the extracted multidimensional features as input, a long short-term memory network based on the attention mechanism is selected as the basic model. The attention mechanism assigns 3-5 times the weight of the features corresponding to sudden increases in light loss and wavelength jumps to the features of the normal time period. The training and validation sets are divided in a 7:3 ratio. An adaptive moment estimator optimizer is used for iterative training. The root mean square error of the validation set does not exceed 5% as the convergence condition. This yields a degradation parameter prediction model that can output predicted values ​​of optical loss and wavelength drift at future moments.

8. The method for constructing a degradation model for a multiphysics-coupled silicon-based optical interconnect chip according to claim 7, characterized in that, The aforementioned time-series analysis outputs the evolution trend curve of chip parameters over operating time, including: Based on the degradation parameter prediction model, time series data are generated using a rolling prediction method. Trend fitting was performed on the prediction results to obtain the evolution trend curves of optical loss and wavelength drift over working time.

9. The method for constructing a degradation model for a multiphysics-coupled silicon-based optical interconnect chip according to claim 1, characterized in that, The formation of the multiphysics coupling degradation model includes: The finite element analysis results and the output of the degradation parameter prediction model are used as input parameters to input the material aging kinetics model. By using the multiphysics coupling coefficients preset in the material aging kinetics model, the environmental stress and chip performance degradation parameters are correlated, the decay rate of chip lifetime over time under different stress conditions is calculated, and finally integrated into a multiphysics coupling degradation model that can directly output the relationship between stress conditions, performance degradation and remaining lifetime.

10. The method for constructing a degradation model for a multiphysics-coupled silicon-based optical interconnect chip according to any one of claims 1-9 is used in the reliability assessment, structural optimization, or lifetime prediction of silicon-based optical interconnect chips.

Images