Semiconductor encapsulation adhesive curing warpage control method and device

Abstract

The application relates to the technical field of semiconductor packaging manufacturing, and provides a semiconductor packaging glue curing warping control method and device, which comprises the following steps: obtaining a constitutive model of dynamic evolution of semiconductor packaging glue material performance with curing degree; based on the constitutive model, performing thermal-mechanical-flow multi-field two-way full coupling transient simulation on a curing process of the semiconductor packaging glue to predict warping deformation; and based on a prediction result of the warping deformation, reversely optimizing curing process parameters. The application effectively solves the problem of poor precision caused by the fact that traditional simulation ignores material dynamic phase change, realizes high-precision prediction of warping evolution in the whole curing process and reverse optimization control of the process, and greatly improves prediction precision and research and development efficiency.

Description

Technical Field

[0001] This application relates to the field of semiconductor packaging manufacturing technology, and more specifically, to a method and apparatus for controlling warpage during the curing of semiconductor packaging adhesive. Background Technology

[0002] As the semiconductor industry rapidly iterates towards advanced processes, advanced packaging technologies such as wafer-level packaging, chiplet heterogeneous integration, and 2.5D / 3D stacked packaging have become mainstream. As a core material for advanced packaging, the curing process of encapsulating adhesives directly determines the reliability and yield of the packaging structure. During the curing process, the adhesive transforms from a liquid to a solid state through gelation, accompanied by multiple effects such as exothermic curing chemical reactions, volume shrinkage, and thermal expansion mismatch, which can easily lead to macroscopic warpage deformation of the wafer or substrate. When the warpage exceeds the control threshold, it can cause problems such as wafer breakage, insufficient back-side grinding precision, dicing misalignment, and subsequent mounting failures, becoming a core bottleneck restricting the improvement of advanced packaging yield. Currently, the industry's warpage control requirements for high-end encapsulating adhesives have reached the ±2μm level, and with the popularization of 12-inch wafers and the increase in multi-chip integration, the difficulty of warpage control continues to increase.

[0003] Current research and development on curing warpage still relies primarily on the traditional model of trial and error and repeated experiments. Researchers adjust the ratios of resin, curing agent, and filler in the formulation, as well as process parameters such as curing heating rate, holding temperature, and cooling rate, to prepare multiple sets of samples for field testing and verification. This model has three major drawbacks: First, the development cycle is long, with the verification cycle for a single batch of formulation and process taking as long as 2 to 3 months, and the complete development cycle for a high-end encapsulating adhesive generally exceeding 12 months; second, the development cost is high, with expensive consumables for wafer-level experiments, and repeated trial and error resulting in a large waste of raw materials and labor time, with the development cost of a single project exceeding ten million yuan; third, the understanding of the mechanism is insufficient, as traditional experiments can only obtain macroscopic results of the final warpage amount, and cannot analyze the dynamic evolution of the temperature field, curing degree field, and stress field during the curing process, making it difficult to locate the core root cause of warpage. Numerical simulation technology is the core path to solving the above pain points, but existing simulation methods have obvious technical shortcomings: First, insufficient coupling dimensions and poor prediction accuracy. Most existing technologies only use simple thermal-structural two-way coupling models, assuming the encapsulating adhesive is a homogeneous solid material, ignoring the flow behavior of the adhesive in the early curing stage, the degree of curing gradient distribution, and the coupling effect of curing chemical shrinkage and thermal expansion. This results in warpage prediction errors generally exceeding 20%, and errors of more than 50% for some complex structures. It can only make qualitative trend judgments and cannot guide precise control at the mass production level. Second, the parameter system is lacking and the model reusability is poor. There is a lack of standardized material parameter calibration methods and model verification processes. Most core input parameters rely on literature values ​​or empirical estimates, and the model cannot be quickly reused under different formulations and processes. Third, it only supports forward verification and lacks reverse optimization capabilities. Existing technologies mostly stay at the forward verification stage of calculating warpage for a given process, and cannot meet the engineering requirements of outputting the optimal process for a given warpage target, making it difficult to form a complete R&D closed loop.

[0004] Due to the aforementioned technological bottlenecks, domestic enterprises have long lagged behind international giants in the research and development of high-end semiconductor encapsulants. Companies such as Henkel, Dow Chemical, and Nitto Denko, leveraging their closed simulation and experimental integrated R&D systems, hold over 70% of the domestic high-end encapsulant market share. Therefore, developing a high-precision, practical, and reverse-optimized simulation control method for encapsulant curing warpage is of significant strategic importance for breaking through technological barriers and achieving domestic substitution of high-end encapsulants. Summary of the Invention

[0005] The purpose of this application is to provide a method and apparatus for controlling warpage during the curing of semiconductor encapsulant, so as to alleviate the technical problems of poor warpage prediction accuracy caused by neglecting the dynamic evolution of material properties during the curing process and insufficient multi-field coupling dimensions in the prior art.

[0006] In a first aspect, the present invention provides a method for controlling warpage during the curing of semiconductor encapsulant, comprising: Obtain a constitutive model of the dynamic evolution of semiconductor encapsulation adhesive material properties with degree of curing; Based on the constitutive model, a thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation of the curing process of the semiconductor encapsulant is performed to predict warpage deformation. Based on the predicted warpage deformation, the curing process parameters are optimized in reverse.

[0007] In an optional implementation, the constitutive model includes: The mapping relationship between thermophysical and mechanical property parameters and degree of curing; The coupled strain relationship between curing shrinkage and thermal expansion.

[0008] In an optional implementation, the constitutive model further includes: An autocatalytic curing kinetic model is used to describe the relationship between the curing reaction rate and the degree of curing and temperature. Viscosity evolution model, used to describe the dynamic changes in viscosity with degree of curing and temperature.

[0009] In an optional implementation, the thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation achieves closed-loop coupling in the following manner: The temperature field drives the change of the curing degree field, the change of the curing degree field updates the material performance parameters, the change of the material performance parameters affects the flow field and stress field, and thus generates warping deformation. The warping deformation is fed back to the temperature field to achieve temperature field redistribution.

[0010] In an optional implementation, the thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation uses an arbitrary Lagrange-Euler method to handle the flow deformation and free surface problems of the semiconductor encapsulant. The time step of the transient simulation is adaptively adjusted according to the curing reaction rate, and the time step is automatically encrypted at the gel point stage.

[0011] In an optional implementation, the reverse optimization curing process parameters include: Rapid prediction of warpage using a surrogate model; The prediction results are globally optimized based on a multi-objective optimization algorithm, and the solidification process parameters are output.

[0012] In an optional implementation, the multi-objective optimization algorithm aims to minimize warpage and curing time, using process parameters and formulation parameters as optimization variables. The output curing process parameters include a curing process curve containing the heating rate, holding temperature and time, and cooling rate, as well as the allowable fluctuation range of each parameter.

[0013] In an optional implementation, the method further includes: A three-level progressive verification process (material level, structural level, and mass production level) is adopted to verify the constitutive model and the predicted results of the warping deformation, and the model parameters are corrected based on the verification deviation.

[0014] In a second aspect, the present invention provides a semiconductor encapsulation adhesive curing warpage control device, comprising: The model building module is used to obtain constitutive models of material properties that evolve dynamically with the degree of curing. The fully coupled simulation module is used to perform a thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation of the curing process of the semiconductor encapsulant based on the constitutive model, and to predict warpage deformation. The reverse optimization module is used to reverse optimize the curing process parameters based on the predicted warping deformation.

[0015] In an optional embodiment, the apparatus further includes: The parameter inversion correction module is used to automatically correct the interface strength and equivalent thermal conductivity parameters based on the verification deviation.

[0016] Beneficial effects: Through full-coupling of thermo-mechanical-fluid processes and simulation of the entire curing process, the maximum error in wafer-level packaging warpage prediction is less than 1.5μm, which can stably meet the high-end packaging control requirements of ±2μm. The prediction accuracy is improved compared to traditional thermo-structural coupling models. The principle is that the constitutive model accurately captures the dynamic phase transition process of materials from liquid to solid state, and the multi-field bidirectional full-coupling mechanism realistically reproduces the synergistic evolution of temperature field, curing degree field and stress field, avoiding the mechanism distortion caused by traditional fixed parameter assumptions.

[0017] Replacing numerous physical trial-and-error experiments, the development cycle of a single encapsulating adhesive is shortened, and the overall R&D cost is reduced. The principle behind this is that the surrogate model and multi-objective optimization algorithm form a closed loop of process reverse output, directly driving R&D optimization. Simultaneously, standardized modeling and verification processes improve the efficiency of building new project models.

[0018] Through precise warpage control and process optimization, the yield of advanced packaging and curing processes is improved, effectively reducing quality issues such as wafer breakage and dicing failures. The principle behind this is that the optimized process curve balances curing shrinkage stress and thermal stress, achieving precise control over warpage and simultaneously improving the consistency of curing within the wafer surface.

[0019] This method has a complete process of experimental calibration, modeling verification, and optimized output. It can be directly embedded into the existing R&D system of enterprises and is applicable to various semiconductor packaging adhesives such as underfill adhesives, molding encapsulation adhesives, and chip bonding adhesives. It can be quickly reused for different packaging structures and formulation systems. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic flowchart of a semiconductor encapsulation adhesive curing warpage control method according to an embodiment of this application; Figure 2 This is a schematic diagram of the multi-physics coupling mechanism in the curing process of the encapsulating adhesive according to an embodiment of this application; Figure 3 This is a comparison chart of the curing degree evolution curves under different heating rates according to the embodiments of this application; Figure 4 This is a simulation cloud diagram of the warpage distribution after wafer curing according to an embodiment of this application; Figure 5 This is a bar chart comparing the warpage and error of embodiments of this application; Figure 6 This is a comparison chart of the curing process curves of embodiments of this application; Figure 7 This is a schematic diagram of a semiconductor encapsulation adhesive curing warpage control device according to an embodiment of this application. Detailed Implementation

[0022] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0023] like Figure 1 As shown, this embodiment provides a method for controlling warpage during the curing of semiconductor encapsulating adhesive. The method includes the following steps: Step S110: Obtain a constitutive model of the dynamic evolution of the semiconductor encapsulation adhesive material properties with curing degree.

[0024] Specifically, the degree of curing is a dimensionless physical quantity describing the extent of the crosslinking reaction of the encapsulating adhesive, typically ranging from 0 to 1. In traditional simulation methods, the encapsulating adhesive is often treated as a homogeneous solid material, and calculations are performed using fixed parameters, neglecting the dynamic phase transition process of the adhesive changing from a liquid to a solid state during the early stages of curing. In this embodiment, the degree of curing is used as the core driving variable to construct a constitutive model of the dynamic evolution of material properties with the degree of curing. This constitutive model can describe the thermophysical properties, mechanical properties, viscosity, and volume shrinkage of the encapsulating adhesive at any degree of curing. For example, when the degree of curing is below the gel point, the material mainly exhibits fluid properties with extremely low shear modulus; when the degree of curing is above the gel point, the material's mechanical properties increase non-linearly with increasing degree of curing. By introducing a degree-dependent mechanism, accurate dynamic material input is provided for subsequent multiphysics calculations, fundamentally avoiding mechanistic distortions caused by fixed parameter assumptions.

[0025] Step S120: Based on the constitutive model, perform a thermal-mechanical-fluid multi-field bidirectional fully coupled transient simulation of the curing process of the semiconductor encapsulation adhesive to predict warping deformation.

[0026] Specifically, such as Figure 2 As shown, the bidirectional fully coupled transient simulation of thermo-mechanical-fluid multi-fields is achieved by constructing a closed-loop feedback logic. The temperature field drives changes in the curing degree field, which in turn updates the material performance parameters. These changes affect the flow and stress fields, leading to warping deformation. This warping deformation is fed back to the temperature field to achieve temperature redistribution. This bidirectional closed-loop coupling mechanism differs from traditional sequential unidirectional transmission simulations. During curing, changes in the temperature field trigger chemical reactions, causing the curing degree to evolve. This evolution in curing degree, in turn, alters parameters such as viscosity and modulus of the material, thus affecting the flow behavior of the adhesive and the stress distribution within the structure. The macroscopic warping deformation caused by stress and flow alters the geometric configuration and heat conduction path, thereby redistributing the temperature field. Through this fully coupled transient simulation, the synergistic mechanism of multiple physics fields throughout the curing process can be realistically reproduced, accurately predicting the warping deformation state of the wafer or substrate at any given time.

[0027] Step S130: Based on the predicted warpage deformation, the curing process parameters are optimized in reverse.

[0028] Specifically, traditional technologies often remain at the forward verification stage of calculating warpage for a given process, failing to proactively drive R&D optimization. This embodiment, after obtaining high-precision warpage deformation prediction results, uses them as feedback input, and an optimization algorithm outputs the optimal curing process parameters in reverse. These curing process parameters can include heating rate, holding temperature and time, and cooling rate. Through reverse optimization, it is possible to find the process combination that minimizes warpage and shortens curing time while meeting mass production control requirements, thus upgrading from a verification tool to an R&D engine. It should be understood that although this embodiment uses the wafer-level encapsulation adhesive curing process as an example, this method is also applicable to the curing warpage control of various semiconductor encapsulation adhesives, such as underfill adhesives and chip adhesives, as long as the logic of multi-field fully coupled simulation based on a dynamic constitutive model and reverse optimization of the process is met.

[0029] Through the above scheme, this embodiment takes the degree of curing as the core driving variable to run through the material evolution and multiphysics calculation, effectively solves the problem of poor accuracy caused by the neglect of dynamic phase transformation of materials in traditional simulation, establishes the technical foundation for high-precision prediction and reverse optimization, and realizes precise control of warpage evolution throughout the curing process.

[0030] In some embodiments, the specific details of the constitutive model for obtaining the dynamic evolution of semiconductor encapsulating adhesive material properties with degree of curing are further elaborated. The constitutive model includes: the mapping relationship between thermophysical and mechanical property parameters and degree of curing; and the coupled strain relationship between curing shrinkage and thermal expansion.

[0031] Specifically, the mapping relationship between thermophysical and mechanical property parameters and degree of cure is used to describe the dynamic evolution of the encapsulating adhesive's physical properties throughout the curing process. In traditional simulations, the encapsulating adhesive is often treated as a homogeneous material with fixed thermal conductivity, specific heat capacity, coefficient of thermal expansion, and elastic modulus. This assumption leads to significant calculation errors near the gel point, where the degree of cure changes drastically. In this embodiment, a continuous mapping relationship is established between core parameters such as thermal conductivity, specific heat capacity, coefficient of thermal expansion, elastic modulus, and Poisson's ratio and the degree of cure. When the degree of cure is below the gel point, the material mainly exhibits fluid properties with extremely low shear modulus, and heat conduction relies primarily on the combined effects of convection and conduction within the adhesive solution. When the degree of cure exceeds the gel point, the material undergoes cross-linking and curing, and the mechanical properties increase non-linearly with increasing degree of cure, while heat conduction becomes purely solid-state conduction. Through staged testing using a thermomechanical analyzer and a dynamic mechanical analyzer, continuous parameter evolution curves can be obtained and fitted as interpolation functions, replacing the fixed constant assumptions of traditional models and providing accurate dynamic material input for multi-field coupled simulations.

[0032] The coupled strain relationship between curing shrinkage and thermal expansion is used to deconstruct the complex deformation mechanism during the curing process. During curing, the encapsulating adhesive undergoes volume shrinkage due to chemical reactions, while temperature changes cause thermal expansion or contraction. This embodiment decomposes the total strain during curing into two parts: chemical shrinkage strain and thermal expansion strain. The chemical shrinkage strain is linearly related to the degree of cure, while the thermal expansion strain is related to temperature and the coefficient of thermal expansion. Physically, before the gel point, the adhesive is in a liquid state, and chemical shrinkage is primarily volume shrinkage. At this point, the material does not have the capacity to withstand shear stress, thus no macroscopic internal stress is generated. After the gel point, the material becomes solid, and chemical shrinkage and thermal expansion work together, constrained by the substrate and other structures, generating internal stress and ultimately driving warping deformation. This coupled strain decomposition allows for precise differentiation of the contribution ratios of thermal stress and curing shrinkage stress, providing mechanistic support for subsequent targeted optimization of the coefficient of thermal expansion matching or curing process curves.

[0033] Furthermore, the constitutive model also includes: an autocatalytic curing kinetic model to describe the relationship between the curing reaction rate and the degree of curing and temperature; and a viscosity evolution model to describe the dynamic changes in viscosity with the degree of curing and temperature.

[0034] Specifically, the autocatalytic curing kinetic model is the core sub-model describing the crosslinking reaction process of the encapsulating adhesive. The curing reaction rate is affected not only by the current temperature but also by the feedback effect of the degree of reaction already completed, i.e., the degree of curing, exhibiting autocatalytic characteristics. This embodiment uses the Kamal autocatalytic model, whose expression is: In the formula, For degree of curing, For reaction time, , The reaction rate constant is... , Let be the reaction order. The reaction rate constant follows the Arrhenius equation, i.e. ,in Pre-exponential factor, For activation energy, Let be the ideal gas constant. The absolute temperature is used. All the above parameters can be obtained by differential scanning calorimetry experiments at different heating rates. This model can accurately calculate the curing degree evolution curve under any temperature-time path, identify key characteristic nodes such as the gel point and the point of complete curing, thus providing a reference time axis for the reaction process in subsequent simulations.

[0035] Viscosity evolution models are used to cover the rheological behavior of encapsulating adhesives throughout their entire lifecycle, from the initial liquid state and gelation to complete solidification. In the early stages of curing, the adhesive viscosity decreases with increasing temperature, but simultaneously increases with increasing degree of curing; these two factors compete with each other. This embodiment uses the Castro-Macosko viscosity model, whose expression is: In the formula, This refers to the viscosity of the encapsulating adhesive at a specific degree of curing and temperature. The initial viscosity is given and its variation with temperature follows the Arrhenius equation. For degree of curing, This refers to the degree of curing at the gel point. , The constants are used for fitting. Multi-temperature isothermal viscosity tests are conducted using a rotational rheometer, and the viscosity variation curves with curing time are recorded, allowing for the calibration of all model parameters. When the degree of curing approaches the gel point, the viscosity approaches infinity, marking the transition of the material from a liquid to a solid state. This model provides accurate rheological input for flow field simulation, ensuring that the simulation system can accurately capture the flow and filling behavior of the adhesive in the early stages of curing.

[0036] To more clearly illustrate the parameter calibration path of this application, the following explanation uses an o-cresol formaldehyde epoxy resin encapsulating adhesive as an example: Approximately 5 to 10 mg of encapsulating adhesive was taken and tested using a differential scanning calorimeter at four heating rates: 5℃ / min, 10℃ / min, 15℃ / min, and 20℃ / min, within a temperature range of 30℃ to 250℃, under a nitrogen atmosphere. Curing exothermic curves at different heating rates were obtained. The experimental data were nonlinearly fitted using the Kamal autocatalytic model to obtain the curing kinetic parameters, including the reaction rate constant. , ,activation energy , Reaction order , Gel point curing degree The complete curing temperature is approximately 175℃. Verification showed that the average deviation between the curing degree curve calculated by the model and the experimental value was 3.2%, meeting the accuracy requirements. Simultaneously, isothermal viscosity tests were conducted using a rotational rheometer at temperatures of 100℃, 120℃, 140℃, and 160℃, with a shear rate of 1Hz. The Castro-Macosko viscosity model parameters were obtained through fitting: initial viscosity pre-exponential factor. Activation energy of viscous flow ,constant , The model's predicted viscosity values ​​deviated by an average of 4.8% from experimental values, accurately describing the viscosity evolution during the curing process. The coefficient of thermal expansion of the fully cured encapsulant was measured using a thermomechanical analyzer to obtain the glass transition temperature. lower than The coefficient of thermal expansion is 42 ppm / ℃, which is higher than that of the standard. The coefficient of thermal expansion was 168 ppm / ℃. A dynamic mechanical analyzer was used to test the storage modulus at different degrees of cure, and the evolution curve of the modulus with degree of cure was obtained by fitting. The thermal conductivity and specific heat capacity at different degrees of cure were tested using the laser scintillation method. All parameters were compiled into tabular interpolation functions and imported into the simulation model for continuous operation.

[0037] It should be understood that although the above embodiments use the Kamal autocatalytic model and the Castro-Macosko viscosity model as examples, in other embodiments, depending on the different reaction mechanisms of the encapsulating adhesive formulation, the autocatalytic curing kinetic model can also adopt other forms such as a phenomenological model or a mechanistic model, and the viscosity evolution model can also adopt other forms such as the double Arrhenius model, as long as it can describe the relationship between reaction rate and degree of cure, temperature, and the dynamic changes of viscosity with degree of cure and temperature. The specific forms and parameter calibration methods of the above models are for illustrative purposes only and are not limiting.

[0038] In some embodiments, the closed-loop mechanism and numerical solution strategy of the thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation are further elaborated in detail. The thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation achieves closed-loop coupling in the following way: the temperature field drives the change of the curing degree field, the change of the curing degree field updates the material performance parameters, the change of the material performance parameters affects the flow field and stress field, thereby generating warping deformation, and the warping deformation is fed back to the temperature field to achieve temperature field redistribution.

[0039] Specifically, such as Figure 2As shown, traditional simulation methods typically employ a unidirectional sequential coupling, where the temperature field is calculated first, and then the temperature field result is used as a load input to calculate the stress field. This unidirectional transmission ignores the real-time interaction between physical fields. In the closed-loop coupling mechanism constructed in this embodiment, the temperature field is the initial driving force for the curing reaction. The increase in temperature causes the encapsulating adhesive to undergo a cross-linking reaction, which in turn produces a change in the curing degree field. The change in the curing degree field directly triggers the updating of material performance parameters. For example, viscosity increases sharply with the degree of curing, and the elastic modulus undergoes a sudden change after the gel point. These dynamically updated material performance parameters simultaneously act on the flow field and the stress field: before the gel point, the low viscosity allows the adhesive to flow and redistribute, alleviating some internal stress; after the gel point, the material becomes solid, and the increase in modulus leads to significant internal thermal stress and curing shrinkage stress, resulting in macroscopic warping deformation. The generation of warping deformation is not the end point of the simulation calculation, but the starting point of the closed-loop feedback. Warpage deformation alters the geometry of the wafer or substrate, causing uneven thickness distribution in the encapsulating adhesive layer. This, in turn, changes the interlayer thermal contact and heat conduction paths, ultimately feeding back into the temperature field and leading to its redistribution. This mechanism of feeding geometric deformation back into the physical field calculation accurately replicates the collaborative evolution of multiple physical fields during curing, avoiding the systematic errors caused by unidirectional transfer models that neglect deformation feedback, thus ensuring high-precision predictions.

[0040] Furthermore, the thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation uses the arbitrary Lagrange-Euler method to handle the flow deformation and free surface problems of the semiconductor encapsulating adhesive; the time step of the transient simulation is adaptively adjusted according to the curing reaction rate, and the time step is automatically densified at the gel point stage.

[0041] Specifically, in the early stages of curing, the encapsulating adhesive is in a liquid state, flowing and moving along with the free surface. If a pure Lagrangian method is used, the mesh will move with the fluid, easily leading to severe mesh distortion near the free surface and causing computational divergence. If a pure Eulerian method is used, the mesh remains stationary, avoiding distortion but making it difficult to accurately track the dynamic boundaries of the flowing free surface. This embodiment employs an arbitrary Lagrangian-Eulerian method. The mesh in this method neither strictly follows material movement like in the Lagrangian method nor is completely fixed like in the Eulerian method. Instead, it allows the mesh to adaptively adjust independently of material deformation. In areas of intense flow, the mesh is locally reconstructed to capture the free surface boundaries, while remaining relatively fixed in areas of less deformation. This effectively avoids mesh distortion and ensures the numerical stability of the coupled calculations of the flow and stress fields.

[0042] Regarding time step control in transient solutions, the curing reaction rate exhibits highly nonlinear characteristics throughout the curing cycle. In the initial and final stages of the reaction, the curing rate is slow, while near the gel point, due to the autocatalytic effect and the appearance of exothermic peaks, the curing reaction rate increases dramatically, causing abrupt changes in material parameters within a very short time. If a fixed time step is used, an excessively large step size will cause the crucial parameter abrupt changes to be missed at the gel point stage, leading to non-convergence; an excessively small step size will significantly increase the overall computation time. This embodiment adaptively adjusts the time step based on the curing reaction rate, using a larger step size in the lower reaction rate stages to improve computational efficiency, and automatically refining the time step size at the gel point stage to precisely capture the abrupt changes in material properties. This adaptive convergence strategy balances the accuracy and efficiency of the simulation calculation, ensuring the reliable implementation of the fully coupled mechanism at the numerical solution level. It should be understood that, in addition to the gel point stage, the adaptive adjustment strategy can also be applied to the vicinity of the glass transition temperature or other stages where material properties change drastically. As long as the logic of dynamically adjusting the step size based on the reaction rate to ensure convergence is met, it is acceptable.

[0043] Furthermore, the thermal-mechanical-fluid multi-field bidirectional fully coupled transient simulation, targeting wafer-level packaging structures, constructs a complete 3D geometric model including silicon wafers, chips, substrates, passivation layers, and encapsulating adhesive layers. It supports both 12-inch wafer-level full-array modeling and single-cell periodic modeling modes. Based on the differences in material properties and deformation gradients of each layer, a differentiated meshing strategy is adopted: the encapsulating adhesive layer and interface layer have finer meshes to ensure the accuracy of stress and flow calculations; homogeneous structures such as wafers and substrates use coarser meshes to balance computational efficiency. Boundary conditions are aligned with actual mass production processes, with furnace temperature profiles set as temperature boundaries, constraints set on the back side of the wafer, and bonding and thermal contacts established between layers to fully reproduce the real operating conditions of mass production curing ovens.

[0044] In some embodiments, the specific implementation path of reverse optimization of curing process parameters is further elaborated in detail. Reverse optimization of curing process parameters includes: rapid prediction of warpage using a surrogate model; global optimization of the prediction results based on a multi-objective optimization algorithm, and outputting the curing process parameters.

[0045] Specifically, in traditional R&D models, simulation technology often remains at the forward verification stage of calculating warpage for a given process, i.e., inputting a set of process parameters and calculating the corresponding warpage amount through high-precision finite element simulation. However, finding the optimal process that minimizes warpage requires traversing a large number of parameter combinations, and each fully coupled transient simulation takes hours or even days, failing to meet the timeliness requirements of global optimization. This embodiment constructs a surrogate model to compress the warpage prediction time from hours to milliseconds while ensuring prediction accuracy. First, the optimal Latin hypercube sampling method is used to generate sample points in the process parameter space. Compared to traditional random sampling, optimal Latin hypercube sampling can distribute sample points more evenly throughout the parameter space, avoiding local overfitting of the surrogate model caused by local clustering. Subsequently, high-precision fully coupled simulations are run on these sample points to obtain the corresponding warpage response values, which are used as training datasets. The surrogate model is constructed using the response surface methodology or radial basis function neural network to establish a nonlinear mapping relationship between process parameters and warpage amount. Response surface methodology (RSM) fits the mathematical relationship between parameters and response using polynomials, making it suitable for scenarios with a moderate number of parameters and moderate nonlinearity. Radial basis function (RBF) neural networks, on the other hand, generate local responses to the input through the radial basis functions of hidden layer nodes, exhibiting strong nonlinear approximation capabilities and making them suitable for complex scenarios with strong coupling of multiple parameters. Through surrogate models, multi-objective optimization algorithms can evaluate tens of thousands of process combinations in a very short time, providing a foundation for large-scale global optimization.

[0046] Furthermore, the multi-objective optimization algorithm aims to minimize warpage and shorten curing time, using process parameters and formulation parameters as optimization variables. The output curing process parameters include curing process curves containing heating rate, holding temperature and time, and cooling rate, as well as the allowable fluctuation range of each parameter.

[0047] Specifically, in curing process optimization, minimizing warpage and minimizing curing time are often conflicting objectives. Shortening curing time usually means increasing the heating rate or holding temperature, but this exacerbates the exothermic peak of the curing reaction, leading to a larger temperature gradient, which in turn increases thermal stress and curing shrinkage stress, resulting in greater warpage. This embodiment employs a non-dominated sorting genetic algorithm (NSGA-II) with an elitist strategy for multi-objective optimization. The algorithm first randomly generates an initial population, with each individual representing a set of process and formulation parameters; it then quickly predicts the warpage and curing time for each individual using a surrogate model; subsequently, it performs non-dominated sorting, retains the Pareto optimal solution set, and generates offspring populations through crossover and mutation operations; after multiple generations of iteration, a set of Pareto optimal solutions is obtained, which constitute the optimal trade-off front between warpage and curing time. Combining the feasibility constraints of mass production processes, such as the maximum heating rate limit of the equipment and the thermal decomposition temperature limit of the material, the optimal comprehensive process scheme is selected from the Pareto optimal solution set. The final output curing process parameters not only include complete curing process curves, such as the heating rate, holding temperature and time, and cooling rate at each stage, but also provide the allowable fluctuation range for each parameter through process robustness analysis. It should be understood that, in addition to minimizing warpage and shortening curing time, objectives such as maximizing process robustness can be introduced based on actual needs, as long as the logic of multi-objective optimization outputs the process window is met. Through this reverse optimization mechanism, simulation has been upgraded from a verification tool to a development engine, directly outputting feasible mass-production process solutions and significantly shortening the development cycle.

[0048] In some implementations, the model validation and parameter correction mechanisms are further elaborated in detail. A three-level progressive validation process—material-level, structural-level, and mass production-level—is used to validate the constitutive model and the predicted warpage deformation, and the model parameters are corrected based on the validation deviations.

[0049] Specifically, because the curing process of semiconductor encapsulation adhesives involves complex chemical and physical changes, relying solely on a single level of verification is insufficient to comprehensively assess the accuracy of the model. This embodiment constructs a progressive verification system from micromaterials to macrostructures. Before conducting verification, a complete experimental test matrix covering curing kinetics, rheological properties, thermophysical properties, mechanical properties, and interfacial properties is established. The sample preparation specifications, equipment parameters, test conditions, and data processing methods for each test are clearly defined. All tests comply with the corresponding ASTM or JEDEC industry standards to ensure the accuracy and repeatability of the test data.

[0050] The first stage is material-level verification, which mainly compares the curing degree curves and viscosity curves calculated by simulation with experimental test results. Experimental data under different temperature conditions are obtained using differential scanning calorimetry and rotational rheology, and then compared with the theoretical curves calculated by the constitutive model. Material-level verification requires an error of less than or equal to 5% to ensure the accuracy of the underlying reaction kinetics and rheological models. If the error exceeds the limit at this stage, the goodness of fit of parameters such as the reaction rate constant or activation energy needs to be re-checked.

[0051] The second level is structural verification, which involves preparing small-sized standard packaged samples, such as chip packages measuring 10mm × 10mm. The amount and morphology of warpage after curing are measured using a laser warpage tester and compared with simulation results. Structural verification requires an error of less than or equal to 10%. This level of verification primarily examines the effectiveness of the multi-field coupling mechanism under realistic geometric constraints, particularly the superposition effect of post-gelation curing shrinkage stress and thermal stress.

[0052] The third level is production-level verification, which involves process verification on a 12-inch wafer production line, comparing the measured and simulated values ​​of the warpage distribution across the entire wafer. Production-level verification requires a maximum warpage error of less than or equal to 1.5 μm. Because large-size wafers are extremely sensitive to temperature gradients and differences in in-plane curing density, this level of verification can expose deviations in the model under extreme boundary conditions.

[0053] Furthermore, when deviations occur, the least squares method and intelligent optimization algorithms are used for parameter inversion to automatically correct parameters that are difficult to measure directly, such as interface strength and equivalent thermal conductivity. During the curing process, the interfacial bonding state between the encapsulant and the wafer / substrate, as well as the anisotropy of thermal conductivity caused by filler sedimentation within the encapsulant layer, are implicit parameters that are difficult to measure directly through conventional experiments. When systematic deviations occur during mass production verification, this embodiment uses the root mean square error between the simulated predicted value and the measured warpage distribution as the objective function, employs the least squares method for preliminary gradient optimization, and combines intelligent optimization algorithms such as genetic algorithms to perform inversion search in the global parameter space. By continuously adjusting the interface strength and equivalent thermal conductivity, the simulated predicted warpage morphology is made to approximate the measured results. The corrected parameters are incorporated into an enterprise-level material parameter library, supporting rapid retrieval of similar formulations in the future, thereby ensuring high-precision matching and reusability of the model under different formulations and processes.

[0054] It should be understood that, in addition to interface strength and equivalent thermal conductivity, the objects of inversion correction can also include interface friction coefficient or residual stress release coefficient, as long as the logic of automatically correcting parameters that are difficult to test directly based on verification deviation is satisfied. The error thresholds and verification levels mentioned above are for interpretation purposes only, not for limitation.

[0055] In some embodiments, the quantitative analysis and sensitivity ranking of warpage influencing factors are further elaborated, providing theoretical guidance for subsequent process optimization through quantitative analysis of the warpage formation mechanism. Specifically, the single-factor influence law analysis system studies the influence of key factors such as adhesive layer thickness, filler content, thermal expansion coefficient matching, curing heating rate, medium-temperature holding time, peak curing temperature, and cooling rate on the amount of warpage. For each factor, while keeping other factors constant, multiple level values ​​are set within a certain range, and a multi-field bidirectional fully coupled transient simulation of thermo-mechanical-fluidic fields is run to extract the maximum warpage after curing, and the correlation curves between each factor and the amount of warpage are plotted to clarify the influence trend and optimal range of each parameter. For example, when the adhesive layer thickness increases, the total curing shrinkage increases, but the stress buffering capacity also increases, and the amount of warpage shows a non-monotonic change characteristic of first decreasing and then increasing with the adhesive layer thickness; when the heating rate increases, the temperature gradient increases, leading to an increase in thermal stress, but the accelerated curing reaction shortens the gelation time, and the influence on the amount of warpage shows a competitive relationship. Multi-factor sensitivity ranking employs orthogonal experimental design combined with range analysis and variance analysis to calculate the contribution of each factor to warpage. First, based on the results of univariate analysis, factors with significant influence are selected, and an orthogonal experimental table is constructed. The warpage response values ​​for each experimental group are calculated through simulation. Then, range analysis is used to calculate the range of each factor; a larger range indicates a more significant impact of the factor on warpage. Simultaneously, variance analysis is used to calculate the F-test value and significance level of each factor, quantifying the contribution of each factor. Sensitivity ranking identifies core and secondary sensitive factors, clarifying the priority of optimization work and avoiding the waste of debugging resources on secondary factors. Warpage stress source decomposition involves post-simulation processing, decomposing the total stress into three components: thermal stress, curing shrinkage stress, and structural mismatch stress, calculating the proportion of each component in the total stress. Thermal stress originates from the difference in thermal expansion coefficients between the encapsulating adhesive and the wafer / substrate, and is dominant during the heating and cooling stages. Curing shrinkage stress originates from volume shrinkage caused by chemical reactions and is dominant during the isothermal curing stage after the gel point. Structural mismatch stress originates from the differences in the mechanical properties of materials in each layer of a multilayer structure and is particularly significant during the cooling stage after complete curing. By identifying the core driving forces of warpage under different encapsulation structures, targeted optimization solutions are developed: when thermal stress is dominant, the focus is on optimizing the matching of thermal expansion coefficients, such as adjusting the filler content or selecting a substrate material with a low thermal expansion coefficient; when curing shrinkage stress is dominant, the focus is on optimizing the curing process curve, such as adjusting the mid-temperature holding time or reducing the peak curing temperature; when structural mismatch stress is dominant, the focus is on optimizing the interlayer interface design or adjusting the thickness ratio of each layer.

[0056] To more clearly illustrate the technical solution and its effects, the following section uses an 8-inch wafer-level chip packaging structure as an application scenario, and combines specific encapsulation adhesive materials, testing equipment, and simulation software to provide a detailed explanation of the semiconductor encapsulation adhesive curing warpage control method of this application. It should be understood that this application scenario is illustrative only and not restrictive.

[0057] Specifically, the encapsulating adhesive raw materials used in this application scenario include o-crestyrene epoxy resin, phenolic resin curing agent, spherical fused silica microparticles, imidazole accelerators, and silane coupling agents. Testing equipment includes a DSC25 differential scanning calorimeter, an MCR 302 rotational rheometer, a TMA Q400 thermomechanical analyzer, a DMA 850 dynamic mechanical analyzer, and an FLX-2320-S laser warpage tester. The simulation software used is Abaqus 2025 finite element analysis software. The encapsulated samples are 8-inch silicon wafer test pieces.

[0058] In the model parameter calibration experiment, approximately 5 to 10 mg of encapsulating adhesive sample was taken and tested using a differential scanning calorimeter at four heating rates: 5℃ / min, 10℃ / min, 15℃ / min, and 20℃ / min, with a temperature range of 30℃ to 250℃ under nitrogen atmosphere protection. Curing exothermic curves at different heating rates were obtained. The experimental data were nonlinearly fitted using the Kamal autocatalytic model to obtain the curing kinetic parameters: reaction rate constant. , ,activation energy , Reaction order , Gel point curing degree The complete curing temperature is approximately 175℃.

[0059] like Figure 3 The figure shows a comparison of the curing degree evolution curves under different heating rates, illustrating the influence of heating rate on the curing reaction process. The average deviation between the curing degree curve calculated by the model and the experimental value is 3.2%, which meets the accuracy requirements. Simultaneously, isothermal viscosity tests were conducted using a rotational rheometer at temperatures of 100℃, 120℃, 140℃, and 160℃, with a shear rate of 1Hz. The Castro-Macosko viscosity model parameters were obtained through fitting: initial viscosity pre-exponential factor. Activation energy of viscous flow ,constant , The model's predicted viscosity values ​​deviated by an average of 4.8% from experimental values, accurately describing the viscosity evolution during the curing process. The coefficient of thermal expansion of the fully cured encapsulant was measured using a thermomechanical analyzer to obtain the glass transition temperature. lower than The coefficient of thermal expansion is 42 ppm / ℃, which is higher than that of the standard. The coefficient of thermal expansion was 168 ppm / ℃. A dynamic mechanical analyzer was used to test the storage modulus at different degrees of cure, and the evolution curve of the modulus with degree of cure was obtained by fitting. The thermal conductivity and specific heat capacity at different degrees of cure were tested using the laser scintillation method. All parameters were compiled into tabular interpolation functions and imported into the simulation model for continuous operation.

[0060] In the simulation model construction, an 8-inch wafer-level chip packaging structure is used as the implementation object to construct the simulation model: wafer thickness. Chip size ,thickness Encapsulation adhesive layer thickness substrate thickness It includes a passivation layer; single-chip cell modeling is selected using periodic boundary conditions; a differentiated meshing strategy is adopted based on the differences in material properties and deformation gradients of each layer; the mesh of the encapsulating adhesive layer and interface layer is refined, while the wafer and substrate use coarser meshes; the total number of meshes is approximately 280,000; and the minimum mesh size of the encapsulating adhesive layer is specified. Initial curing conditions: room temperature ,by Heat up to Insulation , and then Heat up to Insulation Finally Cool to room temperature. Use a user subroutine (UMAT / UDS) in Abaqus software to embed the curing kinetics and viscosity evolution model for a two-way fully coupled transient solution of thermo-mechanical-fluid multi-field.

[0061] To verify the technical effectiveness of this application, Example 1 (forward verification of the fully coupled model of this application), Example 2 (optimized process scheme of this application), Comparative Example 1 (traditional thermal-structural coupling model), and Comparative Example 2 (traditional empirical trial-and-error process) were set up for comparative analysis. In terms of performance testing, warpage was tested using a laser warpage tester to measure the warpage morphology of the 8-inch wafer surface after curing and cooling to room temperature. The height difference between the center point and the edge point of the wafer was recorded as the maximum warpage. Three wafers were tested for each process, and the average value was taken. The degree of curing was tested using a differential scanning calorimeter to measure the residual curing exothermic enthalpy of the cured sample, and the final degree of curing was calculated by comparing it with the total exothermic enthalpy of the fully cured sample. Packaging yield statistics were calculated by counting the number of good products after back-side grinding and dicing processes on 100 cured wafers and calculating the yield percentage. The R&D cycle statistics were the total time from project initiation to process finalization.

[0062] Example 1 uses the thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation model described in this application to calculate the initial curing process and obtain the simulated value of the warpage after wafer curing; simultaneously, an 8-inch wafer package sample is prepared using the same process, and the warpage is measured using a laser warpage tester, recording the maximum warpage value and error. Figure 4 As shown, this is a simulation cloud map of the warpage distribution after curing an 8-inch wafer, presenting the spatial distribution characteristics of warpage deformation within the wafer surface. It allows for intuitive identification of the location of maximum warpage and the overall deformation trend. Figure 5 The bar chart shown is a comparison of warpage and error between the examples and comparative examples, comparing the measured warpage values, simulated values, and predicted errors for each group. The simulated warpage value for Example 1 was 6.82 μm, the measured value was 7.15 μm, the simulation error was 0.33 μm, the curing degree was 96.8%, and the encapsulation yield was 82.3%.

[0063] Example 2 employs the multi-objective optimization method described in this application, with minimizing warpage as the core objective, to optimize the curing process parameters and obtain the optimal process curve; the optimized warpage is verified through simulation and actual measurement, and the results are recorded. Figure 6 The figure shows a comparison of the curing process temperature curves before and after optimization, illustrating the parameter differences in each stage of heating, holding, and cooling, and intuitively presenting the adjustment direction of the process optimization. The simulated value for Example 2 was 1.74 μm, the measured value was 1.89 μm, the simulation error was 0.15 μm, the degree of curing was 97.2%, the encapsulation yield was 96.7%, and the development cycle was 4.5 months.

[0064] Comparative Example 1 employs the industry-standard pure thermo-structural coupling simulation method, neglecting adhesive flow and curing evolution, and using fixed material parameters after complete curing for calculation. The simulation values ​​are compared with the measured values ​​to identify the errors. The simulated value for Comparative Example 1 is 3.26 μm, while the measured value is 7.15 μm, resulting in a simulation error of 3.89 μm.

[0065] Comparative Example 2 uses a mass production process derived from trial and error based on the company's traditional experience, recording the measured warpage, development cycle, and packaging yield. The measured maximum warpage of Comparative Example 2 is 7.03 μm, the curing degree is 95.4%, the packaging yield is 81.5%, and the development cycle is 11 months.

[0066] Please refer to the test results below for details:

[0067] The comparative data above shows that, in terms of prediction accuracy, Example 1, using the fully coupled model of this application, has a warpage error of only 0.33 μm between simulation and actual measurements, far below the accuracy target of ±1.5 μm. In contrast, the traditional thermal-structural model in Comparative Example 1 has an error as high as 3.89 μm, only able to qualitatively judge trends and unable to meet the requirements of high-precision control. This fully demonstrates that this application, by introducing the coupling of curing dynamics and flow field, significantly improves the accuracy of warpage prediction and solves the core pain point of insufficient accuracy in traditional simulation. Regarding optimization effects, after optimization using the method of this application, Example 2 shows that the maximum warpage is reduced to 1.89 μm, stably controlled within the high-end packaging requirement of ±2 μm; the packaging yield increases from 81.5% to 96.7%, an increase of over 15 percentage points, demonstrating significant cost reduction and efficiency improvement. The optimized process curve, by adjusting the duration and heating rate of the mid-temperature holding stage, balances the curing shrinkage stress and thermal stress, achieving precise control of the warpage amount. Regarding R&D efficiency, Example 2 using the method of this application requires only 4.5 months from project initiation to process finalization, while Comparative Example 2 using the traditional trial-and-error model requires 11 months, shortening the R&D cycle by over 59%. Simultaneously, it significantly reduces wafer-level experimental batches, directly lowering R&D material and labor costs. Regarding curing uniformity, the simulation model of this application can simultaneously control curing uniformity. The in-wafer curing deviation in Example 2 is less than or equal to 1.2%, far superior to the 4.7% of the traditional process, ensuring the consistency and stability of product performance.

[0068] This embodiment provides a semiconductor encapsulation adhesive curing warpage control device. For example... Figure 7 As shown, the device includes a model building module 701, a fully coupled simulation module 702, and a reverse optimization module 703.

[0069] Model building module 701 is used to obtain a constitutive model of the dynamic evolution of material properties with degree of curing.

[0070] Specifically, the model building module 701, serving as the data input terminal of the device, has a pre-built standardized parameter calibration interface capable of receiving test data from differential scanning calorimeters, rotational rheometers, thermomechanical analyzers, and dynamic mechanical analyzers. This module uses a built-in fitting algorithm to transform the experimental test data into a self-catalytic curing kinetic model, a viscosity evolution model, and a mapping relationship between thermophysical and mechanical property parameters and the degree of curing. These sub-models are then assembled into a continuously invoked interpolation function. The model building module 701 transmits the generated constitutive model data package to the fully coupled simulation module 702 via an internal bus, providing dynamic material input for subsequent multiphysics calculations.

[0071] The fully coupled simulation module 702 is used to perform bidirectional fully coupled transient simulation of the curing process of semiconductor encapsulation adhesive based on a constitutive model, and to predict warpage deformation.

[0072] Specifically, the fully coupled simulation module 702 is the core computing engine of the device, internally equipped with a finite element solver. This module receives the constitutive model from the model building module 701, along with externally input initial curing process parameters and geometric boundary conditions, to construct a complete three-dimensional geometric model including the silicon wafer, chip, substrate, and encapsulating adhesive layer. During the solution process, the fully coupled simulation module 702 simultaneously calculates the temperature field, curing degree field, flow field, and stress field through closed-loop feedback logic. This module employs an arbitrary Lagrange-Euler method to handle flow deformation and free surface problems, and adaptively adjusts the time step according to the curing reaction rate. The fully coupled simulation module 702 ultimately outputs the predicted warpage deformation of the wafer or substrate at any given time and transmits this prediction to the reverse optimization module 703.

[0073] The reverse optimization module 703 is used to reverse optimize the curing process parameters based on the predicted warpage deformation.

[0074] Specifically, the reverse optimization module 703 receives the high-precision warpage prediction result output by the fully coupled simulation module 702 and uses it as feedback input. This module integrates a surrogate model building unit and a multi-objective optimization unit. The surrogate model building unit generates sample points through optimal Latin hypercube sampling and establishes a nonlinear mapping relationship between process parameters and warpage using response surface methodology or radial basis function neural networks, achieving millisecond-level rapid prediction of warpage. The multi-objective optimization unit aims to minimize warpage and curing time, employing a non-dominated sorting genetic algorithm with an elitist strategy for global optimization. The final output includes a curing process curve containing heating rate, holding temperature and time, cooling rate, and the allowable fluctuation range of each parameter. The reverse optimization module 703 outputs the optimized process scheme to an external display device or mass production control system, upgrading simulation from a verification tool to a research and development engine.

[0075] Furthermore, the device also includes a parameter inversion correction module for automatically correcting the interface strength and equivalent thermal conductivity parameters based on verification deviations.

[0076] Specifically, the parameter inversion correction module communicates with the fully coupled simulation module 702 and the external experimental verification system. During the three-tiered progressive verification process (material-level, structural-level, and mass production-level), if a systematic deviation occurs between the simulation prediction and the measured warpage distribution, the parameter inversion correction module will be triggered. This module uses the root mean square error between the simulation prediction and the measured warpage distribution as the objective function, employs the least squares method for initial gradient optimization, and combines an intelligent optimization algorithm to perform inversion search within the global parameter space. Since the anisotropy of thermal conductivity caused by the interface bonding state and the sedimentation of fillers within the adhesive layer is difficult to measure directly through conventional experiments, the parameter inversion correction module continuously adjusts the interface strength and equivalent thermal conductivity coefficient to make the simulated warpage morphology approximate the measured results. The corrected parameters are automatically transmitted back to the fully coupled simulation module 702 to update the model and simultaneously stored in an enterprise-level material parameter library, supporting rapid retrieval of similar formulations in the future. It should be understood that, in addition to interface strength and equivalent thermal conductivity, the parameter inversion correction module can also be extended to interface friction coefficient or residual stress release coefficient, as long as it satisfies the logic of automatically correcting parameters that are difficult to test directly based on verification deviation. The above division of functional modules is for illustrative purposes only. In actual hardware deployment, these modules can be integrated within the same computing server or distributed across different computing nodes in a local area network, as long as the requirements for data interaction and functional implementation are met.

[0077] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for controlling warpage during the curing of semiconductor encapsulating adhesive, characterized in that, include: Obtain a constitutive model of the dynamic evolution of semiconductor encapsulation adhesive material properties with degree of curing; Based on the constitutive model, a thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation of the curing process of the semiconductor encapsulant is performed to predict warpage deformation. Based on the predicted warpage deformation, the curing process parameters are optimized in reverse.

2. The method according to claim 1, characterized in that, The constitutive model includes: The mapping relationship between thermophysical and mechanical property parameters and degree of curing; The coupled strain relationship between curing shrinkage and thermal expansion.

3. The method according to claim 2, characterized in that, The constitutive model also includes: An autocatalytic curing kinetic model is used to describe the relationship between the curing reaction rate and the degree of curing and temperature. Viscosity evolution model, used to describe the dynamic changes in viscosity with degree of curing and temperature.

4. The method according to claim 1, characterized in that, The thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation achieves closed-loop coupling through the following method: The temperature field drives the change of the curing degree field, the change of the curing degree field updates the material performance parameters, the change of the material performance parameters affects the flow field and stress field, and thus generates warping deformation. The warping deformation is fed back to the temperature field to achieve temperature field redistribution.

5. The method according to claim 4, characterized in that, The thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation uses the arbitrary Lagrange-Euler method to handle the flow deformation and free surface problems of the semiconductor encapsulant; The time step of the transient simulation is adaptively adjusted according to the curing reaction rate, and the time step is automatically encrypted at the gel point stage.

6. The method according to claim 1, characterized in that, The reverse optimization curing process parameters include: Rapid prediction of warpage using a surrogate model; The prediction results are globally optimized based on a multi-objective optimization algorithm, and the solidification process parameters are output.

7. The method according to claim 6, characterized in that, The multi-objective optimization algorithm aims to minimize warpage and curing time, with process parameters and formulation parameters as optimization variables. The output curing process parameters include a curing process curve containing the heating rate, holding temperature and time, and cooling rate, as well as the allowable fluctuation range of each parameter.

8. The method according to claim 1, characterized in that, The method further includes: A three-level progressive verification process (material level, structural level, and mass production level) is adopted to verify the constitutive model and the predicted results of the warping deformation, and the model parameters are corrected based on the verification deviation.

9. A semiconductor encapsulation adhesive curing warpage control device, characterized in that, include: The model building module is used to obtain constitutive models of material properties that evolve dynamically with the degree of curing. The fully coupled simulation module is used to perform a thermo-mechanical-fluid multi-field bidirectional fully coupled transient simulation of the curing process of the semiconductor encapsulant based on the constitutive model, and to predict warpage deformation. The reverse optimization module is used to reverse optimize the curing process parameters based on the predicted warping deformation.

10. The apparatus according to claim 9, characterized in that, The device further includes: The parameter inversion correction module is used to automatically correct the interface strength and equivalent thermal conductivity parameters based on the verification deviation.

Images