Nano twin crystal copper rough surface direct bonding atomic scale simulation method

Abstract

The invention discloses a nano twin copper rough surface direct bonding atomic-scale simulation method, which comprises the following steps: determining an inter-atomic interaction potential function in molecular dynamics through a first principle, constructing a polycrystalline nano twin copper supercell model based on experimental parameters, and generating a rough surface with fractal characteristics; molecular dynamics simulation is adopted to simulate bonding and annealing processes, a bonded polycrystalline nano twin copper model is obtained, and the performance of the interconnection structure is calculated based on the bonded polycrystalline nano twin copper model. The result feedback is used for optimizing the copper-copper direct bonding material and process parameters. According to the method, the atomic-scale polycrystal nano twin copper model is established through simulation, the optimal performance parameter combination of the interconnection structure is obtained, the trial and error cost in tests and production is reduced, the quality of a copper-copper direct bonding interface is improved, and guidance is provided for design and process optimization of high-density packaging.

Description

Technical Field

[0001] The invention relates to the technical field of semiconductor materials and manufacturing, and in particular to an atomic-level simulation method for direct bonding of rough surfaces of nano-twin copper. Background Art

[0002] With the development of semiconductor technology, device size continues to shrink, and traditional two-dimensional integrated circuits (2D ICs) are gradually approaching physical limits in terms of improving integration and performance. In order to solve this problem, three-dimensional integrated circuit (3D IC) technology came into being. By vertically stacking multiple chips, the device integration and signal transmission efficiency are greatly improved. 3D IC technology relies on high-density interconnection between chips, and vertical interconnection is one of the key technologies. Generally, the interconnection technology of chips is mainly implemented by copper bumps and through silicon vias (TSV), which can meet the interconnection requirements under a certain size. However, as the size of devices is further reduced, the reduction in interconnection spacing has put forward more stringent requirements on these traditional interconnection technologies. In micron-level or even nano-level interconnection structures, the performance of copper interconnection materials has become a key factor in determining packaging reliability and electrical performance.

[0003] As device sizes gradually enter the sub-micron or even nanometer level, traditional interconnection technology faces more and more challenges in terms of process and reliability. Micro-bump technology has problems with uneven bonding interfaces and potential electrical short circuits due to the fact that its solder is easily squeezed out during the hot pressing process, which is especially obvious at a pitch of less than 20μm. In addition, the solder itself has poor thermal stability, which can easily lead to a decrease in the reliability of the interconnection layer. Especially in high-density packaging, the use of welding materials will also increase manufacturing complexity. In order to cope with these problems, direct copper-copper bonding technology has gradually attracted attention and has become a key technology for improving electrical performance and packaging reliability in future high-density packaging.

[0004] Direct copper-copper bonding technology utilizes the high electrical conductivity and thermal conductivity of copper materials, and through surface treatment and precise hot-pressing control, achieves reliable interconnection between chips without solder. Compared with traditional welding technology, copper-copper bonding not only reduces the use of materials, but also can further reduce the interconnection spacing to meet packaging requirements of 10μm or even smaller. In recent years, this technology has been mainly implemented in two ways: thermocompression bonding (TCB) and hybrid bonding (HB). TCB technology promotes the diffusion and rearrangement of copper atoms by applying a certain amount of heat and pressure, thereby achieving stable metal bonding; while hybrid bonding combines the advantages of direct copper-copper contact and electrical insulation layer, making it have a broader prospect in applications that meet both mechanical strength and electrical performance requirements.

[0005] Nano-twinned copper exhibits excellent mechanical and electrical properties due to its special crystal structure, especially maintaining low resistivity while improving strength and hardness, making it an ideal interconnect material. However, in direct copper-copper bonding, surface roughness and twin structure have complex effects on the bonding interface. Rough surfaces may lead to incomplete contact between copper atoms and produce holes, which in turn affect the effectiveness and mechanical strength of electrical connections. At the same time, existing bonding process research has not yet been able to effectively deal with the problem of nanoscale surface roughness, and the lack of atomic-level simulation of its bonding process has led to uncontrollable interface performance, which in turn limits its application in high-density packaging. Summary of the invention

[0006] The purpose of the present invention is to provide an atomic-level simulation method for direct bonding of nano-twin copper rough surfaces, which can accurately simulate the influence of nano-twin copper surface roughness and twin structure on the copper-copper direct bonding interface performance, thereby optimizing the high-density interconnection process and improving the copper-copper direct bonding quality of electronic packaging.

[0007] The solution adopted by the present invention to achieve the purpose is: a method for direct bonding atomic level simulation of rough surface of nano twinned copper, which is used for molecular dynamics simulation, and the method comprises the following steps:

[0008] S1: Construct a bulk structure model of nano-twinned copper, perform first-principles calculations and molecular dynamics simulations on the nano-twinned copper structure, compare and analyze characteristic parameters, and determine the appropriate interatomic interaction potential function;

[0009] S2: Based on the polycrystalline and twin parameters obtained from the experiment, a supercell structure model of polycrystalline nano-twinned copper is constructed, and an atomic-level model of polycrystalline nano-twinned copper before bonding is obtained by constructing a rough surface with splitting characteristics;

[0010] S3: Based on the atomic-level model constructed in step S2 and the interatomic interaction potential function determined in step S1, a molecular dynamics method is used to perform bonding simulation, and key parameters in the model bonding process are monitored in real time. After achieving stable bonding, the model is annealed to obtain the polycrystalline nanotwin copper model data after bonding;

[0011] S4: Based on the polycrystalline nanotwin copper model data obtained in S3 and the interatomic interaction potential function determined in S1, the molecular dynamics method is used to perform molecular dynamics simulation of the interconnection performance of polycrystalline nanotwin copper pillars, and the high-reliability polycrystalline nanotwin copper direct bonding process conditions are determined at the atomic level.

[0012] Preferably, in step S1, the characteristic parameters include: lattice constant, density, thermal expansion coefficient, elastic constant, intrinsic stacking fault energy, Poisson's ratio, bulk elastic modulus, Young's modulus, shear model, radial distribution function, and Bain path.

[0013] Preferably, in step S1, suitable interatomic potential functions include: embedded atomic potential function, modified embedded interatomic potential function and machine learning interatomic potential function.

[0014] Preferably, in step S1, the first principle calculation used is based on density functional theory, the functionals include PBE functional and SCAN functional, and the molecular dynamics method first uses empirical potential functions for calculation, the potential functions include EAM potential functions and MEAM potential functions.

[0015] Preferably, in step S1, when the error between the molecular dynamics calculation parameter comparison result and the first-principles calculation is low, the empirical potential function is fine-tuned according to the characteristic parameters and selected as the interatomic interaction potential function for subsequent calculations; when the error is high, a machine learning potential function is constructed based on the first-principles calculation using an active learning method and selected as the interatomic interaction potential function for subsequent calculations.

[0016] Preferably, step S2 includes the following sub-steps:

[0017] S2a, combined with the experimental observation results of nano-twinned copper electroplating on pure copper surface, the twin size and twin density are obtained according to the cross section after focused ion beam treatment, and the parameters are used to model the nano-twinned copper unit cell;

[0018] S2b, based on the polycrystalline distribution, grain size and crystal orientation obtained by electron backscatter diffraction analysis technology, a polycrystalline nano-twin copper supercell model is established on the basis of S2a;

[0019] S2c. Delete a specified number of layers of atoms in the middle of the supercell model established in step S2b to construct a reserved gap before bonding. The entire model is divided into two upper and lower polycrystalline nano-twin copper pillars to be bonded. Based on the Weierstrass-Mandelbrot function, a specified roughness surface with fractal characteristics is constructed on the surfaces to be bonded of the two copper pillars, thereby obtaining a polycrystalline nano-twin copper atomic-level model with a rough surface.

[0020] Preferably, step S3 includes the following sub-steps:

[0021] S3a, dividing the atomic-level model constructed in step S2 into a top fixed layer, a top heating layer, a middle relaxation layer, a bottom heating layer and a bottom fixed layer in sequence;

[0022] S3b, using the interatomic interaction potential function determined in step S1 to perform molecular dynamics simulation, perform energy minimization processing as a whole, use a canonical system to apply different temperatures to the top heating layer and the bottom heating layer to define the temperature gradient during the bonding process, and use a microcanonical system as a whole to simulate the heat transfer and expansion of the upper and lower polycrystalline nano-twin copper columns during the entire bonding process, so as to promote contact between the upper and lower surfaces and self-diffusion to achieve bonding;

[0023] S3c. During the bonding process, the surface roughness, diffusion coefficient, atomic number density distribution, crystal structure, twin structure, pores, internal stress, temperature distribution and energy are monitored in real time. After achieving stable bonding, the model is annealed to obtain the polycrystalline nano-twin copper model data after bonding.

[0024] Preferably, in step S4, the interatomic interaction potential function selected in S1 is used to perform molecular dynamics simulation of interconnection performance on the bonded polycrystalline nanotwin copper model obtained in S3, and the performance simulation performed includes: interconnection structural performance, interconnection thermal performance, fracture mechanical performance and interconnection mechanical performance.

[0025] Preferably, step S4 includes the following sub-steps:

[0026] S4a, based on the stably bonded polycrystalline nano-twinned copper model obtained in step S3, using the molecular dynamics method to simulate its interconnection structure performance, monitor its crystal structure changes, stacking fault twinning pathways, dislocation evolution and grain boundary development;

[0027] S4b, according to the stably bonded polycrystalline nanotwinned copper model obtained in step S3, using molecular dynamics method to simulate its interconnection thermal performance, and monitor its thermal expansion coefficient, glass transition temperature, thermal conductivity and specific heat;

[0028] S4c Based on the stably bonded polycrystalline nanotwinned copper model obtained in S3, the molecular dynamics method is used to simulate its fracture mechanical properties and monitor its stress-strain response, stress evolution, fracture morphology and damage evolution;

[0029] S4d Based on the stably bonded polycrystalline nanotwinned copper model obtained in S3, the molecular dynamics method is used to simulate its interconnection mechanical properties, and its pore and porosity evolution, interface stress-strain response, and interface fracture morphology are monitored;

[0030] S4e. According to the performance monitoring results of steps S4a, S4b, S4c and S4d, the copper-copper direct bonding materials and process parameters are optimized to guide the structural modeling of step S2 and the process simulation of step S3, thereby forming a high-reliability polycrystalline nano-twin copper direct bonding process.

[0031] The present invention proposes a simulation method for the atomic-level rough surface of nano-twin copper, which can accurately model and simulate the influence of surface roughness on bonding performance during copper-copper direct bonding at multiple scales. By simulating the surface morphology of nano-twin copper with atomic-level precision and analyzing the mechanism of action of the twin structure in the copper-copper interface, a reliable theoretical basis is provided for the optimal design of copper interconnection. This method can effectively improve the interface interconnection quality of copper-copper direct bonding, reduce the formation of holes, and further improve the electrical performance and mechanical strength of the interconnection structure, and is suitable for advanced packaging technologies with higher interconnection density.

[0032] The present invention has the following advantages and beneficial effects:

[0033] The present invention constructs an atomic-level model of polycrystalline nanotwinned copper with a rough surface of fractal characteristics, and uses molecular dynamics simulation methods to accurately simulate the influence of key parameters of nanotwinned copper on the performance of copper-copper direct bonding interface. By evaluating the electrical properties, mechanical strength and atomic diffusion behavior of the interface, the high performance of the bonded structure is ensured, and the copper-copper direct bonding process is optimized. This simulation method provides a reliable theoretical basis for the design and optimization of copper interconnect materials in high-density packaging, and has broad engineering application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flow chart of an atomic-level simulation method for direct bonding of a nano-twin copper rough surface provided by an embodiment of the present invention;

[0035] Figure 2 An atomic-level rough surface modeling and visualization model for an atomic-level simulation method for direct bonding of a nano-twin copper rough surface provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] For a better understanding of the present invention, the following examples are provided to further illustrate the present invention, but the present invention is not limited to the following examples.

[0037] Figure 1 This is a flow chart of the present invention. The core of the present invention is to establish a molecular dynamics model of nano-twinned copper and calculate bonding performance, to seek the optimal performance combination, so as to guide the design of high-reliability direct bonding materials and structures. Figure 2 The atomic-level rough surface modeling and visualization model presents the visualization details of structures such as polycrystals, twins and rough surfaces. In particular, it also presents the atomic-level model that processes billions of atoms with the assistance of GPU / CPU acceleration.

[0038] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0039] A method for direct bonding of nano-twinned copper rough surfaces at the atomic level, comprising the following steps:

[0040] S1: Construct a bulk structure model of nanotwins, use first-principles calculations and molecular dynamics simulations to simulate their characteristic parameters, and determine the appropriate interatomic interaction potential function after comparing the errors;

[0041] S2: Based on the experimentally obtained data of polycrystalline and twin structures and bonding model parameters such as copper pillar size and surface roughness, an atomic-level model of polycrystalline nano-twinned copper with a rough surface with parting characteristics before bonding is constructed;

[0042] S3: Based on the pre-bonding model constructed in S2 and the interatomic interaction potential function determined in S1, the molecular dynamics method is used to perform bonding simulation, and the key parameters in the process are monitored in real time. After determining that the bonding is stable, an annealing simulation is performed to obtain a polycrystalline nanotwin copper model after bonding;

[0043] S4: Based on the post-direct bonding model determined in S3, molecular dynamics simulation of the interconnection performance of polycrystalline nanotwin copper pillars is performed using molecular dynamics methods to guide the optimization direction of copper-copper direct bonding materials and process parameters from the atomic level, and determine the high-reliability polycrystalline nanotwin copper direct bonding process conditions.

[0044] In the above embodiment, the characteristic parameters in step S1 include: lattice constant, density, thermal expansion coefficient, elastic constant, intrinsic stacking fault energy, Poisson's ratio, bulk elastic modulus, Young's modulus, shear model, radial distribution function, and Bain path under various temperature and pressure conditions.

[0045] Furthermore, the molecular dynamics method in step S1 first uses empirical potential functions for calculation, and the potential functions include EAM potential functions and MEAM potential functions; the first-principles calculation used is based on density functional theory, and the functionals include PBE functionals and SCAN functionals.

[0046] In a further preferred embodiment, if the error between the comparison result of the molecular dynamics calculation parameters in step S1 and the first principles calculation is low, the empirical potential function is fine-tuned according to the characteristic parameters and selected as the interatomic interaction potential function for subsequent calculations; if the error is high, a machine learning potential function (such as the DeePMD potential function) is constructed based on the first principles calculation using an active learning method and selected as the interatomic interaction potential function for subsequent calculations.

[0047] In step S2, the twin size and twin density are obtained according to the FIB experimental parameters to construct the nano-twin copper unit cell, the polycrystalline distribution, grain size and crystal orientation are obtained according to the EBSD experimental parameters, and the polycrystalline nano-twin copper supercell is constructed based on the nano-twin copper unit cell. The copper column size, twin column height and bonding spacing are set to determine the bonding copper column model, and the rough upper and lower surfaces to be bonded are constructed based on the WM function. The polycrystalline nano-twin copper supercell is expanded by removing the specified atoms to form an atomic-level model of polycrystalline nano-twin copper with a rough surface to be bonded, such as Figure 2 shown.

[0048] Furthermore, using CPU parallel processing and GPU parallel processing can accelerate modeling and expand the model to simulations of billions of atoms.

[0049] In step S3, the atomic-level model of polycrystalline nanotwinned copper with rough surface to be bonded established in S2 is used to perform molecular dynamics simulation of the bonding process. It is necessary to divide the fixed layer, heating layer, and relaxation layer to perform energy minimization calculations. The NVT ensemble is used to set the temperature and temperature gradient in the heating layer, and the bonding time and trajectory result output are defined. The overall model uses the NVE ensemble for bonding simulation, and the annealing simulation is performed under the NVT ensemble.

[0050] Furthermore, the parameters monitored in real time include: surface roughness, diffusion coefficient, atomic number density distribution, crystal structure, twin structure, pores, internal stress distribution, temperature distribution and energy, etc.

[0051] In step S4, the interatomic potential function selected in S1 is used to perform molecular dynamics simulation of interconnection performance on the bonded polycrystalline nanotwin copper model obtained in S3, and the performance simulation performed includes: interconnection structural performance, interconnection thermal performance, fracture mechanical performance and interconnection mechanical performance.

[0052] Furthermore, the optimized copper-copper direct bonding materials and process parameters are divided into structural parameters and process parameters, and the structural parameters are fed back to step S2 to establish an optimized atomic-level model of polycrystalline nano-twin copper on the rough surface to be bonded, including: twin density, twin size, polycrystalline distribution, polycrystalline density, grain size, crystal orientation, copper column size and surface roughness, etc.

[0053] Furthermore, the optimized copper-copper direct bonding materials and process parameters are divided into structural parameters and process parameters, and the process parameters are fed back to step S3 for bonding simulation and annealing treatment, including: bonding temperature or temperature gradient, bonding time, bonding spacing, annealing temperature or temperature gradient, annealing time, etc.

[0054] S4a simulates the performance of the interconnect structure by performing NPT temperature and pressure control simulation, uniaxial tension and compression, nanoindentation and shear simulation on the overall structure, simulating the structural characteristics of the chip interconnect structure under a typical application environment, and monitoring its crystal structure changes, stacking fault twinning pathways, dislocation evolution and grain boundary development.

[0055] S4b simulates the thermal performance of interconnects by performing XYZ proportional changes on the overall structure under multiple temperatures and pressures, performing multiple sets of long-term constant temperature and pressure simulations in various temperature regions, obtaining the number of frames after the final energy balance, and calculating the average value to obtain its thermal expansion coefficient.

[0056] Furthermore, S4b calculated the glass transition temperature using the solid-liquid coexistence method, established the solid-liquid coexistence interface under specified pressure conditions, obtained the equilibrium temperature-pressure correspondence curve through the NPH ensemble, and accurately obtained its glass transition temperature.

[0057] The simulation of fracture mechanical properties of S4c is achieved by applying uniaxial tension to the overall structure, with a strain rate of less than 5.0×10 7 s -1 , constant temperature simulation at different temperatures, monitoring its stress-strain response, stress evolution, fracture morphology and damage evolution, etc.

[0058] S4d simulates the mechanical properties of interconnects by applying uniaxial tension, shear and nanoindentation simulations along the Z direction to the overall structure, and performs constant temperature simulations at different temperatures to monitor the evolution of pores and porosity, interface stress-strain response and interface fracture morphology.

[0059] Furthermore, after the performance is evaluated in step S4, S4e provides the optimized structural parameters and process parameters to steps S2 and S3, establishes the coupling relationship between its material structure and bonding process and interconnection performance, and forms a high-reliability polycrystalline nano-twin copper direct bonding process.

[0060] In a further preferred embodiment, the atomic-level modeling, molecular dynamics bonding simulation and interconnection performance simulation of rough surfaces in the present invention are not only applicable to the copper-copper direct bonding process, but also include the mixed bonding process of interconnection dielectric materials and interconnection metals and the direct bonding process using dissimilar metals.

[0061] The above is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and changes can be made without departing from the principle of the present invention, and these improvements and changes are also regarded as the protection scope of the present invention.

Claims

1. An atomic-level simulation method for direct bonding of rough surfaces of nano-twinned copper, characterized in that: For performing molecular dynamics simulation, the method comprises the following steps: S1: Construct a bulk structure model of nano-twinned copper, perform first-principles calculations and molecular dynamics simulations on the nano-twinned copper structure, compare and analyze characteristic parameters, and determine the appropriate interatomic interaction potential function; S2: Based on the polycrystalline and twin parameters obtained from the experiment, a supercell structure model of polycrystalline nano-twinned copper is constructed, and an atomic-level model of polycrystalline nano-twinned copper before bonding is obtained by constructing a rough surface with splitting characteristics; S3: Based on the atomic-level model constructed in step S2 and the interatomic interaction potential function determined in step S1, a molecular dynamics method is used to perform bonding simulation, and key parameters in the model bonding process are monitored in real time. After achieving stable bonding, the model is annealed to obtain the polycrystalline nanotwin copper model data after bonding; S4: Based on the polycrystalline nanotwin copper model data obtained in S3 and the interatomic interaction potential function determined in S1, the molecular dynamics method is used to perform molecular dynamics simulation of the interconnection performance of polycrystalline nanotwin copper pillars, and the high-reliability polycrystalline nanotwin copper direct bonding process conditions are determined at the atomic level.

2. The atomic level simulation method for direct bonding of nano-twinned copper rough surface according to claim 1, characterized in that: In the step S1, the characteristic parameters include: lattice constant, density, thermal expansion coefficient, elastic constant, intrinsic stacking fault energy, Poisson's ratio, bulk elastic modulus, Young's modulus, shear model, radial distribution function, and Bain path.

3. The atomic level simulation method for direct bonding of nano-twinned copper rough surface as claimed in claim 1, characterized in that: In the step S1, suitable interatomic potential functions include: embedded atomic potential function, modified embedded interatomic potential function and machine learning interatomic potential function.

4. The atomic level simulation method for direct bonding of nano-twinned copper rough surface according to claim 1, characterized in that: In the step S1, the first principle calculation used is based on density functional theory, and the functionals include PBE functional and SCAN functional. The molecular dynamics method first uses empirical potential functions for calculation, and the potential functions include EAM potential functions and MEAM potential functions.

5. The atomic level simulation method for direct bonding of nano-twinned copper rough surface according to claim 1, characterized in that: In step S1, when the error between the molecular dynamics calculation parameter comparison result and the first-principles calculation is low, the empirical potential function is fine-tuned according to the characteristic parameters and selected as the interatomic interaction potential function for subsequent calculations; when the error is high, a machine learning potential function is constructed based on the first-principles calculation using an active learning method and selected as the interatomic interaction potential function for subsequent calculations.

6. The atomic level simulation method for direct bonding of nano-twinned copper rough surface according to claim 1, characterized in that: The step S2 includes the following sub-steps: S2a, combined with the experimental observation results of nano-twinned copper electroplating on pure copper surface, the twin size and twin density are obtained according to the cross section after focused ion beam treatment, and the parameters are used to model the nano-twinned copper unit cell; S2b, based on the polycrystalline distribution, grain size and crystal orientation obtained by electron backscatter diffraction analysis technology, a polycrystalline nano-twin copper supercell model is established on the basis of S2a; S2c. Delete a specified number of layers of atoms in the middle of the supercell model established in step S2b to construct a reserved gap before bonding. The entire model is divided into two upper and lower polycrystalline nano-twin copper pillars to be bonded. Based on the Weierstrass-Mandelbrot function, a specified roughness surface with fractal characteristics is constructed on the surfaces to be bonded of the two copper pillars, thereby obtaining a polycrystalline nano-twin copper atomic-level model with a rough surface.

7. The atomic level simulation method for direct bonding of nano-twinned copper rough surface according to claim 1, characterized in that: The step S3 includes the following sub-steps: S3a, dividing the atomic-level model constructed in step S2 into a top fixed layer, a top heating layer, a middle relaxation layer, a bottom heating layer and a bottom fixed layer in sequence; S3b, using the interatomic interaction potential function determined in step S1 to perform molecular dynamics simulation, perform energy minimization processing as a whole, use a canonical system to apply different temperatures to the top heating layer and the bottom heating layer to define the temperature gradient during the bonding process, and use a microcanonical system as a whole to simulate the heat transfer and expansion of the upper and lower polycrystalline nano-twin copper columns during the entire bonding process, so as to promote contact between the upper and lower surfaces and self-diffusion to achieve bonding; S3c. During the bonding process, the surface roughness, diffusion coefficient, atomic number density distribution, crystal structure, twin structure, pores, internal stress, temperature distribution and energy are monitored in real time. After achieving stable bonding, the model is annealed to obtain the polycrystalline nano-twin copper model data after bonding.

8. The atomic level simulation method for direct bonding of nano-twinned copper rough surface as claimed in claim 1, characterized in that: In step S4, the interatomic interaction potential function selected in S1 is used to perform molecular dynamics simulation of interconnection performance on the bonded polycrystalline nanotwin copper model obtained in S3, and the performance simulation performed includes: interconnection structural performance, interconnection thermal performance, fracture mechanical performance and interconnection mechanical performance.

9. The atomic level simulation method for direct bonding of nano-twinned copper rough surface as claimed in claim 1, characterized in that: The step S4 includes the following sub-steps: S4a, based on the stably bonded polycrystalline nano-twinned copper model obtained in step S3, using the molecular dynamics method to simulate its interconnection structure performance, monitor its crystal structure changes, stacking fault twinning pathways, dislocation evolution and grain boundary development; S4b, according to the stably bonded polycrystalline nanotwinned copper model obtained in step S3, using molecular dynamics method to simulate its interconnection thermal performance, and monitor its thermal expansion coefficient, glass transition temperature, thermal conductivity and specific heat; S4c Based on the stably bonded polycrystalline nanotwinned copper model obtained in S3, the molecular dynamics method is used to simulate its fracture mechanical properties and monitor its stress-strain response, stress evolution, fracture morphology and damage evolution; S4d Based on the stably bonded polycrystalline nanotwinned copper model obtained in S3, the molecular dynamics method is used to simulate its interconnection mechanical properties, and its pore and porosity evolution, interface stress-strain response, and interface fracture morphology are monitored; S4e. According to the performance monitoring results of steps S4a, S4b, S4c and S4d, the copper-copper direct bonding materials and process parameters are optimized to guide the structural modeling of step S2 and the process simulation of step S3, thereby forming a high-reliability polycrystalline nano-twin copper direct bonding process.

Images