Micro-nano structure enhanced semiconductor hetero-interface heat transport simulation method
By employing a semiconductor heterostructure thermal transport simulation method enhanced by micro-nano structures, combined with Monte Carlo and neural network optimization algorithms, the accuracy and efficiency issues in heterostructure thermal transfer design were resolved, thus meeting the thermal management requirements of high-power devices.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF SCI & TECH
- Filing Date
- 2026-04-22
- Publication Date
- 2026-07-10
AI Technical Summary
Existing technologies are insufficient to accurately describe the thermal transport behavior of heterogeneous interfaces at the micro-nano scale, resulting in insufficient accuracy in predicting heat transfer performance, long computation time and low efficiency, and inability to meet the thermal management requirements of high-power devices.
A method for simulating thermal transport at semiconductor heterostructure interfaces using micro/nanostructures is employed. By defining the heterostructure system, interface type, and thermal boundary conditions, and combining a morphology parameter database with first-principles calculations, a phonon parameter library is established. The Monte Carlo method is used to solve the phonon Boltzmann transport equation, and a global search is performed using a neural network surrogate model and a genetic algorithm to optimize the micro/nanostructure design.
It achieves high-precision and high-speed optimization of heterogeneous interface heat transfer, improves the thermal performance and reliability of devices, and solves the shortcomings of traditional methods in terms of computational scale and efficiency.
Smart Images
Figure CN122369744A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of chip thermal management technology and relates to a method for simulating thermal transport at semiconductor heterostructures using micro / nano structures. Background Technology
[0002] With the rapid evolution of third-generation semiconductor devices and advanced packaging technologies, heterostructures such as GaN / SiC and GaN / diamond have become core components of high-power-density devices due to their excellent electrical performance. During operation, these devices often generate heat in the form of micro- and nano-scale hot spots, which must be conducted outwards across multiple material layers and interfaces. However, due to factors such as lattice mismatch between heterostructures, differences in acoustic properties, and interface defects and roughness, heterostructure interfaces often exhibit significant thermal resistance bottlenecks. This leads to heat accumulation in the interface region, resulting in problems such as excessive local temperature rise, performance drift, and even decreased reliability. Therefore, improving the heat transfer efficiency of heterostructure interfaces and reducing interface thermal resistance has become a critical issue that urgently needs to be addressed in the field of thermal management for high-power devices.
[0003] Existing research has found that introducing micro / nanostructures such as nanopillars, nanopores, and nanogrooves into heterojunctions can effectively control the propagation path of phonons in the interface and microstructure regions, thereby improving the interface's thermal conductivity and reducing thermal resistance, providing a feasible approach to solving the aforementioned problems. However, existing technologies still face many challenges when performing heat transfer simulation and optimization design for such heterojunctions containing micro / nanostructures: traditional macroscopic heat transfer models based on Fourier's law are difficult to accurately describe the complex transport behavior of phonons at the micro / nano scale, resulting in insufficient accuracy in predicting heat transfer performance; while microscopic simulation methods such as molecular dynamics (MD) can reproduce the phonon transport process to some extent, they suffer from limitations in computational scale, failing to cover larger structures in actual devices, and their strong dependence on high-precision potential functions also limits the accuracy and universality of their simulation results. Furthermore, current simulation methods generally suffer from long computation times and low efficiency. Coupled with the high cost of high-precision experimental verification, the optimization design of heterogeneous interface micro / nano structures is caught in a dilemma where efficiency and accuracy are difficult to balance. There is an urgent need for a new simulation and optimization method that combines high-precision prediction capabilities with high-efficiency optimization performance to meet the actual needs of thermal management of high-power devices. Summary of the Invention
[0004] The purpose of this invention is to solve the problems of mismatch between model accuracy and computational scale, high optimization time, and low iteration efficiency in the design of heat transfer enhancement at micro / nano structure heterostructure interfaces in the prior art, and to provide a simulation method for heat transport at semiconductor heterostructure interfaces enhanced by micro / nano structures.
[0005] To achieve the above objectives, the present invention employs the following technical solution:
[0006] A method for simulating thermal transport at semiconductor heterostructures using micro / nanostructure enhancement includes:
[0007] Step 1: Identify the semiconductor heterostructure system to be optimized, its interface type, operating conditions, and thermal boundary conditions, and set optimization objectives and evaluation indicators;
[0008] Step 2: Based on the morphology parameter database, the interface micro-nano structure is parametrically described, a three-dimensional geometric model is established and manufacturing constraints are set; the phonon transport parameters of each layer of material are calculated based on first-principles calculations, and a phonon parameter library is constructed.
[0009] Step 3: Under the constraints of the three-dimensional geometric model and phonon parameter library, solve the phonon Boltzmann transport equation based on the Monte Carlo method, simulate the phonon transport process and calculate the thermal parameters;
[0010] Step 4: Sample the topographic parameter vector within the manufacturing constraints and geometrically feasible domain to generate a simulation dataset of structural parameters and thermal response;
[0011] Step 5: Train a neural network surrogate model based on the simulation dataset to establish a fast mapping relationship between topographic parameters and thermal indices; and use the output of the trained neural network surrogate model as a fitness evaluation to perform a global search of the topographic parameter vector using a genetic algorithm to output a set of candidate schemes that meet the constraints.
[0012] Step 6: Apply high-fidelity solution settings to the candidate solution set, re-execute step 3, check the prediction error of the neural network surrogate model, and then select high-confidence candidate solutions; verify the heat transfer performance of the candidate solutions, and output the solution if it meets the preset target; otherwise, backfill the results to step 4 to update the dataset and neural network surrogate model, and iterate until the target is met.
[0013] A further improvement of the present invention is that:
[0014] Furthermore, the semiconductor heterostructure system to be optimized, interface type, operating conditions, and thermal boundary conditions are clearly defined, and optimization objectives and evaluation indicators are set, specifically as follows:
[0015] Semiconductor heterostructure systems include one or more of GaN / SiC, GaN / diamond, and GaN / AlN / GaN; interface types include one or more of flat interfaces, rough interfaces, or interfaces containing intermediate layers; and evaluation indicators include at least interface thermal conductivity. Interfacial thermal resistance Equivalent thermal resistance Hotspots are heating up One or more of the following: thermal boundary conditions, including heat flow boundary, isothermal boundary or heat source boundary.
[0016] Furthermore, the parameterized description of the interface micro / nanostructure based on the morphology parameter database, the establishment of a three-dimensional geometric model, and the setting of manufacturing constraints specifically involve: the interface micro / nanostructure including one or more of the following: nanopillars, nanopores, nanocones, nanogrooves, periodic arrays, aperiodic arrays, and combinations thereof; the morphology and size of the micro / nanostructure are uniformly characterized by a parameter vector, wherein the parameter vector is... ,in, For period or feature spacing, For height / depth, For aperture / linewidth, For duty cycle, For roughness characterization parameters, Inclination angle / taper.
[0017] Furthermore, the step of calculating the phonon transport parameters of each layer of material based on first-principles calculations and constructing a phonon parameter library specifically involves:
[0018] The original cell structure of the target material is obtained and imported into first-principles calculation software to establish an initial calculation model; the cell structure of the initial calculation model is optimized based on the first-principles calculation software to obtain the optimized cell parameters.
[0019] The second-order force constants were calculated based on the optimized unit cell structure, and the phonon dispersion relation and / or phonon density of states were obtained.
[0020] The third-order force constants were calculated based on the optimized cell structure to characterize the anharmonic interactions and three-phonon scattering properties of the material.
[0021] After obtaining the second and third force constants, the Boltzmann transport equation solving software was used to solve the phonon transport behavior of the material, thereby obtaining the intrinsic thermal conductivity and related phonon transport parameters of the material.
[0022] The calculated results of phonon spectrum characteristics and / or intrinsic thermal conductivity are checked for consistency with experimental data. If the consistency does not meet the preset threshold, the calculation settings are adjusted and the above steps are repeated until a set of material micro-phonon characteristic parameters that meet the consistency requirements is obtained.
[0023] Furthermore, the phonon dispersion relation is obtained from the eigenvalue equations of the dynamical matrix, expressed as follows:
[0024]
[0025] in, For phonon frequency, For wave vector, Phonon branch;
[0026] The phonon group velocity is obtained from the dispersion relation:
[0027]
[0028] in, For wave vector Polarization branch phonon group velocity; For the angular frequency of the corresponding phonon, i.e., the phonon dispersion relation Indicates the relationship between wave vectors Find the gradient.
[0029] Furthermore, the Boltzmann transport equation is solved using ShengBTE software; the intrinsic thermal conductivity of the material is calculated as follows:
[0030]
[0031] in, The volume of the material or the corresponding normalized volume. For phonon mode specific heat, For group velocity, For phonon relaxation time;
[0032] The phonon spectral features include, but are not limited to, phonon dispersion relations, the position of characteristic peaks of the density of states, and the speed of sound;
[0033] The phonon characteristic parameters include, but are not limited to: phonon dispersion relation. Group speed Density of states Specific heat Relaxation time and mean free path The phonon characteristic parameters are obtained through first-principles calculations.
[0034] Furthermore, under the constraints of a three-dimensional geometric model and a phonon parameter library, the phonon Boltzmann transport equation is solved using the Monte Carlo method to simulate the phonon transport process and calculate thermal indices. Specifically:
[0035] The phonon Boltzmann transport equation (BTE) was solved using the Monte Carlo method. Statistical simulations were performed on the phonon scattering and interfacial transmission / reflection processes within a heterostructure containing micro / nano structures. The temperature field and heat flow field were obtained, and the interfacial thermal conductivity or interfacial thermal resistance index was inverted.
[0036] The phonon Boltzmann transport equation is expressed as follows:
[0037]
[0038] in, Let be the phonon distribution function. For time; The collision term describes the change in the distribution function caused by collision processes such as phonon scattering, defect scattering, and boundary scattering. The spatial gradient of the phonon distribution function; Represents the group velocity of phonons ; Let be the total time rate of change of the phonon distribution function.
[0039] Furthermore, the method of solving the phonon Boltzmann transport equation using the Monte Carlo method is specifically as follows:
[0040] A geometric model of a heterogeneous structure containing micro / nano structures is imported, and boundary conditions and a computational grid are set. Then, the microscopic phonon characteristic parameters of the material are imported. Based on the set heat source / boundary temperature conditions, initial phonon properties are sampled from the corresponding region according to statistical distribution, and weighted energies are assigned to the phonons. The sampling probability of phonon modes is constructed by combining the density of states to ensure statistical consistency of emitted phonons in frequency and branching. Boundary conditions include, but are not limited to, isothermal boundaries, adiabatic boundaries, symmetric boundaries, and transmission / reflection boundary conditions at interfaces. The grid is used to discretize the computational domain to statistically analyze local temperature and heat flux contributions. Microscopic phonon characteristic parameters include: phonon branching, frequency, group velocity, modal specific heat, relaxation time, or scattering rate. The initial phonon properties include phonon branching, frequency, propagation direction, and initial position.
[0041] Phonon sampling follows a Bose-Einstein distribution:
[0042]
[0043] in, , To reduce Planck's constant, Boltzmann's constant; Absolute temperature;
[0044] After phonon emission, the phonon enters the phonon flight and scattering phase: phonons propagate in the computational domain at group velocity, and their free flight time is determined by the relaxation time and obtained through random sampling. The expression for the free flight time is:
[0045]
[0046] in, Represents the free flight time of phonons. The result is a uniformly random number within the range of (0,1). For phonon relaxation time.
[0047] The position of the phonon during free flight is updated as follows
[0048]
[0049] in, For phonons during free flight time The new position after For phonons at time Location, It is the phonon group velocity;
[0050] When phonons encounter scattering events or boundaries / interfaces during propagation, the corresponding scattering rules are executed: For phonon-phonon scattering, the anharmonic scattering effect is manifested by reheating the phonons or changing their frequency / branch / direction, and the phonon frequency and velocity are updated accordingly; for phonon-boundary scattering, specular reflection or diffuse reflection is implemented according to the boundary type; for phonon-interface scattering, whether the phonons propagate across the interface is determined based on the interface transmittance / reflectance, and the phonon parameter set of the other material is switched during transmission, and the propagation direction is updated during reflection; after each scattering event, the phonon properties are updated and the phonon enters the next free flight stage;
[0051] Throughout the phonon flight process, the energy deposition and heat flux contribution of phonons in each grid cell are statistically analyzed to obtain the local temperature and heat flux distribution. Temperature updates are based on the energy conservation relationship and are deduced from the cell energy deviation. The difference between the phonon energy and the equilibrium energy in the grid is mapped to the temperature correction, and the local energy is made consistent with the target temperature field through iteration. Heat flux statistics are achieved by accumulating the energy transport of phonons in the grid. The grid heat flux contribution is accumulated according to the directional components of phonon energy and velocity to obtain the steady-state or transient heat flux density distribution.
[0052] To control the computational load of a single phonon trajectory and ensure statistical convergence, a dual termination criterion is set: the first is reaching the maximum number of phonon scatterings or encountering an isothermal boundary. When the number of phonon scatterings reaches the preset upper limit or the phonon enters the isothermal boundary and is absorbed / re-emitted, the phonon trajectory is terminated; the second is reaching the maximum number of phonons set in the simulation. When the cumulative number of simulated phonons reaches the preset total, the emission of new phonons stops and the statistical output stage begins; if the maximum number of phonons is not reached, the phonon emission step is returned to continue emitting the next phonon, forming a cyclic sampling.
[0053] Once the termination condition is met, the global temperature and heat flux distribution are summarized, and interface-related heat transfer parameters are calculated, including interface temperature difference, interface heat flux density, and interface thermal conductivity / thermal resistance; the expression for interface thermal conductivity is:
[0054]
[0055] in, For interface thermal conductivity, The heat flux density across the interface, For interface temperature difference;
[0056] Interfacial thermal resistance for
[0057]
[0058] Simultaneously, it outputs the equivalent thermal resistance, local thermal resistance decomposition, and temperature rise and heat flow non-uniformity indicators of the hot spot region containing micro-nano structure heterostructure.
[0059] Furthermore, the sampling of the topographic parameter vector within the manufacturing constraints and geometrically feasible domain to generate a simulation dataset relating structural parameters and thermal response specifically involves:
[0060] Within the manufacturing constraints and geometrically feasible region, for the parameter vector The parameter elements are assigned values to obtain several sets of parameter combinations; for each sampled parameter combination, steps 2 to 3 are automatically repeated to obtain a simulation dataset containing structural parameters, phonon parameters, and thermal response; the thermal response includes at least the interfacial thermal conductivity. Interfacial thermal resistance Equivalent thermal resistance Hotspots are heating up One or more of them.
[0061] Furthermore, a neural network surrogate model is trained based on the simulation dataset to establish a fast mapping relationship between topographic parameters and thermal indices; and the output of the trained neural network surrogate model is used as a fitness evaluation to perform a global search of the topographic parameter vector using a genetic algorithm, outputting a set of candidate solutions that meet the constraints; specifically:
[0062] A neural network surrogate model is trained based on a simulation dataset to establish a fast mapping from input parameters to thermal indices, specifically as follows:
[0063]
[0064] in, Indicates by parameters The determined neural network mapping function, Represents the model input variables, This indicates material and operating condition parameters, including temperature, crystal orientation, layer thickness, and thermal boundary conditions. Evaluation indicators representing predictions;
[0065] The loss function for neural networks uses the mean squared error (MSE), i.e.:
[0066]
[0067] in, Indicates the number of samples; Indicates the first Predicted values for each sample; Indicates the first The true value of each sample.
[0068] The weights and biases of the network are optimized using the backpropagation algorithm to minimize the prediction error; the trained neural network model is used to quickly predict the thermal conductivity or thermal resistance corresponding to different micro / nano structure parameters.
[0069] A genetic algorithm is used to perform a global optimization search on the morphological parameter vector 𝑥, with the output of the surrogate model as the fitness function, to complete multi-objective or single-objective optimization; among them, the single objective is to maximize the interface thermal conductivity. Or minimize interfacial thermal resistance The multiple objectives are to maximize the thermal conductivity of the interface. Minimize equivalent thermal resistance at the same time And satisfy manufacturing and reliability constraints; express the micro / nano structure design variables as
[0070]
[0071] First, an initial population is randomly generated within the specified parameter range, with each individual corresponding to a set of micro / nano structure parameters; then, a neural network surrogate model is invoked to predict heat transfer performance indicators, and an optimization objective function is constructed accordingly.
[0072] When maximizing interface thermal conductivity, take
[0073]
[0074] in, The primary optimization objective is to maximize the interface thermal conductivity G, where G is the interface thermal conductivity.
[0075] When minimizing interface thermal resistance, take
[0076]
[0077] in, The second optimization objective is to minimize the interfacial thermal resistance R, where R is the interfacial thermal resistance.
[0078] Genetic algorithms select individuals based on fitness and generate a new generation through crossover and mutation; the crossover uses a linear combination method.
[0079]
[0080] in, This represents the offspring individuals generated through crossover; and These represent the two parent individuals involved in the crossover; The crossover coefficient, typically between 0 and 1, is used to control the proportion of features inherited by the offspring from the two parents.
[0081] The mutation applies random perturbations to some parameters.
[0082]
[0083] The updated parameters are then adjusted to meet the range constraints; this process is iterated until the convergence condition is met, and the optimal parameters and a set of candidate solutions are output; among them, Indicates the first A design variable or parameter, This represents the amount of random perturbation applied to this parameter. This indicates that parameter variation is achieved through random perturbation.
[0084] Furthermore, the heat transfer performance of the candidate solutions is verified. If the preset target is met, the solution is output; otherwise, the results are backfilled into step 4 to update the dataset and the neural network surrogate model, and the iteration continues until the target is met. Specifically, the heat transfer performance of the optimal candidate structure that has passed the verification is verified. The verification content includes, but is not limited to, the ability to suppress hot spot temperature rise, the reduction of interface thermal resistance, the robustness under different temperatures / operating conditions, and the sensitivity analysis of structural parameters to performance. If the preset target is not met, the verification results are backfilled into the dataset and the surrogate model is updated, and the iteration continues. If the target is met, the optimized micro / nano structure morphology and size scheme and its corresponding thermal performance indicators are output.
[0085] Compared with the prior art, the present invention has the following beneficial effects:
[0086] This invention clarifies the fundamental conditions such as heterogeneous structure systems and interface types, and sets optimization objectives and evaluation indicators. It then parameterizes the interface micro / nano structures based on a morphology parameter database and establishes a three-dimensional geometric model. A phonon parameter library is constructed using first-principles calculations. Subsequently, the phonon Boltzmann transport equation is solved using the Monte Carlo method to simulate phonon transport and calculate thermal indices. Next, a simulation dataset is generated under manufacturing constraints through sampling. A neural network surrogate model is trained based on this dataset to establish a fast mapping between morphology parameters and thermal indices. A genetic algorithm is used for global search to output a set of candidate solutions. Finally, high-fidelity solving is used to verify prediction errors, and highly reliable candidate solutions are selected. The heat transfer performance of the candidate solutions is verified; if the performance does not meet the standards, the results are backfilled to update the dataset and model, and the process is iterated. This invention ensures the credibility and engineering feasibility of candidate solutions, providing an efficient and accurate solution for enhancing heat transfer at semiconductor heterogeneous interfaces, and improving the thermal performance and reliability of devices. Attached Figure Description
[0087] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0088] Figure 1 This is a schematic flowchart of a method for simulating thermal transport at semiconductor heterostructure interfaces using micro / nanostructures, according to the present invention.
[0089] Figure 2 A flowchart illustrating the process of obtaining phonon characteristic parameters using first-principles calculations;
[0090] Figure 3 A flowchart illustrating the process of solving the phonon Boltzmann transport equation using the Monte Carlo method;
[0091] Figure 4 This is a schematic diagram of a micro / nano structure-enhanced semiconductor heterostructure thermal transport simulation system according to the present invention. Detailed Implementation
[0092] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0093] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0094] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0095] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper," "lower," "horizontal," or "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, terms such as "first" and "second" are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0096] Furthermore, the use of the term "horizontal" does not imply that the component must be absolutely horizontal, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.
[0097] In the description of the embodiments of the present invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.
[0098] The present invention will now be described in further detail with reference to the accompanying drawings:
[0099] See Figure 1 This invention discloses a method for simulating thermal transport at semiconductor heterostructures using micro / nano structures, comprising:
[0100] S101: Identify the semiconductor heterostructure system to be optimized, the interface type, the operating conditions and thermal boundary conditions, and set the optimization objectives and evaluation indicators;
[0101] Semiconductor heterostructure systems include, but are not limited to, one or more of GaN / SiC, GaN / diamond, and GaN / AlN / GaN; interface types include one or more of flat interfaces, rough interfaces, or interfaces containing intermediate layers; and evaluation metrics include at least interface thermal conductivity. Interfacial thermal resistance Equivalent thermal resistance Hotspots are heating up One or more of the following: thermal boundary conditions, including but not limited to heat flow boundary, isothermal boundary or heat source boundary.
[0102] S102: Based on the morphology parameter database, the interface micro-nano structure is parametrically described, a three-dimensional geometric model is established and manufacturing constraints are set; the phonon transport parameters of each layer of material are calculated based on first-principles calculations, and a phonon parameter library is constructed.
[0103] The interface micro / nanostructures are parameterized based on a morphology parameter database. A three-dimensional geometric model is established and manufacturing constraints are set. Specifically, the interface micro / nanostructures include one or more of the following: nanopillars, nanopores, nanocones, nanogrooves, periodic arrays, aperiodic arrays, and combinations thereof. The morphology and size of the micro / nanostructures are uniformly characterized by parameter vectors, which are... ,in, For period or feature spacing, For height / depth, For aperture / linewidth, For duty cycle, For roughness characterization parameters, Inclination angle / taper.
[0104] See Figure 2 Phonon transport parameters of each layer of material were calculated based on first-principles calculations, and a phonon parameter library was constructed, specifically as follows:
[0105] The original cell structure of the target material is obtained and imported into first-principles calculation software to establish an initial calculation model; the cell structure of the initial calculation model is optimized based on the first-principles calculation software to obtain the optimized cell parameters.
[0106] The phonon dispersion relation is obtained from the eigenvalue equations of the dynamical matrix, expressed as follows:
[0107]
[0108] in, For phonon frequency, For wave vector, Phonon branch;
[0109] The phonon group velocity is obtained from the dispersion relation:
[0110]
[0111] in, For wave vector Polarization branch phonon group velocity; For the angular frequency of the corresponding phonon, i.e., the phonon dispersion relation Indicates the relationship between wave vectors Find the gradient;
[0112] The Boltzmann transport equation is solved using ShengBTE software; the intrinsic thermal conductivity of the material is calculated as follows:
[0113]
[0114] in, The volume of the material or the corresponding normalized volume. For phonon mode specific heat, For group velocity, For phonon relaxation time;
[0115] The phonon spectral features include, but are not limited to, phonon dispersion relations, the position of characteristic peaks of the density of states, and the speed of sound;
[0116] The phonon characteristic parameters include, but are not limited to: phonon dispersion relation. Group speed Density of states Specific heat Relaxation time and mean free path The phonon characteristic parameters are obtained through first-principles calculations.
[0117] The second-order force constants were calculated based on the optimized unit cell structure, and the phonon dispersion relation and / or phonon density of states were obtained.
[0118] The third-order force constants were calculated based on the optimized cell structure to characterize the anharmonic interactions and three-phonon scattering properties of the material.
[0119] After obtaining the second and third force constants, the Boltzmann transport equation solving software was used to solve the phonon transport behavior of the material, thereby obtaining the intrinsic thermal conductivity and related phonon transport parameters of the material.
[0120] The calculated results of phonon spectrum characteristics and / or intrinsic thermal conductivity are checked for consistency with experimental data. If the consistency does not meet the preset threshold, the calculation settings are adjusted and the above steps are repeated until a set of material micro-phonon characteristic parameters that meet the consistency requirements is obtained.
[0121] S103: Under the constraints of a three-dimensional geometric model and a phonon parameter library, the phonon Boltzmann transport equation is solved based on the Monte Carlo method to simulate the phonon transport process and calculate thermal indices.
[0122] The phonon Boltzmann transport equation (BTE) was solved using the Monte Carlo method. Statistical simulations were performed on the phonon scattering and interfacial transmission / reflection processes within a heterostructure containing micro / nano structures. The temperature field and heat flow field were obtained, and the interfacial thermal conductivity or interfacial thermal resistance index was inverted.
[0123] The phonon Boltzmann transport equation is expressed as follows:
[0124]
[0125] in, Let be the phonon distribution function. For time; The collision term describes the change in the distribution function caused by collision processes such as phonon scattering, defect scattering, and boundary scattering. The spatial gradient of the phonon distribution function; Represents the group velocity of phonons ; Let be the total time rate of change of the phonon distribution function.
[0126] See Figure 3 The phonon Boltzmann transport equation is solved using the Monte Carlo method, specifically as follows:
[0127] A geometric model of a heterogeneous structure containing micro / nano structures is imported, and boundary conditions and a computational grid are set. Then, the microscopic phonon characteristic parameters of the material are imported. Based on the set heat source / boundary temperature conditions, initial phonon properties are sampled from the corresponding region according to statistical distribution, and weighted energies are assigned to the phonons. The sampling probability of phonon modes is constructed by combining the density of states to ensure statistical consistency of emitted phonons in frequency and branching. Boundary conditions include, but are not limited to, isothermal boundaries, adiabatic boundaries, symmetric boundaries, and transmission / reflection boundary conditions at interfaces. The grid is used to discretize the computational domain to statistically analyze local temperature and heat flux contributions. Microscopic phonon characteristic parameters include: phonon branching, frequency, group velocity, modal specific heat, relaxation time, or scattering rate. The initial phonon properties include phonon branching, frequency, propagation direction, and initial position.
[0128] Phonon sampling follows a Bose-Einstein distribution:
[0129]
[0130] in, It is the phonon equilibrium distribution function. To reduce Planck's constant, Boltzmann's constant; Absolute temperature;
[0131] After phonon emission, the phonon enters the phonon flight and scattering phase: phonons propagate in the computational domain at group velocity, and their free flight time is determined by the relaxation time and obtained through random sampling. The expression for the free flight time is:
[0132]
[0133] in, Represents the free flight time of phonons. The result is a uniformly random number within the range of (0,1). For phonon relaxation time.
[0134] The position of the phonon during free flight is updated as follows
[0135]
[0136] in, For phonons during free flight time The new position after For phonons at time Location, It is the phonon group velocity;
[0137] When phonons encounter scattering events or boundaries / interfaces during propagation, the corresponding scattering rules are executed: For phonon-phonon scattering, the anharmonic scattering effect is manifested by reheating the phonons or changing their frequency / branch / direction, and the phonon frequency and velocity are updated accordingly; for phonon-boundary scattering, specular reflection or diffuse reflection is implemented according to the boundary type; for phonon-interface scattering, whether the phonons propagate across the interface is determined based on the interface transmittance / reflectance, and the phonon parameter set of the other material is switched during transmission, and the propagation direction is updated during reflection; after each scattering event, the phonon properties are updated and the phonon enters the next free flight stage;
[0138] Throughout the phonon flight process, the energy deposition and heat flux contribution of phonons in each grid cell are statistically analyzed to obtain the local temperature and heat flux distribution. Temperature updates are based on the energy conservation relationship and are deduced from the cell energy deviation. The difference between the phonon energy and the equilibrium energy in the grid is mapped to the temperature correction, and the local energy is made consistent with the target temperature field through iteration. Heat flux statistics are achieved by accumulating the energy transport of phonons in the grid. The grid heat flux contribution is accumulated according to the directional components of phonon energy and velocity to obtain the steady-state or transient heat flux density distribution.
[0139] To control the computational load of a single phonon trajectory and ensure statistical convergence, a dual termination criterion is set: the first is reaching the maximum number of phonon scatterings or encountering an isothermal boundary. When the number of phonon scatterings reaches the preset upper limit or the phonon enters the isothermal boundary and is absorbed / re-emitted, the phonon trajectory is terminated; the second is reaching the maximum number of phonons set in the simulation. When the cumulative number of simulated phonons reaches the preset total, the emission of new phonons stops and the statistical output stage begins; if the maximum number of phonons is not reached, the phonon emission step is returned to continue emitting the next phonon, forming a cyclic sampling.
[0140] Once the termination condition is met, the global temperature and heat flux distribution are summarized, and interface-related heat transfer parameters are calculated, including interface temperature difference, interface heat flux density, and interface thermal conductivity / thermal resistance; the expression for interface thermal conductivity is:
[0141]
[0142] in, For interface thermal conductivity, The heat flux density across the interface, For interface temperature difference;
[0143] Interfacial thermal resistance for
[0144]
[0145] Simultaneously, it outputs the equivalent thermal resistance, local thermal resistance decomposition, and temperature rise and heat flow non-uniformity indicators of the hot spot region containing micro-nano structure heterostructure.
[0146] S104: Sample the topographic parameter vector within the manufacturing constraints and geometrically feasible domain to generate a simulation dataset between structural parameters and thermal response;
[0147] Within the manufacturing constraints and geometrically feasible region, for the parameter vector The parameter elements are assigned values to obtain several sets of parameter combinations; for each set of sampled parameter combinations, the process from S102 to S103 is automatically repeated to obtain a simulation dataset containing structural parameters, phonon parameters, and thermal response; the thermal response includes at least the interfacial thermal conductivity. Interfacial thermal resistance Equivalent thermal resistance Hotspots are heating up One or more of them.
[0148] S105: Train a neural network surrogate model based on a simulation dataset to establish a fast mapping relationship between topographic parameters and thermal indices; use the output of the trained neural network surrogate model as a fitness evaluation, perform a global search of the topographic parameter vector using a genetic algorithm, and output a set of candidate schemes that meet the constraints.
[0149] A neural network surrogate model is trained based on a simulation dataset to establish a fast mapping from input parameters to thermal indices, specifically as follows:
[0150]
[0151] in, Indicates by parameters The determined neural network mapping function, Represents the model input variables, This indicates material and operating condition parameters, including temperature, crystal orientation, layer thickness, and thermal boundary conditions. Evaluation indicators representing predictions;
[0152] The loss function for neural networks uses the mean squared error (MSE), i.e.:
[0153]
[0154] in, Indicates the number of samples; Indicates the first Predicted values for each sample; Indicates the first The true value of each sample.
[0155] The weights and biases of the network are optimized using the backpropagation algorithm to minimize the prediction error; the trained neural network model is used to quickly predict the thermal conductivity or thermal resistance corresponding to different micro / nano structure parameters.
[0156] A genetic algorithm is used to perform a global optimization search on the morphological parameter vector 𝑥, with the output of the surrogate model as the fitness function, to complete multi-objective or single-objective optimization; among them, the single objective is to maximize the interface thermal conductivity. Or minimize interfacial thermal resistance The multiple objectives are to maximize the thermal conductivity of the interface. Minimize equivalent thermal resistance at the same time And satisfy manufacturing and reliability constraints; express the micro / nano structure design variables as
[0157]
[0158] First, an initial population is randomly generated within the specified parameter range, with each individual corresponding to a set of micro / nano structure parameters; then, a neural network surrogate model is invoked to predict heat transfer performance indicators, and an optimization objective function is constructed accordingly.
[0159] When maximizing interface thermal conductivity, take
[0160]
[0161] in, The primary optimization objective is to maximize the interface thermal conductivity G, where G is the interface thermal conductivity.
[0162] When minimizing interface thermal resistance, take
[0163]
[0164] in, The second optimization objective is to minimize the interfacial thermal resistance R, where R is the interfacial thermal resistance.
[0165] Genetic algorithms select individuals based on fitness and generate a new generation through crossover and mutation; the crossover uses a linear combination method.
[0166]
[0167] in, This represents the offspring individuals generated through crossover; and These represent the two parent individuals involved in the crossover; The crossover coefficient, typically between 0 and 1, is used to control the proportion of features inherited by the offspring from the two parents.
[0168] The mutation applies random perturbations to some parameters.
[0169]
[0170] The updated parameters are then adjusted to meet the range constraints; this process is iterated until the convergence condition is met, and the optimal parameters and a set of candidate solutions are output; among them, Indicates the first A design variable or parameter, This represents the amount of random perturbation applied to this parameter. This indicates that parameter variation is achieved through random perturbation.
[0171] S106: Apply high-fidelity solution settings to the candidate solution set, re-execute S103, check the prediction error of the neural network surrogate model, and then select highly reliable candidate solutions; verify the heat transfer performance of the candidate solutions, and output the solution if it meets the preset target; otherwise, backfill the results to S104 to update the dataset and neural network surrogate model, and iterate until the target is met.
[0172] The optimal candidate structure that passes the review is subjected to heat transfer performance verification. The verification includes, but is not limited to: hot spot temperature rise suppression capability, reduction of interface thermal resistance, robustness under different temperatures / operating conditions, and sensitivity analysis of structural parameters to performance. If the preset target is not met, the verification results are backfilled into the dataset and the surrogate model is updated to continue the iteration. If the target is met, the optimized micro / nano structure morphology and size scheme and its corresponding thermal performance index are output.
[0173] See Figure 4 This invention discloses a micro / nano structure-enhanced semiconductor heterostructure thermal transport simulation system, comprising:
[0174] The defining module defines the semiconductor heterostructure system to be optimized, the interface type, the operating conditions and thermal boundary conditions, and sets the optimization objectives and evaluation indicators.
[0175] The construction module is used to parametrically describe the interface micro-nano structure based on the morphology parameter database, establish a three-dimensional geometric model and set manufacturing constraints; and calculate the phonon transport parameters of each layer of material based on first-principles calculations to construct a phonon parameter library.
[0176] The simulation module, under the constraints of a three-dimensional geometric model and a phonon parameter library, solves the phonon Boltzmann transport equation based on the Monte Carlo method, simulates the phonon transport process, and calculates thermal parameters.
[0177] The sampling module samples the topographic parameter vector within the manufacturing constraints and geometrically feasible domain to generate a simulation dataset between structural parameters and thermal response.
[0178] The output module trains a neural network surrogate model based on the simulation dataset, establishes a fast mapping relationship between morphological parameters and thermal indices, and uses the output of the trained neural network surrogate model as a fitness evaluation to perform a global search of the morphological parameter vector using a genetic algorithm, outputting a set of candidate solutions that meet the constraints.
[0179] The filtering module uses high-fidelity solution settings to check the prediction error of the neural network surrogate model on the candidate solution set, and then filters out highly reliable candidate solutions; the heat transfer performance of the candidate solutions is verified until the preset target is met, and the solution is output.
[0180] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for simulating thermal transport at semiconductor heterostructures using micro / nano structures, characterized in that, include: Step 1: Identify the semiconductor heterostructure system to be optimized, its interface type, operating conditions, and thermal boundary conditions, and set optimization objectives and evaluation indicators; Step 2: Based on the morphology parameter database, the interface micro-nano structure is parametrically described, a three-dimensional geometric model is established and manufacturing constraints are set; the phonon transport parameters of each layer of material are calculated based on first-principles calculations, and a phonon parameter library is constructed. Step 3: Under the constraints of the three-dimensional geometric model and phonon parameter library, solve the phonon Boltzmann transport equation based on the Monte Carlo method, simulate the phonon transport process and calculate the thermal parameters; Step 4: Sample the topographic parameter vector within the manufacturing constraints and geometrically feasible domain to generate a simulation dataset of structural parameters and thermal response; Step 5: Train a neural network surrogate model based on the simulation dataset to establish a fast mapping relationship between topographic parameters and thermal indices; and use the output of the trained neural network surrogate model as a fitness evaluation to perform a global search of the topographic parameter vector using a genetic algorithm to output a set of candidate schemes that meet the constraints. Step 6: Apply high-fidelity solution settings to the candidate solution set, re-execute step 3, check the prediction error of the neural network surrogate model, and then select high-confidence candidate solutions; verify the heat transfer performance of the candidate solutions, and output the solution if it meets the preset target; otherwise, backfill the results to step 4 to update the dataset and neural network surrogate model, and iterate until the target is met.
2. The method for simulating thermal transport at semiconductor heterostructures using micro / nanostructure enhancement according to claim 1, characterized in that, The process involves clearly defining the semiconductor heterostructure system to be optimized, its interface type, operating conditions, and thermal boundary conditions, and setting optimization objectives and evaluation indicators, specifically as follows: Semiconductor heterostructure systems include one or more of GaN / SiC, GaN / diamond, and GaN / AlN / GaN; interface types include one or more of flat interfaces, rough interfaces, or interfaces containing intermediate layers; and evaluation indicators include at least interface thermal conductivity. Interfacial thermal resistance Equivalent thermal resistance Hotspots are heating up One or more of the following: thermal boundary conditions, including heat flow boundary, isothermal boundary or heat source boundary.
3. The method for simulating thermal transport at semiconductor heterostructures using micro / nanostructure enhancement according to claim 2, characterized in that, The method involves parameterizing the interface micro / nanostructure based on a morphology parameter database, establishing a three-dimensional geometric model, and setting manufacturing constraints. Specifically, the interface micro / nanostructure includes one or more of the following: nanopillars, nanopores, nanocones, nanogrooves, periodic arrays, aperiodic arrays, and combinations thereof. The morphology and size of the micro / nanostructure are uniformly characterized by a parameter vector, which is... ,in, For period or feature spacing, For height / depth, For aperture / linewidth, For duty cycle, For roughness characterization parameters, Inclination angle / taper.
4. The method for simulating thermal transport at semiconductor heterostructures using micro / nanostructure enhancement according to claim 3, characterized in that, The phonon transport parameters of each layer of material are obtained based on first-principles calculations, and a phonon parameter library is constructed as follows: The original cell structure of the target material is obtained and imported into first-principles calculation software to establish an initial calculation model; the cell structure of the initial calculation model is optimized based on the first-principles calculation software to obtain the optimized cell parameters. The second-order force constants were calculated based on the optimized unit cell structure, and the phonon dispersion relation and / or phonon density of states were obtained. The third-order force constants were calculated based on the optimized cell structure to characterize the anharmonic interactions and three-phonon scattering properties of the material. After obtaining the second and third force constants, the Boltzmann transport equation solving software was used to solve the phonon transport behavior of the material, thereby obtaining the intrinsic thermal conductivity and related phonon transport parameters of the material. The calculated results of phonon spectrum characteristics and / or intrinsic thermal conductivity are checked for consistency with experimental data. If the consistency does not meet the preset threshold, the calculation settings are adjusted and the above steps are repeated until a set of material micro-phonon characteristic parameters that meet the consistency requirements is obtained.
5. The method for simulating thermal transport at a semiconductor heterostructure enhanced by micro / nano structures according to claim 4, characterized in that, The phonon dispersion relation is obtained from the eigenvalue equations of the dynamical matrix, expressed as follows: in, For phonon frequency, For wave vector, Phonon branch; The phonon group velocity is obtained from the dispersion relation: in, For wave vector Polarization branch phonon group velocity; For the angular frequency of the corresponding phonon, i.e., the phonon dispersion relation Indicates the relationship between wave vectors Find the gradient; The Boltzmann transport equation is solved using ShengBTE software; the intrinsic thermal conductivity of the material is calculated as follows: in, The volume of the material or the corresponding normalized volume. For phonon mode specific heat, For group velocity, For phonon relaxation time; The phonon spectral features include, but are not limited to, phonon dispersion relations, the position of characteristic peaks of the density of states, and the speed of sound; The phonon characteristic parameters include, but are not limited to: phonon dispersion relation. Group speed Density of states Specific heat Relaxation time and mean free path The phonon characteristic parameters are obtained through first-principles calculations.
6. The method for simulating thermal transport at a semiconductor heterostructure enhanced by micro / nano structures according to claim 5, characterized in that, Under the constraints of a three-dimensional geometric model and a phonon parameter library, the phonon Boltzmann transport equation is solved using the Monte Carlo method to simulate the phonon transport process and calculate thermal indices. Specifically: The phonon Boltzmann transport equation (BTE) was solved using the Monte Carlo method. Statistical simulations were performed on the phonon scattering and interfacial transmission / reflection processes within a heterostructure containing micro / nano structures. The temperature field and heat flow field were obtained, and the interfacial thermal conductivity or interfacial thermal resistance index was inverted. The phonon Boltzmann transport equation is expressed as follows: in, Let be the phonon distribution function. For time; The collision term describes the change in the distribution function caused by collision processes such as phonon scattering, defect scattering, and boundary scattering. The spatial gradient of the phonon distribution function; Represents the group velocity of phonons ; Let be the total time rate of change of the phonon distribution function.
7. The method for simulating thermal transport at a semiconductor heterostructure interface using micro / nanostructure enhancement according to claim 6, characterized in that, The Monte Carlo method is used to solve the phonon Boltzmann transport equations, specifically as follows: A geometric model of a heterogeneous structure containing micro / nano structures is imported, and boundary conditions and a computational grid are set. Then, the microscopic phonon characteristic parameters of the material are imported. Based on the set heat source / boundary temperature conditions, initial phonon properties are sampled from the corresponding region according to statistical distribution, and weighted energies are assigned to the phonons. The sampling probability of phonon modes is constructed by combining the density of states to ensure statistical consistency of emitted phonons in frequency and branching. Boundary conditions include, but are not limited to, isothermal boundaries, adiabatic boundaries, symmetric boundaries, and transmission / reflection boundary conditions at interfaces. The grid is used to discretize the computational domain to statistically analyze local temperature and heat flux contributions. Microscopic phonon characteristic parameters include: phonon branching, frequency, group velocity, modal specific heat, relaxation time, or scattering rate. The initial phonon properties include phonon branching, frequency, propagation direction, and initial position. Phonon sampling follows a Bose-Einstein distribution: in, It is the phonon equilibrium distribution function. To reduce Planck's constant, Boltzmann's constant; Absolute temperature; After phonon emission, the phonon enters the phonon flight and scattering phase: phonons propagate in the computational domain at group velocity, and their free flight time is determined by the relaxation time and obtained through random sampling. The expression for the free flight time is: in, Represents the free flight time of phonons. The result is a uniformly random number within the range of (0,1). For phonon relaxation time; The position of the phonon during free flight is updated as follows in, For phonons during free flight time The new position after For phonons at time Position, v g It is the phonon group velocity; When phonons encounter scattering events or boundaries / interfaces during propagation, the corresponding scattering rules are executed: For phonon-phonon scattering, the anharmonic scattering effect is manifested by reheating the phonons or changing their frequency / branch / direction, and the phonon frequency and velocity are updated accordingly; for phonon-boundary scattering, specular reflection or diffuse reflection is implemented according to the boundary type; for phonon-interface scattering, whether the phonons propagate across the interface is determined based on the interface transmittance / reflectance, and the phonon parameter set of the other material is switched during transmission, and the propagation direction is updated during reflection; after each scattering event, the phonon properties are updated and the phonon enters the next free flight stage; Throughout the phonon flight process, the energy deposition and heat flux contribution of phonons in each grid cell are statistically analyzed to obtain the local temperature and heat flux distribution. Temperature updates are based on the energy conservation relationship and are deduced from the cell energy deviation. The difference between the phonon energy and the equilibrium energy in the grid is mapped to the temperature correction, and the local energy is made consistent with the target temperature field through iteration. Heat flux statistics are achieved by accumulating the energy transport of phonons in the grid. The grid heat flux contribution is accumulated according to the directional components of phonon energy and velocity to obtain the steady-state or transient heat flux density distribution. To control the computational load of a single phonon trajectory and ensure statistical convergence, a dual termination criterion is set: the first is reaching the maximum number of phonon scatterings or encountering an isothermal boundary. When the number of phonon scatterings reaches the preset upper limit or the phonon enters the isothermal boundary and is absorbed / re-emitted, the phonon trajectory is terminated; the second is reaching the maximum number of phonons set in the simulation. When the cumulative number of simulated phonons reaches the preset total, the emission of new phonons stops and the statistical output stage begins; if the maximum number of phonons is not reached, the phonon emission step is returned to continue emitting the next phonon, forming a cyclic sampling. Once the termination condition is met, the global temperature and heat flux distribution are summarized, and interface-related heat transfer parameters are calculated, including interface temperature difference, interface heat flux density, and interface thermal conductivity / thermal resistance; the expression for interface thermal conductivity is: in, For interface thermal conductivity, The heat flux density across the interface, For interface temperature difference; Interfacial thermal resistance for Simultaneously, it outputs the equivalent thermal resistance, local thermal resistance decomposition, and temperature rise and heat flow non-uniformity indicators of the hot spot region containing micro-nano structure heterostructure.
8. The method for simulating thermal transport at a semiconductor heterostructure enhanced by micro / nanostructures according to claim 7, characterized in that, The step of sampling the topographic parameter vector within the manufacturing constraints and geometrically feasible domain to generate a simulation dataset between structural parameters and thermal response specifically involves: Within the manufacturing constraints and geometrically feasible region, for the parameter vector The parameter elements are assigned values to obtain several sets of parameter combinations; for each sampled parameter combination, steps 2 to 3 are automatically repeated to obtain a simulation dataset containing structural parameters, phonon parameters, and thermal response; the thermal response includes at least the interfacial thermal conductivity. Interfacial thermal resistance Equivalent thermal resistance Hotspots are heating up One or more of them.
9. The method for simulating thermal transport at a semiconductor heterostructure enhanced by micro / nanostructures according to claim 8, characterized in that, The method involves training a neural network surrogate model based on a simulation dataset to establish a rapid mapping relationship between topographic parameters and thermal indices. The output of the trained neural network surrogate model is then used as a fitness evaluation to perform a global genetic algorithm search on the topographic parameter vectors, outputting a set of candidate solutions that meet the constraints. Specifically: A neural network surrogate model is trained based on a simulation dataset to establish a fast mapping from input parameters to thermal indices, specifically as follows: in, Indicates by parameters The determined neural network mapping function, Represents the model input variables, This indicates material and operating condition parameters, including temperature, crystal orientation, layer thickness, and thermal boundary conditions. Evaluation indicators representing predictions; The loss function for neural networks uses the mean squared error (MSE), i.e.: in, Indicates the number of samples; Indicates the first Predicted values for each sample; Indicates the first The true value of each sample; The weights and biases of the network are optimized using the backpropagation algorithm to minimize the prediction error; the trained neural network model is used to quickly predict the thermal conductivity or thermal resistance corresponding to different micro / nano structure parameters. A genetic algorithm is used to perform a global optimization search on the morphological parameter vector 𝑥, with the output of the surrogate model as the fitness function, to complete multi-objective or single-objective optimization; among them, the single objective is to maximize the interface thermal conductivity. Or minimize interfacial thermal resistance The multiple objectives are to maximize the thermal conductivity of the interface. Minimize equivalent thermal resistance at the same time And satisfy manufacturing and reliability constraints; express the micro / nano structure design variables as First, an initial population is randomly generated within the specified parameter range, with each individual corresponding to a set of micro / nano structure parameters; then, a neural network surrogate model is invoked to predict heat transfer performance indicators, and an optimization objective function is constructed accordingly. When maximizing interface thermal conductivity, take in, The primary optimization objective is to maximize the interface thermal conductivity G, where G is the interface thermal conductivity. When minimizing interface thermal resistance, take in, The second optimization objective is to minimize the interfacial thermal resistance R, where R is the interfacial thermal resistance. Genetic algorithms select individuals based on fitness and generate a new generation through crossover and mutation; the crossover uses a linear combination method. in, This represents the offspring individuals generated through crossover; and These represent the two parent individuals involved in the crossover; The crossover coefficient, typically between 0 and 1, is used to control the proportion of features inherited by the offspring from the two parents. The mutation applies random perturbations to some parameters. The updated parameters are then adjusted to meet the range constraints; this process is iterated until the convergence condition is met, and the optimal parameters and a set of candidate solutions are output; among them, Indicates the first A design variable or parameter, This represents the amount of random perturbation applied to this parameter. This indicates that parameter variation is achieved through random perturbation.
10. The method for simulating thermal transport at a semiconductor heterostructure enhanced by micro / nanostructures according to claim 9, characterized in that, The heat transfer performance of the candidate solutions is verified. If the preset target is met, the solution is output. Otherwise, the result is backfilled into step 4 to update the dataset and neural network proxy model. The iteration continues until the target is met. Specifically, the heat transfer performance of the optimal candidate structure that has passed the verification is verified. The verification content includes, but is not limited to: hot spot temperature rise suppression capability, interface thermal resistance reduction, robustness under different temperatures / operating conditions, and sensitivity analysis of structural parameters to performance. If the preset target is not met, the verification results will be backfilled into the dataset and the surrogate model will be updated to continue the iteration; if the target is met, the optimized micro / nano structure morphology and size scheme and its corresponding thermal performance index will be output.