Simulation design methods, devices, electronic equipment, media and products for gas chambers

By combining photonic Monte Carlo, direct simulated Monte Carlo, and thermal Monte Carlo methods, the gas chamber design was optimized, and the multi-physics coupling effect of the non-dispersive infrared cleaning endpoint detection device in semiconductor manufacturing was solved, achieving high-precision cleaning process endpoint detection.

CN122839467BActive Publication Date: 2026-11-10SHANGHAI CHEYITIAN TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202611318913.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-08-28
Publication Date
2026-11-10
Estimated Expiration
2046-08-28

AI Technical Summary

Technical Problem

In existing technologies, non-dispersive infrared cleaning endpoint detection devices in semiconductor manufacturing suffer from problems such as the coupling effect of infrared radiation, flow, and heat transfer within the gas chamber on detection accuracy, calculation failures caused by gas flow state transitions, and changes in gas concentration distribution during the cleaning process, making it difficult to achieve accurate cleaning process endpoint detection.

Method used

A simulation design method combining photonic Monte Carlo, direct simulation Monte Carlo, and thermal Monte Carlo methods is adopted to simulate infrared radiation, gas flow, and heat exchange. By iteratively optimizing the geometric and material parameters of the gas chamber, the joint solution of multi-physics fields is achieved, and the calibration coefficients are corrected in real time by combining a neural network compensation model.

Benefits of technology

It improves the accuracy and stability of measurements in cleaning processes, overcomes the shortcomings of traditional layered optimization, realizes multi-objective collaborative design of sensor cavities, and enhances measurement accuracy and computational efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122839467B_ABST
    Figure CN122839467B_ABST
Patent Text Reader

Abstract

This invention relates to the field of semiconductor detection technology, and discloses a simulation design method, apparatus, electronic device, medium, and product for a gas chamber. The method includes: constructing a three-dimensional geometric model and dividing it into meshes to configure initial material parameters to obtain a three-dimensional simulation model; simulating infrared radiation transmission using the photonic Monte Carlo method; simulating gas molecule motion using the direct simulation Monte Carlo method; simulating heat exchange between the inner wall of the three-dimensional simulation model and the gas using the thermal Monte Carlo method; iteratively coupling the results of photonic Monte Carlo, direct simulation Monte Carlo, and thermal Monte Carlo until the change in infrared signal intensity obtained after two adjacent iterations is less than or equal to a preset change, with the optimization objectives being to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the concentration field optical detection cross-section. This invention optimizes the design parameters of the gas chamber while considering the nonlinear interactions between the three fields.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of semiconductor testing technology, and specifically to a simulation design method, apparatus, electronic device, medium, and product for a gas chamber. Background Technology

[0002] In the semiconductor manufacturing industry, reactive cleaning gases nitrogen trifluoride (NF3) and chlorine trifluoride (ClF3) are widely used in in-situ cleaning processes for chambers in chemical vapor deposition (CVD) and plasma-enhanced chemical vapor deposition (PECVD). The reactive cleaning gases decompose through plasma or heating to generate fluorine radicals, which react with the silicon-containing coating deposited on the inner wall of the chamber to produce gaseous silicon tetrafluoride (SiF4) as a byproduct, thereby achieving chamber cleaning.

[0003] In cleaning processes, accurately detecting the concentrations of nitrogen trifluoride and chlorine trifluoride, as well as the formation rate of silicon tetrafluoride byproducts, is crucial for optimizing cleaning efficiency, reducing gas consumption, and ensuring process safety. Currently, the concentrations of gases such as nitrogen trifluoride, chlorine trifluoride, and silicon tetrafluoride can be detected using non-dispersive infrared (NDIR) cleaning endpoint detection devices. However, NDIR cleaning endpoint detection devices have the following three problems in their application scenarios in cleaning processes.

[0004] The first issue is the inability to accurately characterize the strong coupling effect of the three physical fields—radiation, flow, and heat transfer—within the gas chamber of the non-dispersive infrared clean endpoint detection device. This affects the simulation design of the gas chamber, and consequently, the accuracy of the concentration detection by the non-dispersive infrared clean endpoint detection device. Specifically, infrared radiation is absorbed and scattered when it propagates in the gas medium, and the absorbed energy is converted into heat; the flow and diffusion of gas molecules affect the concentration distribution; heat exchange occurs between the inner wall of the gas chamber and the gas, generating a temperature gradient; and temperature changes, in turn, affect the infrared absorption intensity by altering the molecular number density and absorption cross-section.

[0005] In related technologies, when simulating the design of gas chambers, the aforementioned physical fields are typically treated separately. Optical design uses ray tracing methods (such as Zemax and TracePro) to optimize the optical path, fluid design uses computational fluid dynamics (CFD) methods (such as ANSYS Fluent) to calculate the flow and concentration fields, and thermal design uses finite element analysis (FEA) methods (such as COMSOL Multiphysics) to analyze the thermal field distribution. This separate-field design method struggles to capture the impact of inter-field interactions on measurement accuracy, neglecting the potential design deviations caused by coupling effects, which in turn affect the sensor's measurement accuracy under dynamic process conditions.

[0006] The second issue is that under actual semiconductor cleaning process conditions, the gas flow state inside the gas chamber may be in a transitional flow regime from rarefied flow to continuous flow, leading to computational failures. When the pressure in the gas chamber drops to several thousand Pascals or even several hundred Pascals, the continuous medium assumption fails, and traditional computational fluid dynamics methods based on the Navier-Stokes equations cannot converge. Although the Direct Simulation Monte Carlo (DSMC) method can solve rarefied gas dynamics problems starting from the Boltzmann equations, it still suffers from the problems mentioned in the first aspect.

[0007] The third aspect concerns the problem of how the chemical reactions of nitrogen trifluoride and chlorine trifluoride in the cleaning process alter the local distribution of gas concentrations. Gas-phase decomposition and surface fluorination not only cause non-uniform consumption along the optical path but also change the uniformity of the detected component at the optical detection cross-section, thus directly introducing measurement bias. Summary of the Invention

[0008] This invention provides a simulation design method, device, electronic equipment, medium, and product for gas chambers, to solve the problem that related technologies lack multi-physics field coupling simulation of gas infrared radiation absorption, gas flow heat transfer, and heat conduction on the inner wall of the chamber, which makes it difficult to ensure the accuracy of cleaning process endpoint detection due to the nonlinear mutual influence between them.

[0009] In a first aspect, the present invention provides a simulation design method for a gas chamber, the method comprising: constructing a three-dimensional geometric model of the gas chamber based on initial geometric parameters; meshing the three-dimensional geometric model and configuring the meshed three-dimensional geometric model based on initial material parameters to obtain a three-dimensional simulation model; simulating the infrared radiation of an infrared light signal emitted by an infrared source through the three-dimensional simulation model to a detector using the photon Monte Carlo method, obtaining the current photon absorption energy and the current infrared signal intensity received by the detector at the current iteration number, wherein the wavelength of the light signal is divided into multiple spectral bands, each corresponding to a current absorption coefficient; simulating the motion of gas molecules within the three-dimensional simulation model under the pressure of the gas chamber using the direct simulation Monte Carlo method, obtaining the current gas concentration field at the current iteration number; and simulating the motion of gas molecules within the three-dimensional simulation model using the thermal Monte Carlo method. The model simulates the heat exchange between the inner wall and the gas in a 3D model to obtain the current temperature field at the current iteration number. If the change between the current infrared signal intensity and the previous infrared signal intensity is greater than a preset change, the current photon absorption energy is used as the heat source for the next iteration to obtain the next temperature field. The next temperature field is used as the boundary condition for the next iteration to obtain the next gas concentration field. The current absorption coefficient is corrected based on the next gas concentration field and the next temperature field, and the next iteration is performed according to the corrected current absorption coefficient to obtain the next infrared signal intensity. If the change is less than or equal to the preset change, the optimization objectives are to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the optical detection section. The initial geometric parameters and / or initial material parameters are adjusted through an optimization algorithm to simulate the target geometric parameters and target material parameters of the gas chamber.

[0010] In an optional implementation, when the current iteration number is the first iteration, the simulation design method further includes: determining the current absorption coefficient based on a preset initial uniform concentration field, a preset initial temperature field, and a first correspondence, wherein the first correspondence is used to characterize the mapping relationship between the concentration field, the temperature field, and the absorption coefficient.

[0011] In one optional implementation, the gas in the three-dimensional simulation model is a reactive clean gas; when simulating the motion of gas molecules in the three-dimensional simulation model by directly simulating the Monte Carlo method, the method further includes: simulating the chemical reaction process of the reactive clean gas by the Monte Carlo method of gas chemical reaction to obtain the current gas concentration field at the current iteration number.

[0012] In one optional implementation, the chemical reaction process of reactive clean gases is simulated using the Monte Carlo method for gas chemical reactions. This includes: within the same gas molecule motion and collision simulation framework, assigning a class identifier, velocity vector, quantity, and internal energy to each gas molecule; and during gas molecule collision processing, determining the probability of a chemical reaction occurring between gas molecules according to the Arrhenius equation, the expression of which is: ; In the formula, Indicates the probability of reaction. Indicates pre-exponential factor, Indicates the activation energy of the reaction. Represents Boltzmann's constant. , Indicates temperature. The collision time step is indicated; based on the reaction probability and the generated random numbers uniformly distributed between 0 and 1, it is determined whether a chemical reaction occurs; if a chemical reaction occurs, the type identifier, quantity, velocity vector, and internal energy of the gas molecules participating in the chemical reaction are updated to obtain the updated type identifier, quantity, velocity vector, and internal energy, and the heat released or absorbed by the chemical reaction is recorded. The updated type identifier, quantity, velocity vector, and internal energy are used to determine the current gas concentration field; the current photon absorbed energy is used as a heat source to perform the next iteration to obtain the next temperature field, including: using the current photon absorbed energy and heat as a heat source to perform the next iteration to obtain the next temperature field.

[0013] In one alternative implementation, determining whether a chemical reaction has occurred based on the reaction probability and a generated random number uniformly distributed between 0 and 1 includes: if the random number is less than the reaction probability, then determining that a chemical reaction has occurred; if the random number is greater than or equal to the reaction probability, then determining that a chemical reaction has not occurred.

[0014] In one optional implementation, the optical signal includes a target optical signal of a target wavelength and a reference optical signal of a reference wavelength. The method further includes: constructing a neural network compensation model and configuring the neural network compensation model in a clean processor. The inputs of the neural network compensation model include time, cavity pressure, target signal intensity corresponding to the target optical signal, reference optical intensity corresponding to the reference optical signal, and generation rate of chemical products formed after the chemical reaction. The output of the neural network compensation model is a correction coefficient for the measured concentration.

[0015] In one optional implementation, the infrared radiation of an infrared light signal emitted from an infrared source reaching a detector via a three-dimensional simulation model is simulated using the photonic Monte Carlo method. This includes: discretizing the wavelength of the light signal into multiple spectral bands and configuring a current absorption coefficient for each band; emitting the light signal and tracing the absorption and scattering paths of each photon in each spectral band within the gas medium until the photon reaches the detector or escapes from the gas chamber simulated by the three-dimensional simulation model; obtaining the current infrared signal intensity by statistically analyzing the photon energies reaching the detector from multiple spectral bands; and obtaining the current photon absorption energy by statistically analyzing the photon energies absorbed by the gas medium from multiple spectral bands.

[0016] In one alternative implementation, the current absorption coefficient is determined using the following formula: ; In the formula, Indicates the first The average absorption coefficient of the first spectral band will be the first band. The average absorption coefficient of the spectral band is used as the first... The current absorption coefficient corresponding to each spectral band. Indicates the bandwidth of the spectral band. Indicating in wavenumber Spectral absorption coefficient at that location, The calculations were performed using the gas concentration and temperature obtained from the previous iteration. Let represent the Planck function.

[0017] In one optional implementation, the optimization algorithm is a single-objective optimization algorithm, with the optimization objectives being to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the optical detection cross section. The optimization algorithm adjusts the initial geometric parameters and / or initial material parameters, including: normalizing the infrared signal intensity and the concentration field variation coefficient to obtain normalized infrared signal intensity and normalized concentration field variation coefficient; constructing a fitness function based on the normalized infrared signal intensity and normalized concentration field variation coefficient; and adjusting the initial geometric parameters and / or initial material parameters according to the fitness function and the single-objective optimization algorithm to obtain the target geometric parameters and target material parameters.

[0018] In one alternative implementation, the single-objective optimization algorithm is a Bayesian optimization algorithm, a genetic algorithm, or a particle swarm optimization algorithm.

[0019] In one optional implementation, the optimization algorithm is a multi-objective optimization algorithm, with the optimization objectives being to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the optical detection cross section. The initial geometric parameters and / or initial material parameters are adjusted through the optimization algorithm, including: adjusting the initial geometric parameters and / or initial material parameters through the multi-objective optimization algorithm with the optimization objectives of maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection cross section to obtain the Pareto front; and selecting target geometric parameters and target material parameters from the Pareto front according to the fitness function.

[0020] In one optional embodiment, the initial geometric parameters include at least one of the following: cavity length, cavity diameter, infrared light source, relative position between the gas chamber and the detector, reflective surface curvature, inlet position, inlet aperture, outlet position, and outlet aperture; the initial material parameters include at least one of the following: infrared reflectivity, absorptivity, thermal conductivity, specific heat capacity, and surface emissivity of the inner wall coating material.

[0021] In one alternative implementation, the pressure in the gas chamber ranges from 1 Torr to 760 Torr.

[0022] In a second aspect, the present invention provides a cleaning endpoint detection device, comprising a gas chamber designed by a simulation design method of a gas chamber according to the first aspect or any of its corresponding embodiments.

[0023] Thirdly, this invention provides a simulation design device for a gas chamber. The device includes: a model building module for constructing a three-dimensional geometric model of the gas chamber based on initial geometric parameters; a mesh generation module for meshing the three-dimensional geometric model and configuring the meshed three-dimensional geometric model based on initial material parameters to obtain a three-dimensional simulation model; a photon Monte Carlo module for simulating the infrared radiation of light signals emitted by an infrared source reaching the detector through the three-dimensional simulation model using the photon Monte Carlo method, obtaining the current photon absorption energy and the current infrared signal intensity received by the detector at the current iteration number, wherein the light signal band is divided into multiple spectral bands, and each spectral band corresponds to a current absorption coefficient; a direct simulation Monte Carlo module for simulating the motion of gas molecules within the three-dimensional simulation model using the direct simulation Monte Carlo method, obtaining the current gas concentration field at the current iteration number; and a thermal Monte Carlo module for... The overheated Monte Carlo method simulates the heat exchange between the inner wall of the 3D simulation model and the gas, obtaining the current temperature field at the current iteration number. The iteration module, if the change between the current infrared signal intensity and the previous infrared signal intensity is greater than a preset change, uses the current photon absorption energy as a heat source for the next iteration to obtain the next temperature field. It then uses the next temperature field as a boundary condition for the next iteration to obtain the next gas concentration field. Based on the next gas concentration field and the next temperature field, it corrects the current absorption coefficient and performs the next iteration according to the corrected current absorption coefficient to obtain the next infrared signal intensity. The optimization design module, if the change is less than or equal to a preset change, stops the iteration and uses the optimization objectives of maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection section. It adjusts the initial geometric parameters and / or initial material parameters through an optimization algorithm to simulate the target geometric parameters and target material parameters of the gas chamber.

[0024] Fourthly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the simulation design method for a gas chamber according to the first aspect or any corresponding embodiment described above.

[0025] Fifthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the simulation design method for a gas chamber according to the first aspect or any corresponding embodiment thereof.

[0026] In a sixth aspect, the present invention provides a computer program product, including computer instructions for causing a computer to execute the simulation design method for a gas chamber according to the first aspect or any corresponding embodiment described above.

[0027] The simulation design method, apparatus, electronic equipment, medium, and product for gas chambers provided by this invention have at least the following advantages: This invention achieves joint solutions for the light field, flow field, and temperature field by simultaneously applying three Monte Carlo methods—photon Monte Carlo, direct simulation Monte Carlo, and thermal Monte Carlo—to the simulation design of a gas chamber. Through a weighted, alternating iterative strategy, the nonlinear interactions between the three fields are accurately simulated without introducing artificially simplified boundary conditions. This allows for optimization of the gas chamber's design parameters while considering these nonlinear interactions, thereby improving measurement accuracy.

[0028] By embedding the gas chemical reaction Monte Carlo module into the direct simulation Monte Carlo module, physical collisions and chemical reactions are handled simultaneously in the same molecular motion simulation. For reactive cleaning gases in semiconductor cleaning processes, accurate simulations of gas-phase decomposition reactions and surface fluorination reactions are achieved, and the nonlinear perturbations of chemical reactions on gas concentration distribution can be predicted.

[0029] By directly simulating the gas concentration field output by the Monte Carlo module to update the absorption coefficient, and feeding the absorption coefficient back to the photonic Monte Carlo module, a closed-loop coupling between chemistry, radiation, flow, and heat transfer is achieved. This enables the simulation model to realistically reflect the non-uniform absorption on the optical path caused by reaction consumption in the cleaning process, providing a theoretical basis for improving measurement accuracy.

[0030] By using a neural network compensation model to correct the calibration coefficients in real time, the measurement accuracy and stability under dynamic cleaning process conditions are improved.

[0031] With the dual optimization objectives of maximizing infrared signal intensity and minimizing the coefficient of variation of the concentration field, the cavity geometry and material parameters are optimized simultaneously. This overcomes the shortcomings of traditional layered optimization, which struggles to balance optical efficiency and concentration uniformity, and achieves multi-objective collaborative design of the sensor cavity.

[0032] Compared to multi-objective optimization, converting multiple objectives into a single objective through weighted summation before parameter optimization eliminates the need to maintain a large number of non-dominated candidate solutions simultaneously, requiring fewer samples and effectively reducing overall simulation computational overhead and improving solution speed. Adjusting the weighting coefficients allows for intuitive control over optimization bias, selectively emphasizing either infrared signal intensity or concentration field uniformity based on process requirements. Attached Figure Description

[0033] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0034] Figure 1 This is a flowchart illustrating a simulation design method for a gas chamber according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a gas chamber according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the coupling and interaction of three simulation modules according to an embodiment of the present invention; Figure 4 This is a flowchart illustrating another simulation design method for a gas chamber according to an embodiment of the present invention; Figure 5 This is a structural block diagram of a simulation design device for a gas chamber according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0035] 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, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0037] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0038] This invention provides a simulation design method, device, electronic equipment, medium, and product for a gas chamber. It simultaneously applies three Monte Carlo methods—Photon Monte Carlo (PMC), Direct Simulation Monte Carlo, and Thermal Monte Carlo (TMC)—to design the gas chamber, achieving joint solutions for the light field, flow field, and temperature field. Through a weighted, alternating iterative strategy, the nonlinear interactions between the three fields are accurately simulated without introducing artificially simplified boundary conditions. Considering these nonlinear interactions, the design parameters (geometric and / or material parameters) of the gas chamber are optimized, improving measurement accuracy.

[0039] The simulation design method for the gas chamber provided by this invention will be described in detail below with reference to the accompanying drawings. The gas chamber can be the gas chamber in a non-dispersive infrared cleaning endpoint detection device.

[0040] Figure 1 This is a flowchart illustrating a simulation design method for a gas chamber according to an embodiment of the present invention. It can be used in electronic devices, which include simulation software running the photonic Monte Carlo method (denoted as the photonic Monte Carlo module), simulation software running the direct simulation Monte Carlo method (denoted as the direct simulation Monte Carlo module), and simulation software running the thermal Monte Carlo method (denoted as the thermal Monte Carlo module). Figure 1 As shown, the process includes the following steps: Step S101: Based on the initial geometric parameters, construct a three-dimensional geometric model of the gas chamber.

[0041] The initial geometric parameters include at least one of the following: cavity length, cavity diameter, infrared light source, relative position between the gas chamber and the detector, reflective surface curvature, inlet position, inlet aperture, outlet position, and outlet aperture. Initial geometric parameters refer to the range of geometric parameters initially determined based on processing requirements.

[0042] For example, a three-dimensional geometric model of the gas chamber can be constructed based on initial geometric parameters in three-dimensional design software (such as SolidWorks) and then imported into simulation software; alternatively, a three-dimensional geometric model of the gas chamber can be constructed directly in the modeling module of the simulation software.

[0043] In one example, the structure of the gas chamber can be as follows: Figure 2As shown, the gas chamber 10 includes a columnar body 11. The body 11 has a receiving cavity 111, a first opening 112 and a second opening 113. The first opening 112 and the second opening 113 are arranged opposite to each other and communicate with the receiving cavity 111. One of the first opening 112 and the second opening 113 is an input port for optical signals, and the other opening is an output port for optical signals. Figure 2 Taking the first opening 112 as the input port of the optical signal and the second opening 113 as the output port of the optical signal as an example.

[0044] The side wall of the main body 11 is also provided with an air inlet 121 and an air outlet 131 that communicate with the receiving cavity 111. The air inlet 121 is the input port of the gas to be measured, and the air outlet 131 is the output port of the gas to be measured. The air inlet 121 is located near the input port of the optical signal, and the air outlet 131 is located near the output port of the optical signal.

[0045] Furthermore, such as Figure 2 As shown, the gas chamber also includes two extensions (first extension 12 and second extension 13) for easy connection to a gas transmission device. The inlet 121 is located in the first extension 12 and communicates with the receiving cavity 111, and the outlet 131 is located in the second extension 13 and communicates with the receiving cavity 111.

[0046] Specifically, see Figure 2 The cavity length can be the length L of the cavity 111 in the optical transmission direction, and the cavity diameter D can be the inner diameter of the cavity 111.

[0047] Step S102: Mesh the three-dimensional geometric model and configure the meshed three-dimensional geometric model based on the initial material parameters to obtain the three-dimensional simulation model.

[0048] The initial material parameters include at least one of the following: infrared reflectance, absorptivity, thermal conductivity, specific heat capacity, and surface emissivity of the inner wall coating material. Initial material parameters refer to the range of material parameters initially determined based on processing requirements.

[0049] After meshing the 3D geometric model, initial material parameters are input through a human-computer interaction device to assign material properties to the meshed 3D geometric model. The 3D geometric model that has been meshed and assigned material properties is the 3D simulation model.

[0050] Specifically, the three-dimensional geometric model is spatially meshed in the simulation software. The mesh can be an unstructured tetrahedral mesh. The mesh can be uniformly or non-uniformly set. For example, local meshing can be performed in areas with large changes in physical quantity gradients, such as the end facing the infrared light source, the end facing the detector, the air inlet, and the air outlet, to provide discretization space for subsequent multiphysics simulations.

[0051] Step S103: Using the photon Monte Carlo method, simulate the infrared radiation of the light signal emitted by the infrared source reaching the detector through the three-dimensional simulation model, and obtain the current photon absorption energy and the current infrared signal intensity received by the detector at the current iteration number.

[0052] The photon Monte Carlo method is a computational optics tool that uses statistical methods to simulate the propagation, scattering, and absorption behavior of photons in complex media. Electronic devices, through a photon Monte Carlo module, simulate the infrared radiation of light signals emitted from an infrared source as they travel through a three-dimensional simulation model to the detector, determining the photon absorption energy and infrared signal intensity after each iteration.

[0053] For example, step S103 above may include the following steps: Step a1: Discretize the optical signal into multiple spectral bands and configure the current absorption coefficient for each spectral band.

[0054] Step a2: emit an optical signal and trace the absorption and scattering paths of each photon in the gas medium in each spectral band until the photon reaches the detector or escapes the gas chamber simulated by the three-dimensional simulation model.

[0055] Step a3: The current infrared signal intensity is obtained by statistically analyzing the photon energies arriving at the detector from multiple spectral bands.

[0056] Step a4: Obtain the current photon absorption energy by statistically analyzing the photon energies absorbed by the gas medium across multiple spectral bands.

[0057] The optical signal is divided into multiple spectral bands, each corresponding to a current absorption coefficient. For example, if the optical signal ranges from 3.5 μm to 4.5 μm, it can be divided into 10 spectral bands, each with a bandwidth of 0.1 μm. The absorption coefficient is an input parameter of the photon Monte Carlo module, determining the probability of photons being absorbed by the gas within the cavity, their free movement distance, and serving as the basis for calculating infrared signal intensity. Optical signals can be emitted in Lambertian mode, with each spectral band operating independently for photon tracking. Each band uses its configured current absorption coefficient, while cavity geometry parameters and wall reflection properties are shared across all bands.

[0058] Specifically, infrared radiation is viewed as a random walk of a large number of photons. The absorption and scattering path of each photon after emission from the infrared source in the gas medium is tracked until it reaches the detector or escapes the gas chamber. By statistically analyzing the photon energy reaching the detector, the photon absorption energy in the gas chamber simulated by the three-dimensional simulation model and the intensity of the infrared signal received by the detector are obtained.

[0059] The simulation uses a non-grey gas radiation model. Based on the characteristic absorption lines of the gas to be measured (such as reactive cleaning gas) in the infrared band, the spectrum is divided into multiple bands. The current absorption coefficient of each band is calculated based on the gas concentration and temperature obtained in the previous iteration.

[0060] For example, the relationship between gas concentration, temperature, and absorption coefficient can be shown as in equation (1):

[0061] In the formula, Indicates temperature and concentration Absorption coefficient and concentration under the given conditions To directly simulate the concentration output by the Monte Carlo module, the concentration is expressed as molecular number density (molecules / cm²). 3 The value is expressed as either a fraction (ppm) or a volume fraction, depending on the temperature. The temperature output by the thermal Monte Carlo module. Indicates the absorption cross section. Denotes Avogadro's constant. .

[0062] Absorption cross section characterizes the wavenumber of a single gas molecule. The effective area for absorbing infrared radiation is expressed in cm². 2 / molecule. The absorption cross section is a function of temperature. As the temperature increases, the molecular energy level population changes (broadening and intensity changes of rotational-vibrational spectral lines), and the absorption cross section changes accordingly. This invention utilizes line-by-line data from the HITRAN2016 database to achieve absorption cross section values ​​at arbitrary temperatures through cubic spline interpolation.

[0063] Furthermore, in the non-gray gas radiation model, the actual absorption coefficient varies drastically with wavelength, as shown in formula (2). In this invention, the average absorption coefficient obtained by weighting by Planck function is used as the absorption coefficient corresponding to the spectral band, which more accurately reflects the radiation absorption process under actual light source conditions.

[0064]

[0065] In the formula, Indicates the first The average absorption coefficient of each spectral band Indicates the bandwidth of the spectral band. Indicating in wavenumber Spectral absorption coefficient at that location, Equivalent to formula (1) , Let represent the Planck function.

[0066] Spectral absorption coefficient Determined by both gas concentration and temperature, it is a function of wavelength (or wavenumber). The underlying data comes from line-by-line absorption cross-sections in the HITRAN2016 database, reflecting the absorption intensity of infrared radiation by gas molecules at a specific wavenumber. Planck function. Indicates temperature Below, blackbody radiation at wavenumber The spectral radiant intensity at a given location. It should be understood that the radiant energy of an infrared source is not uniformly distributed across different wavelengths. The more concentrated the energy in the emission spectrum of the source, the greater its contribution to the detector signal. Therefore, it should be given higher weight when calculating the average absorption coefficient of the spectral band.

[0067] Step S104: By directly simulating the Monte Carlo method, the motion of gas molecules in the three-dimensional simulation model is simulated under the pressure of the gas chamber to obtain the current gas concentration field at the current iteration number.

[0068] The pressure of the gas chamber is set based on the actual situation of the cleaning process, and the pressure of the gas chamber can range from 1 Torr to 760 Torr. The gas concentration field represents the concentration corresponding to each position coordinate of the gas chamber after the current iteration.

[0069] The direct simulation Monte Carlo method is a particle statistical method for simulating rarefied gas flows, primarily applied to high Knudsen number flow fields where the continuum assumption fails. The core principle of this algorithm is to decouple the free motion of particles from their collisions within extremely small time intervals, requiring a mesh size smaller than the mean free path and a time step shorter than the mean collision time. Electronic devices, through the direct simulation Monte Carlo module, simulate the motion of gas molecules within a three-dimensional simulation model under pressure in a gas chamber, determining the gas concentration field after each iteration.

[0070] Specifically, within the direct Monte Carlo simulation framework, the movement, collisions, and interactions of a large number of gas molecules with the walls are simulated to obtain the gas concentration field of the gas chamber simulated by the three-dimensional simulation model.

[0071] Step S105: Using the thermal Monte Carlo method, simulate the heat exchange between the inner wall of the three-dimensional simulation model and the gas to obtain the current temperature field at the current iteration number.

[0072] Here, the current temperature field represents the temperature value corresponding to the coordinates of each position in the gas chamber after the current iteration. The electronic device uses a thermal Monte Carlo module to simulate the heat exchange between the inner wall of the three-dimensional simulation model and the gas, and determines the temperature field after each iteration.

[0073] Specifically, the heat is discretized into a large number of heat packets, and the energy equation is solved by simulating the random walk of these heat packets in a solid mesh. The heat packets diffuse within the cavity's inner wall material and exchange energy according to the heat transfer conditions between the wall and the environment. By statistically analyzing the total energy within each mesh cell, the temperature field distribution within the gas cavity simulated by the three-dimensional simulation model is obtained.

[0074] Step S106: If the change between the current infrared signal intensity and the previous infrared signal intensity is greater than the preset change, the current photon absorption energy is used as the heat source to perform the next iteration to obtain the next temperature field. The next temperature field is used as the boundary condition to perform the next iteration to obtain the next gas concentration field. The current absorption coefficient is corrected based on the next gas concentration field and the next temperature field. The next iteration is performed according to the corrected current absorption coefficient to obtain the next infrared signal intensity.

[0075] The change in the current infrared signal strength and the previous infrared signal strength can be determined by the following formula (3), and the preset change can be 1%.

[0076]

[0077] In the formula, Indicates the amount of change. Indicates the number of iterations. Indicates the first Infrared signal strength after the next iteration Indicates the first Infrared signal strength after the next iteration.

[0078] In the When the next iteration is the current iteration number... Indicates the current infrared signal strength. Indicates the intensity of the previous infrared signal. This indicates the intensity of the next infrared signal.

[0079] The current infrared signal intensity represents the infrared signal intensity obtained in the current iteration, the previous infrared signal intensity represents the infrared signal intensity obtained after the previous iteration, the next infrared signal intensity represents the infrared signal intensity obtained after the next iteration, the next temperature field represents the temperature field obtained after the next iteration, and the laundry gas concentration field represents the gas concentration field obtained after the next iteration.

[0080] The current iteration number, the previous iteration, and the next iteration are relative. If the current iteration number, the previous iteration, and the next iteration are relative, then the current iteration number, the previous iteration, and the next iteration are relative. If the iteration number is the current iteration number, then the... The next iteration is the previous iteration, the th... The next iteration is the next iteration; if the... If the iteration number is the current iteration number, then the... The next iteration is the previous iteration, the th... The next iteration is the next iteration.

[0081] In the Monte Carlo simulation process, the number of iterations is the number of times the three simulation modules (photon Monte Carlo module, direct simulation Monte Carlo module, and thermal Monte Carlo module) are coupled. The increase in the number of iterations also represents the increase in the time step, which is a time-series calculation of the evolution of photon absorption energy, concentration, and temperature over time.

[0082] At the initial moment of the simulation (k=0), the molecular distribution and temperature are in the preset initial state. The gas flow, component diffusion and chemical reaction inside the cavity need to evolve over a period of time to reach dynamic equilibrium. The simulation continues to iterate until the steady-state convergence criterion is met, thus eliminating the transient transition effect.

[0083] In this embodiment, if the change in the current infrared signal intensity and the previous infrared signal intensity is greater than a preset change, it indicates that the iteration has not yet converged and needs to continue. During the iteration process, if... Figure 3 As shown, the electronic device uses the current photon absorption energy output by the photon Monte Carlo module as a volumetric heat source, inputs it into the thermal Monte Carlo module, and iterates a certain number of times. Updated to The thermal Monte Carlo module, based on the condition of the heat source, re-simulates the heat exchange between the inner wall of the 3D simulation model and the gas within a preset time period, and outputs the next temperature field.

[0084] The electronic device uses the next temperature field output by the thermal Monte Carlo module as a boundary condition and inputs it into the direct simulation Monte Carlo module. Based on the next temperature field, the direct simulation Monte Carlo module re-simulates the motion of gas molecules in the three-dimensional simulation model within a preset time period under the pressure of the gas chamber, and outputs the next gas concentration field. After determining the next gas concentration field and the next temperature field, the electronic device can obtain the corrected current absorption coefficient based on the above formulas (1) and (2), and inputs the corrected current absorption coefficient into the photon Monte Carlo module to re-simulate the infrared radiation of the light signal emitted by the infrared source through the three-dimensional simulation model to the detector, and outputs the next infrared signal intensity.

[0085] After obtaining the next infrared signal intensity, use the next infrared signal intensity as the new current infrared signal intensity and execute step S106 or step S107.

[0086] For example, when the current iteration number is the first iteration, the simulation design method further includes: determining the current absorption coefficient based on the preset initial uniform concentration field, the preset initial temperature field and the first correspondence.

[0087] Among them, the initial uniform concentration field and the initial temperature field are preset values, which can be set according to the clean actual measurement environment. The first correspondence is used to characterize the mapping relationship between the concentration field, the temperature field and the absorption coefficient. The first correspondence can be as shown in formula (1), or the first correspondence can be as shown in formula (1) and formula (2).

[0088] Step S107: If the change is less than or equal to the preset change, the optimization objectives are to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the optical detection section. The initial geometric parameters and / or initial material parameters are adjusted by the optimization algorithm to simulate and obtain the target geometric parameters and target material parameters of the gas chamber.

[0089] Among them, the optical detection cross section is a specific cross section of the detector used to receive optical signals, and the concentration field variation coefficient is used to reflect the uniformity of the concentration field. The concentration field variation coefficient can be determined by the following formula (4).

[0090]

[0091] In the formula, This represents the concentration field variation coefficient. This represents the average gas concentration across the optical detection cross section. The standard deviation of the gas concentration on the optical detection cross section reflects the dispersion of the concentration value of each grid cell on the optical detection cross section relative to the average concentration. The smaller the concentration field variation coefficient, the more uniform the concentration field.

[0092] If the change is less than or equal to the preset change, it indicates that the iteration has converged, and the iteration stops. Then, with the optimization objectives of maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection section, the initial geometric parameters and / or initial material parameters are adjusted through an optimization algorithm. For example, the optimization algorithm can be a multi-objective genetic algorithm or a multi-objective particle swarm optimization algorithm, etc.

[0093] When optimizing only the initial geometric parameters, the target geometric parameters are the adjusted initial geometric parameters that satisfy both optimization objectives, and the initial material parameters are the target material parameters; when optimizing only the initial material parameters, the target material parameters are the adjusted initial material parameters that satisfy both optimization objectives, and the initial geometric parameters are the target geometric parameters; when optimizing both the initial geometric parameters and the initial material parameters simultaneously, the target geometric parameters are the adjusted initial geometric parameters that satisfy both optimization objectives, and the target material parameters are the adjusted initial material parameters that satisfy both optimization objectives.

[0094] The simulation design method for the gas chamber provided in this embodiment establishes a three-dimensional geometric model based on initial geometric parameters, and then configures the meshed three-dimensional geometric model based on initial material parameters to obtain a three-dimensional simulation model. The photonic Monte Carlo method is used to simulate infrared radiation transmission, obtaining the spatial energy distribution and the detector's infrared signal intensity. The direct simulation Monte Carlo method is used to simulate the motion of gas molecules, obtaining the gas concentration field. The thermal Monte Carlo method is used to simulate the heat exchange between the inner wall of the chamber and the gas, obtaining the temperature field. Then, the results of photonic Monte Carlo, direct simulation Monte Carlo, and thermal Monte Carlo are alternately and iteratively coupled until the change in infrared signal intensity obtained after two adjacent iterations is less than or equal to a preset change. When the change is less than or equal to the preset change, the initial geometric parameters and / or initial material parameters are adjusted with the dual optimization objectives of maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the concentration field optical detection section, to obtain the target geometric parameters and target material parameters.

[0095] This invention simultaneously applies three Monte Carlo methods—photon Monte Carlo, direct simulation Monte Carlo, and thermal Monte Carlo—to the simulation design of gas chambers, achieving joint solutions for the light field, flow field, and temperature field. Through a weighted, alternating iterative strategy, the nonlinear interactions between the three fields are accurately simulated without introducing artificially simplified boundary conditions. This allows for optimization of the gas chamber design parameters while considering the nonlinear interactions between the three fields, thereby improving measurement accuracy.

[0096] Furthermore, this invention introduces a Chemical Reaction Monte Carlo (CRMC) module into the direct simulation Monte Carlo module for monitoring reactive cleaning gases in semiconductor cleaning processes, while simultaneously predicting the perturbation of gas concentration distribution by chemical reactions.

[0097] Figure 4 This is a flowchart illustrating another simulation design method for a gas chamber according to an embodiment of the present invention, which can be used in electronic devices, including photonic Monte Carlo modules, direct simulation Monte Carlo modules, and thermal Monte Carlo modules. Figure 4 As shown, the process includes the following steps: Step S401: Based on the initial geometric parameters, construct a three-dimensional geometric model of the gas chamber.

[0098] Please see details Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0099] Step S402: Mesh the three-dimensional geometric model and configure the meshed three-dimensional geometric model based on the initial material parameters to obtain the three-dimensional simulation model.

[0100] Please see details Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0101] Step S403: Using the photon Monte Carlo method, simulate the infrared radiation of the light signal emitted by the infrared source reaching the detector through the three-dimensional simulation model, and obtain the current photon absorption energy and the current infrared signal intensity received by the detector at the current iteration number.

[0102] Please see details Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0103] Step S404: Simulate the motion of gas molecules within the three-dimensional simulation model using the direct Monte Carlo simulation method, and simulate the chemical reaction process of reactive clean gas using the gas chemical reaction Monte Carlo method, to obtain the current gas concentration field at the current iteration number.

[0104] Specifically, the direct simulation Monte Carlo method is used to simulate the motion of gas molecules, and a Monte Carlo module for gas chemical reactions is integrated to simulate the chemical reaction process between reactive clean gases and byproducts, thereby obtaining the gas concentration field.

[0105] For example, step S404 above includes steps S4041 to S4044, which can integrate the gas chemical reaction Monte Carlo module into the direct simulation Monte Carlo module, and simulate the chemical reaction process while simulating the motion of gas molecules to obtain a dynamic gas concentration field.

[0106] Step S4041: Within the same gas molecule motion and collision simulation framework, set the type identifier, velocity vector, quantity, and internal energy for each gas molecule.

[0107] Specifically, while assigning a type identifier, velocity, quantity, and internal energy to each gas molecule, at least three reaction pathways are defined: nitrogen trifluoride decomposition to generate active fluorine radicals, chlorine trifluoride reaction with silicon residue to generate silicon tetrafluoride, and adsorption and desorption of silicon tetrafluoride.

[0108] Step S4042: During the gas molecule collision treatment, the reaction probability of chemical reaction between gas molecules is determined according to the Arrhenius equation.

[0109] For example, the expression for determining the reaction probability of a chemical reaction between gas molecules according to the Arrhenius equation can be shown as equation (5):

[0110] In the formula, Indicates the probability of reaction. The pre-exponential factor can be determined based on literature data, for example, , This indicates the activation energy of the reaction, which is determined by the reaction pathway. For example, in a surface fluorination reaction where the reaction pathway is the decomposition of nitrogen trifluoride to generate active fluorine radicals... , Represents Boltzmann's constant. , The temperature is represented by a temperature field provided in real time by the thermal Monte Carlo module; for surface reactions, Take the wall temperature Rather than gas temperature, This indicates the collision time step.

[0111] When two gas molecules are selected as a collision pair, the relative kinetic energy of the two gas molecules is as shown in formula (6). The total collision energy model assumes that the vibrational energy and rotational energy of gas molecules can also contribute to overcoming the reaction energy barrier. Therefore, the total collision energy of gas molecules participating in the chemical reaction in the cavity can be determined by formula (7).

[0112]

[0113] In the formula, Represents relative kinetic energy. Represents reduced quality. , This indicates the mass of the first gas molecule. This indicates the mass of the second gas molecule. This represents the relative velocity of two gas molecules. For surface reactions (molecules colliding with a wall), the collision between the molecule and the wall is considered a virtual collision. It is determined by the average kinetic energy of the gas molecules. This represents the total collision energy. This represents the vibrational energy of the first gas molecule in the collision pair. This represents the vibrational energy of the second gas molecule in the collision pair. This represents the rotational energy of the first gas molecule in the collision pair. This represents the rotational energy of the second gas molecule in the collision pair. The total collision energy includes internal energy.

[0114] Step S4043: Determine whether a chemical reaction has occurred based on the reaction probability and the generated random numbers that are uniformly distributed between 0 and 1.

[0115] For example, if the random number is less than the reaction probability, then a chemical reaction is determined to have occurred; if the random number is greater than or equal to the reaction probability, then a chemical reaction is determined not to have occurred.

[0116] Specifically, reaction probability The value of is between 0 and 1. After determining the reaction probability, a random number is uniformly drawn from the interval [0, 1]. If random number Less than the reaction probability If the random number is 1, then a chemical reaction has occurred; if the random number is 2, then a chemical reaction has occurred. Greater than or equal to the reaction probability If no chemical reaction has occurred, it should be treated as an elastic collision.

[0117] In direct Monte Carlo simulations, most collision pairs are not selected to undergo chemical reactions because the calculated probability of chemical reactions for many collision pairs is... It is zero, or usually very small; for example, the probability of a reaction under typical conditions. Possibly 10 -6 Up to 10 -3 Order of magnitude. If a reaction is only allowed to occur when the probability of reaction is greater than 70%, then almost no reaction will occur, because the calculated reaction probability... The figure is far below 70%, which does not match the actual physical process. Using random number sampling and comparing the reaction probability to determine whether a chemical reaction has occurred is more consistent with reality.

[0118] In step S4044, if a chemical reaction occurs, the type identifier, quantity, velocity vector, and internal energy of the gas molecules participating in the chemical reaction are updated to obtain the updated type identifier, quantity, velocity vector, and internal energy, and the heat released or absorbed by the chemical reaction is recorded.

[0119] The updated species identifier, quantity, velocity vector, and internal energy are used to determine the current gas concentration field.

[0120] This method is applicable to a wide pressure range from rarefied flow to continuous flow, and can obtain the dynamic distribution of velocity field, pressure field, reactive clean gas consumption concentration, and reaction by-product formation concentration under different pressures.

[0121] Step S405: Using the thermal Monte Carlo method, simulate the heat exchange between the inner wall of the three-dimensional simulation model and the gas to obtain the current temperature field at the current iteration number.

[0122] Please see details Figure 1 Step S105 of the illustrated embodiment will not be described again here.

[0123] Step S406: If the change between the current infrared signal intensity and the previous infrared signal intensity is greater than the preset change, the current photon absorption energy and heat are used as the heat source to perform the next iteration to obtain the next temperature field. The next temperature field is used as the boundary condition to perform the next iteration to obtain the next gas concentration field. The current absorption coefficient is corrected based on the next gas concentration field and the next temperature field. The next iteration is performed according to the corrected current absorption coefficient to obtain the next infrared signal intensity.

[0124] Specifically, this step is similar to step S106 above, except that not only is the current photon absorption energy output by the photon Monte Carlo module input to the thermal Monte Carlo module, but the heat output by the gas chemical reaction Monte Carlo module (heat of chemical reaction) is also input to the thermal Monte Carlo module. Based on the heat source and heat, the thermal Monte Carlo module re-simulates the heat exchange between the inner wall of the three-dimensional simulation model and the gas within a preset time period and outputs the next temperature field. Among them, the gas chemical reaction Monte Carlo module updates the chemical reaction rate constant based on the next temperature field, that is, in formula (5). .

[0125] In step S407, if the change is less than or equal to the preset change, the iteration stops, and the initial geometric parameters and / or initial material parameters are adjusted by an optimization algorithm with the optimization objectives of maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection section, so as to simulate and obtain the target geometric parameters and target material parameters of the gas chamber.

[0126] In some optional embodiments, the optimization algorithm is a single-objective optimization algorithm, with the optimization objectives being to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the optical detection cross section. Adjusting the initial geometric parameters and / or initial material parameters through the optimization algorithm (i.e., step S407) may include the following steps: Step b1: Normalize the infrared signal intensity and concentration field variation coefficient to obtain the normalized infrared signal intensity and the normalized concentration field variation coefficient.

[0127] Specifically, the normalized infrared signal intensity can range from 0 to 1, and the normalized concentration field variation coefficient can also range from 0 to 1.

[0128] For example, the infrared signal intensity can be normalized using the following formula (8) and the infrared signal intensity can be normalized using the following formula (9).

[0129]

[0130] In the formula, This represents the normalized infrared signal intensity. This represents the intensity of the infrared signal output by the photon Monte Carlo module after iterative convergence. This represents the lower limit of the detector signal strength within the feasible parameter domain. This represents the upper limit of the detector signal strength within the feasible parameter domain. This represents the concentration field variation coefficient after normalization. This represents the concentration field variation coefficient calculated by directly simulating the gas concentration field output by the Monte Carlo module after iterative convergence. This represents the lower limit of the concentration field variation coefficient within the feasible parameter domain. This represents the upper limit of the concentration field variation coefficient within the feasible parameter domain.

[0131] Step b2: Based on the normalized infrared signal intensity and the normalized concentration field variation coefficient, construct the fitness function.

[0132] For example, the fitness function can be expressed as shown in formula (10):

[0133] In the formula, Indicates fitness. This represents the weighting coefficient corresponding to the infrared signal strength. This represents the weighting coefficient corresponding to the concentration field anomaly coefficient. and The value ranges from 0 to 1, and and The sum is 1. In one example, It can be 0.6. It can be 0.4.

[0134] Step b3: Adjust the initial geometric parameters and / or initial material parameters according to the fitness function and single-objective optimization algorithm to obtain the target geometric parameters and target material parameters.

[0135] For example, a single-objective optimization algorithm can be a Bayesian optimization algorithm, a genetic algorithm, or a particle swarm optimization algorithm.

[0136] Specifically, the initial geometric parameters and / or initial material parameters are adjusted through a single-objective optimization algorithm to find the design parameters that maximize fitness, and these design parameters are denoted as the target geometric parameters and target material parameters.

[0137] Compared to multi-objective optimization, converting multiple objectives into a single objective through weighted summation before parameter optimization eliminates the need to maintain a large number of non-dominated candidate solutions simultaneously, requiring fewer samples and effectively reducing overall simulation computational overhead and improving solution speed. Adjusting the weighting coefficients allows for intuitive control over optimization bias, selectively emphasizing either infrared signal intensity or concentration field uniformity based on process requirements.

[0138] In some alternative embodiments, the optimization algorithm is a multi-objective optimization algorithm, and step S407 may include the following steps: Step c1: With the optimization objectives of maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection cross section, the initial geometric parameters and / or initial material parameters are adjusted through a multi-objective optimization algorithm to obtain the Pareto front.

[0139] The Pareto front is the set of all Pareto optimal solutions in multi-objective optimization, representing the optimal trade-off point when one objective cannot be improved without harming other objectives.

[0140] Step c2: Based on the fitness function, select the target geometric parameters and target material parameters from the Pareto front.

[0141] Specifically, the fitness function can be as shown in formula (10). The Pareto front includes multiple solutions. The fitness of each solution is determined based on the fitness function, and the solution with the largest fitness is used as the target geometric parameters and target material parameters.

[0142] The simulation design method for the gas chamber provided in this embodiment embeds the gas chemical reaction Monte Carlo module into the direct simulation Monte Carlo module, and simultaneously handles physical collisions and chemical reactions in the same molecular motion simulation. For reactive cleaning gases in semiconductor cleaning processes, it achieves accurate simulation of gas phase decomposition reaction and surface fluorination reaction, and can predict the nonlinear perturbation of gas concentration distribution by chemical reaction.

[0143] By directly simulating the gas concentration field output by the Monte Carlo module to update the absorption coefficient, and feeding the absorption coefficient back to the photonic Monte Carlo module, a closed-loop coupling between chemistry, radiation, flow, and heat transfer is achieved. This enables the simulation model to realistically reflect the non-uniform absorption on the optical path caused by reaction consumption in the cleaning process, providing a theoretical basis for improving measurement accuracy.

[0144] With the dual optimization objectives of maximizing infrared signal intensity and minimizing the coefficient of variation of the concentration field, the cavity geometry and material parameters are optimized simultaneously. This overcomes the shortcomings of traditional layered optimization, which struggles to balance optical efficiency and concentration uniformity, and achieves multi-objective collaborative design of the sensor cavity.

[0145] The simulation design method for gas chambers provided by this invention will be described in detail below with specific examples.

[0146] Example 1 The gas chamber designed in this embodiment is used in a 300mm plasma-enhanced chemical vapor deposition chamber for concentration detection during the nitrogen trifluoride cleaning process. The simulation design process of the gas chamber includes the following steps: Step S1: Establish a three-dimensional geometric model and mesh generation, and configure material parameters to obtain a three-dimensional simulation model.

[0147] The gas chamber is cylindrical to accommodate the limited installation space of semiconductor devices. The chamber length ranges from 20mm to 220mm, and the diameter ranges from 5mm to 40mm. This embodiment uses a chamber diameter D=16.0mm and a chamber length L=35.0mm as an example. The infrared light source is a thermal infrared emitter, model AxetrisEMIRS50, with a surface emissivity of 0.92±0.02 and a modulation frequency of 2Hz. The infrared light source is located at the center of one end of the gas chamber. The photodetector is a dual-channel thermopile detector, each channel equipped with 4.0μm and 3.3μm narrowband filters. The photodetector is located at the center of the other end of the gas chamber, forming a through-type optical layout.

[0148] The gas chamber's inlet and outlet use a vacuum radial coupling interface (Swagelok VCR), with both the inlet and outlet apertures being 4.0 mm. The inlet is positioned 12.0 mm from the light source end face, and the outlets are symmetrically positioned 23.0 mm from the light source end face to ensure uniform airflow distribution across the optical detection cross-section (28 mm from the light source end face, i.e., 7 mm in front of the detector).

[0149] The inner wall coating is a gold film deposited by electron beam evaporation. The gold film is 200 nm thick and has a reflectivity ρ=0.95 and an emissivity ε=0.045 in the 4.0 μm band. The substrate material is aluminum alloy 6061-T6 with a thermal conductivity of 167 W / (m·K).

[0150] Unstructured tetrahedral meshes were used to spatially mesh the 3D geometric model, with a maximum mesh size of 0.4 mm. The mesh size was refined to 0.08 mm near the light source emitting surface, detector receiving surface, inlet wall, and outlet wall, resulting in a total of approximately 2.2 million mesh elements. The minimum Jacobian value was >0.3, and the maximum aspect ratio was <10, meeting the requirements of Monte Carlo simulation. The walls were configured with a combination of diffuse and specular reflection, with a specular reflection probability of 0.2 and a diffuse reflection probability of 0.8. The inlet was set as a velocity inlet, with the flow velocity calculated from mass flow rate, pressure, and temperature. The outlet was set as a pressure outlet with a back pressure of 10 Torr.

[0151] Step S2: Simulate infrared radiation transmission using the photon Monte Carlo method.

[0152] The infrared radiation transmission process was simulated based on the Monte Carlo radiation model. The infrared source emitted infrared radiation in Lambertian mode, and the wavelength region selected was the main infrared absorption peak of nitrogen trifluoride at 4.0 μm (i.e., the measurement wavelength λ). m=4.0μm) and the reference band is 3.3μm (i.e., the reference wavelength λ). r =3.3μm), and the absorption of light by nitrogen trifluoride in the reference band is negligible.

[0153] Line-by-line absorption cross-section data of nitrogen trifluoride in the temperature range (300K~550K) and pressure range (1Torr~760Torr) were obtained from the HITRAN2016 database. The 3.5μm~4.5μm band was divided into 10 bands (each with a bandwidth of 0.1μm) using a band model, and the absorption coefficient of each band was determined by the above formulas (1) and (2).

[0154] The total number of analog photons, N0, is 1.5 × 10⁻⁶. 8 Each photon carries an initial energy Q0=P total / N0, Total source power P total =15mW. Mean free path L of photons in a gaseous medium mfp Sampling according to an exponential distribution, L mfp =-ln(ξ) / μ t Where ξ is a uniformly distributed random number in the interval [0,1]. μ s The scattering coefficient is μ, which is negligible for clean gas scattering in this embodiment. s ≈0.

[0155] When a photon propagates to the inner wall of the cavity, it is reflected with a probability of 0.95 and absorbed by the wall with a probability of 0.05, based on the reflectivity ρ=0.95. The reflection type is determined according to the surface roughness parameter. The probability of specular reflection is calculated according to the Phong reflection model. In this embodiment, the specular component is set to 0.2 and the diffuse reflection component to 0.8.

[0156] When a photon reaches the detector's receiving surface, its energy and wavelength are recorded and accumulated in the measurement channel (4.0±0.05μm) and the reference channel (3.3±0.05μm), respectively. To reduce statistical noise, after each batch of simulations, 5×10 7 Photons are batch averaged, with a total of 3 batches. The final detector output signal strength is expressed as voltage: V signal =η·ΣQ photon , where Q photon This represents the photon energy carried by a single photon, and the detector responsivity η = 0.8.

[0157] In the initial iteration (k=0), assuming the gas inside the cavity has a uniform concentration C0=1000ppm and a uniform temperature T0=333K (60℃), the signal strength V of the measurement channel is calculated. meas =0.32μV / (ppm·mW), reference channel signal strength Vref =0.48μV / mW.

[0158] Step S3: Simulate the gas motion process using the direct simulation Monte Carlo method of the integrated gas chemical reaction Monte Carlo module.

[0159] The cleaning gas is a mixture of nitrogen trifluoride and nitrogen (N2), with a volume ratio of nitrogen trifluoride to nitrogen of 1:9 and a nitrogen trifluoride mass flow rate of 50 sccm, corresponding to 50 × 10⁻⁶ under standard conditions. -6 m 3 / min. Absolute working pressure P of the cavity. op =10 Torr (approximately 1333 Pa), initial temperature T init =333K (60℃), the wall temperature is updated in real time by the thermal Monte Carlo module in subsequent step S4.

[0160] The molecular weight of nitrogen trifluoride is 71.00 g / mol, the molecular weight of nitrogen is 28.01 g / mol, and the average molecular weight of the mixed gas is 32.30 g / mol. The mean free path λ of the gas molecules under this operating condition is... mfp ≈0.15mm, cavity characteristic length L char =16mm, Knudsen number (Knudsen number) K n =λ mfp / L char ≈0.0094, at K n When <0.01, it is a continuous flow, and at K n A value >0.1 indicates a rarefied flow. In this embodiment, K... n The value is approximately 0.01 to 0.1, which is in the transition flow region. At this point, the calculation error using computational fluid dynamics methods is relatively large, while the direct simulation Monte Carlo method is still applicable.

[0161] In the simulation process using the direct Monte Carlo method, the total number of simulated molecules N sim =3.0×10 6 Each grid cell contains an average of 13 to 15 simulated molecules. Under conditions of 10 Torr and 333 K, the number of real molecules represented by each simulated molecule, W, is... factor =N real / N sim ≈6.7×10 10 N real Represents the total number of actual molecules. Time step Δt DSMC =1×10 -8 s, less than the average molecular collision interval (approximately 5 × 10⁻⁶). -8 s). Total simulation physics time t total =5×10 -The simulation time is 3 seconds (i.e., 500,000 time steps). The first 10% of the total simulation physical time is used for flow establishment, and the next 90% is used for statistical sampling. The sampling interval is once every 1,000 time steps, resulting in 450 sample points.

[0162] The Gas Chemical Reaction Monte Carlo module is embedded within the Direct Simulation Monte Carlo framework and simulates the following chemical reaction pathways: (1) Gas-phase decomposition reaction; In this embodiment, plasma excitation is assumed to be fully developed at the sampling point, with the free radical concentration determined by the upstream plasma source. No further plasma excitation is performed within the gas chamber, but the recombination reactions of free radicals are considered: F + F + M → F2 + M (M is a third body), and F + NF2 → NF3 (recombination). Actual cleaning consumption mainly occurs inside the chamber, primarily involving the transport of residual gas after the reaction and a small amount of surface reaction.

[0163] (2) Surface fluorination reaction; Trace amounts of silicon residue may be deposited on the inner wall of the gas chamber, carried by the gas flow from the process chamber: 4F(g) + Si(s) → SiF4(g). Surface reaction activation energy Ea surface =0.04eV, the reaction probability can be determined by formula (5).

[0164] (3) Adsorption / desorption of byproduct silicon tetrafluoride: SiF4(g) + surface → SiF4(ads), adsorption probability 0.01, desorption activation energy 0.1eV.

[0165] In the direct Monte Carlo simulation of gas chemical reactions, the following steps are performed sequentially at each time step: (i) Molecular movement, updating positions according to velocity and handling boundary reflection / passage; (ii) Indexing and pairing, performing collision pairing on molecules within the same grid; (iii) Collision handling, calculating collision probabilities and handling elastic collisions and chemical reactions; (iv) Surface reaction handling, determining whether a surface reaction occurs for molecules colliding with the wall based on probability; if a reaction occurs, the molecule is deleted and a silicon tetrafluoride product molecule is added; (v) Sampling statistics, recording the density, velocity, temperature, and concentration fraction of each component for each grid cell. In particular, changes in gas composition caused by chemical reactions affect the subsequent calculation of photon absorption coefficients.

[0166] The steady-state concentration distribution of nitrogen trifluoride can be obtained through direct Monte Carlo simulation of gaseous chemical reactions. At the optical detection section (z=28 mm), the average concentration C of nitrogen trifluoride is... avg=912ppm, nitrogen trifluoride input concentration is 1000ppm, due to consumption reduction of about 8.8%, the concentration field coefficient of variation (CV) is 8.3%. In contrast, if chemical reactions are not considered (i.e., Monte Carlo module without gaseous chemical reactions), the concentration field coefficient of variation is only 4.6%. It can be seen that chemical reactions significantly enhance the non-uniformity of concentration, which is something that traditional simulations cannot capture.

[0167] Step S4: Simulate gas-solid heat exchange using the thermal Monte Carlo method.

[0168] The thermal Monte Carlo module solves the energy equations by simulating random walks of hot carriers (phonons). The cavity's 3D mesh is kept consistent with the photon Monte Carlo module and the direct simulation Monte Carlo module, sharing the same mesh system. Each heat pack carries energy Q. thermal =1×10 -9 J, total number of calories in N thermal =5×10 6 .

[0169] Thermal boundary condition setting: Convective heat transfer coefficient h between the outer wall and the environment ext =7.5W / (m²·K), ambient temperature T amb =300K. The inner wall surface simultaneously receives radiant energy q from the photon Monte Carlo module. rad (i,j,k) and the heat of chemical reaction q from the Monte Carlo module of gaseous chemical reactions chem (i,j,k) represents the index number of the divided mesh cell, i represents the number of the mesh cell in the X direction (or radial direction), j represents the number of the mesh cell in the Y direction (circumferential direction), and k represents the number of the mesh cell in the Z direction (optical path direction).

[0170] At the surface reaction site, the chemical heat released by the formation of each silicon tetrafluoride molecule is approximately ΔH ≈ -320 kJ / mol (exothermic), which is transmitted via q. chem =(ΔH·R surface ) / (N A ·A cell ), where R surface For the surface reaction rate (molecules / m²·s), A cell Let be the wall area corresponding to the grid cell. Total heat flux q total =q rad +q conv +q chem , where q rad The radiative heat flux is calculated based on the photon absorption energy output by the photon Monte Carlo module, q. conv q represents convective heat flux. conv =h gas (T wall-T gas ), h gas T is the gas convective heat transfer coefficient, calculated based on the gas properties and flow velocity output from the direct simulation Monte Carlo module. wall T represents the temperature of the cavity wall. gas This indicates the temperature of the gas inside the cavity.

[0171] In the thermal Monte Carlo simulation, each time step Δt TMC =5×10 -6 s, where the time step is greater than the mean free path time of the heat pack, the heat pack performs a random walk in the solid mesh according to the diffusion equation. The displacement at each time step is s. ,in, Where is the thermal diffusivity, The values ​​are random numbers distributed according to a standard normal distribution. When a heat pack reaches the boundary, it undergoes reflection or absorption / transfer based on the boundary conditions. At the end of each time step, the temperature T(i,j,k)=E is updated based on the total energy of the heat packs within each grid cell. total (i,j,k) / (ρ·c p ·V cell ), where E total The energy accumulated within the unit cell is ρ, where ρ is the material density. The material density of aluminum alloy 6061 is 2700 kg / m³. 3 c p For specific heat capacity, c p =897J / (kg·K), V cell The grid volume is denoted as . The thermal Monte Carlo simulation continues until the temperature field reaches a quasi-steady state. Samples are taken every 1 ms. If the temperature change at 5 consecutive sampling points is less than 0.5 K, the temperature field is considered to have reached a quasi-steady state.

[0172] The final temperature field predicted by the thermal Monte Carlo method: During the stabilization stage of the cleaning process, due to radiative heating and exothermic chemical reactions, the highest temperature of the inner wall is located near the light source, at approximately 530 K (257 °C); the lowest temperature is located near the air outlet, at approximately 420 K (147 °C); the average wall temperature at the optical detection section (z=28 mm) is 485 K (212 °C), and the radial temperature gradient is approximately 15 K / mm.

[0173] Step S5: Alternating iterative coupling of photonic Monte Carlo, direct simulation Monte Carlo / gas chemical reaction Monte Carlo, and thermal Monte Carlo.

[0174] Iterative initialization: k=0, assuming an initial uniform concentration field C 0 =1000ppm, uniform temperature field T 0 =333K, absorption coefficient Determined by formulas (1) and (2). Run photon Monte Carlo (N0 = 1.5 × 10⁻⁶). 8(photon), to obtain the initial radiation energy q rad The detector signal strength is I = 0.32 μV / (ppm·mW).

[0175] The (k+1)th iteration process (k starts from 0): (1) The radiation energy output by the kth photon in Monte Carlo. As a volumetric heat source, it is input to the thermo-Monte Carlo module, while simultaneously receiving the heat of chemical reaction transferred from the gaseous chemical reaction Monte Carlo module. The thermal Monte Carlo algorithm runs for 5000 time steps (corresponding to 25 ms of physical time) to calculate the new temperature field. .

[0176] (2) The new temperature field The Monte Carlo / Gas Chemical Reaction direct simulation module was used as boundary condition input. The direct Monte Carlo simulation ran for 10 runs. 4 At each time step (corresponding to 0.1 ms physical time), during the direct Monte Carlo simulation run, the temperature field is considered frozen because its timescale is much larger than the convergence time of the direct Monte Carlo simulation. The gas chemical reaction Monte Carlo module uses the updated temperature field to calculate the chemical reaction rate constant, and outputs a new gas concentration field after the direct Monte Carlo simulation / gas chemical reaction Monte Carlo run. The gas concentration field includes the concentrations of nitrogen trifluoride, nitrogen, and silicon tetrafluoride.

[0177] (3) Based on formulas (1) and (2), the new gas concentration field is... and new temperature field Used to update the absorption coefficient of the photonic Monte Carlo module , This represents the wavelength of the light signal. Because the gas concentration is no longer spatially uniform, the absorption coefficient becomes spatially distributed. The photon Monte Carlo algorithm is run again, emitting the same 1.5 × 10⁻⁶ wavelength. 8 Photons, resulting in a new detector signal strength I k+1 .

[0178] (4) Calculate the relative change using formula (3). If ΔI < 1%, then convergence is determined and the loop is exited; otherwise, k is incremented and the above steps (1)-(4) are repeated.

[0179] In this embodiment, 20 iterations are required for convergence. The first 5 iterations show a rapid decrease in ΔI (from 35% to 8%), followed by ΔI=2.5% in the 10th iteration, 1.3% in the 15th iteration, and 0.6% in the 20th iteration. The detector signal strength I after convergence is... final=0.54 μV / (ppm·mW). The detector signal intensity measured after convergence is higher than that measured before iteration. This is because the increase in temperature leads to an increase in the absorption cross section, and the average absorption along the optical path increases after the concentration redistribution. The coefficient of variation (CV) of the nitrogen trifluoride concentration field on the optical detection cross section is 4.6%, compared to 8.3% before optimization and 4.6% after convergence. The uniformity is significantly improved after convergence because the coupling of the temperature field and the flow field promotes mixing.

[0180] Step S6: Perform bi-objective iterative optimization of the initial geometric parameters and / or initial material parameters.

[0181] With maximizing the normalized detector signal intensity I after convergence and minimizing the concentration field variation coefficient CV as the dual optimization objectives, a fitness function F'=w1·(I / I) is constructed. max )-w2·(CV / CV min The initial geometric and material parameters to be optimized include: cavity diameter D (8mm~20mm), cavity length L (25mm~45mm), and inlet position d. inlet (8mm~18mm from the end face of the light source), radius of curvature R of the reflecting surface curv (2mm~10mm), inner wall reflectivity ρ (0.85~0.99). The radius of curvature of the reflecting surface is used for the dome's reflecting surface. A Bayesian optimization algorithm combined with a Gaussian process surrogate model was used. Initially, 20 points were sampled, and in each subsequent iteration, the next sampling point was selected according to the expected improvement (EI) criterion. The algorithm converged after 50 target evaluations.

[0182] The optimized design parameters are: D opt =16.5mm, L opt =34.0mm, d inlet_opt =11.0mm, R curv_opt =6.0mm, ρ opt =0.97. The corresponding infrared signal intensity of this design is 0.61 μV / (ppm·mW), and the concentration field coefficient of variation (CV) is 3.9%. In contrast, the design obtained by traditional single-field optimization (optimizing only the optical path) has a signal intensity of only 0.48 μV / (ppm·mW) and CV = 7.2% under three-field coupling simulation evaluation, demonstrating that the multi-field coupling optimization of this invention has significant advantages.

[0183] Example 2 This embodiment verifies the transflow adaptability under different pressure conditions (1 Torr, 100 Torr) and monitors silicon tetrafluoride byproducts.

[0184] In semiconductor cleaning processes, the non-dispersive infrared gas chamber pressure may vary from high vacuum (below 1 Torr) to process pressure (100 Torr-760 Torr) during the cleaning phase. To verify the effectiveness of this invention under different flow conditions, four operating conditions were set: (a) pressure P = 1 Torr, K n (a) ≈0.1, in a rarefied flow; (b) pressure P=10 Torr, K n ≈0.01, consistent with Example 1, in transition flow; (c) Pressure P = 100 Torr, K n ≈0.001, in near-continuous flow; (d) pressure P=760 Torr, K n ≈0.00013, indicating a continuous flow.

[0185] For the cases of P=1 Torr and P=10 Torr, the direct simulation Monte Carlo method is used. For the cases of P=100 Torr and P=760 Torr, due to the extremely frequent molecular collisions, the computational cost of the direct simulation Monte Carlo method is too high. In these cases, a hybrid approach can be used: computational fluid dynamics is used in the mainstream region, while the direct simulation Monte Carlo method or analytical model is still used in the near-wall Knudsen layer. The framework of this invention allows for adaptive selection of the solver based on the local Knudsen number.

[0186] Simulation results show that for 1 Torr, the results obtained by direct simulation Monte Carlo method are accurate; for 10 Torr, direct simulation Monte Carlo method is usable; for 100 Torr, computational fluid dynamics and slip boundary methods are usable; and for 760 Torr, standard computational fluid dynamics is usable. In step S3, this invention automatically switches the solver according to the inlet pressure conditions, ensuring applicability across a wide pressure range.

[0187] Under the condition of P=10 Torr, the concentration distribution of silicon tetrafluoride (SiF) can be output through the Monte Carlo module of gaseous chemical reaction. After the cleaning process injects nitrogen trifluoride, the SiF concentration in the chamber changes over time as follows: In the initial stage (0 ms–5 ms), nitrogen trifluoride enters the chamber, and the SiF concentration is almost zero; subsequently (5 ms–20 ms), surface reaction produces SiF, and the SiF concentration increases monotonically; after 20 ms, the nitrogen trifluoride concentration decreases, the SiF formation rate slows down, and the concentration tends to reach an equilibrium value of approximately 1.8 × 10⁻⁶. 17 molecules / m³, which is equivalent to approximately 5 ppm volume fraction.

[0188] It is worth noting that silicon tetrafluoride has a strong absorption peak at 9.7 μm, which is far from the 4.0 μm absorption peak of nitrogen trifluoride. Therefore, a dual-channel approach (nitrogen trifluoride at 4.0 μm and silicon tetrafluoride at 9.7 μm) can be used to simultaneously monitor reactants and products. The photonic Monte Carlo module of this invention supports simultaneous simulation at multiple wavelengths; only the absorption coefficient and detector channel for the corresponding wavelengths need to be increased.

[0189] By simultaneously monitoring nitrogen trifluoride consumption and silicon tetrafluoride formation, the cleaning endpoint can be determined more accurately. When the nitrogen trifluoride concentration drops below 20% of its initial value and the silicon tetrafluoride concentration begins to decrease, it indicates that the silicon residue on the wall surface has been largely depleted, signaling the end of the cleaning process. For dynamic compensation, the nitrogen trifluoride reading can be corrected in real time using the silicon tetrafluoride formation rate, reducing the measurement error of the nitrogen trifluoride concentration.

[0190] This embodiment demonstrates the adaptability of the method of the present invention over a wide pressure range and its ability to simultaneously monitor multiple components.

[0191] Optionally, the optical signal includes a target optical signal of the target wavelength and a reference optical signal of the reference wavelength. The simulation design method for the gas chamber also includes: constructing an Artificial Neural Network (ANN) compensation model and configuring the neural network compensation model in the processor of the cleaning endpoint detection device.

[0192] The inputs to the neural network compensation model include time, cavity pressure, target signal intensity corresponding to the target light signal (i.e., signal intensity corresponding to the measurement channel), reference light intensity corresponding to the reference light signal (i.e., signal intensity corresponding to the reference channel), and the generation rate of the chemical products formed after the chemical reaction. The output of the neural network compensation model is the correction coefficient for the measured concentration.

[0193] The neural network compensation model employs a three-layer feedforward neural network, consisting of an input layer, a first hidden layer, a second hidden layer, and an output layer. The input layer comprises five neurons, corresponding to time t, cavity pressure P(t), and reference channel signal intensity V, respectively. ref (t), Measurement channel signal strength V meas (t) and the rate of formation of chemical products after the chemical reaction. Generation rate It can be obtained through independent measurement via the 9.7μm channel or predicted by the model. The first hidden layer consists of 64 neurons, corresponding to the activation function, which can be the Rectified Linear Unit (ReLU). The second hidden layer consists of 32 neurons, corresponding to the activation function. The output layer consists of 1 neuron, corresponding to the correction coefficient K. corr .

[0194] Using a converged three-field coupling model, the deviation between the original and actual concentrations of nitrogen trifluoride output during the entire semiconductor cleaning process (0s–180s) is simulated. The original readings are those without considering chemical perturbations, while the actual concentrations are the true concentrations after considering chemical perturbations.

[0195] After the simulation, the input features include: time t, cavity pressure P(t), and reference channel signal strength V. ref (t), Measurement channel signal strength V meas (t) and the rate of formation of chemical products after the chemical reaction. The output target is the correction factor K between the true concentration and the original reading. corr A total of 200 sets of simulation data under different process conditions (pressure ramp, flow rate change, temperature fluctuation) were generated. Each set of data contained 1,800 time points (one sample per second), and a total of 360,000 training samples were used to train the neural network to obtain the neural network compensation model.

[0196] During training, the loss function was Mean Squared Error (MSE), the optimizer was the Adaptive Moment Estimator (Adam), the learning rate was 0.001, the batch size was 256, and the number of training epochs was 500. 80% of the generated training samples were used for training, 10% for validation, and 10% for testing.

[0197] After training, the Mean Absolute Percentage Error (MAPE) on the test set was 1.8%, and the maximum error was 4.2%. The resulting neural network compensation model was integrated into the processor. While calculating the raw readings, the processor also acquired time t, cavity pressure P(t), and reference channel signal strength V. ref (t), Measurement channel signal strength V meas (t) and the rate of formation of chemical products after the chemical reaction. Input the neural network compensation model, and obtain the correction coefficient K from the output of the neural network compensation model. corr It can output the corrected nitrogen trifluoride concentration C in real time. corr =C raw ×K corr C raw This represents the original concentration obtained from the calculation.

[0198] A neural network compensation model was applied to a set of independently generated simulation data, which simulated a ramp process in a cleaning process where pressure gradually decreased from 10 Torr to 1 Torr. Simulation results show that without compensation, the maximum error occurs during the rapid pressure change phase; after compensation, the maximum error throughout the process can be reduced, meeting the accuracy requirements for semiconductor cleaning endpoint detection. In contrast, simple linear compensation (based solely on pressure) provides only a limited reduction in the maximum error and is insufficient to meet the accuracy requirements for semiconductor cleaning endpoint detection. This invention provides a pre-trained neural network compensation model for cleaning signal processing, which can further reduce measurement errors.

[0199] This invention improves the measurement accuracy and stability of the sensor under dynamic cleaning process conditions by correcting the calibration coefficients in real time through a neural network compensation model.

[0200] Example 3 This embodiment applies the gas chamber simulation design method provided by the present invention to the cleaning process of chlorine trifluoride in an atomic layer deposition equipment. Compared with nitrogen trifluoride, chlorine trifluoride has higher reactivity and can react with silicon at room temperature without plasma excitation, but it is also more corrosive, requiring higher standards for the gas chamber material.

[0201] The cleaning process parameters for the atomic layer deposition chamber include: chlorine trifluoride flow rate of 30 sccm, nitrogen flow rate of 200 sccm, total pressure of 50 Torr, and chamber temperature of 200℃ (473K). The cleaning unit is installed in the exhaust bypass, and the sampling flow rate is controlled at 10 sccm via a proportional valve. The gas to be measured is chlorine trifluoride, and the main interfering gases are the reaction byproducts chlorine fluoride, chlorine, and silicon tetrafluoride. The main infrared absorption peaks of chlorine trifluoride are located at 11.2 μm and 12.1 μm, with the reference band selected at 10.5 μm (no absorption). Due to the strong corrosiveness of chlorine trifluoride, the inner wall material of the chamber is replaced with a corrosion-resistant nickel (Ni)-based alloy and plated with a nickel-phosphorus (Ni-P) alloy reflective layer, with a reflectivity of 0.92.

[0202] In the simulation design process, the coupling process of photonic Monte Carlo, direct simulated Monte Carlo, and thermal Monte Carlo is the same as in Example 1, but the material parameters and chemical reaction channels are updated. The gas chemical reaction Monte Carlo module adds a new chemical reaction channel: 2ClF3(g) + Si(s) → SiF4(g) + 2ClF(g), with activation energy Ea. surface =0.02eV, the reaction is exothermic ΔH≈-420kJ / mol. Due to the high reactivity of chlorine trifluoride, the surface reaction probability is about an order of magnitude higher than that of nitrogen trifluoride, thus the concentration perturbation is more significant. Direct Monte Carlo simulation results show that, at the optical detection cross section, the coefficient of variation (CV) of the chlorine trifluoride concentration field reaches 12.5%, compared to 4.2% in the absence of reaction, indicating severe inhomogeneity.

[0203] Optimization objective: Maximize the infrared signal intensity in the 11.2μm band while minimizing the concentration field variation coefficient. Optimization variables include the cavity shell material, inner wall reflectivity, inlet location, and sampling flow rate. The optimized design parameters are: cavity length L = 25mm (shortening the optical path to reduce absorption saturation due to rapid reaction consumption), cavity diameter D = 12mm, inlet distance from the light source end face 8mm, inner wall reflectivity ρ = 0.92, and sampling flow rate 15sccm. Through the optimization algorithm, the infrared signal intensity received by the cleaning endpoint detection device can be maximized while ensuring a uniform gas concentration distribution.

[0204] A prototype was fabricated according to the optimized design parameters and validated on a 300mm atomic layer deposition equipment. Fourier transform infrared spectroscopy (FTIR) was used as a reference standard to compare the measurement errors before and after optimization of the method of this invention. The signal intensity before optimization was 0.23 μV / (ppm·mW), and the signal intensity after optimization was 0.28 μV / (ppm·mW).

[0205] During the dynamic cleaning process, the chlorine trifluoride concentration decreased from 500 ppm to 50 ppm within 120 seconds, and the optimized average measurement error was better than that of the traditional design. The average deviation between the predicted and experimental values ​​from the three-field coupled simulation was small, which was better than the average deviation of the traditional single-field simulation, verifying the accuracy and practicality of the method of this invention.

[0206] The present invention also provides a cleaning endpoint detection device, comprising a gas chamber designed by the simulation design method of the gas chamber provided in any of the above embodiments.

[0207] This invention also provides a simulation design apparatus for a gas chamber, which is used to implement the above embodiments and preferred embodiments, and will not be repeated hereafter. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0208] This embodiment provides a simulation design device for a gas chamber, such as... Figure 5 As shown, it includes: Model building module 501 is used to build a three-dimensional geometric model of the gas chamber based on the initial geometric parameters; The mesh generation module 502 is used to mesh the three-dimensional geometric model and configure the meshed three-dimensional geometric model based on the initial material parameters to obtain the three-dimensional simulation model. The photon Monte Carlo module 503 is used to simulate the infrared radiation of the light signal emitted by the infrared source through the three-dimensional simulation model to the detector using the photon Monte Carlo method. It obtains the current photon absorption energy and the current infrared signal intensity received by the detector at the current iteration number. The light signal is divided into multiple spectral bands, and each spectral band corresponds to a current absorption coefficient. The Direct Monte Carlo Simulation Module 504 is used to simulate the motion of gas molecules within a three-dimensional simulation model using the direct Monte Carlo simulation method, and obtain the current gas concentration field at the current iteration number. The thermal Monte Carlo module 505 is used to simulate the heat exchange between the inner wall of the three-dimensional simulation model and the gas using the thermal Monte Carlo method, and obtain the current temperature field at the current iteration number. The iteration module 506 is used to use the current photon absorption energy as a heat source to perform the next iteration to obtain the next temperature field, use the next temperature field as a boundary condition to perform the next iteration to obtain the next gas concentration field, and correct the current absorption coefficient based on the next gas concentration field and the next temperature field, and perform the next iteration according to the corrected current absorption coefficient to obtain the next infrared signal intensity. The optimization design module 507 stops the iteration if the change is less than or equal to the preset change. It takes maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection section as the optimization objectives, and adjusts the initial geometric parameters and / or initial material parameters through the optimization algorithm to simulate the target geometric parameters and target material parameters of the gas chamber.

[0209] In some alternative embodiments, the apparatus further includes: The first determining module is used to determine the current absorption coefficient when the current iteration number is the first iteration, based on a preset initial uniform concentration field, a preset initial temperature field, and a first correspondence relationship, wherein the first correspondence relationship is used to characterize the mapping relationship between the concentration field, the temperature field, and the absorption coefficient.

[0210] In some optional implementations, the gas within the three-dimensional simulation model is a reactive cleaning gas; the apparatus further includes: The Gas Chemical Reaction Monte Carlo module is used to simulate the chemical reaction process of reactive clean gases using the gas chemical reaction Monte Carlo method, and obtain the current gas concentration field at the current iteration number.

[0211] In some alternative implementations, directly simulating the Monte Carlo module includes: The configuration unit is used to set the type identifier, velocity vector, quantity, and internal energy for each gas molecule within the same gas molecule motion and collision simulation framework. The probability determination unit is used to determine the probability of a chemical reaction between gas molecules during gas molecule collision processing, based on the Arrhenius equation, which is expressed as follows: ; In the formula, Indicates the probability of reaction. Indicates pre-exponential factor, Indicates the activation energy of the reaction. Represents Boltzmann's constant. , Indicates temperature. Indicates the collision time step; The judgment unit is used to determine whether a chemical reaction will occur based on the reaction probability and the generated random numbers that are uniformly distributed between 0 and 1. The update unit is used to update the type identifier, quantity, velocity vector and internal energy of the gas molecules participating in the chemical reaction if a chemical reaction occurs, to obtain the updated type identifier, quantity, velocity vector and internal energy, and to record the heat released or absorbed by the chemical reaction. The updated type identifier, quantity, velocity vector and internal energy are used to determine the current gas concentration field. The iteration module includes: The iterative unit is used to take the energy and heat absorbed by the current photon as a heat source and perform the next iteration to obtain the next temperature field.

[0212] In some optional implementations, the determination unit includes: The first determining subunit is used to determine if a chemical reaction occurs if the random number is less than the reaction probability. The second determining subunit is used to determine that no chemical reaction has occurred if the random number is greater than or equal to the reaction probability.

[0213] In some optional embodiments, the optical signal includes a target optical signal of a target wavelength and a reference optical signal of a reference wavelength, and the apparatus further includes: The module is used to build a neural network compensation model and configure the neural network compensation model in a clean processor. The inputs of the neural network compensation model include time, cavity pressure, target signal intensity corresponding to the target light signal, reference light intensity corresponding to the reference light signal, and generation rate of chemical products formed after the chemical reaction. The output of the neural network compensation model is the correction coefficient for the measured concentration.

[0214] In some alternative implementations, the photonic Monte Carlo module 503 includes: Discrete unit, used to discretize the optical signal into multiple spectral bands and configure the current absorption coefficient for each spectral band; The tracking unit is used to emit optical signals and track the absorption and scattering paths of each photon in the gas medium in multiple photons in each spectral band until the photon reaches the detector or escapes the gas chamber simulated by the three-dimensional simulation model. The first statistical unit is used to obtain the current infrared signal intensity by statistically analyzing the photon energies arriving at the detector from multiple spectral bands; The second statistical unit is used to obtain the current photon absorption energy by statistically analyzing the photon energy absorbed by multiple spectral bands in the gas medium.

[0215] In some alternative implementations, the apparatus further includes a coefficient determination model; The coefficient determination model is used to determine the current absorption coefficient using the following formula: ; In the formula, Indicates the first The average absorption coefficient of the first spectral band will be the first band. The average absorption coefficient of the spectral band is used as the first... The current absorption coefficient corresponding to each spectral band. Indicates the bandwidth of the spectral band. Indicating in wavenumber Spectral absorption coefficient at that location, The calculations were performed using the gas concentration and temperature obtained from the previous iteration. Let represent the Planck function.

[0216] In some optional implementations, the optimization algorithm is a single-objective optimization algorithm, and the optimization design module includes: The first processing unit is used to normalize the infrared signal intensity and concentration field variation coefficient to obtain the normalized infrared signal intensity and the normalized concentration field variation coefficient. The first building unit is used to construct a fitness function based on the normalized infrared signal intensity and the normalized concentration field variation coefficient. The first adjustment unit is used to adjust the initial geometric parameters and / or initial material parameters according to the fitness function and the single-objective optimization algorithm to obtain the target geometric parameters and target material parameters.

[0217] In some alternative implementations, the single-objective optimization algorithm is a Bayesian optimization algorithm, a genetic algorithm, or a particle swarm optimization algorithm.

[0218] In some optional implementations, the optimization algorithm is a multi-objective optimization algorithm, and the optimization design module includes: The second adjustment unit is used to adjust the initial geometric parameters and / or initial material parameters through a multi-objective optimization algorithm with the optimization objectives of maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection section, so as to obtain the Pareto front. The screening unit is used to select target geometric parameters and target material parameters from the Pareto front based on the fitness function.

[0219] In some optional implementations, the initial geometric parameters include at least one of the following: cavity length, cavity diameter, infrared light source, relative position between the gas chamber and the detector, reflective surface curvature, inlet position, inlet aperture, outlet position, and outlet aperture; the initial material parameters include at least one of the following: infrared reflectivity, absorptivity, thermal conductivity, specific heat capacity, and surface emissivity of the inner wall coating material.

[0220] In some alternative implementations, the pressure in the gas chamber ranges from 1 Torr to 760 Torr.

[0221] The gas chamber simulation design apparatus provided in this embodiment of the invention can execute the gas chamber simulation design method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0222] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0223] like Figure 6 As shown, the electronic device includes a memory 610 and a processor 620. The memory 610 stores a computer program, and the processor 620 is configured to run the computer program to perform the steps in any of the above-described embodiments of the simulation design method for a gas chamber. Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0224] Furthermore, the electronic device also includes a communication interface 630 for communicating with other devices or communication networks. Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0225] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the simulation design method for the gas chamber shown in the above embodiments is implemented.

[0226] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0227] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A simulation design method for a gas chamber, characterized in that, The gas chamber is used in a non-dispersive infrared cleaning endpoint detection device, and the method includes: Based on the initial geometric parameters, a three-dimensional geometric model of the gas chamber is constructed; The three-dimensional geometric model is meshed, and the meshed three-dimensional geometric model is configured based on the initial material parameters to obtain a three-dimensional simulation model; The infrared radiation of the light signal emitted by the infrared source is simulated through the three-dimensional simulation model to the detector using the photon Monte Carlo method. The current photon absorption energy and the current infrared signal intensity received by the detector are obtained at the current iteration number. The light signal is divided into multiple spectral bands, and each spectral band corresponds to a current absorption coefficient. By directly simulating the Monte Carlo method, the motion of gas molecules within the three-dimensional simulation model is simulated under the pressure of the gas chamber, and the current gas concentration field at the current iteration number is obtained. The thermal Monte Carlo method is used to simulate the heat exchange between the inner wall of the three-dimensional simulation model and the gas, and the current temperature field at the current iteration number is obtained. If the change between the current infrared signal intensity and the previous infrared signal intensity is greater than a preset change, then the current photon absorption energy is used as a heat source to perform the next iteration to obtain the next temperature field. The next temperature field is used as a boundary condition to perform the next iteration to obtain the next gas concentration field. The current absorption coefficient is corrected based on the next gas concentration field and the next temperature field. The next iteration is performed according to the corrected current absorption coefficient to obtain the next infrared signal intensity. If the change is less than or equal to the preset change, then the optimization objectives are to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the optical detection section. The initial geometric parameters and / or the initial material parameters are adjusted by an optimization algorithm to simulate and obtain the target geometric parameters and target material parameters of the gas chamber.

2. The method according to claim 1, characterized in that, When the current iteration number is the first iteration, the simulation design method further includes: The current absorption coefficient is determined based on a preset initial uniform concentration field, a preset initial temperature field, and a first correspondence, wherein the first correspondence is used to characterize the mapping relationship between the concentration field, the temperature field, and the absorption coefficient.

3. The method according to claim 1, characterized in that, The gas within the three-dimensional simulation model is a reactive, clean gas; when simulating the motion of gas molecules within the three-dimensional simulation model using the direct Monte Carlo simulation method, the method further includes: The Monte Carlo method for gas chemical reactions is used to simulate the chemical reaction process of reactive clean gases, and the current gas concentration field at the current iteration number is obtained.

4. The method according to claim 3, characterized in that, The method of simulating the chemical reaction process of reactive clean gases using the Monte Carlo method of gaseous chemical reactions includes: Within the same framework for simulating the motion and collision of gas molecules, each gas molecule is assigned a type identifier, velocity vector, quantity, and internal energy. During gas molecule collision processing, the probability of a chemical reaction occurring between gas molecules is determined according to the Arrhenius equation, which is expressed as follows: ; In the formula, Indicates the probability of reaction. Indicates pre-exponential factor, Indicates the activation energy of the reaction. Represents Boltzmann's constant. , Indicates temperature. Indicates the collision time step; Based on the reaction probability and the generated random numbers that are uniformly distributed between 0 and 1, it is determined whether a chemical reaction occurs; If a chemical reaction occurs, the type identifier, quantity, velocity vector, and internal energy of the gas molecules participating in the chemical reaction are updated to obtain the updated type identifier, quantity, velocity vector, and internal energy, and the heat released or absorbed by the chemical reaction is recorded. The updated type identifier, quantity, velocity vector, and internal energy are used to determine the current gas concentration field. The step of using the energy absorbed by the current photon as a heat source to obtain the next temperature field in the next iteration includes: Using the current photon absorption energy and the heat as a heat source, the next temperature field is obtained in the next iteration.

5. The method according to claim 4, characterized in that, The step of determining whether a chemical reaction occurs based on the reaction probability and a randomly generated number uniformly distributed between 0 and 1 includes: If the random number is less than the reaction probability, then a chemical reaction is determined to have occurred. If the random number is greater than or equal to the reaction probability, then it is determined that no chemical reaction has occurred.

6. The method according to claim 3, characterized in that, The optical signal includes a target optical signal at a target wavelength and a reference optical signal at a reference wavelength. The method further includes: A neural network compensation model is constructed and configured in a clean processor. The inputs of the neural network compensation model include time, cavity pressure, target signal intensity corresponding to the target light signal, reference light intensity corresponding to the reference light signal, and generation rate of chemical products formed after the chemical reaction. The output of the neural network compensation model is a correction coefficient for the measured concentration.

7. The method according to any one of claims 1 to 6, characterized in that, The simulation of infrared radiation from an infrared light source reaching the detector via the three-dimensional simulation model using the photon Monte Carlo method includes: The optical signal is discretized into multiple spectral bands, and a current absorption coefficient is configured for each spectral band; The optical signal is emitted, and the absorption and scattering paths of each photon in the gas medium are tracked among multiple photons in each spectral band until the photon reaches the detector or escapes the gas chamber simulated by the three-dimensional simulation model. The current infrared signal intensity is obtained by statistically analyzing the photon energies arriving at the detector from the multiple spectral bands. The current photon absorption energy is obtained by statistically analyzing the photon energies absorbed by the gaseous medium across the multiple spectral bands.

8. The method according to claim 7, characterized in that, The current absorption coefficient is determined using the following formula: ; In the formula, Indicates the first The average absorption coefficient of the first spectral band will be the first band. The average absorption coefficient of the spectral band is used as the first... The current absorption coefficient corresponding to each spectral band. Indicates the bandwidth of the spectral band. Indicating in wavenumber Spectral absorption coefficient at that location, The calculations were performed using the gas concentration and temperature obtained from the previous iteration. Let represent the Planck function.

9. The method according to any one of claims 1 to 6, characterized in that, The optimization algorithm is a single-objective optimization algorithm, with the optimization objectives being to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the optical detection cross section. The algorithm adjusts the initial geometric parameters and / or the initial material parameters, including: The infrared signal intensity and the concentration field variation coefficient are normalized to obtain the normalized infrared signal intensity and the normalized concentration field variation coefficient. A fitness function is constructed based on the normalized infrared signal intensity and the normalized concentration field variation coefficient. Based on the fitness function and the single-objective optimization algorithm, the initial geometric parameters and / or the initial material parameters are adjusted to obtain the target geometric parameters and target material parameters.

10. The method according to claim 9, characterized in that, The single-objective optimization algorithm is a Bayesian optimization algorithm, a genetic algorithm, or a particle swarm optimization algorithm.

11. The method according to any one of claims 1 to 6, characterized in that, The optimization algorithm is a multi-objective optimization algorithm, with the optimization objectives being to maximize the infrared signal intensity and minimize the concentration field variation coefficient of the optical detection cross section. The algorithm adjusts the initial geometric parameters and / or the initial material parameters, including: With the optimization objectives of maximizing infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection section, the Pareto front is obtained by adjusting the initial geometric parameters and / or the initial material parameters through a multi-objective optimization algorithm. The target geometric parameters and target material parameters are selected from the Pareto front based on the fitness function.

12. The method according to any one of claims 1 to 6, characterized in that, The initial geometric parameters include at least one of the following: cavity length, cavity diameter, infrared light source, relative position between the gas chamber and the detector, reflective surface curvature, inlet position, inlet aperture, outlet position, and outlet aperture; the initial material parameters include at least one of the following: infrared reflectivity, absorptivity, thermal conductivity, specific heat capacity, and surface emissivity of the inner wall coating material.

13. The method according to any one of claims 1 to 6, characterized in that, The pressure in the gas chamber ranges from 1 Torr to 760 Torr.

14. A cleaning endpoint detection device, characterized in that, It includes a gas chamber, which is designed using the simulation design method for gas chambers according to any one of claims 1 to 13.

15. A simulation design device for a gas chamber, characterized in that, The device includes: The model building module is used to construct a three-dimensional geometric model of the gas chamber based on the initial geometric parameters; The mesh generation module is used to generate a mesh for the three-dimensional geometric model and configure the meshed three-dimensional geometric model based on the initial material parameters to obtain a three-dimensional simulation model. The photon Monte Carlo module is used to simulate the infrared radiation of an infrared light signal emitted by an infrared source through the three-dimensional simulation model to the detector using the photon Monte Carlo method. It obtains the current photon absorption energy and the current infrared signal intensity received by the detector at the current iteration number. The light signal is divided into multiple spectral bands, and each spectral band corresponds to a current absorption coefficient. The direct simulation Monte Carlo module is used to simulate the motion of gas molecules within the three-dimensional simulation model using the direct simulation Monte Carlo method, and to obtain the current gas concentration field at the current iteration number. The thermal Monte Carlo module is used to simulate the heat exchange between the inner wall of the three-dimensional simulation model and the gas using the thermal Monte Carlo method, and to obtain the current temperature field at the current iteration number. The iteration module, if the change between the current infrared signal intensity and the previous infrared signal intensity is greater than a preset change, is used to use the current photon absorption energy as a heat source to perform the next iteration to obtain the next temperature field, use the next temperature field as a boundary condition to perform the next iteration to obtain the next gas concentration field, and correct the current absorption coefficient based on the next gas concentration field and the next temperature field, and perform the next iteration according to the corrected current absorption coefficient to obtain the next infrared signal intensity; The optimization design module stops the iteration if the change is less than or equal to the preset change. It then adjusts the initial geometric parameters and / or the initial material parameters through an optimization algorithm with the optimization objectives of maximizing the infrared signal intensity and minimizing the concentration field variation coefficient of the optical detection section, so as to simulate and obtain the target geometric parameters and target material parameters of the gas chamber.

16. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the simulation design method for the gas chamber according to any one of claims 1 to 13.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to execute the simulation design method for the gas chamber according to any one of claims 1 to 13.

18. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the simulation design method for a gas chamber according to any one of claims 1 to 13.

Citation Information

Patent Citations

  • Gas activation measurement method and system

    CN111323806A

  • MEMS device multi-physical field simulation method based on Monte Carlo-finite element

    CN119476160A