Self-adaptive control method and system for remote plasma source

Abstract

The invention discloses an adaptive control method and system for a remote plasma source, and relates to the related technical field of plasma control, and the method comprises the steps: carrying out the multi-stage control mining of the remote plasma source according to the semiconductor etching task information; performing multi-stage cooperative control analysis according to the multi-stage control trigger space; according to the etching risk factors, an etching risk assessment channel is built, and simulation optimization is carried out on the plasma source control first group; performing comprehensive risk optimization on the plasma source control second group according to the etching comprehensive risk constraint; and according to the plasma source, controlling the third group to carry out multi-round variation joint optimization, obtaining a plasma source control strategy and carrying out adaptive control on the remote plasma source. The technical problems of insufficient multi-stage etching process cooperative control, weak etching risk dynamic evaluation capability and low parameter optimization efficiency in the prior art are solved, and the technical effects of improving the adaptability, stability and process yield of remote plasma source control are achieved.

Description

Technical Field

[0001] The present application relates to the technical field related to plasma control, and in particular to an adaptive control method and system for a remote plasma source. Background Art

[0002] In semiconductor manufacturing, plasma etching is a key step in achieving high-precision micro-nanostructure fabrication. With the continuous reduction in semiconductor device size and increasing process complexity, traditional plasma etching control faces numerous challenges, such as insufficient process stability, difficulty ensuring etching uniformity, and an increase in dynamic interference factors during the etching process. Remote plasma sources can reduce substrate damage and improve plasma uniformity, and are widely used in advanced semiconductor manufacturing. However, remote plasma sources have complex dynamic characteristics, and their control parameters, such as power, gas flow, and pressure, directly affect etching quality and process yield. Existing plasma control systems are difficult to adapt to the needs of different etching tasks, especially in multi-stage etching processes, where control objectives at each stage may conflict, leading to a narrowing of the process window and increased etching risk. Furthermore, random disturbances during the etching process, such as gas composition fluctuations and equipment state drift, can affect the stability and consistency of the etching process, resulting in poor performance of the final semiconductor device.

[0003] Therefore, in the current related technologies, there are technical problems such as insufficient coordinated control of multi-stage etching processes, weak dynamic assessment capabilities of etching risks, and low parameter optimization efficiency. Summary of the Invention

[0004] This application solves the technical problems existing in the prior art, such as insufficient coordinated control of multi-stage etching processes, weak dynamic assessment capability of etching risks, and low parameter optimization efficiency, by providing an adaptive control method and system for a remote plasma source, thereby achieving the technical effect of improving the adaptability, stability and process yield of remote plasma source control.

[0005] The application provides a self-adaptive control method of a remote plasma source, which comprises the following steps: performing multi-stage control excavation on the remote plasma source according to semiconductor etching task information, establishing a multi-stage control trigger space; performing multi-stage collaborative control analysis on the remote plasma source according to the multi-stage control trigger space based on the semiconductor etching task information, obtaining a first group of plasma source control; building an etching risk evaluation channel according to an etching risk factor, and performing analog optimization on the first group of plasma source control according to the etching risk evaluation channel, to establish a second group of plasma source control; performing comprehensive risk optimization on the second group of plasma source control according to an etching comprehensive risk constraint, to obtain a third group of plasma source control; performing multi-round variation joint optimization on the third group of plasma source control based on the multi-stage control trigger space, the etching risk evaluation channel and the etching comprehensive risk constraint, to obtain a plasma source control strategy, and performing self-adaptive control on the remote plasma source according to the plasma source control strategy.

[0006] In a possible implementation, the self-adaptive control method of the remote plasma source further performs the following processing: loading an etching morphology distortion risk record set, an etching endpoint drift risk record set and a device performance degradation risk record set according to the etching risk factor; training an etching morphology distortion risk evaluation model according to the etching morphology distortion risk record set; training an etching endpoint drift risk evaluation model according to the etching endpoint drift risk record set; training a device performance degradation risk evaluation model according to the device performance degradation risk record set; and performing loss minimization distillation on the etching morphology distortion risk evaluation model, the etching endpoint drift risk evaluation model and the device performance degradation risk evaluation model, to generate the etching risk evaluation channel.

[0007] In a possible implementation, the self-adaptive control method of the remote plasma source further performs the following processing: performing virtual control on the remote plasma source according to each scheme in the first group of plasma source control based on the semiconductor etching task information, to obtain each scheme simulation data; inputting the each scheme simulation data into the etching risk evaluation channel, to obtain a plurality of etching risk evaluation results; constructing an etching risk constraint based on the etching risk factor, and performing abnormal analysis on the plurality of etching risk evaluation results according to the etching risk constraint, to obtain a plurality of etching risk abnormal analysis results; and screening the second group of plasma source control according to the plurality of etching risk abnormal analysis results.

[0008] In a possible implementation, the adaptive control method of the remote plasma source further performs the following processing: extracting a plasma source control n-th scheme according to the plasma source control second group, and calling an n-th etching risk assessment result corresponding to the plasma source control n-th scheme, n being a positive integer; performing weight distribution according to the etching risk factors to establish an etching comprehensive risk assessment model; inputting the n-th etching risk assessment result into the etching comprehensive risk assessment model to obtain an n-th etching comprehensive risk coefficient; and adding the plasma source control n-th scheme to the plasma source control third group if the n-th etching comprehensive risk coefficient satisfies the etching comprehensive risk constraint.

[0009] In a possible implementation, the adaptive control method of the remote plasma source further performs the following processing: performing mutation optimization on the plasma source control third group according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a first round mutation optimization space; performing mutation optimization on the first round mutation optimization space according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a second round mutation optimization space; and continuing to perform mutation optimization on the second round mutation optimization space according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a P-th round mutation optimization space, P being a positive integer greater than 2; and performing etching comprehensive risk minimization optimization according to the plasma source control third group, the first round mutation optimization space, the second round mutation optimization space and the P-th round mutation optimization space to generate the plasma source control strategy.

[0010] In a possible implementation, the adaptive control method of the remote plasma source further performs the following processing: performing mutation value evaluation on the plasma source control third group to determine a mutation value coefficient of each scheme; and performing mutation adjustment on the plasma source control third group based on the multi-stage control trigger space and according to the mutation value coefficient of each scheme to establish a mutation adjustment first group; performing simulation optimization on the mutation adjustment first group according to the etching risk assessment channel to establish a mutation adjustment second group; and performing comprehensive risk optimization on the mutation adjustment second group according to the etching comprehensive risk constraint to obtain the first round mutation optimization space.

[0011] In a possible implementation, the adaptive control method of the remote plasma source further performs the following processing: disassembling the semiconductor etching task information to obtain a multi-stage etching task; performing control sample retrieval on the remote plasma source according to the multi-stage etching task to obtain a control retrieval set of each stage; and performing trigger feature carding according to the control retrieval set of each stage to generate the multi-stage control trigger space.

[0012] In a possible implementation, the adaptive control method of the remote plasma source further performs the following processing: performing data cleaning on etching task data of the remote plasma source to generate the semiconductor etching task information.

[0013] In a possible implementation, the adaptive control method of the remote plasma source further performs the following processing: the etching risk factors include etching profile distortion risk, etching endpoint drift risk, and device performance degradation risk.

[0014] The application further provides an adaptive control system of a remote plasma source, which comprises: a multi-stage control mining module configured to perform multi-stage control mining on a remote plasma source according to semiconductor etching task information, and establish a multi-stage control trigger space; a collaborative control analysis module configured to perform multi-stage collaborative control analysis on the remote plasma source based on the semiconductor etching task information and the multi-stage control trigger space, and obtain a first group of plasma source control; a simulation optimization module configured to build an etching risk evaluation channel according to etching risk factors, and perform simulation optimization on the first group of plasma source control according to the etching risk evaluation channel, and establish a second group of plasma source control; a comprehensive risk optimization module configured to perform comprehensive risk optimization on the second group of plasma source control according to etching comprehensive risk constraints, and obtain a third group of plasma source control; and a variation joint optimization module configured to perform multi-round variation joint optimization on the third group of plasma source control based on the multi-stage control trigger space, the etching risk evaluation channel, and the etching comprehensive risk constraints, obtain a plasma source control strategy, and perform adaptive control on the remote plasma source according to the plasma source control strategy.

[0015] The application provides an adaptive control method and system of a remote plasma source, which performs multi-stage control mining on a remote plasma source according to semiconductor etching task information, performs multi-stage collaborative control analysis according to a multi-stage control trigger space, builds an etching risk evaluation channel according to etching risk factors, performs simulation optimization on a first group of plasma source control, performs comprehensive risk optimization on a second group of plasma source control according to etching comprehensive risk constraints, performs multi-round variation joint optimization on a third group of plasma source control, obtains a plasma source control strategy, and performs adaptive control on the remote plasma source. The technical problems of insufficient multi-stage etching process collaborative control, weak etching risk dynamic evaluation capability, and low parameter optimization efficiency in the prior art are solved, and the technical effects of improving the adaptability, stability, and process yield of remote plasma source control are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. In the present application, flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, various steps can be processed in reverse order or simultaneously according to needs. Meanwhile, other operations can be added to these processes, or one or more steps of operations can be removed from these processes.

[0017] Figure 1 A flowchart of an adaptive control method of a remote plasma source provided by an embodiment of the present application.

[0018] Figure 2 A structural schematic diagram of an adaptive control system of a remote plasma source provided by an embodiment of the present application.

[0019] Legend: multi-stage control excavation module 10, cooperative control analysis module 20, simulation optimization module 30, comprehensive risk optimization module 40, and variation joint optimization module 50. DETAILED DESCRIPTION

[0020] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the present application can be implemented according to the content of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described.

[0021] In order to make the purposes, technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by those of ordinary skill in the art without making creative labor belong to the scope of protection of the present application.

[0022] In the following description, "some embodiments" are referred to, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict, and the term "first\second" referred to is only to distinguish similar objects, and does not represent a specific order for the objects. The terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0023] The embodiments of the present application provide a self-adaptive control method of a remote plasma source, as shown in the method comprises: Figure 1 Step S100, according to the semiconductor etching task information, a multi-stage control excavation of the remote plasma source is performed, and a multi-stage control trigger space is established.

[0024] Step S100 further comprises data cleaning of the etching task data of the remote plasma source, and generating the semiconductor etching task information.

[0025] Preferably, the original etching task data of the remote plasma source is collected in real time, including plasma spectrum, radio frequency power, gas flow, chamber pressure, temperature, etc., wherein the original etching task data usually contains noise, outliers or missing values, and direct use for control optimization may lead to wrong decisions, the etching task data of the remote plasma source is data cleaned, and high-quality and reliable semiconductor etching task information is extracted, specifically, sliding average filtering, low-pass filtering or wavelet denoising is used to smooth high-frequency noise; 3σ principle and box plot are used to remove obvious outliers, and interpolation repair is performed on key parameters; if the data is less missing, mean value is used for supplement, if the data is seriously missing, LSTM time series prediction is used to estimate reasonable values; and then the semiconductor etching task information is generated by standardization, mainly including etching material, target etching depth, side wall angle, selection ratio and other etching process target parameters, equipment state and process constraints such as plasma source power, gas flow limit, chamber pressure set temperature safety threshold, real-time monitoring features such as plasma spectrum characteristics and etching rate, so as to ensure data reliability.

[0026] ​Preferably, the multi-stage control mining of the remote plasma source according to the semiconductor etching task information refers to intelligently decomposing the entire etching process into multiple etching process stages with different control requirements by analyzing the semiconductor etching task information. Specifically, the etching process stages are intelligently divided, such as the initial stabilization period, the main etching period, the over-etching control period, the endpoint detection period, etc. The optimal control requirements of each stage are mined, such as the need for higher plasma density in the initial stage to quickly break through the surface layer, the need for accurate control of etching rate and uniformity in the main etching stage, and the need to reduce energy in the over-etching stage to avoid damaging the underlying material. Then, the relevance between stages is established to ensure smooth switching of etching control parameters and avoid process instability due to sudden changes. The adjustable parameter range corresponding to each etching process stage is obtained, such as the power range and gas ratio range, and a time threshold, etching depth monitoring signal, or plasma spectral characteristics is set as a trigger condition. When the condition is met, the control strategy is automatically switched to the next stage, and finally a multi-stage control trigger space is obtained, which can dynamically adjust the control strategy in different etching process stages to dynamically adapt to different etching tasks and improve etching quality.

[0027] Further, step S100 further includes step S110 of disassembling the semiconductor etching task information to obtain a multi-stage etching task; step S120 of retrieving control samples of the remote plasma source according to the multi-stage etching task to obtain a control retrieval set for each stage; and step S130 of combing trigger features according to the control retrieval set for each stage to generate the multi-stage control trigger space.

[0028] Preferably, according to the material stack structure of the etching object and the process target, the entire etching process is divided into multiple key stages, including the initial stabilization period, the main etching period, the over-etching control period, and the endpoint detection period. Then, a multi-stage etching task is obtained, each stage defines clear physical indicators such as plasma emission spectrum characteristics, etching depth progress, and control targets such as power range and gas ratio. According to the multi-stage etching task, similar etching tasks of the remote plasma source are retrieved from the historical process library, and a control retrieval set for each stage is extracted, including device parameters such as radio frequency power, gas flow, and chamber pressure, monitoring data such as spectral peak value and interferometer signal change rate, and results such as etching morphology quality and electrical test performance. Then, trigger features are combed according to the control retrieval set for each stage, i.e., the key criteria for stage switching are determined, including time triggering and signal triggering. Then, controllable parameter boundaries and optimization targets are established for each stage, and finally a multi-stage control trigger space is formed, which can dynamically adjust the control strategy in different etching process stages to dynamically adapt to different etching tasks and improve etching quality.

[0029] Step S200 , based on the semiconductor etching task information and according to the multi-stage control trigger space, a multi-stage coordinated control analysis is performed on the remote plasma source to obtain a first plasma source control group.

[0030] Preferably, the entire etching process is deconstructed into multiple key sub-stages based on the semiconductor etching task information and the stage division rules defined in the multi-stage control trigger space, and then the remote plasma source is subjected to multi-stage collaborative control analysis, that is, for each sub-stage, combined with its process goals, such as rapid stabilization in the initial stage, rate control in the main etching stage, etc., the corresponding parameter constraint range is extracted from the multi-stage control trigger space; then, multi-objective optimization algorithms such as genetic algorithms and particle swarm optimization are used to solve the parameter combination group that optimizes the overall etching effect under the premise of meeting the process constraints of each stage, and perform preliminary verification and screening on it to eliminate parameters that obviously do not meet the requirements, and finally form the first group of plasma source control, which includes multiple feasible control strategies, and each strategy contains a complete stage parameter sequence to ensure the accuracy of risk assessment.

[0031] In step S300 , an etching risk assessment channel is constructed according to the etching risk factor, and a simulation optimization is performed on the first plasma source control group according to the etching risk assessment channel to establish a second plasma source control group.

[0032] Step S300 further includes that the etching risk factors include etching morphology distortion risk, etching endpoint drift risk and device performance degradation risk.

[0033] Preferably, the etching risk factors include the risk of etching morphology distortion, the risk of etching endpoint drift, and the risk of device performance degradation. Specifically, the risk of etching morphology distortion refers to the risk that the three-dimensional structures such as grooves, through-holes, and sidewalls after etching deviate from the target morphology, which is manifested as sidewall tilt, bottom roughness, and critical dimension deviation. It may be caused by plasma inhomogeneity, gas chemical ratio imbalance, RF power / bias parameter mismatch, etc.; etching morphology distortion may directly lead to endpoint drift. The risk of etching endpoint drift refers to the risk of deviation between the actual etching stop time and the theoretical endpoint, resulting in over-etching or under-etching, where over-etching refers to etching through the target layer and damaging the underlying material, and under-etching refers to residual unetched material, which may be caused by etching rate fluctuations or material inhomogeneity; etching endpoint drift will aggravate device performance degradation. The risk of device performance degradation refers to the potential risk of damage to the electrical characteristics of the device due to the etching process, which manifests itself in surface damage, charge accumulation, and contamination introduction. Among them, surface damage refers to lattice defects caused by plasma high-energy particle bombardment, charge accumulation refers to the injection of charges in the plasma into the gate oxide layer, causing threshold voltage drift, and contamination introduction refers to the residual etching by-products leading to increased leakage current, which may be caused by excessive plasma ion energy, incomplete post-etching cleaning, or insufficient gas purity. Device performance degradation can be inferred from abnormal etching morphology through electrical testing.

[0034] Preferably, the risk of etching profile distortion, the risk of etching endpoint drift, and the risk of device performance degradation are converted into quantifiable indicators, a multi-dimensional risk prediction model is constructed, and specifically, the etching profile is simulated by plasma simulation software (such as COMSOL), the parameters such as the sidewall angle deviation and the bottom roughness are calculated, the prediction model is established based on machine learning to analyze real-time optical microscope images, and the profile anomaly probability is predicted; the LSTM time series model is trained based on historical data to predict the etching rate trend, and the endpoint misjudgment probability model is constructed in combination with the statistical characteristics of the optical emission spectrum signal; the ion bombardment energy distribution is calculated by the plasma sheath model to predict the lattice damage depth, and the mapping relationship between the damage coefficient and the process parameters is established in association with the electrical test data; and then the prediction models are fused to generate the etching risk evaluation channel.

[0035] Preferably, the first group of plasma source control is simulated and optimized according to the etching risk evaluation channel, and specifically, the multi-stage parameter combinations in the first group of plasma source control are input into the etching process simulator to simulate the whole etching process, the simulated profile data, endpoint time, surface damage and other simulation results are obtained, the simulation results are input into the etching risk evaluation channel, the risk scores of the parameter combinations are calculated, the parameter combination solutions balanced in risk and performance are screened by applying the Pareto optimization, and the risk weights are allocated according to the task requirements, such as the memory etching priority endpoint risk and the logic device priority profile risk, then the parameter combinations of each stage are reordered, the parameter combinations with high risk are eliminated, and finally the low-risk and high-performance control parameter combinations that are reserved are integrated to form the second group of plasma source control, wherein each parameter combination meets the cooperative constraints of etching profile distortion, etching endpoint drift and device performance degradation.

[0036] Further, step S300 further comprises step S310 of loading an etching profile distortion risk record set, an etching endpoint drift risk record set and a device performance degradation risk record set according to the etching risk factors; step S320 of training an etching profile distortion risk evaluation model according to the etching profile distortion risk record set; step S330 of training an etching endpoint drift risk evaluation model according to the etching endpoint drift risk record set; step S340 of training a device performance degradation risk evaluation model according to the device performance degradation risk record set; and step S350 of performing loss minimization distillation on the etching profile distortion risk evaluation model, the etching endpoint drift risk evaluation model and the device performance degradation risk evaluation model to generate the etching risk evaluation channel.

[0037] Preferably, according to the etching risk factor, the etching topography distortion risk record set, the etching endpoint drift risk record set and the device performance degradation risk record set are loaded, wherein the etching topography distortion risk record set contains topography measurement data in historical processes, such as scanning electron microscope images, atomic force microscope roughness data, sidewall angle, etc., and abnormal cases are marked, such as sidewall inclination > 5°, bottom roughness exceeding the standard, etc.; the etching endpoint drift risk record set stores etching endpoint detection data, such as optical emission spectrum timing, interferometer signal, actual endpoint deviation value from theoretical endpoint; the device performance degradation risk record set is associated with electrical test results, such as current-voltage characteristic curve, gate oxide leakage current, threshold voltage drift, and the relationship with process parameters.

[0038] Preferably, a prediction model is constructed according to convolutional neural network or random forest, trained with data in the etching topography distortion risk record set to obtain an etching topography distortion risk assessment model capable of assessing topography distortion probability and key abnormal types; an LSTM time series model is constructed, trained with data in the etching endpoint drift risk record set to obtain an etching endpoint drift risk assessment model capable of predicting endpoint drift risk level; a performance prediction model is constructed based on gradient boosting tree, trained with data in the device performance degradation risk record set to obtain a device performance degradation risk assessment model capable of predicting output period performance reliability score; then the three risk assessment models are fused into an etching risk assessment channel through loss minimization distillation, i.e. taking the three risk assessment models as teacher models, taking multilayer perceptron or small neural network as student model, minimizing the KL divergence of student model output and teacher model overall output, and retaining key features of each risk factor, such as topography sensitive parameters, endpoint timing pattern, to finally obtain an etching risk assessment channel capable of outputting comprehensive risk score and ensuring accuracy and real-time performance of risk assessment.

[0039] Further, step S300 further comprises step S360, based on the semiconductor etching task information, virtually controlling the remote plasma source according to the first group of plasma source control schemes to obtain simulation data of each scheme; step S370, inputting the simulation data of each scheme into the etching risk assessment channel to obtain multiple etching risk assessment results; step S380, based on the etching risk factor, constructing etching risk constraint, and performing abnormal analysis on the multiple etching risk assessment results according to the etching risk constraint to obtain multiple etching risk abnormal analysis results; step S390, screening the first group of plasma source control schemes according to the multiple etching risk abnormal analysis results to obtain the second group of plasma source control schemes.

[0040] Preferably, the remote plasma source is virtually controlled according to each scheme in the first plasma source control group, wherein the first plasma source control group includes multiple candidate control schemes, each scheme includes a multi-stage parameter combination, and based on the semiconductor etching task information, the parameter combination of each control scheme is input into an etching process simulator, such as a COMSOL plasma component, to simulate the dynamic response and etching results of the remote plasma source, and obtain simulation data for each scheme, including etching morphology data such as 3D etching profile, sidewall angle, and bottom roughness, time series data such as spectral sequence, etching rate curve, endpoint time, and device performance impact data such as surface damage depth and electrical parameter prediction value. The simulation data of each scheme is then used as input data and input into the etching risk assessment channel for risk assessment, and multiple etching risk assessment results are quantitatively output, including an etching morphology distortion risk coefficient, an etching endpoint drift risk coefficient, and a device performance degradation risk coefficient.

[0041] Preferably, etching risk constraints are constructed based on etching risk factors, i.e., risk thresholds are set according to etching process requirements, including etching profile distortion risk ≤ 0.2, such as sidewall angle deviation ≤ ±2°; etching endpoint drift risk ≤ 0.3, such as overetching ≤ 5% of the target depth; and device performance degradation risk ≤ 0.15, such as gate oxide leakage current variation ≤ 5%. Multiple etching risk assessment results are then analyzed for anomalies based on the etching risk constraints. This involves comparing the etching risk assessment results with the etching risk constraints and marking anomalies. If the etching risk constraints are met, the results are considered normal and labeled normal; if the etching risk constraints are not met, the results are considered abnormal and labeled abnormal, ultimately obtaining multiple etching risk anomaly analysis results. The first group of plasma source control strategies is then screened based on the multiple etching risk anomaly analysis results. Candidate control schemes corresponding to the abnormal labels are eliminated, while candidate control schemes corresponding to the normal labels are retained to form the second group of plasma source control strategies. All strategies have passed virtual verification and risk compliance checks.

[0042] Step S400 , performing comprehensive risk optimization on the second plasma source control group according to the etching comprehensive risk constraint to obtain a third plasma source control group.

[0043] Step S400 further includes step S410, extracting the nth plasma source control scheme according to the second plasma source control group, and retrieving the nth etching risk assessment result corresponding to the nth plasma source control scheme, where n is a positive integer; step S420, performing weight allocation according to the etching risk factors, and establishing an etching comprehensive risk assessment model; step S430, inputting the nth etching risk assessment result into the etching comprehensive risk assessment model to obtain the nth etching comprehensive risk coefficient; step S440, if the nth etching comprehensive risk coefficient meets the etching comprehensive risk constraint, adding the nth plasma source control scheme to the third plasma source control group.

[0044] Preferably, the second group of plasma source control schemes is optimized according to an etching comprehensive risk constraint, wherein the etching comprehensive risk constraint is a comprehensive risk of etching topography distortion risk, etching endpoint drift risk and device performance degradation risk, and a preset global risk threshold, such as a comprehensive risk coefficient ≤ 0.25. Specifically, an nth etching risk evaluation result corresponding to an nth plasma source control scheme is obtained from the second group of plasma source control schemes, wherein n is a positive integer, representing the number of plasma source control schemes, and the nth etching risk evaluation result includes an etching topography distortion risk coefficient 0.18, an etching endpoint drift risk coefficient 0.22 and a device performance degradation risk coefficient 0.12. Then, according to the etching risk factors and the etching task requirements, the weights of the etching topography distortion risk, the etching endpoint drift risk and the device performance degradation risk are obtained, and the etching topography distortion risk evaluation model, the etching endpoint drift risk evaluation model and the device performance degradation risk evaluation model are fused to determine the etching comprehensive risk evaluation model. The nth etching risk evaluation result is input into the etching comprehensive risk evaluation model to output a determined nth etching comprehensive risk coefficient. If the nth etching comprehensive risk coefficient meets the etching comprehensive risk constraint, the nth plasma source control scheme is added to the third group of plasma source control schemes. If the nth etching comprehensive risk coefficient does not meet the etching comprehensive risk constraint, the nth plasma source control scheme is marked as a high-risk scheme and eliminated. Finally, the third group of plasma source control schemes is obtained, wherein all the schemes meet the comprehensive risk constraint, so as to ensure the precise control of the etching risk.

[0045] In step S500, based on the multi-order control trigger space, the etching risk evaluation channel and the etching comprehensive risk constraint, a multi-round variation joint optimization is performed according to the third group of plasma source control schemes to obtain a plasma source control strategy, and the remote plasma source is adaptively controlled according to the plasma source control strategy.

[0046] The step S500 further comprises a step S510 of performing mutation optimization on the third group of plasma source control according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a first round mutation optimization space; a step S520 of performing mutation optimization on the first round mutation optimization space according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a second round mutation optimization space; a step S530 of continuously performing mutation optimization on the second round mutation optimization space according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a Pth round mutation optimization space, P being a positive integer greater than 2; and a step S540 of performing etching comprehensive risk minimization optimization according to the third group of plasma source control, the first round mutation optimization space, the second round mutation optimization space and the Pth round mutation optimization space to generate the plasma source control strategy.

[0047] Preferably, the mutation optimization is a parameter space exploration method based on evolutionary algorithm, the core of which is to perform controllable random disturbance on the parameters of the existing strategy, fuse the advantage parameters of different strategies, and retain the mutation scheme with improved risk-performance comprehensive score. Specifically, the parameter disturbance range of the control scheme in the third group of plasma source control is limited within the device safety threshold based on the constraint of the multi-stage control trigger space, such as power adjustment ± 10%, and then the mutated control scheme is input into the etching risk assessment channel for evaluation, the comprehensive risk coefficient of the mutation scheme is calculated, and then compared with the etching comprehensive risk constraint, all the mutated control schemes that meet the etching comprehensive risk constraint and have improved performance are retained to form the first round mutation optimization space.

[0048] Preferably, then the control schemes in the first round mutation optimization space are subjected to mutation optimization again to reduce the parameter mutation amplitude of the control scheme, and the mutated control schemes that meet the etching comprehensive risk constraint are retained to establish the second round mutation optimization space. In this way, the mutation optimization is performed based on the output mutation optimization space of the previous round to perform more refined directional optimization and output the Pth round mutation optimization space, wherein P is a positive integer greater than 2, representing the number of mutation optimization; then all the candidate control schemes in the third group of plasma source control and the first round mutation optimization space, the second round mutation optimization space and the Pth round mutation optimization space are combined to perform etching comprehensive risk minimization optimization to find the candidate control scheme with the minimum etching comprehensive risk as the plasma source control strategy. Finally, through the real-time feedback-optimization cycle, the generated optimal control strategy is dynamically applied to the remote plasma source, so that the plasma source can autonomously adapt to process fluctuations, device state changes and material differences, and always maintain optimal etching performance, thereby improving the adaptability, stability and process yield of the remote plasma source control.

[0049] Further, step S510 further comprises step S511, evaluating the variation value of the third group of plasma source control, determining the variation value coefficient of each scheme; step S512, based on the multi-stage control trigger space, adjusting the third group of plasma source control according to the variation value coefficient of each scheme, establishing a first group of variation adjustment; step S513, simulating and optimizing the first group of variation adjustment according to the etching risk evaluation channel, establishing a second group of variation adjustment; step S514, comprehensively optimizing the second group of variation adjustment according to the etching comprehensive risk constraint, obtaining the first round of variation optimization space.

[0050] Preferably, the variation value of the control scheme in the third group of plasma source control is evaluated, that is, the influence weight of each control parameter on the etching effect is analyzed and the risk reduction space after parameter adjustment is predicted, and then the variation value coefficient of each scheme is generated, wherein the smaller the etching comprehensive risk coefficient, the greater the variation value coefficient of the control scheme, and the more the variation quantity; then the control parameters of the scheme whose variation value coefficient is greater than 0.6 in the third group of plasma source control are disturbed by ±10%, and part of the parameters are exchanged by selecting strategies with similar risk scores, and the stage switching logic of the multi-stage control trigger space is strictly followed to form the first group of variation adjustment. Then, the multiple control schemes in the first group of variation adjustment are input into the etching process simulator to obtain the virtual etching result, and then the risk is evaluated through the etching risk evaluation channel, the schemes with single risk exceeding the threshold are removed, and multiple control schemes meeting the risk threshold are retained to form the second group of variation adjustment. Then, the comprehensive risk optimization of the second group of variation adjustment is performed, that is, the multiple control schemes in the second group of variation adjustment are compared with the etching comprehensive risk constraint, and the control schemes meeting the comprehensive risk optimization are combined to form the first round of variation optimization space.

[0051] In the foregoing, with reference to Figure 1 A remote plasma source adaptive control method according to an embodiment of the present application is described in detail. Next, with reference to Figure 2 A remote plasma source adaptive control system according to an embodiment of the present application will be described.

[0052] The remote plasma source adaptive control system according to the embodiment of the present application is used to solve the technical problems of insufficient multi-stage etching process cooperative control, weak etching risk dynamic evaluation capability and low parameter optimization efficiency in the prior art, and achieves the technical effects of improving the adaptability, stability and process yield of remote plasma source control. As shown in Figure 2 The remote plasma source adaptive control system comprises a multi-stage control excavation module 10, a cooperative control analysis module 20, a simulation optimization module 30, a comprehensive risk optimization module 40 and a variation joint optimization module 50.

[0053] The multi-stage control mining module 10 is used to perform multi-stage control mining on the remote plasma source based on the semiconductor etching task information and establish a multi-stage control trigger space; the collaborative control analysis module 20 is used to perform multi-stage collaborative control analysis on the remote plasma source based on the semiconductor etching task information and the multi-stage control trigger space to obtain a first plasma source control group; the simulation optimization module 30 is used to build an etching risk assessment channel based on the etching risk factor, and perform simulation optimization on the first plasma source control group based on the etching risk assessment channel to establish a second plasma source control group; the comprehensive risk optimization module 40 is used to perform comprehensive risk optimization on the second plasma source control group based on the etching comprehensive risk constraint to obtain a third plasma source control group; the variation joint optimization module 50 is used to perform multiple rounds of variation joint optimization based on the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint according to the third plasma source control group to obtain a plasma source control strategy, and perform adaptive control of the remote plasma source according to the plasma source control strategy.

[0054] The specific configuration of the simulation optimization module 30 will be described in detail below. The simulation optimization module 30 further includes: loading an etching profile distortion risk record set, an etching endpoint drift risk record set, and a device performance degradation risk record set based on the etching risk factor; training an etching profile distortion risk assessment model based on the etching endpoint drift risk record set; training a device performance degradation risk assessment model based on the device performance degradation risk record set; and performing loss minimization distillation on the etching profile distortion risk assessment model, the etching endpoint drift risk assessment model, and the device performance degradation risk assessment model to generate the etching risk assessment channel.

[0055] The specific configuration of the simulation optimization module 30 will be described in detail below. The simulation optimization module 30 further includes: based on the semiconductor etching task information, virtually controlling the remote plasma source according to each scheme within the first plasma source control group to obtain simulation data for each scheme; inputting the simulation data for each scheme into the etching risk assessment channel to obtain multiple etching risk assessment results; constructing etching risk constraints based on the etching risk factors, and performing anomaly analysis on the multiple etching risk assessment results according to the etching risk constraints to obtain multiple etching risk anomaly analysis results; and screening the first plasma source control group according to the multiple etching risk anomaly analysis results to obtain the second plasma source control group.

[0056] The specific configuration of the comprehensive risk optimization module 40 will be described in detail below. The comprehensive risk optimization module 40 further comprises: extracting an nth etching risk assessment result corresponding to an nth plasma source control scheme according to the second group of plasma source control schemes, n being a positive integer; performing weight distribution according to the etching risk factors to establish an etching comprehensive risk assessment model; inputting the nth etching risk assessment result into the etching comprehensive risk assessment model to obtain an nth etching comprehensive risk coefficient; and adding the nth plasma source control scheme to the third group of plasma source control schemes if the nth etching comprehensive risk coefficient meets the etching comprehensive risk constraint.

[0057] The specific configuration of the variation joint optimization module 50 will be described in detail below. The variation joint optimization module 50 further comprises: performing variation optimization on the third group of plasma source control schemes according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a first round of variation optimization space; performing variation optimization on the first round of variation optimization space according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a second round of variation optimization space; and continuing to perform variation optimization on the second round of variation optimization space according to the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint to establish a Pth round of variation optimization space, P being a positive integer greater than 2; and performing etching comprehensive risk minimization optimization according to the third group of plasma source control schemes, the first round of variation optimization space, the second round of variation optimization space, …, and the Pth round of variation optimization space to generate the plasma source control strategy.

[0058] The specific configuration of the variation joint optimization module 50 will be described in detail below. The variation joint optimization module 50 further comprises: performing variation value evaluation on the third group of plasma source control schemes to determine a variation value coefficient of each scheme; adjusting the third group of plasma source control schemes based on the multi-stage control trigger space according to the variation value coefficient of each scheme to establish a variation adjustment first group; performing simulation optimization on the variation adjustment first group according to the etching risk assessment channel to establish a variation adjustment second group; and performing comprehensive risk optimization on the variation adjustment second group according to the etching comprehensive risk constraint to obtain the first round of variation optimization space.

[0059] The specific configuration of the multi-stage control mining module 10 will be described in detail below. The multi-stage control mining module 10 further comprises: disassembling the semiconductor etching task information to obtain a multi-stage etching task; performing control sample retrieval on the remote plasma source according to the multi-stage etching task to obtain a set of control retrieval of each stage; and performing trigger feature carding according to the set of control retrieval of each stage to generate the multi-stage control trigger space.

[0060] The following will describe in detail the specific configuration of the collaborative control analysis module 20. The collaborative control analysis module 20 further includes: performing data cleaning on the etching task data of the remote plasma source to generate the semiconductor etching task information.

[0061] The following will further describe the specific configuration of the simulation optimization module 30. The simulation optimization module 30 further includes: the etching risk factors include etching morphology distortion risk, etching endpoint drift risk and device performance degradation risk.

[0062] An adaptive control system for a remote plasma source provided by an embodiment of the present invention can execute an adaptive control method for a remote plasma source provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0063] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention.

[0064] The above specific embodiments do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the scope of protection of this application.

Claims

1. A method for adaptively controlling a remote plasma source, characterized in that: The method comprises: Based on the semiconductor etching task information, multi-stage control mining of remote plasma source is carried out to establish a multi-stage control trigger space; Based on the semiconductor etching task information, performing multi-stage collaborative control analysis on the remote plasma source according to the multi-stage control trigger space to obtain a first plasma source control group; According to the etching risk factor, an etching risk assessment channel is established, and according to the etching risk assessment channel, a simulation optimization is performed on the first plasma source control group to establish a second plasma source control group; Performing comprehensive risk optimization on the second plasma source control group according to the etching comprehensive risk constraint to obtain a third plasma source control group; Based on the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint, multiple rounds of variation joint optimization are performed according to the plasma source control third group to obtain a plasma source control strategy, and the remote plasma source is adaptively controlled according to the plasma source control strategy.

2. The adaptive control method of a remote plasma source according to claim 1, wherein: According to the etching risk factors, an etching risk assessment channel is established, including: Loading an etching profile distortion risk record set, an etching endpoint drift risk record set, and a device performance degradation risk record set according to the etching risk factor; Training an etching profile distortion risk assessment model according to the etching profile distortion risk record set; Training an etching endpoint drift risk assessment model based on the etching endpoint drift risk record set; Training a device performance degradation risk assessment model based on the device performance degradation risk record set; Loss minimization distillation is performed on the etching profile distortion risk assessment model, the etching endpoint drift risk assessment model, and the device performance degradation risk assessment model to generate the etching risk assessment channel.

3. The adaptive control method of a remote plasma source according to claim 1, wherein: Simulating and optimizing the first plasma source control group according to the etching risk assessment channel to establish a second plasma source control group includes: Based on the semiconductor etching task information, the remote plasma source is virtually controlled according to each scheme in the first plasma source control group to obtain simulation data of each scheme; Inputting the simulation data of each scheme into the etching risk assessment channel to obtain multiple etching risk assessment results; Based on the etching risk factor, an etching risk constraint is constructed, and an abnormality analysis is performed on the multiple etching risk assessment results according to the etching risk constraint to obtain multiple etching risk abnormality analysis results; The plasma source control first group is screened according to the plurality of etching risk abnormality analysis results to obtain the plasma source control second group.

4. The adaptive control method of a remote plasma source according to claim 1, wherein: The second plasma source control group is optimized based on the comprehensive etching risk constraint to obtain a third plasma source control group, including: Extracting an nth plasma source control scheme according to the second plasma source control group, and retrieving an nth etching risk assessment result corresponding to the nth plasma source control scheme, where n is a positive integer; Perform weight allocation according to the etching risk factors to establish an etching comprehensive risk assessment model; Inputting the nth etching risk assessment result into the etching comprehensive risk assessment model to obtain the nth etching comprehensive risk coefficient; If the nth comprehensive etching risk coefficient satisfies the comprehensive etching risk constraint, the nth plasma source control scheme is added to the third plasma source control group.

5. The adaptive control method of a remote plasma source according to claim 1, wherein: Based on the multi-stage control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint, multiple rounds of mutation joint optimization are performed according to the plasma source control third group to obtain a plasma source control strategy, including: Performing mutation optimization on the third group of plasma source control according to the multi-stage control trigger space, the etching risk assessment channel, and the etching comprehensive risk constraint to establish a first round of mutation optimization space; Performing mutation optimization on the first round of mutation optimization space according to the multi-stage control trigger space, the etching risk assessment channel, and the etching comprehensive risk constraint to establish a second round of mutation optimization space; Continue to perform mutation optimization on the second round of mutation optimization space according to the multi-stage control trigger space, the etching risk assessment channel, and the etching comprehensive risk constraint to establish a P-th round of mutation optimization space, where P is a positive integer greater than 2; The plasma source control strategy is generated by performing etching comprehensive risk minimization optimization based on the plasma source control third group, the first round variation optimization space, the second round variation optimization space...the Pth round variation optimization space.

6. The adaptive control method of a remote plasma source according to claim 5, wherein: Performing mutation optimization on the third group of plasma source control according to the multi-stage control trigger space, the etching risk assessment channel, and the etching comprehensive risk constraint to establish a first round of mutation optimization space, including: Performing variation value evaluation on the third group of plasma source control to determine the variation value coefficient of each scheme; Based on the multi-stage control trigger space, according to the variation value coefficient of each scheme, the plasma source control third group is variably adjusted to establish the variation adjustment first group; Performing simulation optimization on the first variation adjustment group according to the etching risk assessment channel to establish a second variation adjustment group; The second mutation adjustment group is subjected to comprehensive risk optimization according to the comprehensive etching risk constraint to obtain the first round of mutation optimization space.

7. The adaptive control method of a remote plasma source according to claim 1, wherein: Based on the semiconductor etching task information, multi-stage control mining of the remote plasma source is carried out to establish a multi-stage control trigger space, including: Disassembling the semiconductor etching task information to obtain a multi-stage etching task; Performing control sample retrieval on the remote plasma source according to the multi-stage etching task to obtain a control retrieval set for each stage; Trigger features are sorted out according to the control retrieval sets of each stage to generate the multi-stage control trigger space.

8. The adaptive control method of a remote plasma source according to claim 1, wherein: The etching task data of the remote plasma source is cleaned to generate the semiconductor etching task information.

9. The adaptive control method of a remote plasma source according to claim 1, wherein: The etching risk factors include etching morphology distortion risk, etching endpoint drift risk and device performance degradation risk.

10. An adaptive control system for a remote plasma source, characterized in that: The system is used to implement the adaptive control method of a remote plasma source according to any one of claims 1 to 9, and the system comprises: Multi-stage control mining module, used to perform multi-stage control mining on remote plasma sources based on semiconductor etching task information and establish a multi-stage control trigger space; A collaborative control analysis module is configured to perform multi-stage collaborative control analysis on the remote plasma source based on the semiconductor etching task information and the multi-stage control trigger space to obtain a first plasma source control group; A simulation optimization module is used to build an etching risk assessment channel according to the etching risk factor, and simulate and optimize the first plasma source control group according to the etching risk assessment channel to establish a second plasma source control group; A comprehensive risk optimization module is used to perform comprehensive risk optimization on the second plasma source control group according to the etching comprehensive risk constraint to obtain a third plasma source control group; A variation joint optimization module is used to perform multiple rounds of variation joint optimization based on the multi-order control trigger space, the etching risk assessment channel and the etching comprehensive risk constraint according to the plasma source control third group to obtain a plasma source control strategy, and adaptively control the remote plasma source according to the plasma source control strategy.

Images