Atomic layer etching control system, method, storage medium, and etching apparatus

Abstract

The application relates to the technical field of semiconductor detection, and discloses an atomic layer etching control system, an atomic layer etching control method, a storage medium and etching equipment. The system comprises an etching module, a detection module, a feedback adjustment module and a control module. The etching module comprises a reaction chamber, an electrostatic chuck, a bias power supply unit, a plasma generating unit, a gas injection system and a gas extraction system. The detection module comprises a plurality of sampling units arranged at different spatial positions in the reaction chamber, an infrared spectrum analysis unit and a temperature and pressure inversion unit, which are used for extracting gas samples from different spatial positions and performing analysis, so that the gas concentration, temperature and pressure at the positions corresponding to the sampling units are determined, and spatial distribution information is generated. The feedback adjustment module generates a process adjustment instruction according to the spatial distribution information. The control module outputs the instruction to drive an actuator. The application can realize real-time sensing of the concentration spatial distribution difference inside the reaction chamber and perform partition compensation, so that the etching uniformity in the wafer is controlled at the atomic layer level.

Description

Technical Field

[0001] This invention relates to the field of semiconductor inspection technology, specifically to atomic layer etching control systems, methods, storage media, and etching equipment. Background Technology

[0002] ALE (Atomic Layer Etching) achieves single-atom-layer-level control of material removal by breaking down the etching process into self-limiting surface modification steps and selective removal steps, making it a key process in advanced semiconductor manufacturing.

[0003] Related ALE process detection typically employs single-point tail exhaust NDIR (Non-Dispersive Infrared Spectroscopy) to analyze byproducts. However, in actual mass production, the ALE process suffers from the following problems: In high aspect ratio structures, reactant transport to the bottom of the structure is limited, leading to inconsistent etching rates between the top and bottom, making morphology control difficult, and single-point detection cannot distinguish the reaction progress at different depths; Ion bombardment in the etching removal step may introduce cumulative lattice damage to the substrate, which is difficult to detect in real time; In the surface modification step, the spatially uneven distribution of gas concentration leads to inconsistent modification saturation times in different regions, and global detection cannot determine the saturation state of each region; When etching the interface of multilayer material structures, the global endpoint detection signal is lagging and spatially uneven, making it difficult to achieve synchronous termination across the entire wafer; As production batches increase, the slow drift of the chamber conditions leads to a decrease in process uniformity, while conventional periodic calibration methods result in significant yield losses.

[0004] Furthermore, the lack of closed-loop linkage between relevant detection methods and process actuators based on spatial distribution information makes it difficult to provide differentiated real-time compensation for the aforementioned spatial non-uniformity issues. Summary of the Invention

[0005] This invention provides an atomic layer etching control system, method, storage medium, and etching equipment to solve the problem of uneven etching caused by the non-uniformity of the reaction chamber space in the atomic layer etching process.

[0006] In a first aspect, the present invention provides an atomic layer etching control system for use in the semiconductor field, used for real-time detection and control via non-dispersive infrared spectroscopy, comprising: an etching module including a reaction chamber and associated electrostatic chuck, bias power supply unit, plasma generation unit, gas injection system, and pumping system, wherein the etching module is used to perform a cyclic process of atomic layer etching; and a detection module including multiple sampling units disposed at different spatial locations within the reaction chamber, an infrared spectroscopy analysis unit, and a temperature and pressure inversion unit; wherein the multiple sampling units are used to extract gas samples from different spatial locations within the reaction chamber, and the infrared spectroscopy analysis unit is used to perform infrared spectral analysis on the gas samples to determine the characteristics of each sampling unit. The gas concentration at each sampling unit location is calculated, and spatial distribution information of the concentration is generated based on the location information of each sampling unit. A feedback adjustment module, which uses an embedded simulation unit to feed back and control the etching process parameters, including a transmission simulation unit, a reaction kinetics simulation unit, and a damage simulation unit, is used to calculate the etching state of different regions based on the spatial distribution information of the concentration provided by the detection module, and generate spatially differentiated process adjustment instructions. A control module communicates with the etching module, detection module, and feedback adjustment module, and outputs the process adjustment instructions to the electrostatic chuck, bias power supply unit, and / or gas injection system to adjust the spatial uniformity of the etching process within the reaction chamber.

[0007] Beneficial Effects: The atomic-layer etching control system provided by this invention, by setting up a spatially perceptible non-dispersive infrared spectroscopy device, including multiple sampling units and an infrared spectroscopy analysis unit, collects gas samples at different spatial locations within the reaction chamber using multiple sampling units. Infrared spectroscopy analysis determines the gas concentration at each location and generates spatial distribution information. The feedback adjustment module generates spatially differentiated process adjustment commands based on this spatial distribution information. The control module outputs these process adjustment commands to the electrostatic chuck and / or gas injection system, achieving closed-loop control based on spatial distribution information. Compared to traditional single-point tail-out detection methods that only obtain global average concentration, this invention can sense the spatial distribution differences of concentration within the reaction chamber and perform differentiated compensation for different regions. This effectively overcomes the spatial non-uniformity problem caused by uneven distribution of parameters such as reactant concentration, ion energy, and temperature within the chamber, controlling the etching uniformity within the wafer at the atomic-layer level, significantly improving device consistency and yield.

[0008] In one optional embodiment, the electrostatic chuck is radially divided into at least a central region, an intermediate region, and an edge region. The electrostatic chuck includes: a temperature control unit, with each region having an independent temperature control unit, used to independently adjust the temperature of different radial regions of the wafer in response to spatially differentiated process adjustment commands output by the control module; a bias power supply unit, used to independently supply bias voltage to each region of the electrostatic chuck, and the bias voltage amplitude of each region is independently adjustable to a range of 20eV to 200eV corresponding to ion energy, and the pulse duty cycle of each region is independently adjustable to a range of 5% to 100%; the control module is also configured to independently adjust the bias voltage parameters for different regions to independently adjust the etching rate of different radial regions of the wafer; a plasma generation unit, communicatively connected to the control module, used to adjust the plasma parameters in the reaction chamber in response to the process adjustment commands; and a evacuation unit, communicatively connected to the control module, used to adjust the pressure and by-product discharge rate in the reaction chamber in response to the process adjustment commands.

[0009] Beneficial effects: This invention divides the electrostatic chuck radially into at least a central region, an intermediate region, and an edge region, and sets up an independent temperature control unit for each region. This allows for independent temperature adjustment of different radial regions of the wafer in response to differentiated process adjustment commands output by the control module. This compensates for local reaction rate differences caused by uneven gas flow or plasma distribution, synchronizing the growth of the modified layer in each region and improving surface modification uniformity. Simultaneously, the bias power supply unit independently supplies bias voltage to each region, while the plasma generation unit and the exhaust unit adjust plasma parameters, chamber pressure, and exhaust rate in response to process adjustment commands. The coordinated operation of multiple actuators further improves etching uniformity.

[0010] In one optional embodiment, the plurality of sampling units are disposed at different spatial locations within the reaction chamber and connected to the infrared spectroscopy analysis unit via independent pipelines to obtain gas concentration, temperature, and pressure data at different spatial locations within the reaction chamber. The plurality of sampling units include at least: a first sampling unit disposed above the central region of the electrostatic chuck for collecting gas samples from the etching front of the central region; a second sampling unit disposed above the edge region of the electrostatic chuck for collecting gas samples from the edge region; a third sampling unit disposed below the plasma generation unit for collecting modified gas decomposition products in the plasma; and a fourth sampling unit disposed at the exhaust port of the reaction chamber. The system is used to collect gas samples from the exhaust port area and use the concentration of the gas samples from the exhaust port area as a background reference value. The first sampling unit and the second sampling unit are located at the same height layer, allowing the feedback adjustment module to detect radial spatial distribution differences by comparing the data measured by the two units. The first sampling unit and the third sampling unit are located at different height layers, allowing the feedback adjustment module to determine concentration changes at different height positions by comparing the data measured by the two units. The concentration values ​​of the first sampling unit, the second sampling unit, and the third sampling unit are corrected based on the background reference value measured by the fourth sampling unit. The data collected by the multiple sampling units are used together to generate the spatial distribution information of the concentration.

[0011] Beneficial effects: The first and second sampling units are located at the same height level, allowing the feedback adjustment module to detect radial spatial distribution differences by comparing their measured data; the first and third sampling units are located at different height levels, allowing the feedback adjustment module to determine concentration changes at different height locations by comparing their measured data; the concentration measured by the fourth sampling unit serves as a background reference value, enabling the feedback adjustment module to correct the concentration values ​​of the first to third sampling units. The four sampling units collectively acquire spatial distribution information inside the reaction chamber from three dimensions: radial distribution, longitudinal distribution, and background correction, providing a reliable data foundation for the feedback adjustment module to subsequently perform differentiated compensation for different regions.

[0012] In one optional embodiment, each sampling unit includes: a sampling pipeline; a microplate disposed at the inlet of the sampling pipeline to limit the sampling flow rate; a heating jacket wrapped around the outside of the sampling pipeline to prevent byproduct condensation; and a fast switching valve for sequentially activating each sampling unit according to a predetermined time sequence. The control system further includes a multi-channel gate valve group, the input end of which is connected to the outlet end of all the sampling pipelines, and the output end of which is connected to the infrared spectroscopy analysis unit. The fast switching valve sequentially activates each sampling unit according to a predetermined time sequence, allowing gas samples from each sampling unit to sequentially enter the multi-channel gate valve group, which then sequentially delivers the gas samples to the infrared spectroscopy analysis unit to achieve time-division measurement.

[0013] Beneficial Effects: This invention sets up a sampling pipeline, an inlet microporous plate, an external heating jacket, and a quick-switching valve in each sampling unit, and converges all sampling pipeline outlets to the infrared spectroscopy analysis unit through a multi-channel gate valve group. Each sampling unit achieves in-situ sampling through a sampling port pre-set in the cavity wall and a sampling probe extending into the cavity. A microporous plate is set at the inlet of the sampling pipeline to limit the sampling flow rate, so that the total extraction volume is much smaller than the total flow rate of the process gas, avoiding local airflow disturbance; the pipeline is wrapped with a heating jacket to prevent by-product condensation and adsorption. In terms of sampling strategy, the first, second, and third sampling units adopt a time-sharing circulation gas extraction mode, while the fourth sampling unit (exhaust port) maintains continuous gas extraction to provide a continuous and stable background reference value. Specifically, each sampling unit maintains continuous micro-flow gas extraction through the microporous plate when the quick-switching valve is off, and each sampling pipeline has an independent bypass exhaust channel, so that the extracted gas sample is continuously discharged through the bypass when it does not enter the detection flow path, ensuring that the gas sample always maintains dynamic flow in each pipeline. The infrared light source and detector can cover multiple optical paths, and the gas channels of the four infrared spectral detection channels are arranged in a 2x2 matrix to ensure detection accuracy. Meanwhile, the total continuous sampling flow rate of each microplate is much smaller than the total process gas flow rate, and the rapid switching valve action only changes the direction of a small amount of gas flow, without impacting the chamber pressure balance. Without altering the stability of the original process window and gas flow field, it provides the feedback adjustment module with spatially discernible, high-fidelity concentration spatial distribution information, ensuring the accuracy and reliability of the concentration information upon which the feedback adjustment is based. Simultaneously, it controls the disturbance of the original process window caused by sampling within an acceptable range, without affecting key process states such as etching rate, plasma density, and gas flow field.

[0014] In one optional embodiment, the infrared spectroscopy analysis unit includes: a broadband infrared light source or a wavelength-tunable laser for emitting infrared light; a gas absorption cell, which is a long-path cell, for ensuring sufficient interaction between the infrared light and the gas sample entering the cell; a detector, which is a thermoelectrically cooled mercury cadmium telluride infrared detector, for receiving the infrared light absorbed by the gas sample and converting it into an electrical signal; and a signal processing unit, electrically connected to the detector, for calculating the gas concentration at the corresponding position of each sampling unit based on the electrical signal, and using multiple linear regression to eliminate the overlap between adjacent absorption peaks. Cross interference is detected, and the concentration is sent to the feedback adjustment module. The temperature and pressure inversion unit, connected to the detector, receives the electrical signal generated by the detector, inverts the local temperature by measuring the full width at half maximum (FWHM) of the absorption peak of the impurity gas in the infrared spectrum corresponding to the electrical signal, and calculates the local pressure by measuring the Lorentz linewidth and combining it with the inverted local temperature, and sends the local temperature and local pressure to the feedback adjustment module. The feedback adjustment module is also configured to receive the spatial distribution information of gas concentration, local temperature and local pressure, and to generate the process adjustment command.

[0015] Beneficial Effects: This invention integrates a broadband infrared light source or wavelength-tunable laser, a long-path gas absorption cell, a thermoelectrically cooled mercury cadmium telluride infrared detector, and a signal processing unit into the infrared spectroscopy analysis unit. The long-path absorption cell improves the detection sensitivity of trace gases; the thermoelectrically cooled detector ensures signal stability; the signal processing unit calculates the concentration of each component using the Lambert-Beer law and eliminates cross-interference between adjacent absorption peaks using multiple linear regression, enabling simultaneous and accurate measurement of the concentration of multiple byproducts, providing reliable data support for multi-material interface detection and selective etching control. This invention also includes a temperature and pressure inversion unit in the detection module. This unit is connected to the detector, receives the electrical signal generated by the detector, inverts the local temperature by measuring the full width at half maximum (FWHM) of the impurity gas absorption peak, and calculates the local pressure by measuring the Lorentz linewidth and combining it with the inverted local temperature. This feature enables the same NDIR device to simultaneously acquire spatial distribution information of concentration, temperature, and pressure without adding additional sensors. The data are from the same source, have consistent dimensions, and are synchronized in time and space, which facilitates comprehensive judgment by the feedback adjustment module. In particular, it provides real-time input for the calculation of damage indication function based on the rate of change of molecular rotational temperature in ion damage suppression.

[0016] In one optional embodiment, the infrared spectroscopy analysis unit further includes: a plurality of narrowband filters, each narrowband filter corresponding to a characteristic absorption peak of a characteristic by-product, the characteristic by-product including at least one of SiF4, SiCl4, GeCl4, GaCl2, AlCl3 or COF2.

[0017] Beneficial Effects: This invention incorporates multiple narrowband filters within the infrared spectroscopy analysis unit. Each narrowband filter corresponds to a characteristic absorption peak of a specific byproduct, including at least one of SiF4, SiCl4, GeCl4, GaCl2, AlCl3, or COF2. This feature enables the system to selectively detect corresponding characteristic byproducts for different etching materials and process scenarios. For example, it can detect SiCl4 in silicon (Si) etching, GeCl4 in germanium (Ge) or SiGe etching, GaCl2 in GaN (gallium nitride) etching, AlCl3 in aluminum (Al) or Al2O3 etching, and COF2 in silicon-containing films (such as Si, SiO2, SiN) etching. This allows for precise monitoring of the etching process in different material systems and enhances the system's process compatibility.

[0018] In one optional embodiment, the feedback adjustment module includes: a transport simulation unit for calculating the transport attenuation of reactants at different spatial locations based on the spatial distribution information; a reaction kinetics simulation unit for calculating the growth rate and etching removal rate of the surface modification layer based on the spatial distribution information; a damage simulation unit for calculating the damage index accumulated by ion bombardment in the crystal lattice based on the spatial distribution information; and an instruction generation unit for generating the spatially differentiated process adjustment instruction based on the calculation results of the transport simulation unit, the reaction kinetics simulation unit, and the damage simulation unit; wherein the etching state includes the transport state determined by the transport simulation unit, the surface modification saturation degree determined by the reaction kinetics simulation unit, the degree of cumulative damage from ion bombardment determined by the damage simulation unit, and the material interface position.

[0019] Beneficial Effects: This invention incorporates a transport simulation unit, a reaction kinetics simulation unit, a damage simulation unit, and an instruction generation unit within the feedback regulation module. The transport simulation unit calculates the transport attenuation of reactants within the structure based on the spatial distribution of concentration. The reaction kinetics simulation unit calculates the growth rate of the modified layer and the etching removal rate, and determines the surface modification saturation level. The damage simulation unit calculates the damage index accumulated by ion bombardment in the crystal lattice and determines the degree of accumulated damage. The instruction generation unit integrates the calculation results from each unit to generate differentiated process regulation instructions. This feature enables digital twin predictive control based on a physical model, allowing for advanced simulation and optimization of the etching process. It predicts transport limitations, modification saturation, damage accumulation, and other states in advance, upgrading the control system from passive response to active prediction, further improving the accuracy and stability of process control.

[0020] In an optional implementation, the feedback adjustment module is further configured to: compare the rise time of byproduct concentration measured by different sampling units, calculate the effective diffusion coefficient using the concentration difference of sampling units at different vertical heights, and automatically increase the chamber pressure or increase the carrier gas flow rate in the surface modification step when the effective diffusion coefficient is lower than a preset threshold, and use closed-loop saturation determination to control the termination of the modification step.

[0021] Beneficial Effects: The feedback adjustment module of this invention is configured to compare the rise time of byproduct concentration measured by different sampling units, calculate the effective diffusion coefficient using the concentration difference between sampling units at different vertical heights, and automatically increase the chamber pressure or carrier gas flow rate in the surface modification step when the effective diffusion coefficient is lower than a preset threshold. A closed-loop saturation determination is used to control the termination of the modification step. This feature directly addresses the problem of limited reactant transport to the bottom in high aspect ratio structures (aspect ratio ≥ 40:1). By real-time diagnosis of transport efficiency and adaptive adjustment of process parameters (increasing pressure to increase the molecular mean free path and increasing lightweight carrier gas to enhance diffusion), it effectively improves the uniformity of trench etching at the bottom of the structure.

[0022] In an optional implementation, the feedback adjustment module is further configured to: estimate the transient temperature rise of the local lattice during etching by measuring the rotational temperature distribution of by-product molecules, and reduce the bias amplitude of the corresponding region or switch to pulse bias mode based on the comparison result of the product of the transient temperature rise and the bias energy with the damage threshold.

[0023] Beneficial Effects: The feedback adjustment module of this invention is configured to estimate the transient temperature rise of the local lattice during etching by measuring the rotational temperature distribution of by-product molecules. Based on the comparison between the product of this transient temperature rise and the bias energy and the damage threshold, the bias amplitude in the corresponding region is reduced or switched to pulsed bias mode. This feature addresses the problem of cumulative lattice damage induced by ion bombardment in the prior art by utilizing the sensitivity of molecular rotational temperature to ion energy, achieving real-time online sensing and immediate suppression of damage risk. Experiments have shown that in GaN HEMT (High Electron Mobility Transistor) gate trench etching, this method can reduce gate leakage current by more than 90% and interface state density by an order of magnitude, significantly improving device reliability.

[0024] In an optional implementation, the feedback adjustment module is further configured to: record the residual by-product concentration of each sampling unit at the end of the purging step, establish a dynamic baseline, and initiate active cleaning or extended purging for the area when the residual concentration in a certain area exceeds a preset multiple of the dynamic baseline.

[0025] Beneficial effects: The feedback adjustment module of this invention is also configured to record the residual byproduct concentration of each sampling unit at the end of the purging step, establish a dynamic baseline, and initiate active cleaning or extended purging for that area when the residual concentration in a certain area exceeds a preset multiple of the dynamic baseline. This feature addresses the spatial non-uniform evolution problem of chamber drift in the prior art. By tracking the changing trend of residual concentration at each port over a long period, it can distinguish between local drift (such as polymer accumulation at the exhaust port) and global drift (such as a decrease in pump speed) and take differentiated compensation measures.

[0026] In an optional implementation, the feedback adjustment module is further configured to: calculate the interface indicator factor of each sampling unit during multi-layer structure etching; determine that the interface switching is complete when the interface indicator factor of all sampling units is lower than the threshold; and for hysteresis regions, perform selective etching of the region by independently controlling the multi-region bias voltage.

[0027] Beneficial effects: The feedback adjustment module of this invention is further configured to calculate the interface indicator factor of each sampling unit during multi-layer structure etching. Interface switching is only determined to be complete when the interface indicator factors of all ports are below a threshold. For lagging regions, selective etching is performed by independently controlling the bias voltage of multiple regions. This feature addresses the problem of achieving full-wafer synchronous termination in multi-material interface etching in the background technology. By using multi-point determination, it ensures that the process is terminated only after all regions of the entire wafer have completed interface switching, avoiding the over-etching at the center or edge residue caused by signal lag and spatial non-uniformity in traditional global endpoint detection.

[0028] Secondly, the present invention provides an atomic layer etching control method, applied to the atomic layer etching control system of the first aspect above or any corresponding embodiment thereof, comprising: acquiring gas samples by means of multiple sampling units disposed at different spatial positions in the reaction chamber, performing infrared spectral analysis on each gas sample to determine the gas concentration at the corresponding position of each sampling unit, and generating spatial distribution information of the concentration; generating spatially differentiated process adjustment instructions by means of a feedback adjustment module based on the spatial distribution information of the concentration; receiving the process adjustment instructions by means of a control module, and respectively controlling the electrostatic chuck and / or the gas injection system to execute the process adjustment instructions, so as to adjust the spatial uniformity of the atomic layer etching process in the reaction chamber.

[0029] Beneficial effects: The atomic layer etching control method provided by this invention generates concentration spatial distribution information through a detection module, and the feedback adjustment module generates spatially differentiated process adjustment instructions based on this information. The control module receives and outputs these instructions, realizing a closed-loop control process based on spatial distribution information. This method can compensate in real time for etching non-uniformity caused by the non-uniformity of the internal space of the reaction chamber in the etching process, overcoming the fundamental defect of traditional global feedback control that cannot distinguish spatial differences, and improving the etching uniformity within the wafer surface.

[0030] In one optional embodiment, the atomic layer etching process includes a surface modification step, a first purging step, an etching removal step, and a second purging step performed sequentially in a cycle; the control method further includes: during each of the steps, acquiring gas samples from each sampling unit in a time-division manner according to a preset scanning sequence, wherein a complete scanning cycle of the scanning sequence is less than the duration of any of the steps.

[0031] Beneficial Effects: The method of this invention clearly defines the atomic layer etching process as including surface modification, first purging, etching removal, and second purging in a sequential cycle. During each process, gas samples from each sampling unit are acquired time-divisionally according to a preset scanning sequence, and the complete scanning cycle is shorter than the duration of any single process. This feature ensures the ability to capture dynamic changes within each half-step, enabling the detection data to reflect key transient characteristics such as changes in consumption rate during surface modification and peak times of byproducts during etching removal, providing real-time evidence for closed-loop saturation determination and transport limitation diagnosis.

[0032] In an optional implementation, the control method further includes: running a standard atomic layer etching recipe using a patternless wafer, recording reference characteristic quantities for each sampling unit, including saturation time in the surface modification step, peak time and peak concentration of byproducts in the etching removal step, and residual concentration at the end of the purging step; and storing the reference characteristic quantities in a database.

[0033] Beneficial effects: This step is the initial calibration step, which provides a benchmark for deviation detection in subsequent production. It enables the feedback adjustment module to determine whether there is an anomaly by comparing it with the reference value. For example, the quantitative evaluation of rise time delay in transmission limitation judgment depends on this reference database.

[0034] In an optional implementation, if the aspect ratio of the etched structure is greater than or equal to 40:1, the control method further includes: in the etching removal step, comparing the rise time of the byproduct concentration measured by different sampling units and comparing it with a reference value; if the rise time delay exceeds a first threshold and the delay difference between different ports exceeds a second threshold, then it is determined that there is a transmission limitation; if it is determined that there is a transmission limitation, at least one of the following compensation instructions is generated by the feedback adjustment module: increasing the chamber pressure of surface modification, increasing the proportion of light carrier gas in the modified gas, and switching the termination mode of surface modification from a fixed duration to a closed-loop saturation determination mode.

[0035] Beneficial Effects: When the aspect ratio of the etched structure is greater than or equal to 40:1, the method of this invention compares the rise time of the byproduct concentration measured by different sampling units with a reference value. If the delay exceeds a threshold, it is determined that there is a transport limitation. The feedback adjustment module then generates compensation commands to increase the pressure, add a light carrier gas, or switch the termination mode to closed-loop saturation judgment, which are then executed by the control module. This method directly addresses the problem of limited reactant transport to the bottom in high aspect ratio structures. It diagnoses the degree of transport limitation through the observable indicator of rise time delay and automatically performs multi-mode compensation. Experiments have shown that it can significantly improve the etching uniformity at the bottom of the trench.

[0036] In one optional implementation, the closed-loop saturation determination step includes: real-time detection of the concentration of modified gas decomposition products in each sampling unit, and termination of the surface modification when the concentration change rate of all sampling units is lower than a preset threshold.

[0037] Beneficial effects: The closed-loop saturation determination in the method of this invention is specifically implemented by real-time detection of the concentration of modified gas decomposition products in each sampling unit. Surface modification is terminated only when the concentration change rate at all ports is lower than a preset threshold. This ensures that all areas of the entire wafer reach modification saturation before entering the next cycle, avoiding insufficient etching due to local unsaturation or plasma pre-damage introduced by oversaturation. The modification time fluctuation is controlled within ±0.1s, and the surface roughness RMS is stabilized below 0.2nm.

[0038] In an optional embodiment, the control method further includes: during the surface modification step, acquiring in real time the consumption rate of the modified gas or its decomposition products at the plurality of sampling units; when the consumption rate at all sampling units is lower than a preset threshold, determining that the entire region is saturated and terminating the surface modification; when the time difference between each sampling unit reaching saturation exceeds a preset duration, generating a process adjustment command through the feedback adjustment module to increase the flow rate of the gas injection system corresponding to the hysteresis region.

[0039] Beneficial effects: The method of this invention acquires the consumption rate of the modified gas or its decomposition products in each sampling unit in real time during surface modification. When the consumption rate of all ports is lower than a threshold, the entire region is determined to be saturated. For lagging regions where the saturation time difference exceeds a preset time, the feedback adjustment module generates instructions, which are output by the control module to automatically increase the flow rate of the corresponding gas injection system. This feature achieves spatially differentiated adaptive adjustment. Through closed-loop feedback, it dynamically compensates for the saturation time difference caused by uneven gas flow field distribution, making the modification saturation of different regions tend to be synchronized, and significantly improving the uniformity of wafer surface modification.

[0040] In an optional embodiment, the control method further includes: during the etching removal step, acquiring in real time the molecular rotation temperature of the etching byproducts at the plurality of sampling units; calculating the damage indication function at the corresponding position of each sampling unit based on the rate of change of the molecular rotation temperature; and when the damage indication function at any position exceeds a preset safety threshold, generating a process adjustment command through the feedback adjustment module to reduce the bias amplitude of the electrostatic chuck region corresponding to that position, switch to pulse bias mode, and / or increase the local temperature of that region.

[0041] Beneficial effects: The method of this invention acquires the molecular rotation temperature of etching byproducts in each sampling unit in real time during etching removal, calculates a damage indicator function based on its rate of change, and when a safety threshold is exceeded, the feedback adjustment module generates adjustment commands to reduce the bias voltage amplitude, switch to pulse bias mode, and / or increase the local temperature, which are then executed by the control module. This method utilizes the sensitivity of molecular rotation temperature to ion energy to perceive damage risk in real time, achieving a sub-millisecond rapid response and effectively suppressing cumulative lattice damage.

[0042] In one alternative implementation, the damage indication function is proportional to the rate of change of the molecular rotational temperature and proportional to the difference between the current bias energy and the preset minimum energy required for complete removal of the modified layer.

[0043] Beneficial effects: The method of this invention defines a damage indicator function that is proportional to the rate of change of molecular rotational temperature and also proportional to the difference between the current bias energy and the minimum energy required to completely remove the modified layer. This function form is based on a physical model, converting excess ion energy into lattice heat and quantifying damage risk into a calculable index. This allows the controller to set a clear safety threshold, avoiding reliance on empirical judgment and improving the reliability and repeatability of damage suppression.

[0044] In an optional implementation, when the etched structure includes at least a stacked structure of a first material layer and a second material layer, the control method further includes: acquiring the concentrations of the first and second byproducts at the plurality of sampling units in real time; calculating the interface indicator factor at each sampling unit, wherein the interface indicator factor is the ratio of the concentration of the second byproduct to the sum of the concentrations of the first and second byproducts; determining that the interface switching of the entire region is completed and terminating the etching when the interface indicator factor at all sampling units is lower than a preset interface threshold; generating a process adjustment command through the feedback adjustment module to apply an additional etching bias pulse to the electrostatic chuck at the corresponding position of the hysteresis region, and outputting the process adjustment command through the control module until the interface indicator factor at all sampling units is lower than the interface threshold.

[0045] Beneficial effects: When the etched structure contains multilayered materials, this invention acquires the concentrations of the first and second byproducts in real time, calculates the interface indicator factor, and determines that the interface switching is complete only when the interface indicator factor of all ports is below the threshold. The feedback adjustment module then generates additional etching bias pulse commands for the hysteresis region, which are executed by the control module. This method, through multi-point independent determination and region-selective etching, controls the interface etching depth difference within ±0.5nm, solving the problem of synchronous termination of the entire wafer in multilayer material interface etching.

[0046] In an optional implementation, the control method further includes: recording the residual byproduct concentration at the multiple sampling units at the end of the atomic layer etching cycle; establishing a dynamic baseline for each sampling unit based on the residual byproduct concentration; determining that contaminant accumulation exists in the chamber region corresponding to the sampling unit when the residual byproduct concentration at any sampling unit continuously exceeds a preset multiple of its corresponding dynamic baseline; generating corresponding compensation commands based on the drift mode: if the residual concentration at only the port near the exhaust port exceeds the threshold, generating a local purging pulse or an in-situ short cleaning command; if the residual concentration at all ports simultaneously exceeds the threshold, generating a comprehensive in-situ cleaning command; if the residual concentration at only the near-field port increases and is related to the bias voltage, generating a surface charge neutralization or extended purging command.

[0047] Beneficial effects: The method of this invention records the residual byproduct concentration of each sampling unit at the end of the purging process, establishes a dynamic baseline for each port, and determines contaminant accumulation when the residual concentration continuously exceeds a preset multiple of the baseline, and generates corresponding compensation instructions based on the drift pattern. This method achieves spatial adaptive compensation for chamber drift, can distinguish different types of drift and take precise measures, extends the average cleaning interval of the chamber, and improves equipment utilization.

[0048] In an optional implementation, the control method further includes: accumulating all spatiotemporal data collected in multiple atomic layer etching batches, the spatiotemporal data including concentration-time series, temperature-time series, pressure-time series, process parameter records, and actuator action records of each sampling unit; training a hybrid neural network model using the spatiotemporal data and corresponding offline etching results, the hybrid neural network model including a convolutional neural network for extracting spatial distribution features and a long short-term memory network for extracting temporal dynamic features; deploying the trained model in a feedback adjustment module for online prediction of etching depth distribution and / or damage index, and optimizing process parameters under constraints; and incrementally learning and fine-tuning the model using new data after each predetermined number of wafers are produced.

[0049] Beneficial Effects: This invention accumulates all spatiotemporal data from multiple batches, trains a hybrid neural network model incorporating CNN and LSTM, and deploys the model in a feedback adjustment module for online prediction of etching depth distribution and damage index. It also optimizes process parameters under constraints and supports incremental learning fine-tuning. This digital twin self-optimization method utilizes big data and deep learning technologies to upgrade process control from rule-based feedback to prediction-based optimization, significantly reducing the frequency of manual adjustments and providing an intelligent tool for rapid process development of new products.

[0050] Thirdly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the atomic layer etching control method of the second aspect or any corresponding embodiment described above.

[0051] Fourthly, the present invention provides a semiconductor atomic layer etching apparatus, including the atomic layer etching control system of the first aspect above or any corresponding embodiment thereof.

[0052] This invention acquires the gas concentration distribution at different spatial locations within the reaction chamber using a spatially perceptual nondispersive infrared spectroscopy device. A feedback adjustment module generates spatially differentiated process adjustment commands based on this spatial distribution information. The control module outputs these commands to drive the actuators to execute them, thereby achieving closed-loop optimization control of the spatial uniformity of the etching process. This enables the etching uniformity within the wafer to be controlled at the atomic level, significantly improving device consistency and yield. Attached Figure Description

[0053] 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.

[0054] Figure 1 This is a structural block diagram of the atomic layer etching control system according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the first process of the atomic layer etching control method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a second process for the atomic layer etching control method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the third process of the atomic layer etching control method according to an embodiment of the present invention; Figure 5This is a schematic diagram of the fourth process of the atomic layer etching control method according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the fifth process of the atomic layer etching control method according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the sixth process of the atomic layer etching control method according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the seventh process of the atomic layer etching control method according to an embodiment of the present invention; Figure 9 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0055] 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.

[0056] 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.

[0057] 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.

[0058] Reference Figure 1As shown, this embodiment provides an atomic layer etching control system, including: an etching module 101, a detection module 102, a feedback adjustment module 103, and a control module 104. The etching module 101 includes a reaction chamber, and an electrostatic chuck and a gas injection system disposed within the reaction chamber. The detection module 102 includes multiple sampling units disposed at different spatial locations within the reaction chamber, an infrared spectroscopy analysis unit, and a temperature and pressure inversion unit; the multiple sampling units are used to extract gas samples from different spatial locations within the reaction chamber, and the infrared spectroscopy analysis unit is used to perform infrared spectral analysis on the gas samples to determine the gas concentration at the corresponding location of each sampling unit, and to generate spatial distribution information of the gas concentration based on the location information of each sampling unit. The feedback adjustment module 103 is used to calculate the etching state of different regions based on the spatial distribution information of the concentration provided by the detection module 102, and to generate spatially differentiated process adjustment commands. The control module 104 is communicatively connected to the etching module 101, the detection module 102, and the feedback adjustment module 103. It is used to receive the process adjustment instructions generated by the feedback adjustment module 103 and output the process adjustment instructions to the electrostatic chuck, the bias power supply unit, and / or the gas injection system, respectively, so as to adjust the spatial uniformity of the etching process in the reaction chamber.

[0059] To acquire gas concentration, temperature, and pressure data at different locations within the reaction chamber, the detection module 102 is equipped with four sampling units, respectively located above the center area of ​​the electrostatic chuck (first sampling unit), above the edge area (second sampling unit), below the plasma generation unit (third sampling unit), and at the exhaust port (fourth sampling unit). Each sampling unit achieves in-situ sampling through sampling ports pre-installed on the chamber wall and sampling probes extending into the chamber. A microporous plate is installed at the inlet of the sampling pipeline to limit the sampling flow rate, ensuring that the total extracted volume is much smaller than the total process gas flow rate, thus avoiding local airflow disturbance. A heating jacket is wrapped around the outside of the pipeline to prevent byproduct condensation and adsorption. In terms of sampling strategy, the first, second, and third sampling units adopt a time-sharing cyclic gas extraction mode, while the fourth sampling unit (exhaust port) maintains continuous gas extraction to provide a continuous and stable background reference value. Specifically, each sampling unit maintains a continuous micro-flow of gas through a micro-orifice plate when the fast-switching valve is off. Each sampling pipeline also has an independent bypass exhaust channel, ensuring that the extracted gas sample is continuously discharged via the bypass when it does not enter the detection flow path, thus maintaining dynamic flow of the gas sample within each pipeline. When a spatial point (such as the first, second, or third unit) needs to be measured according to a predetermined sequence, the fast-switching valve for that point opens instantaneously, controlling a very short sampling pulse. The gas sample at that point at the current moment is instantaneously switched into the infrared spectral detection channel of the multi-channel gate valve group via an independent interface, and then the valve closes. The unit immediately resumes bypass exhaust after the pulse ends, awaiting the next measurement cycle. The remaining unselected sampling units continue to maintain bypass exhaust during this period, unaffected by the switching. Because the gas sample volume introduced by a single sampling pulse is extremely small, and each pipeline is in bypass exhaust mode for the vast majority of the time, the four optical detection channels do not interfere with each other. Combined with full-process insulation by the heating jacket to reduce adsorption sites, this completely avoids cross-interference of subsequent signals by residual adsorption from previous gas. The infrared light source and detectors cover multiple optical paths, and the gas channels of the four infrared spectral detection channels are arranged in a 2x2 matrix to ensure detection accuracy. Simultaneously, the total continuous sampling flow rate of each microplate is much smaller than the total process gas flow rate, and the rapid switching valve action only changes the direction of a small amount of gas flow, without impacting the chamber pressure balance. Each sampling unit has its own completely independent, fully conductive gas sample transmission pipeline, directly connected to the corresponding infrared spectral detection channel (or integrated multi-path parallel detection chamber). Each detection channel can independently perform continuous spectral analysis on its received gas sample. Under this parallel architecture, the background reference value continuously measured at the exhaust port by the fourth sampling unit can be used to correct the time-division measurement data of the first to third sampling units in real time, simultaneously eliminating the influence of chamber background fluctuations and pipeline transmission losses on concentration determination.Meanwhile, since the first three units are time-sharing short-interval cyclic gas extraction and the fourth unit is continuous gas extraction, the total flow rate of multi-path time-sharing and continuous sampling is still much smaller than the process gas flow rate. Moreover, each pipeline is independently constant flow connected, avoiding the impact of single-path switching on the chamber pressure balance. Thus, without changing the original process window and the stability of the airflow field, it provides the feedback adjustment module 103 with spatially discernible high-fidelity concentration spatial distribution information, ensuring that the concentration information on which the feedback adjustment is based is accurate and reliable. At the same time, it controls the disturbance of the sampling behavior on the original process window within an acceptable range, without affecting key process states such as etching rate, plasma density and airflow field.

[0060] The aforementioned atomic-level etching control system utilizes a spatially perceptible non-dispersive infrared spectroscopy device, including multiple sampling units, an infrared spectroscopy analysis unit, and a temperature and pressure inversion unit. Multiple sampling units collect gas samples at different spatial locations within the reaction chamber. Infrared spectroscopy analysis determines the gas concentration, temperature, and pressure at each location, generating spatial distribution information. A feedback adjustment module 103 generates spatially differentiated process adjustment commands based on this spatial distribution information. A control module 104 receives and outputs these process adjustment commands, achieving closed-loop control based on spatial distribution information. Compared to traditional single-point tail-out detection methods that only obtain global average concentration, this invention can sense the spatial distribution differences of concentration, temperature, and pressure within the reaction chamber and perform differentiated compensation for different regions. This effectively overcomes the spatial non-uniformity problem caused by uneven distribution of reactant concentration, ion energy, and temperature within the chamber, controlling wafer etching uniformity at the atomic level and significantly improving device consistency and yield.

[0061] Specifically, the reaction chamber is a vacuum chamber with heating and cooling channels on its walls to maintain the inner wall temperature at 80-150°C to prevent byproduct adhesion. An electrostatic chuck is mounted at the bottom of the chamber to hold the wafer. The gas injection system is arranged in a ring and connected to a mass flow controller (MFC) and sources of modifying, etching, and dilution gases. The modifying gas sources include Cl2 and BCl3, while the etching gas sources include Ar and He.

[0062] As a specific implementation of an electrostatic chuck, it is radially divided into at least a central region, a middle region, and an edge region. The electrostatic chuck includes a temperature control unit, with each region having an independent temperature control unit. The temperature control unit responds to spatially differentiated process adjustment commands output by the control module 104, independently adjusting the temperature of different radial regions of the wafer.

[0063] The aforementioned multi-zone temperature-controlled electrostatic chuck, through independent temperature control in each zone, can compensate for local reaction rate differences caused by uneven gas flow or plasma distribution, making the growth of modified layers in each zone tend to be synchronized, thereby improving the uniformity of surface modification and avoiding local over-modification or under-modification.

[0064] As a variation of the electrostatic chuck, the temperature control unit of each area can be composed of independent heating elements and thermocouples. The feedback adjustment module 103 sends proportional-integral-derivative (PID) control signals to each area through the control module 104 based on the local temperature distribution fed back by the detection module 102, so that the temperature of each area can be independently adjusted within the range of -50℃ to 400℃, and the temperature difference control accuracy between areas is better than ±0.1℃.

[0065] In some embodiments of the present invention, the etching module 101 further includes a bias power supply unit electrically connected to the electrostatic chuck. The bias power supply unit is used to independently supply bias voltage to each region of the electrostatic chuck, and the bias voltage amplitude of each region is independently adjustable to a range of 20 eV to 200 eV corresponding to ion energies, and the pulse duty cycle of each region is independently adjustable to a range of 5% to 100%. The feedback adjustment module 103 is also configured to independently adjust the bias voltage parameters for different regions, and outputs through the control module 104 to independently adjust the etching rate of different radial regions of the wafer.

[0066] The aforementioned multi-zone bias power supply unit, by independently adjusting the bias amplitude and pulse duty cycle for each zone, can directly compensate for the ion flux attenuation in the edge region, ensuring uniform etching depth within the wafer surface. For example, in silicon trench etching with an aspect ratio of 40:1, when a significant time delay in the rise of byproduct concentration is detected in the edge region, the feedback adjustment module 103 can generate a command to increase the bias amplitude in the edge region, and the control module 104 outputs this command to synchronize the etching rate with the central region.

[0067] As a variation of the bias power supply unit, the bias power supply unit can be multiple independent radio frequency (RF) power supplies, each connected to an electrostatic chuck area; or it can be a multi-channel output power supply that provides adjustable bias to each area through a power divider. The pulse duty cycle can be adjusted by controlling the switching sequence of the power supply, with a frequency range of 2-60MHz.

[0068] In one embodiment, the gas injection system includes multiple radial zones, each with an independent gas channel and mass flow controller, for responding to spatially differentiated process adjustment commands output by the control module 104 to independently adjust the process gas concentration in different radial zones of the wafer.

[0069] The aforementioned multi-zone gas injection system can compensate for concentration gradients caused by factors such as inlet distribution and chamber geometry by independently adjusting the gas flow rate in each radial zone. For example, when the detection module 102 detects that the concentration of modified gas decomposition products in the edge region is lower than that in the center region, the feedback adjustment module 103 can generate a command to increase the MFC flow rate in the edge radial zone, and the control module 104 outputs this command to make the gas concentration distribution on the wafer surface more uniform.

[0070] As a variation of the gas injection system, the number of radial regions in the gas injection system can be three, including a central region, an intermediate region, and an edge region, or five or more, depending on the wafer size and process requirements. The gas channels in each radial region can be connected to different gas sources, or they can share the same gas source through a switching valve.

[0071] As a further improvement to the etching module 101, the etching module 101 also includes a plasma generation unit and a vacuum unit. The plasma generation unit is communicatively connected to the control module 104 and is used to adjust the plasma parameters in the reaction chamber in response to process adjustment commands. The vacuum unit is also communicatively connected to the control module 104 and is used to adjust the pressure and by-product discharge rate in the reaction chamber in response to process adjustment commands.

[0072] The aforementioned plasma generation and pumping units enable the control module 104 to not only adjust the bias voltage and gas injection, but also to participate in the spatial uniformity regulation from two dimensions: plasma excitation conditions and chamber environment. For example, when a transport limitation is detected in a high aspect ratio structure, the feedback regulation module 103 can generate a command to close the throttle valve to increase the chamber pressure. The control module 104 outputs this command, for example, increasing it from 10 mTorr to 35 mTorr, thereby increasing the molecular mean free path and improving the diffusion of reactants to the bottom of the structure.

[0073] As a variation of the plasma generating unit, the plasma generating unit can be an inductively coupled plasma (ICP) type with a frequency of 13.56MHz and a maximum power of 3000W; or a capacitively coupled plasma (CCP) type with a frequency of 400kHz. The feedback adjustment module 103 can adjust the plasma power and the phase distribution of the coil, thereby changing the spatial distribution of the plasma density.

[0074] In some embodiments of the present invention, the spatial arrangement of multiple sampling units within the reaction chamber is as follows: a first sampling unit is positioned above the central region of the electrostatic chuck for collecting gas samples from the etching front of the central region; a second sampling unit is positioned above the edge region of the electrostatic chuck for collecting gas samples from the edge region; a third sampling unit is positioned below the plasma generating unit for collecting modified gas decomposition products in the plasma; and a fourth sampling unit is positioned at the exhaust port of the reaction chamber for collecting gas samples from the exhaust port region, using the gas sample concentration in the exhaust port region as a background reference value. Specifically, the first and second sampling units are located at the same height level, allowing the feedback adjustment module 103 to detect radial spatial distribution differences by comparing the data measured by both; the first and third sampling units are located at different height levels, allowing the feedback adjustment module 103 to determine concentration changes at different height positions by comparing the data measured by both; the concentration values ​​of the first to third sampling units are corrected based on the background reference value measured by the fourth sampling unit, and the data collected by multiple sampling units are used together to generate spatial distribution information of the concentration.

[0075] The first and second sampling units are located at the same height level, allowing the feedback adjustment module 103 to detect radial spatial distribution differences by comparing their measured data. The first and third sampling units are located at different height levels, allowing the feedback adjustment module 103 to determine concentration changes at different height locations by comparing their measured data. The concentration measured by the fourth sampling unit serves as a background reference value, which the feedback adjustment module 103 can use to correct the concentration values ​​of the first to third sampling units. The four sampling units jointly acquire spatial distribution information inside the reaction chamber from three dimensions: radial distribution, longitudinal distribution, and background correction, providing a reliable data foundation for the feedback adjustment module 103 to perform differentiated compensation for different regions.

[0076] As one specific embodiment of the detection module 102, it includes multiple sampling units and an infrared spectroscopy analysis unit. The multiple sampling units are located at different spatial positions within the reaction chamber, and each sampling unit is used to extract a gas sample from the reaction chamber. The infrared spectroscopy analysis unit is connected to the multiple sampling units via pipelines and is used to measure the infrared spectrum of the gas sample extracted by each sampling unit in a time-division manner. Based on the infrared spectrum, it determines the gas concentration at the corresponding location of each sampling unit, and simultaneously determines the local temperature and local pressure at the corresponding location of each sampling unit based on the spectral characteristics of the infrared spectrum. The feedback adjustment module 103 is also configured to receive the spatial distribution information of gas concentration, local temperature, and local pressure determined by the infrared spectroscopy analysis unit, and generate process adjustment commands.

[0077] The aforementioned detection module 102, through multiple sampling units and time-division multiplexed infrared spectroscopy analysis, achieves real-time spatial resolution reconstruction of the concentration field, temperature field, and pressure field inside the reaction chamber, providing a data foundation for differentiated closed-loop control.

[0078] As a variation of the sampling unit implementation, the number of sampling units can be four, respectively located in the near field above the wafer center, the near field above the wafer edge, near the plasma source, and near the exhaust port. Alternatively, there can be six, eight, or more units to increase spatial resolution. Each sampling unit can spatially correspond one-to-one with the radial regions of the electrostatic chuck and the gas injection system; for example, the first sampling unit corresponds to the central region, and the second sampling unit corresponds to the edge region.

[0079] In some specific embodiments, each sampling unit includes: a sampling pipeline; a microplate disposed at the inlet of the sampling pipeline to limit the sampling flow rate; a heating jacket wrapped around the outside of the sampling pipeline to prevent byproduct condensation; and a fast-switching valve for sequentially activating each sampling unit according to a predetermined time sequence. The control system also includes a multi-channel gate valve group, the input of which is connected to the outlet of all sampling pipelines, and the output of which is connected to the infrared spectroscopy analysis unit. The fast-switching valve sequentially activates each sampling unit according to a predetermined time sequence, allowing the gas samples from each sampling unit to sequentially enter the multi-channel gate valve group. The multi-channel gate valve group then sequentially delivers the gas samples to the infrared spectroscopy analysis unit to achieve time-division measurement.

[0080] The above-mentioned sampling unit structure uses a microplate to limit the sampling flow rate to within 1-3 sccm, resulting in less than 0.05% disturbance to the chamber pressure; the heating jacket temperature is controlled at 120±10℃ to prevent condensation and blockage by high-boiling-point byproducts such as SiCl4 and AlCl3; and the combination of a fast switching valve and a multi-channel selector valve group allows a single infrared spectroscopy analysis unit to serve multiple sampling units, reducing system complexity and cost.

[0081] As a modified implementation of the sampling unit, the sampling pipeline is a 316L stainless steel capillary tube with an inner diameter of 0.8 mm; the pore size of the microplate is 5 μm; the fast switching valve is a piezoelectric valve with an opening and closing time ≤10 ms; the multi-channel valve group has 16 channels and an internal volume of less than 10 μL. The heating jacket can be of the resistance wire heating type or fluid heating type, and each sampling pipeline has independent temperature control.

[0082] In some specific embodiments of the present invention, the infrared spectroscopy analysis unit includes: an infrared light source, which may be a broadband infrared light source or a wavelength-tunable laser, for emitting infrared light; a gas absorption cell, which is a long optical path cell, for enabling the infrared light to fully interact with the gas sample entering the cell; a detector, which is a thermoelectrically cooled mercury cadmium telluride infrared detector, for receiving the infrared light absorbed by the gas sample and converting it into an electrical signal; and a signal processing unit, electrically connected to the detector, for calculating the gas concentration at the corresponding position of each sampling unit based on the electrical signal, and using multiple linear regression to eliminate cross-interference between adjacent absorption peaks, and sending the concentration to the feedback adjustment module 103.

[0083] The aforementioned infrared spectroscopy analysis unit features a long-path cell that enhances the detection sensitivity of trace gases, achieving a detection limit of 1 ppm; a thermoelectrically cooled detector ensures signal stability; and a signal processing unit calculates concentrations and eliminates cross-interference using Lambert-Beer's law, enabling simultaneous and accurate measurement of multiple byproduct concentrations. The optical path of the long-path cell can be 1.5 m.

[0084] As a variation of the infrared spectroscopy analysis unit, the infrared light source can be a wide-tuned pulsed quantum cascade laser (QCL) with a wavelength range of 5-12 μm and a spectral resolution ≤2 cm⁻¹. -1 The pulse repetition frequency is 1MHz; alternatively, a thermal radiation infrared light source combined with a filter wheel can be used, which is less expensive. The gas absorption cell can be a Herriott-type cell with gold-plated inner walls to prevent adsorption; the cell volume is 180mL, and the sample gas passes through at a flow rate of 100sccm during measurement. The thermal radiation infrared light source can be SiC ceramic, etc.

[0085] As an improved embodiment of the detection module 102, a temperature and pressure inversion unit is also included. The temperature and pressure inversion unit is connected to the detector and is used to receive the electrical signal generated by the detector, invert the local temperature by measuring the full width at half maximum (FWHM) of the absorption peak of the impurity gas in the infrared spectrum corresponding to the electrical signal, and calculate the local pressure by measuring the Lorentz linewidth and combining it with the inverted local temperature, and send the local temperature and local pressure to the feedback adjustment module 103.

[0086] The aforementioned temperature and pressure inversion unit is connected to the detector and receives the electrical signals generated by the detector. It simultaneously acquires spatial distribution information of concentration, temperature, and pressure without adding additional sensors, ensuring that the data are from the same source, have consistent dimensions, and are synchronized in time and space. For example, by measuring CO2 at 668 cm⁻¹... -1 The half-width at half-maximum (FWHM) of the nearby absorption peaks is used to invert the gas temperature based on the square root relationship between FWHM and temperature; then the Lorentz linewidth is measured, and the local pressure is calculated in conjunction with the temperature.

[0087] As a variation of the temperature and pressure inversion unit, the impurity gas can be residual water vapor (H2O) or intentionally added tracer gas (such as CO2). In the near-field region, the gas temperature can be approximately equal to the wafer surface temperature, with an error of <10°C. The temperature and pressure inversion unit can be integrated into the signal processing unit or it can be a separate digital signal processor. It should be noted that the near-field region is the space region 5-15 mm above the wafer surface.

[0088] In some specific embodiments of the present invention, the infrared spectroscopy analysis unit further includes multiple narrowband filters. Each narrowband filter corresponds to a characteristic absorption peak of a characteristic byproduct, including SiF4 (1030 cm⁻¹). -1 ), SiCl4 (620cm) -1 ), GeCl4 (460cm) -1 ), GaCl2 (415cm) -1 AlCl3 (625cm) -1 ) or COF2 (1930cm) -1 At least one of the following.

[0089] The aforementioned narrowband filters enable the system to selectively detect corresponding characteristic byproducts for different etching materials and process scenarios. For example, it can detect SiCl4 in silicon (Si) etching, GeCl4 in germanium (Ge) or SiGe etching, GaCl2 in GaN (gallium nitride) etching, AlCl3 in aluminum (Al) or Al2O3 etching, and COF2 in silicon-containing films (such as Si, SiO2, SiN) etching. By switching the filter wheel or tunable filters, the detection target can be quickly switched in different process steps, enhancing the system's process compatibility.

[0090] As a variation of a narrowband filter, the half-width at half-maximum (WHM) of the narrowband filter is approximately 25 cm. -1 Each filter corresponds to a characteristic peak of a byproduct. Acousto-optic tunable filters (AOTF) or liquid crystal tunable filters (LCTF) can also be used to replace mechanical filter wheels, achieving rapid wavelength switching without mechanical movement.

[0091] In some specific embodiments of the present invention, the feedback adjustment module 103 includes: a transport simulation unit, used to calculate the transport attenuation of reactants at different spatial locations based on spatial distribution information; a reaction kinetics simulation unit, used to calculate the growth rate and etching removal rate of the surface modification layer based on spatial distribution information to determine the surface modification saturation level; a damage simulation unit, used to calculate the damage index accumulated by ion bombardment in the lattice based on spatial distribution information to determine the degree of ion bombardment cumulative damage; and an instruction generation unit, used to generate spatially differentiated process adjustment instructions based on the calculation results of the transport simulation unit, the reaction kinetics simulation unit, and the damage simulation unit; wherein, the etching state includes the transport state determined by the transport simulation unit, the surface modification saturation level determined by the reaction kinetics simulation unit, the degree of ion bombardment cumulative damage determined by the damage simulation unit, and the material interface position.

[0092] The aforementioned simulation units constitute a digital twin engine. The transport simulation unit, based on the Knudsen diffusion model or Monte Carlo ion trajectory simulation, calculates the reactant concentration decay factor within the high aspect ratio structure; the reaction kinetics simulation unit, based on the Langmuir adsorption model, calculates the modified layer growth rate and etching removal rate, and determines the surface modification saturation level; the damage simulation unit, based on the Kinchin-Pease collision cascade theory, calculates the cumulative damage index and determines the cumulative damage level caused by ion bombardment. The instruction generation unit integrates the outputs of each unit to predict the etching state in advance, achieving proactive predictive control.

[0093] It should be noted that the above simulation units can be implemented by software modules running on the controller, or by hardware acceleration using a Field-Programmable Gate Array (FPGA). The model parameters of each unit are calibrated through offline experiments and continuously optimized through incremental learning during production.

[0094] As a further configuration of the feedback adjustment module 103, the feedback adjustment module 103 is also configured to: compare the rise time of the by-product concentration measured by different sampling units, and calculate the effective diffusion coefficient using the concentration difference of sampling units at different vertical heights; when the effective diffusion coefficient is lower than a preset threshold, automatically increase the chamber pressure or increase the carrier gas flow rate in the surface modification step, and use closed-loop saturation judgment to control the termination of the modification step.

[0095] One specific example of a preset threshold is 90% of the theoretical Knudsen diffusion coefficient. For instance, in trench etching with an aspect ratio of 60:1, the effective diffusion coefficient D is estimated by the SiCl4 concentration difference between the near-wafer port and the port near the top electrode. eff If D eff=0.12m² / s, lower than the theoretical Knudsen diffusion coefficient of 0.28m. 2 / S 90% (i.e. 0.196m) 2 If the speed is less than or equal to 10 mTorr, a severe transmission limitation is determined. In response, the feedback adjustment module 103 automatically generates instructions to gradually increase the surface modification pressure from 10 mTorr to 35 mTorr, increase the proportion of light carrier gas (Ar) in the modified gas from 0% to 30%, and switch the modification termination method from a fixed duration to a closed-loop saturation determination method. The control module 104 outputs these instructions. The rise time and D are then reassessed after compensation. eff Continue until the requirements are met. Experiments have shown that this method can improve the uniformity of etching depth in trenches with an aspect ratio of 60:1 from ±25% to within ±5%.

[0096] Wherein, the effective diffusion coefficient D eff It can be estimated using the concentration difference and vertical distance between the near-wafer port and the port near the top electrode. When D eff Compensation is triggered when the concentration falls below 90% of the theoretical Knudsen diffusion coefficient. In the closed-loop saturation determination, the concentration change rate threshold can be set to below 2% of the initial maximum value.

[0097] The feedback adjustment module 103 is also configured to: estimate the transient temperature rise of the local lattice during etching by measuring the rotational temperature distribution of by-product molecules, and reduce the bias amplitude of the corresponding region or switch to pulse bias mode based on the comparison result of the product of the transient temperature rise and the bias energy with the damage threshold.

[0098] For example, in the etching of the gate trench in GaN HEMT, the rotation temperature T of GaCl2 molecules is measured. rot Calculate the damage indicator function: , Where E bias E represents the currently applied bias energy. th E is the minimum energy required to completely remove the preset modified layer. th =55eV. When D indicator When the safety threshold of 250 a.u. is exceeded, the feedback adjustment module 103 generates a command within 10 ms to reduce the bias voltage of the corresponding area from 90V to 70V and switch the continuous bias voltage to pulse mode (40% duty cycle). The control module 104 outputs this command. Simultaneously, the chuck temperature in this area can be increased by 5-10°C to promote annealing. When D indicator After five consecutive cycles below the threshold, the feedback adjustment module 103 generates instructions to gradually restore the bias voltage (increasing by 2V per cycle) and duty cycle, and the control module 104 outputs these instructions. This method can reduce the gate leakage current by more than 90%.

[0099] In addition, the chuck temperature in this area can be increased by 5-10°C simultaneously to promote in-situ annealing of point defects. After the damage indication function remains within the safe range for 5 consecutive cycles, the feedback adjustment module 103 gradually generates instructions for recovery bias and duty cycle by increasing by 2V per cycle, and the control module 104 outputs these instructions.

[0100] The feedback adjustment module 103 is also configured to: record the concentration of residual byproducts in each sampling unit at the end of the purging step, establish a dynamic baseline, and initiate active cleaning or extended purging for the area when the residual concentration in a certain area exceeds the preset multiple of the dynamic baseline.

[0101] For example, in Al2O3 etching, the exponentially weighted moving average baseline B is maintained for each port. i and standard deviation σ i Forgetting factor α = 0.1, warning threshold set to B. i +3σ i If the residual concentration at only the port near the exhaust port consistently exceeds the threshold, it is determined to be localized polymer accumulation. The feedback adjustment module 103 generates a command to add an Ar pulse (150 sccm, 0.5 s) during the second purge of each cycle, and the control module 104 outputs this command. If the residual concentration at all ports simultaneously exceeds the threshold, the feedback adjustment module 103 generates a command for comprehensive in-situ cleaning (O2 / NF3 = 10:1, 500 W, 120 s), and the control module 104 outputs this command. If the residual concentration at only the near-field port increases and is related to the bias voltage, the feedback adjustment module 103 generates a command to add low-power argon plasma during the first purge to neutralize the surface charge, and the control module 104 outputs this command. After continuous production of 1000 wafers, the process capability index C... pk It can be improved from 0.67 to over 1.33.

[0102] Optionally, in some implementations, the dynamic baseline employs an exponentially weighted moving average with a forgetting factor α = 0.1. For bias-dependent drift modes, i.e., when only the near-field port residual concentration increases and is correlated with the bias magnitude, a low-power argon plasma (50 W, no bias, 0.5 s) can be added during the first purge to neutralize the surface charge.

[0103] The feedback adjustment module 103 is also configured to: calculate the interface indicator factor of each sampling unit in the multi-layer structure etching; when the interface indicator factor of all sampling units is lower than the threshold, determine that the interface switching is completed; for the hysteresis region, perform selective etching of the region by independently controlling the multi-zone bias voltage.

[0104] For example, in Si / SiGe superlattice etching, the interface indicator factor The threshold is set to 0.1. When the central area port RGe The value dropped to 0.08 (below the threshold) while the edge port R Ge When the bias voltage is still 0.28, the feedback adjustment module 103 generates an instruction to reduce the bias voltage of the center region to 0V to pause etching, maintain a bias voltage of 75V for the edge region, and add two additional etching pulses. The control module 104 outputs this instruction, with each etching pulse lasting 0.2s and an interval of 0.1s, until the edge region R... Ge The etching depth is reduced to 0.06, and then the entire etching sequence is terminated. This method can control the difference in interface etching depth within ±0.5 nm.

[0105] Optionally, in some embodiments, the interface threshold can be set to 0.1 for the SiGe / Si interface and 0.05 for other materials. The hysteresis region can be independently subjected to 1-5 regionalized auxiliary pulses, with adjustable pulse width and interval.

[0106] According to an embodiment of the present invention, an embodiment of an atomic layer etching control method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0107] This embodiment provides an atomic layer etching control method, which can be used in the atomic layer etching control system of any of the above embodiments, and is executed by the feedback adjustment module 103 and the control module 104. Figure 2 This is a flowchart of an atomic layer etching control method according to an embodiment of the present invention, with reference to... Figure 2 As shown, the process includes the following steps: Step S201: Gas samples are acquired by multiple sampling units set at different spatial locations in the reaction chamber, and infrared spectral analysis is performed on each gas sample to determine the gas concentration at the corresponding location of each sampling unit, thereby generating spatial distribution information of the concentration.

[0108] In this step, process-related gases refer to reactants, byproducts, or dilution gases associated with the atomic layer etching process, such as Cl2 and BCl3 in the surface modification step, and byproducts SiCl4, GeCl4, and GaCl2 in the etching removal step.

[0109] Concentration spatial distribution information refers to the gas concentration values ​​and their spatial coordinate correspondence at each location, which are collected and analyzed by the detection module 102 through multiple sampling units set at different spatial locations in the reaction chamber.

[0110] Specifically, the receiving process includes: the control detection module 102 acquires gas samples in a time-division manner through multiple sampling units set at different spatial locations in the reaction chamber, and performs infrared spectral analysis on each gas sample to determine the concentration.

[0111] For example, the detection module 102 detects and acquires concentration data measured at the near-field port above the wafer center, the near-field port above the edge, the port near the plasma source, and the port near the exhaust port. The host computer algorithm integrates these data into a concentration distribution covering the radial distribution of the wafer. After receiving this information, the feedback adjustment module 103 can identify which regions have high concentrations and which regions have low concentrations, providing a basis for subsequent differential adjustments. This method allows one infrared spectroscopy analysis unit to serve multiple sampling units, reducing hardware costs while ensuring the time synchronization and comparability of data from each sampling point, providing a foundation for generating accurate spatial distribution information.

[0112] In step S202, the feedback adjustment module 103 generates spatially differentiated process adjustment instructions based on the spatial distribution information of the concentration.

[0113] In this step, spatially differentiated process adjustment instructions refer to control parameters set separately for different spatial locations. For example, the bias amplitude is set to 80V for the central area and 90V for the edge area, or the gas injection system flow rate is 30sccm for the central area and 50sccm for the edge area.

[0114] Specifically, when the concentration spatial distribution information shows that the concentration of modified gas in the edge region is significantly lower than that in the center region, the feedback adjustment module 103 generates an instruction to increase the flow rate of the gas injection system in the edge region. After the instruction is executed, the gas concentration in the edge region increases, and the concentration distribution on the wafer surface tends to be more uniform, thereby suppressing the etching rate difference caused by the uneven gas flow field.

[0115] In step S203, the control module 104 receives the process adjustment command and controls the electrostatic chuck and / or gas injection system to execute the process adjustment command to adjust the spatial uniformity of the atomic layer etching process in the reaction chamber.

[0116] Meanwhile, the atomic layer etching process in this embodiment includes sequentially cyclical surface modification, first purging, etching removal, and second purging. During each process, gas samples from each sampling unit are acquired in a time-sharing manner according to a preset scanning sequence, and a complete scan cycle of the scanning sequence is less than the duration of any single process.

[0117] For example, if the surface modification step lasts 1.2s and the scan cycle is 200ms, then 6 complete scans can be completed during this step, which can capture the entire process of the concentration of modified gas decomposition products gradually decreasing from the initial value to the plateau.

[0118] The atomic layer etching control method provided in this embodiment generates concentration spatial distribution information through the detection module 102, and the feedback adjustment module 103 generates spatially differentiated process adjustment instructions based on this information. The control module 104 receives and outputs these instructions. Corresponding to the system, this method realizes a closed-loop control process based on spatial distribution information. It can compensate in real time for etching non-uniformity caused by the non-uniformity of the internal space of the reaction chamber in the etching process, overcomes the fundamental defect of traditional global feedback control that cannot distinguish spatial differences, and improves the etching uniformity within the wafer plane.

[0119] In some embodiments of the present invention, the atomic layer etching control method further includes the following steps: Step a1: Run the standard atomic layer etching recipe using a patternless wafer, record the reference characteristic values ​​of each sampling unit, including the saturation time in surface modification, the peak time and peak concentration of byproducts in etching removal, and the residual concentration at the end of purging, and store the reference characteristic values ​​in the database.

[0120] In this step, a patternless wafer refers to a flat wafer with no etched patterns on its surface, used to establish a baseline.

[0121] Standard atomic layer etching (ALT) formulations refer to validated default combinations of process parameters.

[0122] The reference characteristic is the benchmark value for deviation detection in subsequent production.

[0123] For example, 100 ALE cycles are run on the wafer, while the byproduct concentration curves at each port are monitored. The time required for the concentration of decomposition products to reach a plateau during surface modification at the center port (e.g., 0.35 s), the time required for the SiCl4 concentration to rise from the background value to 10% of the peak value during etching removal (e.g., 0.12 s), the peak concentration (e.g., 15 ppm), and the residual concentration at the end of the second purging (e.g., 1 ppm) are recorded. These values ​​are stored in a database. During production, if the rise time delay of a batch of wafers exceeds 0.15 s, an anomaly is identified, triggering a compensation mechanism.

[0124] This step is the initial calibration step, which provides a benchmark for deviation detection in subsequent production. This allows the feedback adjustment module 103 to determine whether there is an anomaly by comparing it with the reference value. For example, the quantitative evaluation of rise time delay in transmission limitation judgment depends on this reference database.

[0125] Reference Figure 3 As shown, in some embodiments of the present invention, the atomic layer etching control method has a transport limitation diagnosis and compensation mechanism for high aspect ratio structures. When the aspect ratio of the etched structure is greater than or equal to 40:1, the atomic layer etching control method further includes the following steps: Step b1: In the etching removal step, compare the rise time of byproduct concentration measured by different sampling units and compare it with the reference value. In this step, rise time refers to the time it takes for the byproduct concentration to rise from the background value to 10% of the peak value after the bias is turned on.

[0126] Step b2: If the rise time delay exceeds the first threshold and the delay difference between different ports exceeds the second threshold, then it is determined that there is a transmission limitation. In this step, the first threshold can be set to 0.15 s, and the second threshold can be set to 0.2 s. Transport limitation refers to the obstruction of reactant diffusion to the bottom of the high aspect ratio structure.

[0127] For example, in Si trench etching with an aspect ratio of 60:1, the rise time of SiCl4 measured at the center near-field port is 0.4s, which is 0.15s later than the reference value of 0.25s. The rise time at the edge near-field port is 0.6s, which is 0.35s later. The difference in delay between the two is 0.2s, which meets the judgment condition.

[0128] Simultaneously, the effective diffusion coefficient D is calculated using the concentration difference between the central near-field port and the port near the upper electrode. eff =0.12 m² / s, which is lower than 90% of the theoretical Knudsen diffusion coefficient of 0.28 m² / s (0.196 m² / s), further confirming a severe transport limitation. Accordingly, the feedback regulation module 103 generates an instruction: increase the chamber pressure from 10 mTorr to 35 mTorr, increase the He proportion in the modified gas from 0% to 30%, and switch the modification termination method from a fixed 1.2 s to a closed-loop saturation determination. The control module 104 outputs this instruction.

[0129] Step b3: If a transmission limitation is determined to exist, at least one of the following compensation instructions is generated: increasing the chamber pressure for surface modification, increasing the proportion of light carrier gas in the modified gas, and switching the termination method of surface modification from a fixed duration to a method based on closed-loop saturation determination.

[0130] In this step, the closed-loop saturation determination step includes: real-time detection of the concentration of modified gas decomposition products in each sampling unit, and termination of surface modification when the concentration change rate of all sampling units is lower than a preset threshold.

[0131] Specifically, at the start of surface modification, the initial concentration C of the decomposition products at each port was recorded. 0,i As modification proceeds, the concentration of decomposition products decreases due to consumption. Calculate the instantaneous consumption rate: .

[0132] When the central port R consumeWhen the plasma velocity drops from 20 ppm / s to 1 ppm / s, and at the edge port from 18 ppm / s to 0.9 ppm / s, both below 5% (1 ppm / s) of the initial maximum value (20 ppm / s), it is determined that all regions have reached saturation. The feedback adjustment module 103 generates a command to shut off the plasma and stop the introduction of the modifying gas, and the control module 104 outputs this command. This method ensures that all regions of the entire wafer reach modification saturation before entering the next cycle, avoiding insufficient etching due to local unsaturation or plasma pre-damage introduced by oversaturation. It controls the modification time fluctuation within ±0.1 s and stabilizes the surface roughness RMS below 0.2 nm.

[0133] In the above steps, when the aspect ratio of the etched structure is greater than or equal to 40:1, the rise time of the byproduct concentration measured by different sampling units is compared with a reference value. If the delay exceeds a threshold, it is determined that there is a transport limitation. The feedback adjustment module 103 then generates compensation commands to increase the pressure, add light carrier gas, or switch the termination mode to closed-loop saturation judgment, which are then executed by the control module 104. This method directly addresses the problem of limited reactant transport to the bottom in high aspect ratio structures. It diagnoses the degree of transport limitation through the observable indicator of rise time delay and automatically performs multi-mode compensation. Experiments have shown that it can significantly improve the etching uniformity at the bottom of the trench.

[0134] Reference Figure 4 As shown, based on the above embodiments, the atomic layer etching control method of the present invention further includes the following steps: Step c1: In the surface modification step, the consumption rate of the modified gas or its decomposition products at multiple sampling units is acquired in real time. Step c2: When the consumption rate at all sampling units is lower than the preset threshold, it is determined that the modification of the entire region is saturated, and the surface modification is terminated. In step c3, when the time difference between the saturation of each sampling unit exceeds a preset time, the feedback adjustment module 103 generates a process adjustment command to increase the flow rate of the gas injection system corresponding to the hysteresis region, and the control module 104 outputs the command.

[0135] In steps c1 to c3, the lag region refers to the region with a later saturation time. In some specific examples, the preset duration can be set to 0.2 s. For example, in a surface modification, the center port reaches saturation at 0.8 s, and the edge port reaches saturation at 1.1 s, a time difference of 0.3 s exceeding the preset duration of 0.2 s. The feedback adjustment module 103 records this deviation and, at the start of the next modification cycle, generates an instruction to increase the gas injection system flow rate corresponding to the edge region by 10% (e.g., from 30 sccm to 33 sccm), which is output by the control module 104. Simultaneously, the plasma power distribution can be fine-tuned to increase the concentration of gas decomposition products in the edge region. Through this adaptive adjustment, the saturation time difference caused by uneven gas flow field distribution is dynamically compensated through closed-loop feedback, making the saturation of different regions more synchronized, and significantly improving the uniformity of wafer surface modification.

[0136] Reference Figure 5 As shown, based on the above embodiments, the atomic layer etching control method of the present invention further includes the following steps: Step d1: In the etching removal step, the molecular rotation temperature of the etching byproducts at multiple sampling units is acquired in real time. In this step, the molecular rotation temperature is obtained by inverting the intensity ratio of the P-branch and R-branch in the infrared absorption spectrum of the by-product, reflecting the instantaneous thermal effect generated by ion bombardment.

[0137] Step d2: Calculate the damage indication function at the corresponding position of each sampling unit based on the rate of change of molecular rotation temperature. In this step, the damage indicator function can be specifically expressed as:

[0138] Among them, T rot dT is the molecular rotation temperature. rot / dt is the rate of change of molecular rotational temperature with time t, E bias E represents the currently applied bias energy. th The minimum energy required to completely remove the preset modified layer.

[0139] For example, in GaN HEMT gate trench etching, the minimum energy E required for complete removal of the modified layer is... th =55eV, safety threshold set to 250a.u.

[0140] Step d3: When the damage indication function at any location exceeds the preset safety threshold, the feedback adjustment module 103 generates a process adjustment command to reduce the bias amplitude of the electrostatic chuck area corresponding to that location, switch to pulse bias mode, and / or increase the local temperature of that area, and the control module 104 outputs the command.

[0141] The damage indicator function is proportional to the rate of change of molecular rotational temperature and also proportional to the difference between the current bias energy and the preset minimum energy required for complete removal of the modified layer. This function, based on a physical model, converts excess ion energy into lattice heat and quantifies damage risk into a calculable index. This allows the feedback adjustment module 103 to set a clear safety threshold, avoiding reliance on empirical judgment and improving the reliability and repeatability of damage suppression.

[0142] In a specific example, during the initial cycle, the molecular rotation temperature increased from 25°C to 80°C at a rate of 5.5°C / s. Bias energy 90 eV, calculate D indicator =5.5×(90-55)=192.5, which is below the threshold. During the 21st cycle, the molecular rotation temperature in the edge region rapidly increased from 25℃ to 150℃ within 0.1s. D indicator =1250×(90-55)=43750, far exceeding the threshold. The feedback adjustment module 103 generates instructions to reduce the bias voltage in this region from 90V to 70V, switch the continuous bias voltage to pulse mode (40% duty cycle), and raise the chuck temperature from 20℃ to 40℃. The control module 104 outputs this instruction. Two seconds later, dT rot / dt falls back, D indicator When the voltage drops to 180V, the feedback adjustment module 103 generates a command to gradually restore the bias voltage to 80V with a duty cycle of 60%, and the control module 104 outputs this command. This method reduces the gate-drain current from 15µA / mm to 1.2µA / mm, and the interface state density is reduced by an order of magnitude.

[0143] In the methods described in steps d1 to d3 above, the risk of damage is detected in real time by utilizing the sensitivity of molecular rotation temperature to ion energy, which can achieve a rapid response at the sub-millisecond level and effectively suppress cumulative lattice damage.

[0144] Reference Figure 6 As shown, in some embodiments of the present invention, when the etched structure includes at least a stacked structure of a first material layer and a second material layer, the atomic layer etching control method further includes the following steps: Step e1: Real-time acquisition of the concentrations of the first and second byproducts at multiple sampling units; In this step, the first material layer and the second material layer refer to two different materials in the stacked structure, such as SiGe and Si. The first byproduct comes from the first material layer, such as SiGe etching producing SiCl4 and GeCl4, while the second byproduct specifically comes from the second material layer, such as only SiGe producing GeCl4, while the Si layer does not produce it.

[0145] Step e2: Calculate the interface indicator factor at each sampling unit. The interface indicator factor is the ratio of the concentration of the second byproduct to the sum of the concentrations of the first and second byproducts. Step e3: When the interface indicator factor at all sampling units is lower than the preset interface threshold, it is determined that the interface switching of the entire area has been completed and the etching is terminated. In step e4, when the interface indicator factor at some sampling units is lower than the interface threshold while other sampling units are still higher than the interface threshold, the feedback adjustment module 103 generates a process adjustment command to apply an additional etching bias pulse to the electrostatic chuck at the corresponding position of the hysteresis region, and the control module 104 outputs the command until the interface indicator factor at all sampling units is lower than the interface threshold.

[0146] In steps e1 to e4, specifically, the interface threshold can be set to 0.1.

[0147] For example, in Si / SiGe superlattice etching, the first 8 periods R Ge ≈0.35±0.02. During the 9th etching cycle, the center port R... Ge It started to decrease at 0.6s, dropped to 0.08 at 0.9s, below 0.1, while the edge port R Ge The value remains at 0.28. The feedback adjustment module 103 generates a command to reduce the center region bias voltage to 0V, maintain 75V for the edge region, and add two additional etching pulses (each 0.2s, with a 0.1s interval). The control module 104 outputs this command. During this period, the edge port R is monitored in real time. Ge After the bias voltage is reduced to 0.06, the edge region bias is stopped, and the etching of the entire wafer is terminated.

[0148] Reference Figure 7 As shown, in some embodiments of the present invention, the atomic layer etching control method further includes the following steps: At the end of the atomic layer etching cycle, the concentration of residual byproducts at multiple sampling units was recorded; Step f1: Establish a dynamic baseline for each sampling unit based on the concentration of residual byproducts; Step f2: When the concentration of residual byproducts at any sampling unit continuously exceeds the preset multiple of its corresponding dynamic baseline, it is determined that there is contaminant accumulation in the chamber area corresponding to that sampling unit. Step f3: Generate corresponding compensation commands based on the drift mode: if the residual concentration at only the port near the exhaust port exceeds the threshold, generate a local purging pulse or a short in-situ cleaning command; if the residual concentration at all ports exceeds the threshold simultaneously, generate a comprehensive in-situ cleaning command; if the residual concentration at only the near-field port increases and is related to the bias voltage, generate a surface charge neutralization or extended purging command.

[0149] In steps f1 to f3, the dynamic baseline uses an exponentially weighted moving average. Standard deviation ; Among them, C res,i (n) represents the residual byproduct concentration measured at the i-th sampling port in the n-th cycle, B i (n) and σ i (n) represents the dynamic baseline and standard deviation of the i-th sampling port, respectively, and α is the forgetting factor, which is set to 0.1. The preset factor is usually 3 times the standard deviation.

[0150] In a specific example, during Al2O3 etching, for wafers 201-400, only the port C near the vent is etched. res It gradually increased from 3 ppm to 12 ppm, exceeding B. i +3σ i This is determined to be a localized accumulation of byproducts. The feedback adjustment module 103 generates a command to add an Ar pulse to the exhaust port-side gas injection system during the second purge of each cycle. The control module 104 outputs this command. res Restored to 4-5 ppm. Chips 600-800, all ports C. res If the concentration rises to 15-18 ppm, indicating overall contamination or a decrease in pump speed, the feedback adjustment module 103 generates an instruction for automatic batch-to-batch in-situ cleaning (O2 / NF3=10:1, 200 sccm, 500W, 90s). The control module 104 outputs this instruction and uses NDIR to detect CO2 and SiF4 concentrations. Production resumes after the concentrations drop to baseline. After adaptive compensation, C... pk The efficiency was improved from 0.71 to 1.42, and the average cleaning interval was extended from 200 pieces to 400-600 pieces. This method achieves spatial adaptive compensation for chamber drift, can distinguish different types of drift and take precise measures, thus extending the average cleaning interval of the chamber and improving equipment utilization.

[0151] Reference Figure 8 As shown, in some embodiments of the present invention, the atomic layer etching control method further includes the following steps: Step g1: Accumulate all spatiotemporal data collected from multiple atomic layer etching batches. The spatiotemporal data includes the concentration-time series, temperature-time series, pressure-time series, process parameter records, and actuator action records of each sampling unit. Step g2: Using spatiotemporal data and the corresponding offline etching results, train a hybrid neural network model. The hybrid neural network model includes a convolutional neural network for extracting spatial distribution features and a long short-term memory network for extracting temporal dynamic features. Step g3 involves deploying the trained model to the feedback adjustment module 103 for online prediction of etching depth distribution and / or damage index, and for optimizing process parameters under constraints. After producing a predetermined number of wafers, the model is incrementally learned and fine-tuned using new data.

[0152] In steps g1 to g3, spatiotemporal data refers to the concentration, temperature, and pressure values, along with corresponding process parameters, continuously recorded over time for each sampling unit. Offline etching results include wafer dot etching depth contour maps and surface roughness. A hybrid neural network model combines a Convolutional Neural Network (CNN) to extract spatial distribution features and a Long Short-Term Memory (LSTM) network to extract temporal dynamic features. Constraints include a bias voltage range of 59–120V and a pressure range of 1–500 mTorr.

[0153] In a specific example, after accumulating data from 100 batches, the system trains a model with a prediction error of less than 2% for the etching depth distribution. When producing new products, such as GaN HEMTs, the feedback adjustment module 103 uses the model to perform Bayesian optimization under constraints, outputting recommended multi-zone bias, gas flow, and temperature parameters, which are then output by the control module 104. Every 100 wafers produced, the model weights are fine-tuned using new data, continuously improving prediction accuracy and significantly reducing the frequency of manual adjustments. This digital twin self-optimization method utilizes big data and deep learning technologies to upgrade process control from rule-based feedback to prediction-based optimization, significantly reducing the frequency of manual adjustments and providing an intelligent tool for rapid process development of new products.

[0154] This embodiment also provides a semiconductor atomic layer etching apparatus, including the atomic layer etching control system of any of the above embodiments. The apparatus can be an ICP-type or CCP-type plasma etching machine, with a flange interface (e.g., CF16 high-vacuum blind flange) pre-reserved on the reaction chamber wall for installing a sampling unit. This apparatus integrates spatially resolved infrared spectroscopy detection capability and spatially differentiated closed-loop control capability, enabling it to sense the spatial distribution of concentration, temperature, and pressure inside the chamber in real time, and automatically adjust actuators such as multi-zone bias voltage, multi-zone airflow, and local temperature control. It can be widely used in advanced semiconductor manufacturing fields such as three-dimensional NAND flash memory (3D NAND), FinFET, Gate-All-Around Field-Effect Transistor (GAAFET), and GaN power devices.

[0155] As an improved implementation, the device may also include a host computer display screen connected to the feedback adjustment module 103 and / or the control module 104, which displays a concentration spatial distribution map and / or a damage risk spatial distribution map in real time; the user can click on any position in the distribution map, and the system will automatically display the historical process parameters and real-time detection data corresponding to that position.

[0156] The semiconductor atomic layer etching apparatus provided in this embodiment of the invention can execute the atomic layer etching control system provided in any embodiment of the invention, and has corresponding functional modules and beneficial effects. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0157] This invention also provides a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the atomic layer etching control method described above. This non-transitory computer-readable storage medium possesses all the beneficial effects of the atomic layer etching control method described above, which will not be elaborated further here.

[0158] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described display system control method and has all the beneficial effects of the above-described atomic layer etching control method. Further details are omitted here.

[0159] This invention also provides an electronic device, including a memory and a processor. The memory stores a computer program; the processor executes the computer program in the memory to implement the steps of the atomic layer etching control method described above. This electronic device has all the beneficial effects of the atomic layer etching control method described above, which will not be elaborated further here.

[0160] Reference Figure 9 As shown, the electronic device may include a processing system (e.g., a central processing unit, a graphics processing unit, etc.) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage system 908 into a random access memory (RAM) 903. The random access memory 903 also stores various programs and data required for the operation of the electronic device. The processing system 901, the read-only memory 902, and the random access memory 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0161] Typically, the following systems can be connected to I / O interface 905: input systems 906 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output systems 907 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage systems 908 including, for example, magnetic tapes, hard disks, etc.; and communication systems 909. Communication system 909 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although... Figure 9 Electronic devices with various systems are shown, but it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented alternatively. Figure 9 Each box shown can represent a system, or multiple systems as needed.

[0162] In particular, according to some embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication system 909, or installed from a storage system 908, or installed from a read-only memory 902. When the computer program is executed by a processing system 901, it performs the functions defined in the methods of some embodiments of the present invention.

[0163] It should be noted that, in some embodiments of the present invention, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. The computer-readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination thereof, and the present invention does not specifically limit this. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0164] In some embodiments of the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in connection with an instruction execution system, system, or device. In some embodiments of the present invention, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may transmit, propagate, or transmit a program for use by or in connection with an instruction execution system, system, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0165] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol, such as Hypertext Transfer Protocol (HTTP), and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (LANs), wide area networks (WANs), the Internet (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0166] The aforementioned computer-readable medium may be included in the aforementioned electronic device, or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs.

[0167] Computer program code for performing operations of some embodiments of the present invention can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0168] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function.

[0169] It should also be noted that in some alternative implementations, the functions marked in the box may occur in a different order than those marked in the attached figures.

[0170] For example, two consecutively represented blocks can actually be executed in substantially parallel order, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, as well as combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified functions or operations, or using a combination of dedicated hardware and computer instructions.

[0171] The units described in some embodiments of the present invention can be implemented in software or in hardware.

[0172] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0173] 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. An atomic layer etching control system, applied in the semiconductor field, for real-time detection and control using non-dispersive infrared spectroscopy, characterized in that, include: The etching module includes a reaction chamber and an associated electrostatic chuck, bias power supply unit, plasma generation unit, gas injection system and evacuation system. The etching module is used to perform a cyclic process of atomic layer etching. The cyclic process of atomic layer etching includes a surface modification step, a first purging step, an etching removal step and a second purging step performed sequentially. The detection module includes multiple sampling units, an infrared spectroscopy analysis unit, and a temperature and pressure inversion unit, all located at different spatial positions within the reaction chamber. The multiple sampling units are used to extract gas samples from different spatial positions within the reaction chamber, and the infrared spectroscopy analysis unit is used to perform infrared spectroscopy analysis on the gas samples to determine the gas concentration at each sampling unit location and generate spatial distribution information of the concentration based on the location information of each sampling unit. The infrared spectroscopy analysis unit includes: a detector, which is a thermoelectrically cooled mercury cadmium telluride infrared detector, used to receive infrared light absorbed by the gas sample and convert it into an electrical signal; and a temperature and pressure inversion unit connected to the detector, used to receive the electrical signal generated by the detector, invert the local temperature by measuring the full width at half maximum (FWHM) of the absorption peak of the impurity gas in the infrared spectrum corresponding to the electrical signal, and calculate the local pressure by measuring the Lorentz linewidth and combining it with the inverted local temperature. The feedback adjustment module, which uses an embedded simulation unit to feed back and control the adjustment of etching process parameters, includes a transmission simulation unit, a reaction kinetics simulation unit, and a damage simulation unit. It is used to receive the spatial distribution information of the gas concentration, the local temperature, and the local pressure, calculate the etching state of different regions, and generate spatially differentiated process adjustment instructions. The control module is communicatively connected to the etching module, the detection module, and the feedback adjustment module, and outputs the process adjustment commands to the electrostatic chuck, the bias power supply unit, and / or the gas injection system, respectively, to adjust the spatial uniformity of the etching process in the reaction chamber. The feedback adjustment module is also configured to: In the etching removal step, the molecular rotation temperature of the etching byproducts at the multiple sampling units is acquired in real time. The molecular rotation temperature is obtained by inverting the intensity ratio of the P-branch and R-branch in the infrared absorption spectrum of the byproducts. Based on the rate of change of the molecular rotation temperature over time, a damage indication function is calculated for the corresponding position of each sampling unit. The damage indication function is proportional to the rate of change of the molecular rotation temperature and proportional to the difference between the current bias energy and the preset minimum energy required to completely remove the modified layer. When the damage indication function at any location exceeds a preset safety threshold, a process adjustment command is generated to reduce the bias amplitude of the electrostatic chuck area corresponding to that location, switch to pulse bias mode, and / or increase the local temperature of that area.

2. The control system according to claim 1, characterized in that, The electrostatic chuck is radially divided into at least a central region, an intermediate region, and an edge region. The electrostatic chuck includes a temperature control unit, with each region having an independent temperature control unit, used to respond to spatially differentiated process adjustment commands output by the control module to independently adjust the temperature of different radial regions of the wafer. The bias power supply unit is used to independently supply bias voltage to each region of the electrostatic chuck, and the bias voltage amplitude of each region is independently adjustable to the range of 20eV to 200eV corresponding ion energy, and the pulse duty cycle of each region is independently adjustable to the range of 5% to 100%; the control module is also configured to independently adjust the bias voltage parameters for different regions to independently adjust the etching rate of different radial regions of the wafer. The plasma generating unit is communicatively connected to the control module and is used to adjust the plasma parameters in the reaction chamber in response to the process adjustment command. The extraction unit is communicatively connected to the control module and is used to adjust the pressure and by-product discharge rate within the reaction chamber in response to the process adjustment command.

3. The control system according to claim 1, characterized in that, The multiple sampling units are set in different spatial locations within the reaction chamber and connected to the infrared spectroscopy analysis unit through independent pipelines to obtain gas concentration, temperature, and pressure data at different spatial locations inside the reaction chamber. The plurality of sampling units include at least: The first sampling unit is located above the central region of the electrostatic chuck and is used to collect gas samples from the etching front surface of the central region. The second sampling unit is located above the edge region of the electrostatic chuck and is used to collect gas samples from the edge region. The third sampling unit is located below the plasma generating unit and is used to collect the decomposition products of modified gases in the plasma. The fourth sampling unit is located at the exhaust port of the reaction chamber and is used to collect gas samples in the exhaust port area and use the concentration of the gas samples in the exhaust port area as a background reference value. The first sampling unit and the second sampling unit are located at the same height layer, so that the feedback adjustment module can detect the difference in radial spatial distribution by comparing the data measured by the two units; the first sampling unit and the third sampling unit are located at different height layers, so that the feedback adjustment module can determine the concentration change at different height positions by comparing the data measured by the two units; the concentration values ​​of the first sampling unit, the second sampling unit, and the third sampling unit are corrected based on the background reference value measured by the fourth sampling unit, and the data collected by the multiple sampling units are used together to generate the spatial distribution information of the concentration.

4. The control system according to claim 1, characterized in that, Each of the sampling units includes: Sampling tubing; A microplate is placed at the inlet of the sampling pipeline to limit the sampling flow rate; A heating jacket is used to wrap around the sampling pipeline to prevent byproducts from condensing. A quick-switching valve is used to sequentially activate each sampling unit according to a predetermined timing sequence; The control system also includes a multi-channel gate valve group, the input end of which is connected to the outlet end of all the sampling pipelines, and the output end of which is connected to the infrared spectroscopy analysis unit. The fast switching valve sequentially activates each sampling unit according to a predetermined time sequence, allowing the gas samples from each sampling unit to enter the multi-channel gate valve group in sequence. The multi-channel gate valve group then sequentially delivers the gas samples to the infrared spectroscopy analysis unit to achieve time-division measurement.

5. The control system according to claim 1, characterized in that, The infrared spectroscopy analysis unit also includes: Infrared light source, used to emit infrared light; A gas absorption cell, wherein the gas absorption cell is a long optical path cell, used to enable the infrared light to fully interact with the gas sample entering the cell; The signal processing unit, electrically connected to the detector, is used to calculate the gas concentration at the corresponding position of each sampling unit based on the electrical signal, and to use multiple linear regression to eliminate cross-interference between adjacent absorption peaks, and then send the concentration to the feedback adjustment module.

6. The control system according to claim 1, characterized in that, The infrared spectroscopy analysis unit also includes: Multiple narrowband filters, each corresponding to a characteristic absorption peak of a characteristic byproduct, wherein the characteristic byproduct includes at least one of SiF4, SiCl4, GeCl4, GaCl2, AlCl3, or COF2.

7. The control system according to claim 1, characterized in that, The feedback adjustment module includes: The transport simulation unit is used to calculate the transport attenuation of reactants at different spatial locations based on the spatial distribution information. The reaction kinetics simulation unit is used to calculate the growth rate and etching removal rate of the surface modification layer based on the spatial distribution information. The damage simulation unit is used to calculate the damage index accumulated in the lattice by ion bombardment based on the spatial distribution information. The instruction generation unit is used to generate the spatially differentiated process adjustment instructions based on the calculation results of the transmission simulation unit, the reaction kinetics simulation unit, and the damage simulation unit. The etching state includes the transport state determined by the transport simulation unit, the surface modification saturation degree determined by the reaction kinetics simulation unit, the cumulative damage degree from ion bombardment determined by the damage simulation unit, and the material interface position.

8. The control system according to claim 1, characterized in that, The feedback adjustment module is also configured to: The rise time of byproduct concentration measured by different sampling units is compared, and the effective diffusion coefficient is calculated using the concentration difference of sampling units at different vertical heights. When the effective diffusion coefficient is lower than a preset threshold, the chamber pressure in the surface modification step is automatically increased or the carrier gas flow rate is increased, and closed-loop saturation judgment is used to control the termination of the modification step. The transient temperature rise of the local lattice during etching is estimated by measuring the rotational temperature distribution of by-product molecules. Based on the comparison between the product of the transient temperature rise and the bias energy and the damage threshold, the bias amplitude of the corresponding region is reduced or switched to pulse bias mode. At the end of the purging step, the concentration of residual byproducts in each sampling unit is recorded to establish a dynamic baseline. When the residual concentration in a certain area exceeds the preset multiple of the dynamic baseline, active cleaning or extended purging is initiated for that area. In multi-layer structure etching, the interface indicator factor of each sampling unit is calculated. When the interface indicator factor of all sampling units is lower than the threshold, the interface switching is determined to be complete. For hysteresis regions, selective etching of the region is performed by independently controlling the bias voltage of multiple regions.

9. An atomic layer etching control method, applied to the atomic layer etching control system according to any one of claims 1 to 8, characterized in that, include: Gas samples are acquired by multiple sampling units located at different spatial positions in the reaction chamber, and infrared spectral analysis is performed on each gas sample to determine the gas concentration at the corresponding position of each sampling unit, thereby generating spatial distribution information of the concentration. The feedback adjustment module generates spatially differentiated process adjustment instructions based on the concentration spatial distribution information. The control module receives the process adjustment command and controls the electrostatic chuck and / or the gas injection system to execute the process adjustment command, so as to adjust the spatial uniformity of the atomic layer etching process in the reaction chamber.

10. The control method according to claim 9, characterized in that, The atomic layer etching process includes a surface modification step, a first purging step, an etching removal step, and a second purging step, which are performed sequentially in a cycle. The control method further includes: during each of the steps, acquiring gas samples from each sampling unit in a time-division manner according to a preset scanning sequence, wherein a complete scanning cycle of the scanning sequence is less than the duration of any of the steps.

11. The control method according to claim 10, characterized in that, The control method further includes: Using patternless wafers, standard atomic layer etching recipes were run, and reference characteristic quantities for each sampling unit were recorded, including saturation time in the surface modification step, peak time and peak concentration of byproducts in the etching removal step, and residual concentration at the end of the purging step. The reference feature values ​​are then stored in the database.

12. The control method according to claim 10, characterized in that, If the aspect ratio of the etched structure is greater than or equal to 40:1, the control method further includes: In the etching removal step, the rise time of byproduct concentration measured by different sampling units is compared with the reference value; If the rise time delay exceeds the first threshold and the delay difference between different ports exceeds the second threshold, then a transmission limitation is determined to exist. If a transmission limitation is determined, the feedback adjustment module generates at least one of the following compensation instructions: increasing the chamber pressure for surface modification, increasing the proportion of light carrier gas in the modified gas, and switching the termination method of surface modification from a fixed duration to a method based on closed-loop saturation determination.

13. The control method according to claim 12, characterized in that, The closed-loop saturation determination includes: The concentration of modified gas decomposition products in each sampling unit is monitored in real time. When the concentration change rate of all sampling units is lower than a preset threshold, the surface modification is terminated.

14. The control method according to claim 10, characterized in that, The control method further includes: During the surface modification step, the consumption rate of the modified gas or its decomposition products at the multiple sampling units is acquired in real time. When the consumption rate at all sampling units is lower than a preset threshold, the entire area modification is determined to be saturated, and the surface modification is terminated. When the time difference between the saturation of each sampling unit exceeds a preset duration, the feedback adjustment module generates a process adjustment command to increase the flow rate of the gas injection system corresponding to the lag region.

15. The control method according to claim 9, characterized in that, When the etched structure comprises at least a stacked structure of a first material layer and a second material layer, the control method further includes: The concentrations of the first and second byproducts at the multiple sampling units are acquired in real time. Calculate the interface indicator factor at each sampling unit, where the interface indicator factor is the ratio of the concentration of the second byproduct to the sum of the concentrations of the first and second byproducts; When the interface indicator factor at all sampling units is lower than the preset interface threshold, it is determined that the interface switching of the entire area has been completed and the etching is terminated. When the interface indicator factor at some sampling units is lower than the interface threshold while other sampling units are still higher than the interface threshold, the feedback adjustment module generates a process adjustment command to apply an additional etching bias pulse to the electrostatic chuck at the corresponding position of the hysteresis region, and the control module outputs the process adjustment command until the interface indicator factor at all sampling units is lower than the interface threshold.

16. The control method according to claim 10, characterized in that, The control method further includes: At the end of the atomic layer etching cycle, the concentration of residual byproducts at the plurality of sampling units is recorded; A dynamic baseline is established for each sampling unit based on the concentration of the residual byproducts. When the concentration of residual byproducts at any sampling unit continuously exceeds the preset multiple of its corresponding dynamic baseline, it is determined that there is contaminant accumulation in the chamber area corresponding to that sampling unit. Based on the drift pattern, generate corresponding compensation commands: if the residual concentration at only the port near the exhaust port exceeds the threshold, generate a local purge pulse or an in-situ short cleaning command. If the residual concentration at all ports synchronously exceeds the threshold, a full in-situ cleaning command is generated. If only the residual concentration at the near-field port increases and is related to the bias voltage, a surface charge neutralization or extended purge command is generated.

17. The control method according to claim 9, characterized in that, The control method further includes: Accumulate all spatiotemporal data collected from multiple atomic layer etching batches. The spatiotemporal data includes concentration-time series, temperature-time series, pressure-time series, process parameter records, and actuator action records for each sampling unit. Using the spatiotemporal data and the corresponding offline etching results, a hybrid neural network model is trained. The hybrid neural network model includes a convolutional neural network for extracting spatial distribution features and a long short-term memory network for extracting temporal dynamic features. The trained model is deployed in the feedback adjustment module to predict the etching depth distribution and / or damage index online, and to optimize process parameters under constraints. After each predetermined number of wafers are produced, the model is incrementally learned and fine-tuned using new data.

18. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the atomic layer etching control method according to any one of claims 9 to 17.

19. A semiconductor atomic layer etching apparatus, characterized in that, The atomic layer etching control system includes any one of claims 1 to 8.

Images