Etching process monitoring method, parameter determination method, device and electronic equipment

Abstract

This invention relates to the field of semiconductor manufacturing process technology, and discloses a monitoring method, parameter determination method, apparatus, and electronic device for etching processes. The method includes: acquiring current three-dimensional point cloud data of the etching region in the current acquisition cycle; determining the current etching depth and current sidewall angle based on the current three-dimensional point cloud data, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region; determining the depth change rate based on the current etching depth and multiple historical etching depths corresponding to multiple consecutive historical acquisition cycles, wherein the depth change rate is the rate of change of etching depth over time; determining whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles corresponding to multiple consecutive historical acquisition cycles; and determining whether the etching endpoint has been reached based on the depth change rate and whether the sidewall angle is in a stable state. This invention can accurately determine the endpoint under complex etching conditions and reduce false alarms.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing process technology, and specifically to a method for monitoring etching processes, a method for determining parameters, an apparatus, and an electronic device. Background Technology

[0002] In semiconductor device manufacturing, etching is one of the key process steps, and its quality directly determines the device's performance and stability. During etching, endpoint detection is performed to monitor the etching process and stop etching at the endpoint to obtain precise etching coverage.

[0003] In related technologies, laser interferometry, emission spectroscopy, or impedance monitoring are used to determine the etching endpoint. Laser interferometry determines the etching endpoint by monitoring the interference signal of the single-point reflected light intensity changing with the etching depth; emission spectroscopy indirectly determines whether the etching endpoint has been reached by analyzing the light intensity changes of a specific wavelength in the plasma; and impedance monitoring indirectly determines the endpoint by detecting changes in plasma impedance.

[0004] Laser interferometry relies on the intensity of reflected light at a single point, which can easily lead to global misjudgment due to local over-etching. Emission spectroscopy and impedance monitoring reflect the macroscopic state changes of the entire reaction chamber. When the etching endpoint only occurs in a local pattern area, the macroscopic signal change is extremely weak and can be easily overwhelmed by fluctuations in process conditions (such as small changes in gas pressure and power), resulting in low sensitivity and a high false alarm rate. Summary of the Invention

[0005] This invention provides a method for monitoring etching processes, a method for determining parameters, an apparatus, and an electronic device to solve the problem of low accuracy of etching endpoints obtained by laser interferometry, emission spectroscopy, and impedance monitoring.

[0006] In a first aspect, the present invention provides a method for monitoring an etching process, the method comprising: during the execution of the etching process, acquiring current three-dimensional point cloud data of the etching region in the current acquisition cycle; determining the current etching depth and the current sidewall angle based on the current three-dimensional point cloud data, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region; determining the depth change rate based on the current etching depth and multiple historical etching depths corresponding one-to-one with multiple consecutive historical acquisition cycles, wherein the depth change rate is the rate of change of etching depth over time; determining whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles corresponding one-to-one with multiple consecutive historical acquisition cycles; and determining whether the etching endpoint has been reached based on the depth change rate and whether the sidewall angle is in a stable state.

[0007] In one optional implementation, determining whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles corresponding one-to-one with multiple consecutive historical acquisition cycles includes: determining the angle change rate based on the standard deviation of the current sidewall angle and the standard deviation of each historical sidewall angle, wherein the angle change rate is the change rate of the standard deviation of the sidewall angle over time; if the angle change rate is less than or equal to 0 and the standard deviation of the current sidewall angle is less than a second preset threshold, then the sidewall angle is in a stable state; determining whether the etching endpoint has been reached based on the change rate of the etching depth over time and whether the sidewall angle is in a stable state includes: if the change rate of the depth is less than a first preset threshold and the sidewall angle is in a stable state, then it is determined that the etching endpoint has been reached.

[0008] In one optional implementation, determining whether the etching endpoint has been reached based on the rate of change of etching depth over time and whether the sidewall angle is in a stable state includes: if the rate of change of depth is less than a first preset threshold and the sidewall angle is in a stable state, then updating the number of candidate endpoints; if the number of candidate endpoints after updating is less than a preset number, then re-determining whether the rate of change of depth and the sidewall angle are in a stable state based on the target three-dimensional point cloud data of the etching area under the target acquisition cycle, wherein the duration of the target acquisition cycle is less than the duration of the current acquisition cycle; if the number of candidate endpoints after updating is greater than or equal to the preset number, then determining that the etching endpoint has been reached.

[0009] In one optional implementation, the etched region includes a dense region, an isolated region, and a test pattern region. The depth change rate includes the dense depth change rate of the dense region, the isolated depth change rate of the isolated region, and the test depth change rate of the test pattern region. Whether the sidewall angle is in a stable state includes whether the sidewall angle of the dense region, the isolated region, and the test pattern region is in a stable state. Determining whether the etching endpoint has been reached based on the etching depth change rate over time and whether the sidewall angle is in a stable state includes: if the dense depth change rate is less than a first preset threshold and the sidewall angle of the dense region is in a stable state, then the dense region is determined to have reached the etching endpoint; if the isolated depth change rate is less than the first preset threshold and the sidewall angle of the isolated region is in a stable state, then the isolated region is determined to have reached the etching endpoint; if the test depth change rate is less than the first preset threshold and the sidewall angle of the test pattern region is in a stable state, then the test pattern region is determined to have reached the etching endpoint.

[0010] In one alternative implementation, before determining whether the etching endpoint has been reached, the method further includes: determining a first preset threshold based on the material of the etched region; or, determining the first preset threshold based on the historical average rate of the etched region.

[0011] In one optional implementation, acquiring the current three-dimensional point cloud data of the etched area during the current acquisition cycle includes: acquiring the current three-dimensional point cloud data from an optical measurement system, wherein the optical measurement system is used to output multiple projection lights of different wavelengths, the projection lights of different wavelengths are used to form stripes of different spatial frequencies, or each projection light is used to form a set of stripes of different spatial frequencies, and the different spatial frequencies are in a multiple relationship.

[0012] In one alternative implementation, multiple projection lights of different wavelengths are all different from the characteristic emission spectra of the plasma in the etching process.

[0013] In one alternative implementation, the number of projection lights is three, and the three projection lights with different wavelengths include a 450nm projection light, a 650nm projection light, and an 850nm projection light.

[0014] In one alternative implementation, when the radio frequency power switch in the semiconductor process equipment is off or during the period of minimum plasma afterglow in the semiconductor process equipment, the optical measurement system outputs multiple projection lights of different wavelengths to detect the current three-dimensional point cloud data.

[0015] In an optional implementation, the method further includes: determining a current derived feature for characterizing the degree of defect growth based on the current three-dimensional point cloud data, wherein the defect includes at least one of a target trench located at the bottom of the etched trench and sidewall roughness; when the defect includes a target trench, the current derived feature includes the current trench residual peak value; when the defect includes sidewall roughness, the current derived feature includes the energy of the target frequency band in the current sidewall power spectral density; the current trench residual peak value is the maximum difference between the target trench and the theoretical depth in the current acquisition cycle; inputting the current derived feature into an anomaly detection model, and determining the predicted next derived feature based on the output of the anomaly detection model; determining the actual next derived feature based on the next three-dimensional point cloud data of the etched area in the next acquisition cycle; determining the degree of defect based on the predicted next derived feature and the actual next derived feature; and outputting a warning message based on the degree of defect when the degree of defect is greater than a fourth preset threshold.

[0016] In one optional implementation, when the current derived feature includes the current trench residual peak value, the current derived feature used to characterize the degree of defect growth is determined based on the current three-dimensional point cloud data, including: fitting and determining the theoretical contour line of the bottom of the etching trench based on the first point cloud data of the sidewall of the etching trench extracted from the current three-dimensional point cloud data; determining the difference distribution between the theoretical contour line and the second point cloud data of the bottom of the etching trench extracted from the current three-dimensional point cloud data; in the target area near the sidewalls of the etching trench, if the difference between the theoretical contour line and the second point cloud data is less than 0, and the absolute value of the difference is greater than or equal to a third preset threshold, then a target trench exists at the bottom of the etching trench; and determining the current trench residual peak value based on the difference between the theoretical contour line and the second point cloud data in the target area.

[0017] In one optional implementation, when the current derived feature includes the energy of the target frequency band in the current sidewall power spectral density, determining the current derived feature for characterizing the degree of defect growth based on the current three-dimensional point cloud data includes: determining the current sidewall power spectral density based on the contour line of the etched trench sidewall extracted from the current three-dimensional point cloud data; and determining the energy of the target frequency band based on the current sidewall power spectral density, wherein the target frequency band is a frequency greater than 0.01 nm. -1 The frequency band.

[0018] Secondly, the present invention provides a method for determining parameters of an etching process, the method comprising: during the execution of the etching process, acquiring current three-dimensional point cloud data of the etching region in the current acquisition cycle; determining the current etching depth and the current sidewall angle based on the current three-dimensional point cloud data, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region; determining the depth change rate based on the current etching depth and multiple historical etching depths corresponding one-to-one with multiple consecutive historical acquisition cycles, wherein the depth change rate is the rate of change of etching depth over time; and determining the current sidewall angle and multiple historical acquisition cycles... Multiple historical sidewall angles are used to determine whether the sidewall angles are in a stable state. Based on the depth change rate and whether the sidewall angles are in a stable state, it is determined whether the etching endpoint has been reached. After determining that the etching endpoint has been reached, the current 3D point cloud data of the current acquisition cycle is stored in the process database, which includes multiple 3D point cloud data arranged in time. Based on the process database, a process model is trained, whereby the process model is used to characterize the correspondence between process features and process parameters. The process features are features extracted from the 3D point cloud data that reflect the execution state of the etching process. Based on the process model, the optimal combination of process parameters that meets the preset optimization objective is determined by searching in the process parameter space through an optimization algorithm.

[0019] Thirdly, the present invention provides a monitoring device for an etching process. The monitoring device includes: an acquisition module for acquiring current three-dimensional point cloud data of the etching region during the etching process, in the current acquisition cycle; a parameter determination module for determining the current etching depth and the current sidewall angle based on the current three-dimensional point cloud data, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region; a rate of change determination module for determining the depth rate of change based on the current etching depth and multiple historical etching depths corresponding to multiple consecutive historical acquisition cycles, wherein the depth rate of change is the rate of change of etching depth over time; a state judgment module for determining whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles corresponding to multiple consecutive historical acquisition cycles; and an endpoint determination module for determining whether the etching endpoint has been reached based on the depth rate of change and whether the sidewall angle is in a stable state.

[0020] Fourthly, the present invention provides a parameter determination device for an etching process. The device includes: an acquisition module for acquiring current three-dimensional point cloud data of the etching region during the etching process, within the current acquisition cycle; a parameter determination module for determining the current etching depth and current sidewall angle based on the current three-dimensional point cloud data, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region; a rate of change determination module for determining the depth rate of change based on the current etching depth and multiple historical etching depths corresponding to multiple consecutive historical acquisition cycles, wherein the depth rate of change is the rate of change of etching depth over time; and a state judgment module for determining the state based on the current sidewall angle and multiple consecutive historical... The system employs a multi-stage acquisition module to determine whether the sidewall angles are stable, corresponding to multiple historical sidewall angles for each acquisition cycle. An endpoint determination module determines whether the etching endpoint has been reached based on the depth change rate and whether the sidewall angles are stable. A storage module stores the current 3D point cloud data for the current acquisition cycle into a process database after determining that the etching endpoint has been reached. The process database includes multiple 3D point cloud datasets arranged chronologically. A training module trains a process model based on the process database. This model represents the correspondence between process features and process parameters. The process features are extracted from the 3D point cloud data and reflect the etching process execution state. A parameter optimization module searches the process parameter space using an optimization algorithm based on the process model to determine the optimal combination of process parameters that meets the preset optimization objectives.

[0021] Fifthly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the etching process monitoring method of the first aspect or any corresponding embodiment described above, or to perform the etching process parameter determination method of the second aspect or any corresponding embodiment described above.

[0022] In a sixth aspect, the present invention provides a computer-readable storage medium storing computer instructions, the computer instructions being used to cause a computer to perform a monitoring method for an etching process according to the first aspect or any corresponding embodiment thereof, or a parameter determination method for an etching process according to the second aspect or any corresponding embodiment thereof.

[0023] In a seventh aspect, the present invention provides a computer program product, including computer instructions, which are used to cause a computer to execute the monitoring method for the etching process of the first aspect or any corresponding embodiment described above, or the parameter determination method for executing the etching process of the second aspect or any corresponding embodiment described above.

[0024] The etching process monitoring method, parameter determination method, apparatus, and electronic equipment provided by this invention have at least the following advantages: This invention uses features extracted from 3D point cloud data to determine whether the etching endpoint has been reached during the etching process. Compared to laser interferometry, emission spectroscopy, and impedance monitoring, this method enables direct and accurate tracking of key morphological parameters during etching, thus accurately determining the endpoint under complex etching conditions. Furthermore, directly determining the endpoint based on 3D morphological parameters avoids the errors of indirect methods, reducing the endpoint misjudgment rate by more than 50% compared to laser interferometry.

[0025] Moreover, this invention comprehensively judges the etching endpoint based on whether the depth change rate and sidewall angle are in a stable state. This can weaken the influence of random errors in single depth measurement, identify the evolution characteristics of the process nearing the endpoint in advance, predict changes in the etching state, accurately lock down time, effectively reduce the invalid etching time at the end of the etching process, reduce process defects such as bottom roughness deterioration and trench corrosion, reduce material waste caused by over-etching, reduce rework costs, and improve equipment utilization.

[0026] When determining the etching endpoint, the initial fulfillment of the endpoint conditions (depth change rate less than the first preset threshold and sidewall angle in a stable state) is not directly taken as the etching endpoint. Instead, it is used as a candidate trigger signal. The endpoint conditions are re-measured to determine if they can be met again. Only if the endpoint conditions are still met is the etching process finally determined to have reached the etching endpoint, further reducing the probability of false triggering. After triggering the candidate endpoint signal, the acquisition cycle is shortened for confirmation measurement, which can avoid misjudging the endpoint due to accidental fluctuations.

[0027] After determining the current 3D point cloud data, this invention does not calculate whether the depth change rate and sidewall angle of the entire etched area are in a stable state. Instead, it divides the etched area into different measurement areas (dense area, isolated area, and test pattern area) and detects the parameters of each area to determine the etch endpoint. This can quantify the load effect (depth ratio of dense area to isolated area), improve the accuracy of etch endpoint determination, and also detect local anomalies earlier (such as microgrooves appearing in an isolated area).

[0028] This invention employs a multi-wavelength strategy combining fundamental frequency and harmonic frequency to detect three-dimensional topographic data. Through the principle of multi-frequency heterodyne, it can effectively solve the phase ambiguity problem of steep sidewalls and deep trench bottoms, and achieve unambiguous measurement from submicron to tens of micron depths.

[0029] When measuring three-dimensional topography data, three projection lights are used, with three different wavelengths including 450nm, 650nm and 850nm projection lights. This can further expand the equivalent wavelength and optimize the penetration capability of signals at different depths (shallow, medium and deep), accurately measuring etched trenches with aspect ratios of 10:1 or even 20:1 or higher.

[0030] When the etching endpoint is not reached, the system also determines the current derived features from the current 3D point cloud data to characterize the degree of defect growth. These derived features are then used for defect early warning, enabling effective early warning at the initial stage of defect initiation, such as microgrooves and sidewall roughness. This avoids batch scrapping and improves etching process control accuracy and product yield. The defect early warning time of this invention can be reduced by 10% to 20% compared to traditional detection methods.

[0031] After determining that the etching endpoint has been reached, the current 3D point cloud data of the current acquisition cycle is also stored in the process database. The accumulated 3D point cloud data can provide process engineers with an intuitive etching process, and the process model trained based on the process database can help optimize the formulation of the next generation of etching processes, thereby improving the processing quality and yield of the etching process. Attached Figure Description

[0032] 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.

[0033] Figure 1 This is a schematic flowchart of a monitoring method for an etching process according to an embodiment of the present invention; Figure 2 This is a schematic flowchart of a monitoring method for another etching process according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating the process of monitoring whether the etching process has reached the etching endpoint according to the present invention. Figure 4 This is a schematic flowchart of a monitoring method for another etching process according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the process for defect early warning based on the ARIMA model according to an embodiment of the present invention; Figure 6 This is a flowchart illustrating a method for determining parameters of an etching process according to an embodiment of the present invention. Figure 7 This is a schematic diagram of the monitoring and process optimization of the etching process according to an embodiment of the present invention; Figure 8 This is a schematic diagram comparing the etching depth versus time curves of the laser interferometry method and the present invention; Figure 9 This is a schematic diagram comparing the endpoint determination error time of the laser interferometry method and the present invention; Figure 10 This is a schematic diagram of the bottom contour of the etched trench under normal conditions according to an embodiment of the present invention; Figure 11 This is a schematic diagram of the bottom contour of the microgrooves according to an embodiment of the present invention; Figure 12 This is a schematic diagram comparing the time it takes for the laser interferometry method and the present invention to discover the defect; Figure 13 This is a schematic diagram comparing batch yield before and after optimization of etching process parameters according to an embodiment of the present invention; Figure 14 This is a structural block diagram of a monitoring device for an etching process according to an embodiment of the present invention; Figure 15 This is a structural block diagram of a parameter determination device for an etching process according to an embodiment of the present invention; Figure 16This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0034] 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.

[0035] 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.

[0036] This invention provides a monitoring method, parameter determination method, device, and electronic device for etching processes. It extracts etching depth and sidewall angle from three-dimensional point cloud data of the etched surface to determine the etching endpoint. This enables direct and accurate tracking of key morphological parameters in the etching process. Furthermore, it comprehensively determines the etching endpoint based on the rate of change of etching depth and the stability of sidewall angle, which can accurately determine the endpoint under complex etching conditions and reduce false alarms.

[0037] Furthermore, this invention also provides defect early warning based on derived features representing defect development trends extracted from 3D point cloud data. This allows for effective early warning at the initial stage of defect initiation, such as microgrooves and sidewall roughness, thereby improving etching process control accuracy and product yield. Simultaneously, this invention stores the detected 3D point cloud data in a process database for subsequent optimization and adjustment of process formulations, achieving adaptive process control.

[0038] According to embodiments of the present invention, an embodiment of a method for monitoring an etching process and an embodiment of a method for determining parameters of an etching process are provided. It should be noted that the steps shown in the flowcharts in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although in Figure 4 The logical order is shown, but in some cases, the steps shown or described may be performed in a different order than that shown here.

[0039] This embodiment provides a method for monitoring etching processes, which can be used in electronic devices such as tablets, computers, or servers. Figure 1 This is a schematic flowchart of a monitoring method for an etching process according to an embodiment of the present invention, as shown below. Figure 1 As shown, the process includes the following steps: Step S101: During the etching process, acquire the current three-dimensional point cloud data of the etching area in the current acquisition cycle.

[0040] Specifically, when processing wafers using etching technology, the optical measurement system acquires three-dimensional point cloud data (three-dimensional topography data) of the wafer processing area according to an acquisition cycle. The time interval of the acquisition cycle can be 20s or 30s, etc. The electronic device obtains the current three-dimensional point cloud data of the etched area from the optical measurement system under the current acquisition cycle. The current three-dimensional point cloud data is the three-dimensional point cloud data acquired in the acquisition cycle corresponding to the current moment (i.e., the current acquisition cycle).

[0041] Step S102: Determine the current etching depth and the current sidewall angle based on the current 3D point cloud data.

[0042] The current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region. For example, if the etching region includes 3 etching trenches, then the current etching depth includes the etching depth corresponding to the 1st etching trench, the etching depth corresponding to the 2nd etching trench, and the etching depth corresponding to the 3rd etching trench.

[0043] Etching depth can be defined as the difference between the average height of the top of the etching trench and the average height of the bottom of the etching trench. Sidewall angle can be defined as the angle between the tangent to the sidewall of the etching trench and the horizontal plane at the top of the etching trench. After acquiring the current 3D point cloud data, the sidewall contour point cloud data can be fitted with a straight line or a quadratic curve. Then, multiple test points are taken from each of the two sidewalls of the etching trench, and the average angle between the tangent at each test point and the horizontal plane at the top of the etching trench is determined as the sidewall angle.

[0044] Step S103: Determine the depth change rate based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles.

[0045] Among them, the depth change rate is the rate of change of etching depth over time, the historical acquisition period is the acquisition period before the current acquisition period, and the historical etching depth is the etching depth corresponding to the historical acquisition period.

[0046] Specifically, the mean (or median) of the etching depths corresponding to all etching trenches included in the current etching depth is calculated to obtain the current depth value. For each historical etching depth, the mean (or median) of the etching depths corresponding to all etching trenches is calculated to obtain multiple consecutive historical depth values. The current depth value and the depth-time series composed of multiple consecutive historical depth values ​​are then fitted to obtain a depth curve that changes over time. The first derivative of the depth curve is then calculated. It was determined to be the rate of change of depth.

[0047] Individual trenches may experience random fluctuations due to local particles or measurement noise. Determining the depth change rate based on the mean (or median) of the etched area is more representative of the overall etching process of that area, and the change rate calculated after smoothing is more reliable.

[0048] The time window, consisting of the current acquisition cycle and multiple consecutive historical acquisition cycles, can be dynamically adjusted according to the etching rate. In other words, the number of acquisition cycles included in the time window is adjusted based on the etching rate.

[0049] For example, the number of acquisition cycles is inversely proportional to the etching rate. When the etching rate is high (greater than the preset etching rate), the number of acquisition cycles in the time window is small, such as including 3 acquisition cycles (1 current acquisition cycle and 2 historical acquisition cycles); when the etching rate is low (less than or equal to the preset etching rate), the number of acquisition cycles in the time window is large, such as including 5 to 7 acquisition cycles.

[0050] Furthermore, the original current etching depth contains measurement noise and process noise. Before determining the depth curve over time, the depth time series can be filtered, for example, by using Savitzky-Golay smoothing differential filtering. Then, based on the filtered depth time series, the depth curve over time can be determined, improving the accuracy of the calculated depth change rate.

[0051] Savitzky-Golay filtering can smooth noise while maintaining the shape and width of the signal, thereby further improving the accuracy of the calculated first derivative (rate of change).

[0052] Step S104: Determine whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles.

[0053] The historical sidewall angle is the sidewall angle corresponding to the historical acquisition cycle. The standard deviation can be used to determine whether the sidewall angle is in a stable state. The standard deviation can reflect the consistency of the sidewall angle of the etching trench within the etching area. When the standard deviation and its difference are very small, the sidewall angle of the etching trench can be considered to be stable.

[0054] Specifically, determine the current angular standard deviation of the sidewall angles corresponding to all etch trenches included in the current etching depth. , This represents the sidewall angle corresponding to the i-th etched trench. , It is an integer. This indicates the total number of etched trenches in the etched area; it determines the historical angle standard deviation of the sidewall angles corresponding to all etched trenches included in the historical etch depth; if the difference between the current angle standard deviation and any two of the multiple consecutive historical angle standard deviations is less than or equal to the preset difference, and the current angle standard deviation and the multiple consecutive historical angle standard deviations are all less than or equal to the preset standard deviation, then it can be determined that the sidewall angle is in a stable state.

[0055] If the difference between the current angle standard deviation and any two of the multiple consecutive historical angle standard deviations is greater than the preset difference, or if the difference between the current angle standard deviation and any one of the multiple consecutive historical angle standard deviations is less than the preset standard deviation, then it can be determined that the sidewall angle is not yet in a stable state.

[0056] Step S105: Determine whether the etching endpoint has been reached based on whether the depth change rate and sidewall angle are in a stable state.

[0057] Specifically, this invention considers the etching endpoint not as a simple achievement of a target depth, but rather as the moment when a physical process transitions. For etching processes (such as the Bosch process), when the etching front reaches the etching stop layer, the etching rate will decrease sharply due to a material transition, such as from silicon to silicon dioxide. (Reduced), and at the same time, due to the different material selection ratios, the balance between passivation layer consumption and redeposition on the sidewalls is disrupted, and the sidewall angle tends to stabilize. (Become smaller). Therefore... and The combined changes are a strong physical signal that the etching endpoint has arrived.

[0058] When the rate of change in depth is less than the first preset threshold When the sidewall angle is stable, it can be determined that the etching endpoint has been reached, and the semiconductor process equipment ends the etching process; when the depth change rate is greater than or equal to the first preset threshold... If the sidewall angle is not in a stable state, it can be determined that the etching endpoint has not been reached, and the semiconductor process equipment continues to perform the etching process.

[0059] The etching process monitoring method provided in this embodiment uses features extracted from three-dimensional point cloud data to determine whether the etching endpoint has been reached during the etching process. Compared with laser interferometry, emission spectroscopy, and impedance monitoring, this method can directly and accurately track key morphological parameters of the etching process, thereby accurately determining the endpoint under complex etching conditions. Furthermore, directly determining the endpoint based on three-dimensional morphological parameters avoids the errors of indirect methods, reducing the endpoint misjudgment rate by more than 50% compared to laser interferometry.

[0060] Moreover, this embodiment comprehensively judges the etching endpoint based on whether the depth change rate and sidewall angle are in a stable state. This can weaken the influence of random errors in single depth measurement, identify the evolution characteristics of the process nearing the endpoint in advance, predict changes in the etching state, accurately lock down time, effectively reduce the invalid etching time at the end of the etching process, reduce process defects such as bottom roughness deterioration and trench corrosion, reduce material waste caused by over-etching, reduce rework costs, and improve equipment utilization.

[0061] This embodiment also provides another embodiment of the etching process monitoring method, which can be used in electronic devices such as tablet computers, computers, or servers. Figure 2 This is a schematic flowchart of a monitoring method for another etching process according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: During the etching process, acquire the current three-dimensional point cloud data from the optical measurement system.

[0062] The optical measurement system outputs multiple projection lights of different wavelengths. The projection lights of different wavelengths are used to form stripes of different spatial frequencies, or each projection light is used to form multiple stripes of different spatial frequencies, with the different spatial frequencies being multiples of each other.

[0063] Specifically, the optical measurement system outputs at least two projection lights with different wavelengths. During measurement, one approach is for each projection light to project and form multiple fringes with different spatial frequencies. For example, the optical measurement system includes three projection lights (a first projection light, a second projection light, and a third projection light). The first projection light projects and forms fringes with a spatial frequency of f (fundamental frequency) and fringes with a spatial frequency of 2f (harmonic frequency). The second projection light also projects and forms fringes with a spatial frequency of f and fringes with a spatial frequency of 2f. The third projection light also projects and forms fringes with a spatial frequency of f and fringes with a spatial frequency of 2f. Another approach is to combine multiple projection lights with different wavelengths to form fringes with different spatial frequencies. For example, the optical measurement system includes three projection lights. The first projection light projects and forms fringes with a spatial frequency of f (fundamental frequency). The second projection light projects and forms fringes with a spatial frequency of 2f (second harmonic frequency). The third projection light projects and forms fringes with a spatial frequency of 3f (third harmonic frequency).

[0064] Spatial frequency refers to the spatial period of structured light fringes (or the number of fringes per unit length). In structured light 3D measurement, the projected fringe pattern has a specific spatial period. , corresponding to a frequency The fundamental frequency can refer to the reference spatial frequency, which has a relatively large equivalent wavelength (synthetic wavelength), enabling unambiguous measurements over a wide range, but with lower accuracy. The harmonic frequency refers to the spatial frequency that is an integer multiple of the fundamental frequency, corresponding to a smaller synthetic wavelength, and has higher measurement accuracy.

[0065] The equivalent composite wavelength corresponding to the fundamental frequency The wavelength must be greater than the maximum depth (or maximum height difference) of the etched trench to ensure unambiguous phase unfolding. For example, if the maximum depth of the etched trench is 50 μm, then the equivalent synthesized wavelength corresponding to the selected fundamental frequency... The phase measurement accuracy corresponding to the harmonics is high, but it is prone to 2π ambiguity. The approximate depth is first solved by the wide range of the fundamental frequency, and then the fine phase of the harmonics is used for high-precision correction.

[0066] Assuming the spatial period corresponding to the fundamental frequency is The spatial period corresponding to the frequency multiplication is m is an integer representing the multiple of the harmonic relative to the fundamental frequency, which can be 2, 3, or 4, etc. The equivalent combined wavelength of the fundamental frequency and the harmonic can be shown in formula (1):

[0067] By choosing a suitable m, it is possible to make Covers the entire measurement depth range. For example, if the maximum depth is 50 μm, it can be taken as... If m=2, then The requirements are met.

[0068] The optical measurement system employs a multi-wavelength strategy combining fundamental frequency and harmonic frequency, which can obtain multiple independent phase measurements. These phase values ​​at different spatial frequencies satisfy a linear relationship. By using the multi-frequency heterodyne principle, an equivalent composite wavelength much larger than a single spatial period can be synthesized, enabling unambiguous measurements from submicron to tens of micron depths.

[0069] In single-frequency structured light measurement, the initial phase is calculated using a phase-shifting algorithm. The true phase is contained within the interval [0, 2π). , where integer To determine the true altitude, it must be determined from the unknown information. This is known as unwrapping. For steep sidewalls or deep trench bottoms, the height variation between adjacent pixels may exceed one fringe period (Λ / 2), causing traditional spatial phase unwrapping algorithms (such as path tracking) to fail and resulting in phase jumps.

[0070] This invention projects multiple fringes of different spatial frequencies (periods) based on the multi-frequency heterodyne principle, allowing for the acquisition of their respective enveloping phases. For a spatial period of The fundamental frequency and spatial period are The frequency harmonics can be used to construct a composite phase as shown in formula (2), and the equivalent wavelength corresponding to this composite phase is... It can be determined by formula (1).

[0071]

[0072] because The depth of the etch can be designed to be very large (e.g., 100 μm), while the depth of the etch to be measured may only be 50 μm. Therefore, the number of integer cycles in the synthesized phase... It only has 0 or 1, making it easy to uniquely determine. Once determined... This allows us to deduce the true phase of each single frequency, thereby obtaining a high-precision depth value.

[0073] The height difference between adjacent pixels on steep sidewalls can exceed the phase change range of a single short-period fringe, making tracking impossible. However, the period of the synthesized wavelength can be much larger than this height difference, so the synthesized phase will not jump, thus providing globally unambiguous guidance. The signal is weak and noisy at the bottom of deep trenches, and traditional spatial unwrapping is easily affected by noise and fails. Multi-frequency heterodyne methods utilize measurements at different frequencies in the time dimension (i.e., multiple projections), do not rely on spatial neighborhood information, and are more robust to noise and local shadows.

[0074] Therefore, the optical measurement system adopts a multi-wavelength strategy combining fundamental frequency and harmonic frequency to detect three-dimensional topographic data. Through the principle of multi-frequency heterodyne, it can effectively solve the phase ambiguity problem of steep sidewalls and deep trench bottoms, and realize unambiguous measurement from submicron to tens of micron depth range.

[0075] During measurement, using a scheme where each projection light projects multiple stripes of different spatial frequencies can obtain more wrapped phase, thus achieving a wider equivalent wavelength through multi-frequency heterodyne and stronger robustness. However, because multiple sets of stripes are projected, the measurement time is longer. Using a scheme where multiple projection lights of different wavelengths are combined to form stripes of different spatial frequencies results in fewer sets of projected stripes (the number of stripe sets is equal to the number of projection lights), and the measurement speed is faster.

[0076] This invention does not specifically limit the number of projection lights. The optical measurement system can output two or three projection lights of different wavelengths. When the number of projection lights is two, it can measure etched trenches with an aspect ratio of less than 10:1. When the number of projection lights is three, the equivalent wavelength can be further expanded, and the penetration capability of signals at different depths (shallow, medium, and deep) can be optimized, enabling the measurement of etched trenches with an aspect ratio of more than 10:1, or even more than 20:1.

[0077] Optionally, when the number of projection lights is three, the three projection lights with different wavelengths include a 450nm projection light, a 650nm projection light, and an 850nm projection light. The etched trenches can be divided into a shallow layer near the top, a middle layer, and a deep layer near the bottom from top to bottom. The 450nm projection light is used to measure the 3D point cloud data of the shallow layer, the 650nm projection light is used to measure the 3D point cloud data of the middle layer, and the 850nm projection light is used to measure the 3D point cloud data of the deep layer.

[0078] This embodiment combines projection light of different wavelengths, utilizing the strong penetrating power of long wavelength (850nm) in silicon to obtain effective signals at the bottom of deep trenches. At the same time, it is combined with a multi-frequency heterodyne algorithm to achieve reliable measurement of high aspect ratio etched trenches.

[0079] In some alternative embodiments, multiple projection lights of different wavelengths are different from the characteristic emission spectra of the plasma in the etching process.

[0080] Specifically, the optical measurement system includes a projection module that generates multiple projection lights of different wavelengths. The projection module uses a narrow-band filtered digital micromirror device (DMD) for projection, and the wavelength range of the projection pattern strictly avoids the characteristic emission lines of the plasma (such as the fluorine atom spectral lines used for silicon etching), thereby suppressing the interference of plasma background light.

[0081] Narrowband filters are filters whose full width at half maximum (FWHM) is small enough relative to their center wavelength to effectively block characteristic emission lines of plasma and other stray light. The bandwidth (FWHM) of a narrowband filter is ≤10nm. For example, if the projection wavelength is 650nm, a bandpass filter with a center wavelength of 650nm and a bandwidth of ±5nm (i.e., FWHM 10nm) would be selected.

[0082] Plasma characteristic emission lines (such as fluorine atom lines at 703.7 nm and 777.2 nm) typically have narrow bandwidths (<1 nm) but are surrounded by a continuous background. Using filters with a FWHM ≤ 10 nm can attenuate the characteristic spectral peaks by more than three orders of magnitude (depending on the filter's optical density) while preserving the projection signal. For near-infrared wavelengths (such as 850 nm), plasma emission in semiconductor etching processes is relatively weak, and the bandwidth can be appropriately widened to 15 nm, but it is still considered a narrow band.

[0083] Conventional color filters or broadband filters (FWHM>50nm) cannot effectively distinguish between projected light and plasma emission, while the narrowband filter (FWHM≤10nm) defined in this invention can achieve a background suppression ratio greater than 100:1.

[0084] The filter can be a hard-film interference filter, which has good temperature stability and is suitable for the high-temperature environment of the etching chamber. The filter is mounted on a motor-driven filter wheel and automatically switches according to the wavelength of the current projected light.

[0085] Furthermore, the optical measurement system also includes an imaging module, which is equipped with a bandpass filter corresponding to the wavelength of the projected light and a global shutter camera. For deep trench measurements, the imaging module uses a short-wave infrared (SWIR) camera, which has better penetration into silicon materials, to obtain the effective signal at the bottom of the trench.

[0086] Optionally, when the RF power switch in the semiconductor process equipment is off or during the period of minimum plasma afterglow in the semiconductor process equipment, the optical measurement system outputs multiple projection lights of different wavelengths to detect the current 3D point cloud data. Triggering the measurement when the plasma light intensity is weak can further reduce background light noise.

[0087] Step S202: Determine the current etching depth and the current sidewall angle based on the current 3D point cloud data.

[0088] Please see details Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0089] Step S203: Determine the depth change rate based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles.

[0090] Please see details Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0091] Step S204: Determine whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles.

[0092] Specifically, step S204 may include the following steps: Step S2041: Determine the rate of change of angle based on the standard deviation of the current sidewall angle and the standard deviation of each historical sidewall angle.

[0093] The rate of change of angle is the rate of change of the standard deviation of the sidewall angle over time.

[0094] Specifically, the standard deviation corresponding to the current sidewall angle and the standard deviation corresponding to each historical sidewall angle are arranged in chronological order to obtain the angle standard deviation time series; the angle standard deviation time series is fitted to obtain the standard deviation curve that changes with time, and the first derivative (slope) of the standard deviation curve is determined as the angle change rate.

[0095] In step S2042, if the rate of change of angle is less than or equal to 0 and the standard deviation of the current sidewall angle is less than the second preset threshold, then the sidewall angle is in a stable state.

[0096] Specifically, the standard deviation of the sidewall angle continuously decreases and its absolute value is less than the second preset threshold. For example, the standard deviation of the sidewall angle decreases for three consecutive times and is less than the second preset threshold. It can be assumed that the sidewall angle of the entire etched area has entered a stable plateau period, that is, the sidewall angle is in a stable state.

[0097] This invention determines whether the standard deviation of the sidewall angle is continuously decreasing by determining whether the rate of change of the angle is less than or equal to 0. If the rate of change of the angle is less than or equal to 0, it can be determined that the standard deviation of the sidewall angle is continuously decreasing; if the rate of change of the angle is greater than 0, it can be determined that the standard deviation of the sidewall angle is not continuously decreasing.

[0098] It should be noted that if the angle change rate is greater than 0 or the standard deviation of the current sidewall angle is greater than or equal to the second preset threshold, the sidewall angle is not in a stable state. At this time, wait for the next three-dimensional point cloud data to be detected in the next acquisition cycle, and use the acquired next three-dimensional point cloud data as new three-dimensional point cloud data to re-execute steps S202 to S204.

[0099] In step S205, if the depth change rate is less than the first preset threshold and the sidewall angle is in a stable state, then it is determined that the etching endpoint has been reached.

[0100] Specifically, if the rate of change in depth is greater than or equal to the first preset threshold If the sidewall angle is not in a stable state, it can be determined that the etching endpoint has not been reached. The semiconductor process equipment continues to perform the etching process, waits for the next three-dimensional point cloud data to be detected in the next acquisition cycle, and uses the acquired next three-dimensional point cloud data as new three-dimensional point cloud data to re-execute steps S202 to S205 to monitor whether the etching endpoint has been reached.

[0101] For example, the first preset threshold This is not a fixed value; in some embodiments, the first preset threshold can be determined based on the material of the etched region. For example, the electronic device may have a pre-stored first ratio of the theoretical etching rate to the actual etching rate at the interface between the etched layer and the etch stop layer (such as Si-Oxide). It can be set to 80% of the first ratio.

[0102] In other alternative embodiments, a first preset threshold can be determined based on the historical average rate of the etched region. For example, the product of the historical average rate and a scaling factor (such as 0.1) can be used as the first preset threshold. .

[0103] For example, to reduce false positives, the step of determining whether the etching endpoint has been reached based on the rate of change of etching depth over time and whether the sidewall angle is in a stable state (i.e., step S105) may include the following steps: Step a1: If the depth change rate is less than the first preset threshold and the sidewall angle is in a stable state, then update the number of candidate endpoints.

[0104] Specifically, if the depth change rate is less than the first preset threshold and the sidewall angle is in a stable state, it means that the etching has reached the required etching endpoint, and the number of candidate endpoints is incremented by 1 (counter + 1) as the updated number of candidate endpoints.

[0105] Step a2: If the number of candidate endpoints after the update is less than the preset number, then based on the target three-dimensional point cloud data of the etched area under the target acquisition cycle, re-determine whether the depth change rate and sidewall angle are in a stable state.

[0106] The target acquisition period is shorter than the current acquisition period. For example, if the current acquisition period is 30 seconds, the target acquisition period could be 10 seconds.

[0107] The target 3D point cloud data is the 3D point cloud data under the target acquisition cycle. The process of redetermining the depth change rate based on the target 3D point cloud data is similar to the process of determining the depth change rate based on the current 3D point cloud data, and will not be repeated here. The process of redetermining whether the sidewall angle is in a stable state based on the target 3D point cloud data is similar to the above step S204, and will not be repeated here.

[0108] Step a3: If the number of updated candidate endpoints is greater than or equal to the preset number, then the etching endpoint is determined to have been reached.

[0109] The preset number of times can be 2 or 3, etc.

[0110] In this embodiment, the initial fulfillment of the endpoint condition (depth change rate less than a first preset threshold and sidewall angle in a stable state) is not directly taken as the etching endpoint. Instead, it is taken as a trigger candidate signal. The measurement is repeated to determine whether the endpoint condition can be met again. Only if the endpoint condition is still met is the etching process finally determined to have reached the etching endpoint, further reducing the probability of false triggering. After the candidate endpoint signal is triggered, the acquisition cycle is shortened for confirmation measurement, which can avoid the situation of misjudging the endpoint due to accidental fluctuations.

[0111] It should be understood that the target acquisition cycle is a temporary adjustment and is only used for the confirmation phase. After confirmation, the original interval will be restored or the process will be terminated.

[0112] The process of monitoring whether the etching process of the present invention has reached the etching endpoint is described with a preset number of times of 2.

[0113] like Figure 3 As shown, the current etching depth is determined based on the current 3D point cloud data. and current sidewall angle Subsequently, based on the etching depth corresponding to a time window containing multiple acquisition cycles, a linear fitting solution was obtained. And the standard deviation of the sidewall angle is calculated based on the sidewall angle corresponding to the time window. Then, determine Is it less than the first preset threshold? If so, then continue to determine the standard deviation. Is it less than the second preset threshold? If not, wait for the 3D point cloud data of the next acquisition cycle (the next measurement point) to redetermine the current etching depth and the current sidewall angle.

[0114] at standard deviation Less than the second preset threshold When the endpoint candidate count is reached, increment it by 1 and check if the value is greater than or equal to 2. If so, trigger the endpoint signal to confirm that the etching endpoint has been reached; otherwise, wait for the 3D point cloud data of the next acquisition cycle (the next measurement point) to re-determine the current etching depth and the current sidewall angle. Within the standard deviation... Greater than or equal to the second preset threshold At the same time, it also waits for the three-dimensional point cloud data of the next acquisition cycle (the next measurement point) to redetermine the current etching depth and the current sidewall angle.

[0115] In some optional embodiments, the etched region includes a dense region, an isolated region, and a test pattern region; the depth change rate includes the dense depth change rate of the dense region, the isolated depth change rate of the isolated region, and the test depth change rate of the test pattern region; whether the sidewall angle is in a stable state includes whether the sidewall angle of the dense region is in a stable state, whether the sidewall angle of the isolated region is in a stable state, and whether the sidewall angle of the test pattern region is in a stable state; in this case, the above step S105 includes: Step b1: If the rate of change of dense depth is less than the first preset threshold and the sidewall angle of the dense region is in a stable state, then it is determined that the dense region has reached the etching endpoint.

[0116] Step b2: If the rate of change of isolated depth is less than the first preset threshold and the sidewall angle of the isolated region is in a stable state, then it is determined that the isolated region has reached the etching endpoint.

[0117] Step b3: If the rate of change of test depth is less than the first preset threshold and the sidewall angle of the test pattern area is in a stable state, then it is determined that the test pattern area has reached the etching endpoint.

[0118] Specifically, after determining the current three-dimensional point cloud data, this embodiment does not calculate whether the depth change rate and sidewall angle of the entire etched area are in a stable state. Instead, it divides the etched area into different measurement areas (dense area, isolated area, and test pattern area), monitors the etch endpoint of the etch process according to the measurement area, and improves the accuracy of the etch endpoint judgment.

[0119] The dense and isolated regions refer to areas on the wafer to be etched that contain functional devices. Dense regions specifically refer to areas with a high number of trenches per unit area and small spacing (e.g., linewidth / spacing ratio ≤ 1:1) within the functional device area, typically located in the core circuitry of the chip. Isolated regions refer to areas with sparse trenches and large spacing (e.g., spacing > 10 times the linewidth) within the functional device area, often located at the chip edge or in the test structure section. The test pattern area is located in areas on the wafer to be etched that do not contain functional devices, such as dicing grooves or dedicated test areas. The test pattern area can refer to a specially designed reference pattern area with specific dimensions and spacing, used to monitor process stability.

[0120] For example, based on the coordinate information of the etched graphic design layout (GDS), the etched area can be automatically aligned and divided into dense areas, isolated areas, and test graphic areas.

[0121] In etching processes, pattern density affects the local etching rate, a phenomenon known as the loading effect (RIE-lag). Due to faster reactant consumption and varying ion flux distribution, densely packed regions typically have lower etching rates than isolated regions. Therefore, if only the overall average depth is measured, the densely packed regions may not have reached the target depth while the isolated regions have already been etched. By dividing the etching process into regions and separately detecting the parameters of each region to determine the etching endpoint, the loading effect (the depth ratio of densely packed to isolated regions) can be quantified. This allows for the determination of whether each measurement region has reached its endpoint (e.g., using the slowest region as the benchmark), and also enables earlier detection of local anomalies (such as micro-trenches in an isolated region).

[0122] It should be noted that an isolated region may contain only one etched trench, in which case the standard deviation is 0, and the deviation between the current sidewall angle and the historical sidewall angle is directly used as the angle change rate. In specific implementation, for measurement areas with fewer than three etched trenches, the standard deviation of the sidewall angle is not calculated; instead, the deviation between two adjacent sidewall angles at each time point is used as the angle change rate.

[0123] This embodiment also provides another embodiment of an etching process monitoring method, which can be used in electronic devices such as tablet computers, computers, or servers. Figure 4 This is a schematic flowchart of a monitoring method for another etching process according to an embodiment of the present invention, as shown below. Figure 4 As shown, the process includes the following steps: Step S401: During the etching process, acquire the current three-dimensional point cloud data of the etching area in the current acquisition cycle.

[0124] Please see details Figure 1 Step S101 of the illustrated embodiment or Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0125] Step S402: Determine the current etching depth and current sidewall angle based on the current 3D point cloud data.

[0126] Please see details Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0127] Step S403: Determine the depth change rate based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles.

[0128] Please see details Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0129] Step S404: Determine whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles.

[0130] Please see details Figure 1 Step S104 of the illustrated embodiment or Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0131] Step S405: Determine whether the etching endpoint has been reached based on whether the depth change rate and sidewall angle are in a stable state.

[0132] Please see details Figure 1 Step S105 of the illustrated embodiment will not be described again here.

[0133] Step S406: Based on the current 3D point cloud data, determine the current derived features used to characterize the degree of defect growth.

[0134] The defects include at least one of the target trench located at the bottom of the etched trench and the sidewall roughness. When the defects include the target trench, the current derived feature includes the current trench residual peak value. When the defects include the sidewall roughness, the current derived feature includes the energy of the target frequency band in the current sidewall power spectral density. The current trench residual peak value is the maximum difference between the target trench and the theoretical depth in the current acquisition cycle.

[0135] Target trenches refer to micro-grooves, which are V-shaped localized deep pits that appear at the bottom of the trench near the sidewall. Morphologically, they are characterized by localized depth deviations at specific locations; sidewall roughness is manifested as high-frequency fluctuations along the sidewall contour line.

[0136] Specifically, the local depression depth of the microtrench is typically 20 nm to 200 nm, with a relative value ranging from 3% to 10% of the total depth of the current etched trench. The horizontal extension range of the microtrench is generally less than or equal to 1 μm, with a typical value of 0.3 μm to 0.8 μm. The microtrench is located close to the sidewall, and the distance between the microtrench and the sidewall of the etched trench is less than or equal to 0.5 μm. Furthermore, the sidewalls of the microtrench are distributed in a linear pattern parallel to the sidewalls of the etched trench.

[0137] When the current derived feature includes the current trench residual peak value, step S406 above may include the following steps: Step c1: Based on the first point cloud data of the sidewall of the etched trench extracted from the current three-dimensional point cloud data, fit and determine the theoretical contour line of the bottom of the etched trench.

[0138] Specifically, firstly, the cross-sectional profile of a single etched trench is extracted from the three-dimensional point cloud. Multiple cross-sections are taken along the length y of the etched trench (e.g., one cross-section is taken every 1μm). Each cross-section can obtain a two-dimensional contour line. The horizontal coordinate x of the two-dimensional contour line is the horizontal position, and the vertical coordinate z is the height. Then, based on each two-dimensional contour line, the sidewall region and bottom region of the etched trench are identified.

[0139] In this 2D contour, locations with drastic height changes represent the sidewalls, while those with gentler changes represent the bottom. Regions are defined by calculating the gradient (slope). Scanning from one side of the etched trench to the other, the system searches for inflection points where the slope changes from positive to negative (or vice versa). These inflection points correspond to the boundaries between the sidewalls and the bottom. The boundary is defined as the location where the absolute value of the slope decreases from greater than 0.5 (steep) to less than 0.1 (flat). This allows us to determine the extent of the left and right sidewalls and the bottom region (the area between the left and right boundaries).

[0140] Next, the bottom contour is fitted using point cloud data within the bottom region. The fitting model is determined based on the target topography of the etching process. For example, if the bottom is flat and straight, the fitting model can be a straight-line fitting model, using the least squares method to fit a horizontal straight line z=c (or a slightly tilted straight line z=ax+b) as the initial bottom contour, where b is a compensation coefficient used to compensate for wafer tilt. If the bottom is arc-shaped or V-shaped, the fitting model can be a parabolic fitting model, fitting a quadratic function as the initial bottom contour.

[0141] After obtaining the initial bottom contour, outliers (residuals > 3 standard deviations) are removed to improve the robustness of the fit. Then, the fit is re-fitted to obtain the ideal bottom contour function. The function for the bottom contour line corresponding to all two-dimensional contour lines. The combination of these elements constitutes the theoretical contour line.

[0142] Step c2: Determine the distribution of the difference between the theoretical contour line and the second point cloud data extracted from the bottom of the etched trench from the current 3D point cloud data.

[0143] Specifically, the distribution of the difference can be characterized by the following formula (3). :

[0144] In the formula, This indicates the height of the second point cloud data extracted from the bottom of the etched trench from the current 3D point cloud data.

[0145] Step c3: In the target area near the sidewalls of the etched trench, if the difference between the theoretical contour line and the second point cloud data is less than 0, and the absolute value of the difference is greater than or equal to the third preset threshold, then a target trench exists at the bottom of the etched trench.

[0146] The target area refers to the area from the sidewall of the etched trench to within 0.5 μm of the sidewall of the etched trench. The third preset threshold is the critical threshold for whether the bottom trench is the target trench, which can be the product of the current depth of the etched trench and the scaling factor (e.g., 8%).

[0147] Optionally, within the target area, if the difference between the theoretical contour line and the second point cloud data is less than 0, the absolute value of the difference is greater than or equal to a third preset threshold, and the sidewalls of the target trench exhibit a linear distribution parallel to the sidewalls of the etched trench, then it is determined to be a microgroove bud. This method, compared to calculating the global depth deviation to determine the target trench, can more comprehensively and accurately locate microgroove defects.

[0148] Step c4: Determine the current peak value of the trench residual based on the difference between the theoretical contour line and the second point cloud data within the target area.

[0149] Specifically, the absolute value of the difference between the theoretical contour line and the second point cloud data within the target area is determined as the current peak value of the trench residual.

[0150] When the current derived feature includes the energy of the target frequency band in the current sidewall power spectral density, step S406 above may include the following steps: Step d1: Determine the current sidewall power spectral density based on the contour line of the etched trench sidewall extracted from the current 3D point cloud data.

[0151] Step d2: Determine the energy of the target frequency band based on the current sidewall power spectral density.

[0152] The target frequency band is defined as a frequency greater than 0.01 nm. -1 The frequency band. Power spectral density (PSD) is used to analyze the energy distribution of sidewall profile roughness at different spatial frequencies, which can reveal the physical causes of roughness, such as high-frequency roughness corresponding to passivation layer inhomogeneity, and low-frequency roughness corresponding to mask edge roughness.

[0153] Specifically, firstly, the contour lines of the sidewalls of the etched trenches are extracted. From the current 3D point cloud data, the contour lines of a single sidewall of an etched trench are extracted, and multiple positions are taken along the trench length direction y (e.g., a cross-section every 0.5 μm). Each cross-section yields a 2D sidewall contour. Alternatively, a 1D height sequence can be extracted along the trench length direction. The sidewall contour lines are represented as a 1D function z(y), and the sampling interval of the 1D function z(y) is... Evenly distributed, such as .

[0154] Then, the sidewall profile is processed by removing the trend term and using a window function. The trend term removal is used to remove DC and tilt issues caused by etching inhomogeneities or wafer tilt. The sidewall profile may contain overall tilt or curvature; these low-frequency components interfere with roughness analysis and must be removed first. After obtaining the one-dimensional function z(y) characterizing the profile, a polynomial fitting (usually a first or second-order polynomial) is performed on z(y) to obtain the trend term p(y). The difference between the one-dimensional function z(y) and the trend term p(y) is determined as the detrended profile. Where h(y) represents the roughness fluctuation of the sidewall, with a mean of zero.

[0155] Window functions are used to reduce spectral leakage. Specifically, a window function (such as a Hanning window or a Hamming window) is applied to the trended contour h(y) to reduce spectral leakage caused by finite-length signals. The Hanning window function... As shown in formula (4), the contour line processed by the window function is: .

[0156]

[0157] In the formula, Indicates the length of the outline.

[0158] The contour line processed by the window function is a time-domain signal. Then, the time-domain signal is converted into a frequency-domain signal by performing a Discrete Fourier Transform (DFT) on the contour line processed by the window function using formula (5).

[0159]

[0160] In the formula, This represents the frequency domain signal corresponding to the contour line after the Discrete Fourier Transform. For frequency domain frequency index, Indicates the sampling point index. Indicates the total number of sampling points. This represents the time-domain signal corresponding to the windowed contour line.

[0161] Then, based on the frequency domain signal corresponding to the contour line, the power spectral density of the sidewall is determined. The power spectral density represents the power per unit spatial frequency. For a one-dimensional profile, it can be determined by the following formula (6):

[0162] In the formula, Indicates spatial frequency, , .

[0163] After determining the current sidewall power spectral density, it is plotted as a double logarithmic curve of log(PSD) versus log(frequency). Based on the spatial frequency range, the current sidewall power spectral density can be divided into high-frequency, mid-frequency, and low-frequency bands, as shown in Table 1. The target frequency band refers to the high-frequency band, and the energy of the target frequency band... ,in, .

[0164] Table 1 Frequency band distribution of current sidewall power spectral density

[0165] When the energy in the target frequency band exceeds the preset energy, a sidewall roughness warning is triggered. In a double logarithmic coordinate system, a smaller absolute value of the PSD descent slope (less than the preset slope, such as -1 or -3) indicates that the roughness is expanding to higher frequencies, predicting a deterioration in sidewall quality. The spatial frequency corresponding to the obvious peak in the PSD curve is used to identify periodic roughness, such as the fan-shaped structure of the Bosch process.

[0166] This embodiment uses PSD analysis to predict sidewall roughness, which is more revealing of the physical causes of roughness than a single Rq value.

[0167] Step S407: Input the current derived feature into the anomaly detection model, and determine the next predicted derived feature based on the output of the anomaly detection model.

[0168] For example, the anomaly detection model can be an Autoregressive Integrated Moving Average (ARIMA) model or other machine learning models. The input of the anomaly detection model is the current derived feature, and the output is the predicted next derived feature, which is the derived feature corresponding to the next 3D point cloud data collected in the next acquisition cycle.

[0169] ARIMA models are a class of classic statistical models commonly used in time series analysis and forecasting. The structure is RIMA(p,d,q), where p represents the autoregressive order, the linear dependence of the series on its past p time values; d represents the differencing order, the number of differencing operations required to make a non-stationary series stationary; and q represents the moving average order, the dependence of the series on random errors (white noise) over the past q time values.

[0170] In this invention, the ARIMA model is used to model and predict the time series of defect-sensitive derived features in one step. By using the ARIMA model to predict the next derived feature, the ARIMA parameters can reflect the dynamic characteristics of defect evolution, resulting in stronger interpretability. Online prediction only requires simple linear recursion, meeting the real-time control requirements of semiconductor processes, with low computational load and a single prediction time of less than 10ms. Furthermore, it has good adaptability; by periodically re-estimating parameters through a sliding window, it can track slow process drift.

[0171] Step S408: Determine the actual next derived feature based on the next three-dimensional point cloud data of the etched area in the next acquisition cycle.

[0172] Specifically, the process of determining the actual next derived feature based on the next 3D point cloud data is similar to step S406 above, and will not be repeated here.

[0173] Step S409: Determine the degree of defect based on the predicted next derived feature and the actual next derived feature.

[0174] Step S410: When the defect level is greater than the fourth preset threshold, output early warning information according to the defect level.

[0175] Specifically, when the actual next derived feature deviates from the confidence interval (e.g., 95%) of the predicted next derived feature, an abnormal trench trend can be identified. The degree of defect is directly proportional to the degree of deviation between the predicted and actual next derived features; the greater the deviation, the greater the defect. If the defect degree exceeds a fourth preset threshold, it indicates an abnormal trench trend, requiring an early warning to be issued before the defect actually forms.

[0176] The warning information is divided into three levels. When the defect severity is greater than the fourth preset threshold and the target trench depth is less than or equal to the first depth value, the warning information is yellow, indicating that the etching trench trend may be abnormal and should be checked in time. When the defect severity is greater than the fourth preset threshold and the target trench depth is greater than the first depth value but less than the second depth value, the warning information is orange, indicating that there are micro-grooves at the bottom of the trench and the process parameters need to be adjusted to return the process to the target state. When the defect severity is greater than the fourth preset threshold and the target trench depth is greater than the second depth value or the predicted sidewall roughness is greater than the preset roughness, the warning information is red, indicating that there are defects in the etching trench and the etching process needs to be suspended.

[0177] In this embodiment, current derived features used to characterize the degree of defect growth are determined from the current 3D point cloud data, and defect early warning is performed using these derived features. This allows for effective early warning at the initial stage of defect initiation, such as microgrooves and sidewall roughness, avoiding batch scrap and improving etching process control accuracy and product yield. Specifically, the defect early warning time in this embodiment is 10% to 20% faster than traditional detection methods.

[0178] The defect early warning process will be explained using the ARIMA model as an example. Figure 5 As shown, after obtaining the current derived feature, it is input into the ARIMA model. Based on the output of the ARIMA model, the predicted next derived feature can be obtained, and the actual next derived feature is also obtained. Then, the deviation (residual) between the predicted next derived feature and the actual next derived feature is calculated, and it is determined whether the residual is greater than 1 / 3. If yes, an anomaly warning is triggered; otherwise, the ARIMA model is updated based on the current derived feature and the actual next derived feature, and then the actual next derived feature is used as the current derived feature. The above process is repeated until the etching process is completed.

[0179] This embodiment also provides a method for determining etching process parameters, which can be used in electronic devices such as tablet computers, computers, or servers. Figure 6 This is a flowchart illustrating a method for determining parameters of an etching process according to an embodiment of the present invention, as shown below. Figure 6 As shown, the process includes the following steps: Step S601: During the etching process, acquire the current three-dimensional point cloud data of the etching area in the current acquisition cycle.

[0180] Please see details Figure 1 Step S101 of the illustrated embodiment or Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0181] Step S602: Determine the current etching depth and the current sidewall angle based on the current 3D point cloud data.

[0182] Please see details Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0183] Step S603: Determine the depth change rate based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles.

[0184] Please see details Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0185] Step S604: Determine whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles.

[0186] Please see details Figure 1 Step S104 of the illustrated embodiment or Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0187] Step S605: Determine whether the etching endpoint has been reached based on whether the depth change rate and sidewall angle are in a stable state.

[0188] Please see details Figure 1 Step S105 of the illustrated embodiment will not be described again here.

[0189] Step S606: After determining that the etching endpoint has been reached, store the current 3D point cloud data for the current acquisition cycle into the process database.

[0190] The process database includes multiple 3D point cloud data arranged by time.

[0191] Step S607: Train the process model based on the process database.

[0192] The process model is used to characterize the correspondence between process features and process parameters. Process features are those extracted from 3D point cloud data that reflect the execution status of the etching process. Process parameters include bias power and gas flow ratio, etc.

[0193] Process characteristics are used to quantify the health, uniformity, defect tendency, and controllability of the etching process. For example, process characteristics may include at least one of the following: the shape of the etching rate decay curve over time (linear or saturated), the rate difference in different measurement regions, the critical depth at which the target trench appears, the target energy growth rate in the sidewall PSD, and the second derivative of the defect density.

[0194] The shape of the etching rate decay curve over time indicates the gradual consumption of the photoresist mask or the stability of the chamber state. A saturation type (sudden drop in rate) indicates the arrival of the stop layer or the appearance of a micro-loading effect. The rate difference in different measurement areas is used to quantify the loading effect. Excessive difference indicates a process window offset, requiring adjustment of gas flow rate or power. The critical depth at which the target trench appears characterizes the process selectivity and sidewall protection capability. The larger the critical depth, the more the process can tolerate deeper etching without producing micro-trenches, and the better the process robustness. The target energy growth rate in the sidewall PSD characterizes the uniformity of passivation layer deposition. Rapid growth indicates insufficient sidewall protection and may produce fan-shaped roughness. The second derivative of the defect density characterizes the degree of chamber contamination or component aging. A positive second derivative indicates the entry into the accelerated failure stage, requiring predictive maintenance.

[0195] After storing the 3D point cloud data, process features are extracted from the 3D point cloud data and the corresponding process parameters are determined. With the process parameters as input and the process features as output, the process model is trained using methods such as Gaussian process regression to establish the mapping relationship between process parameters and process features.

[0196] Step S608: Based on the process model, the optimal combination of process parameters that meets the preset optimization objective is determined by searching in the process parameter space using an optimization algorithm.

[0197] The preset optimization objective can be a single objective or multiple objectives. For example, preset optimization objectives may include reaching the target depth in the shortest time, maximizing the critical depth of the target trench, and minimizing the rate difference in the measurement area. The optimization algorithm can be a Bayesian optimization algorithm, a genetic algorithm, or other optimization algorithms.

[0198] Specifically, by searching the process parameter space through Bayesian optimization, a Pareto optimal solution that satisfies the preset optimization objective is found, and the Pareto optimal solution is used as the optimal combination of process parameters for the next etching process.

[0199] In this embodiment, after determining that the etching endpoint has been reached, the current 3D point cloud data of the current acquisition cycle is stored in the process database. The accumulated 3D point cloud data can provide process engineers with an intuitive etching process, and the process model trained based on the process database can help optimize the formulation of the next generation of etching processes, thereby improving the processing quality and yield of the etching process.

[0200] For example, such as Figure 7 As shown, the etching process monitoring method and etching process parameter determination method provided by the present invention can include four core steps: Step e1: Real-time three-dimensional topography monitoring. During the etching process, a multi-wavelength optical measurement system is used to perform full-field three-dimensional topography measurement of the etched area at fixed time intervals (e.g., every 30 seconds) to obtain parameters such as trench depth, sidewall angle, and bottom surface roughness.

[0201] Step e2, endpoint determination: Calculate the depth change rate based on the extracted trench depth, and calculate the sidewall angle change based on the sidewall angle. Determine the stability of the sidewall angle based on the sidewall angle change, and then comprehensively determine whether the endpoint condition is met. When the depth change rate is lower than the set threshold Thr1 and the sidewall angle remains stable in multiple consecutive measurements (…), the endpoint is reached. At Thr2, an end signal is sent to stop etching.

[0202] Step e3, defect early warning: If the endpoint condition is not met, abnormal features such as microgrooves and abnormally increased sidewall roughness are detected in real time based on the 3D topography data. The defect development trend is predicted through time series analysis. When a defect bud is detected, an early warning is issued, and the defect type and defect location are recorded. If no defect bud is detected, the full-field 3D topography measurement process continues, repeating steps e1 to e3 above.

[0203] Step e4, process optimization feedback: After the endpoint signal is issued, the detection results are stored in the process database for subsequent optimization and adjustment of the process formula, so as to realize adaptive process control.

[0204] This embodiment directly performs endpoint detection based on three-dimensional morphology parameters, avoiding the errors of indirect methods and improving the accuracy of endpoint detection. Compared with traditional laser interferometry, the false endpoint detection rate is reduced by more than 50%, reducing material waste caused by over-etching, lowering rework costs, and improving equipment utilization. It can issue early warnings at the initial stage of defect formation (10% to 20% of the process time earlier than traditional detection), avoiding batch scrap. The accumulated morphology evolution data provides process engineers with an intuitive understanding of the etching process, which can assist in optimizing next-generation process formulations.

[0205] The etching process monitoring method provided by this invention is applied in the deep trench etching process of 12-inch silicon wafers. The target etching depth is 10 μm, the aspect ratio of the etching trench is 20:1, and the semiconductor process equipment is an inductively coupled plasma (ICP) device. The specific process parameters are: power 1500W, pressure 10 mTorr, and gas SF6 / C4F8 circulation. The measurement is set to perform a full-field three-dimensional measurement every 30 seconds, and the etching area is divided into multiple measurement areas (dense area, isolated area, and test pattern area).

[0206] Figure 8 A comparison graph of the etching depth versus time at the monitoring endpoint using laser interferometry and this invention is provided. Figure 8 It can be seen that the etching depth curve obtained by this invention is basically consistent with the actual etching depth, resulting in higher accuracy for endpoint detection. Among other things, Figure 8 In this context, data1 represents the location of the actual endpoint.

[0207] Figure 9 A comparison chart of the endpoint determination error time of the laser interferometry method and the present invention is provided, from... Figure 9 It can be seen that the average endpoint judgment error of the present invention is 2.6s, while the average endpoint judgment error of the laser interferometry method is 19.7s, and the endpoint judgment error is reduced by 7.6 times.

[0208] When performing defect warning, the bottom contour of the etching trench can normally be as follows: Figure 10 As shown, the bottom profile where microgrooves appear can be as follows: Figure 11 As shown.

[0209] Figure 12 A comparison chart showing the time it took for the defect to be discovered using laser interferometry and this invention is provided. Figure 12 It can be seen that the traditional laser interferometry method detects defects on average in about 3 seconds, while the present application detects defects on average more than 4 seconds earlier than the traditional method, which can provide early warning in the early stages of defect formation and avoid mass scrapping. Figure 12 In this context, data1 represents the traditional average defect warning time, and data2 represents the average defect warning time of the present invention compared to the traditional solution.

[0210] Figure 13 A comparative diagram of batch yield before and after etching process parameter optimization is provided. Before the optimization, the average yield of 25 batches was 71.2%, and after the optimization, the average yield of 25 batches was 88.0%, an improvement of 16.8 percentage points.

[0211] This embodiment also provides a monitoring device for the etching process and a parameter determination device for the etching process. This device is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0212] This embodiment provides a monitoring device for etching processes, such as... Figure 14 As shown, it includes: The acquisition module 1401 is used to acquire the current three-dimensional point cloud data of the etched area during the current acquisition cycle during the etching process. The parameter determination module 1402 is used to determine the current etching depth and the current sidewall angle based on the current three-dimensional point cloud data. The current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region. The rate of change determination module 1403 is used to determine the rate of change of depth based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles, wherein the rate of change of depth is the rate of change of etching depth over time. The status judgment module 1404 is used to determine whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles. The endpoint determination module 1405 is used to determine whether the etching endpoint has been reached based on the depth change rate and whether the sidewall angle is in a stable state.

[0213] This embodiment also provides a device for determining the parameters of an etching process, such as... Figure 15 As shown, it includes: The acquisition module 1401 is used to acquire the current three-dimensional point cloud data of the etched area during the current acquisition cycle during the etching process. The parameter determination module 1402 is used to determine the current etching depth and the current sidewall angle based on the current three-dimensional point cloud data. The current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region. The rate of change determination module 1403 is used to determine the rate of change of depth based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles, wherein the rate of change of depth is the rate of change of etching depth over time. The status judgment module 1404 is used to determine whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles. The endpoint determination module 1405 is used to determine whether the etching endpoint has been reached based on the depth change rate and whether the sidewall angle is in a stable state. Storage module 1501 is used to store the current three-dimensional point cloud data of the current acquisition cycle into the process database after determining that the etching endpoint has been reached. The process database includes multiple three-dimensional point cloud data arranged by time. Training module 1502 is used to train a process model based on the process database. The process model is used to characterize the correspondence between process features and process parameters. The process features are features extracted from three-dimensional point cloud data that reflect the execution state of the etching process. The parameter optimization module 1503 is used to search and determine the optimal combination of process parameters that meets the preset optimization objectives in the process parameter space based on the process model and through optimization algorithms.

[0214] The etching process monitoring device provided in this embodiment of the invention can execute the etching process monitoring method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. The etching process parameter determination device provided in this embodiment of the invention can execute the etching process parameter determination method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0215] Figure 16 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0216] Embodiments of this application also provide an electronic device, such as... Figure 16 As shown, the electronic device includes a memory 1610 and a processor 1620. The memory 1610 stores a computer program, and the processor 1620 is configured to run the computer program to perform the steps in the monitoring method embodiment of any of the above-described etching processes, or to perform the steps in the parameter determination method embodiment of any of the above-described etching processes. Figure 16 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0217] Furthermore, the electronic device also includes a communication interface 1630 for communicating with other devices or communication networks.

[0218] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium after being downloaded via a network. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the etching process monitoring method or etching process parameter determination method shown in the above embodiments is implemented.

[0219] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0220] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for monitoring an etching process, characterized in that, The method includes: During the etching process, acquire the current 3D point cloud data of the etched area in the current acquisition cycle; Based on the current 3D point cloud data, the current etching depth and the current sidewall angle are determined, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region. Based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles, the depth change rate is determined, wherein the depth change rate is the rate of change of etching depth over time. Based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles, determine whether the sidewall angle is in a stable state. The etching endpoint is determined based on whether the depth change rate and the sidewall angle are in a stable state.

2. The method according to claim 1, characterized in that, The step of determining whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles corresponding one-to-one with multiple consecutive historical acquisition cycles includes: The angle change rate is determined based on the standard deviation of the current sidewall angle and the standard deviation of each historical sidewall angle, wherein the angle change rate is the rate of change of the standard deviation of the sidewall angle over time. If the rate of change of the angle is less than or equal to 0 and the standard deviation of the current sidewall angle is less than the second preset threshold, then the sidewall angle is in a stable state. The step of determining whether the etching endpoint has been reached based on the rate of change of the etching depth over time and whether the sidewall angle is in a stable state includes: If the depth change rate is less than a first preset threshold and the sidewall angle is in a stable state, then the etching endpoint is determined to have been reached.

3. The method according to claim 2, characterized in that, The step of determining whether the etching endpoint has been reached based on the rate of change of the etching depth over time and whether the sidewall angle is in a stable state includes: If the depth change rate is less than the first preset threshold and the sidewall angle is in a stable state, then update the number of candidate endpoints. If the number of candidate endpoints after the update is less than the preset number, then based on the target three-dimensional point cloud data of the etched area under the target acquisition cycle, it is re-determined whether the depth change rate and sidewall angle are in a stable state, wherein the duration of the target acquisition cycle is less than the duration of the current acquisition cycle. If the number of candidate endpoints after the update is greater than or equal to the preset number, then the etching endpoint is determined to have been reached.

4. The method according to claim 1, characterized in that, The etched area includes a dense area, an isolated area, and a test pattern area. The depth change rate includes the dense depth change rate of the dense area, the isolated depth change rate of the isolated area, and the test depth change rate of the test pattern area. Whether the sidewall angle is in a stable state includes whether the sidewall angle of the dense area is in a stable state, whether the sidewall angle of the isolated area is in a stable state, and whether the sidewall angle of the test pattern area is in a stable state. The step of determining whether the etching endpoint has been reached based on the rate of change of the etching depth over time and whether the sidewall angle is in a stable state includes: If the rate of change of the dense depth is less than a first preset threshold and the sidewall angle of the dense region is in a stable state, then it is determined that the dense region has reached the etching endpoint. If the rate of change of the isolated depth is less than the first preset threshold and the sidewall angle of the isolated region is in a stable state, then it is determined that the isolated region has reached the etching endpoint. If the rate of change of the test depth is less than the first preset threshold and the sidewall angle of the test pattern area is in a stable state, then it is determined that the test pattern area has reached the etching endpoint.

5. The method according to any one of claims 2 to 4, characterized in that, Before determining whether the etching endpoint has been reached, the method further includes: The first preset threshold is determined based on the material of the etched region; or, the first preset threshold is determined based on the historical average rate of the etched region.

6. The method according to any one of claims 1 to 4, characterized in that, The acquisition of the current 3D point cloud data of the etched area in the current acquisition cycle includes: The current three-dimensional point cloud data is acquired from the optical measurement system, wherein the optical measurement system is used to output multiple projection lights of different wavelengths, and the projection lights of different wavelengths are used to form stripes of different spatial frequencies, or each projection light is used to form a set of stripes of different spatial frequencies, and the different spatial frequencies are in a multiple relationship.

7. The method according to claim 6, characterized in that, Multiple projected lights of different wavelengths are different from the characteristic emission spectrum of the plasma in the etching process.

8. The method according to claim 7, characterized in that, There are three projection lights, each with a different wavelength: 450nm, 650nm, and 850nm.

9. The method according to claim 6, characterized in that, When the radio frequency power switch in the semiconductor process equipment is in the off state or during the period when the plasma afterglow in the semiconductor process equipment is at its minimum, the optical measurement system outputs multiple projection lights of different wavelengths to detect the current three-dimensional point cloud data.

10. The method according to any one of claims 1 to 4, characterized in that, The method further includes: Based on the current 3D point cloud data, a current derived feature is determined to characterize the degree of defect growth. The defect includes at least one of a target trench located at the bottom of the etched trench and sidewall roughness. When the defect includes the target trench, the current derived feature includes the current trench residual peak value. When the defect includes the sidewall roughness, the current derived feature includes the energy of the target frequency band in the current sidewall power spectral density. The current trench residual peak value is the maximum difference between the target trench and the theoretical depth in the current acquisition period. The current derived feature is input into the anomaly detection model, and the next predicted derived feature is determined based on the output of the anomaly detection model. Based on the next 3D point cloud data of the etched area in the next acquisition cycle, determine the actual next derived feature; The degree of defect is determined based on the predicted next derived feature and the actual next derived feature; When the degree of defect exceeds a fourth preset threshold, a warning message is output based on the degree of defect.

11. The method according to claim 10, characterized in that, When the current derived feature includes the current trench residual peak value, the step of determining the current derived feature for characterizing the degree of defect growth based on the current 3D point cloud data includes: Based on the first point cloud data of the sidewall of the etched trench extracted from the current three-dimensional point cloud data, the theoretical contour line of the bottom of the etched trench is fitted and determined. Determine the distribution of the difference between the theoretical contour line and the second point cloud data extracted from the bottom of the etched trench from the current three-dimensional point cloud data; In the target area near the sidewalls of the etched trench, if the difference between the theoretical contour line and the second point cloud data is less than 0, and the absolute value of the difference is greater than or equal to the third preset threshold, then the target trench exists at the bottom of the etched trench. The current trench residual peak value is determined based on the difference between the theoretical contour line and the second point cloud data within the target area.

12. The method according to claim 10, characterized in that, When the current derived feature includes the energy of the target frequency band in the current sidewall power spectral density, determining the current derived feature for characterizing the degree of defect growth based on the current three-dimensional point cloud data includes: The current sidewall power spectral density is determined based on the contour line of the etched trench sidewall extracted from the current three-dimensional point cloud data. Based on the current sidewall power spectral density, the energy of the target frequency band is determined, wherein the target frequency band is a frequency greater than 0.01 nm. -1 The frequency band.

13. A method for determining parameters of an etching process, characterized in that, The method includes: During the etching process, acquire the current 3D point cloud data of the etched area in the current acquisition cycle; Based on the current 3D point cloud data, the current etching depth and the current sidewall angle are determined, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region. Based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles, the depth change rate is determined, wherein the depth change rate is the rate of change of etching depth over time. Based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles, determine whether the sidewall angle is in a stable state. Whether the etching endpoint has been reached is determined based on whether the depth change rate and the sidewall angle are in a stable state; After determining that the etching endpoint has been reached, the current three-dimensional point cloud data of the current acquisition cycle is stored in the process database, which includes multiple three-dimensional point cloud data arranged by time. Based on the process database, a process model is trained, wherein the process model is used to characterize the correspondence between process features and process parameters, and the process features are features extracted from the three-dimensional point cloud data that reflect the execution state of the etching process. Based on the process model, the optimal combination of process parameters that meets the preset optimization objective is determined by searching in the process parameter space using an optimization algorithm.

14. A monitoring device for an etching process, characterized in that, The monitoring device includes: The acquisition module is used to acquire the current three-dimensional point cloud data of the etched area during the current acquisition cycle during the etching process. The parameter determination module is used to determine the current etching depth and the current sidewall angle based on the current three-dimensional point cloud data, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region. The rate of change determination module is used to determine the depth change rate based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles, wherein the depth change rate is the rate of change of etching depth over time. The status judgment module is used to determine whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles. The endpoint determination module is used to determine whether the etching endpoint has been reached based on whether the depth change rate and the sidewall angle are in a stable state.

15. A parameter determination device for an etching process, characterized in that, The parameter determination device includes: The acquisition module is used to acquire the current three-dimensional point cloud data of the etched area during the current acquisition cycle during the etching process. The parameter determination module is used to determine the current etching depth and the current sidewall angle based on the current three-dimensional point cloud data, wherein the current etching depth is the set of etching depths corresponding to each etching trench in the etching region, and the current sidewall angle is the set of sidewall angles corresponding to each etching trench in the etching region. The rate of change determination module is used to determine the depth change rate based on the current etching depth and multiple historical etching depths that correspond one-to-one with multiple consecutive historical acquisition cycles, wherein the depth change rate is the rate of change of etching depth over time. The status judgment module is used to determine whether the sidewall angle is in a stable state based on the current sidewall angle and multiple historical sidewall angles that correspond one-to-one with multiple consecutive historical acquisition cycles. The endpoint determination module is used to determine whether the etching endpoint has been reached based on whether the depth change rate and the sidewall angle are in a stable state. The storage module is used to store the current three-dimensional point cloud data of the current acquisition cycle into the process database after determining that the etching endpoint has been reached. The process database includes multiple three-dimensional point cloud data arranged in time. The training module is used to train a process model based on the process database. The process model is used to characterize the correspondence between process features and process parameters. The process features are features extracted from the three-dimensional point cloud data that reflect the execution state of the etching process. The parameter optimization module is used to search and determine the optimal combination of process parameters that meets the preset optimization objective in the process parameter space based on the process model and through optimization algorithms.

16. An electronic device, characterized in that, include: The system includes a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the monitoring method for the etching process according to any one of claims 1 to 12, or the parameter determination method for the etching process according to claim 13.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the monitoring method of the etching process according to any one of claims 1 to 12, or the parameter determination method of the etching process according to claim 13.

18. A computer program product, characterized in that, Includes computer instructions, which are used to cause a computer to perform the monitoring method of the etching process according to any one of claims 1 to 12, or to perform the parameter determination method of the etching process according to claim 13.

Images