Sampling frequency calculation method, system and device and storage medium

Abstract

The invention provides a method, a system and equipment for calculating the sampling frequency of a spectrograph and a storage medium, and relates to the technical field of semiconductor detection, and a time window set of a reference parameter and a key event obtained by querying a historical knowledge base provides a prior basis for configuration of the sampling frequency; a real-time adaptation factor is generated based on the collected spectral data in each sampling period, and the real-time change of the current working condition can be reflected; the target sampling frequency is configured by combining the time of the timer and the time window set matching result of the key event, the reference parameter and the real-time adaptation factor, the sampling frequency can be adaptively configured based on the key event, and distribution of sampling resources according to needs is achieved; the target sampling frequency is used for the next sampling period to form a continuous closed-loop updating mechanism, so that the optimal sampling strategy is dynamically generated without continuous manual intervention, and reliable monitoring of the process core quality forming process is guaranteed while the overall efficiency of data acquisition is greatly improved finally.

Description

Technical Field

[0001] This application relates to the field of semiconductor detection technology, and in particular to a method, system, device and storage medium for calculating the sampling frequency of a spectrometer. Background Technology

[0002] In semiconductor epitaxial growth, key parameters such as film thickness, composition, and interface quality have a decisive impact on device performance. To achieve real-time monitoring and process traceability of the epitaxial growth process, the industry typically introduces in-situ optical monitoring methods within the reaction chamber. Among these, in-situ monitoring based on reflectance spectroscopy is widely used in epitaxial process development and mass production process control due to its non-contact nature, fast response, and rich information content. In practical applications, spectrometers need to continuously acquire reflectance spectral data during epitaxial growth to support subsequent feature extraction, process determination, and anomaly diagnosis.

[0003] Traditional sampling strategies commonly include fixed-frequency sampling and adaptive sampling based on simple empirical rules. Fixed-frequency sampling is simple to implement, but it struggles to account for significant spectral changes between different process steps, often resulting in data redundancy in non-critical stages and insufficient sampling resolution in critical stages. While adaptive sampling based on empirical rules can switch sampling frequencies step-by-step to some extent, its rules are usually manually preset and statically fixed, offering limited adaptability to formulation iterations, equipment state drift, or abnormal operating conditions, and incurring high maintenance costs. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, device and storage medium for calculating the sampling frequency of a spectrometer, so as to overcome the deficiency of traditional spectrometers in lacking adaptive scheduling capability for different processes.

[0005] Firstly, this application proposes a method for calculating the sampling frequency of a spectrometer, applicable to a semiconductor epitaxial growth detection device, wherein the semiconductor epitaxial growth detection device includes a spectrometer; the method includes:

[0006] In response to a process change signal, initialize the timer and determine the current process; Based on the current process, the corresponding baseline parameters and time window set of key events are obtained by querying the preset historical knowledge base; For each sampling period during the current process, acquire the spectral data sampled by the spectrometer based on the current sampling frequency, and generate a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data; The timer time corresponding to the current sampling period is matched with the set of time windows, and the target sampling frequency is configured according to the matching result, the benchmark parameter and the real-time adaptation factor. The target sampling frequency is used as the sampling frequency for the next sampling period of the spectrometer.

[0007] In one embodiment, the benchmark parameters include benchmark statistics and benchmark sampling frequency; Methods for pre-setting a historical knowledge base include: Acquire historical testing data, which includes at least one of the following: raw reflectance spectral time-series data, process time-series log, wafer endpoint testing data, and device electrical test data; Key data is extracted from the historical detection data, including process sequences that meet the screening criteria and their corresponding spectral characteristic curves; Identify key events in the key data to obtain the time window set and stable time interval of the key events; Calculate a sample set of the rate of change of the spectral characteristic curves over time for each process series, and perform statistics on the sample set of the rate of change to obtain a baseline statistic; The baseline sampling frequency is obtained by statistically analyzing the historical sampling frequencies within the aforementioned stable time interval. The baseline sampling frequency, baseline statistics, and time window set of key events are stored according to a preset data structure to obtain a historical knowledge base.

[0008] In one embodiment, identifying key events in the key data and obtaining a set of time windows and a stable time interval for the key events includes: Perform time-series alignment and pattern clustering operations on the key data to obtain multiple change templates corresponding to each process sequence; parse each change template, identify the key events within it, and obtain a time window set for the key events corresponding to each change template; the time interval outside the time window set is denoted as the stationary time interval.

[0009] In one embodiment, the step of performing time-series alignment and pattern clustering operations on the key data to obtain multiple variation templates corresponding to each process sequence includes: Obtain the spectral characteristic curves of different batches under the same process sequence; A dynamic time warping algorithm is used to perform nonlinear time axis alignment on each of the spectral feature curves to obtain a set of feature curves. Cluster analysis is performed on the feature curve set to divide the feature curve set into a preset number of pattern categories; for each pattern category, the feature curves belonging to that pattern category are extracted representatively to generate corresponding typical change pattern templates, thereby obtaining multiple change templates.

[0010] In one embodiment, the step of parsing each of the change templates, identifying key events within them, and obtaining a time window set for the key events corresponding to each change template includes: Numerical differentiation is performed on each of the aforementioned change templates to obtain a sequence of the rate of change and a sequence of the acceleration of change that characterize the change template as a function of time within the process. Local extreme points are detected in the rate of change sequence to determine the critical moment when the change in the change template reaches its fastest or slowest point. In the changing acceleration sequence, the turning point where the sign changes is detected is used to determine the key moment when the changing trend in the changing template changes; The key moments corresponding to the local extreme points and the key moments corresponding to the turning points are taken as key event points; and with each key event point as the center, the corresponding key event time window is formed by expanding in a preset direction, thereby obtaining the time window set of key events corresponding to each change template.

[0011] In one embodiment, the benchmark statistic includes the arithmetic mean and standard deviation; The calculation of the rate of change of the spectral characteristic curve over time for each process series, and the statistical analysis of the rate of change sample set to obtain a baseline statistic, includes: Obtain the spectral characteristic curves and their time information of different batches under the same process sequence; Based on the spectral feature difference between adjacent sampling points and the sampling time interval, the rate of change of spectral features at each sampling point is calculated to form a sample set of the rate of change under the process sequence. Statistical analysis is performed on the sample set of change rates to calculate the arithmetic mean that characterizes the normal degree of change in the process sequence and the standard deviation that characterizes the normal fluctuation range of the process sequence.

[0012] In one embodiment, the benchmark parameter includes a benchmark statistic; The step of acquiring spectral data sampled by the spectrometer based on the current sampling frequency, and generating a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data, includes: Acquire the spectral data obtained by the spectrometer from two consecutive samplings based on the current sampling frequency; Extract the characteristic wavelength values ​​and their corresponding sampling time intervals from two continuous spectral data, and calculate the instantaneous rate of change based on the characteristic wavelength values ​​and the sampling time interval; Based on two continuous spectral data, the correlation between the two is calculated on a preset key band to obtain a stability index characterizing the stability of the spectral shape; wherein the preset key band includes the stop band edge band. Based on the degree of deviation of the instantaneous rate of change from the benchmark statistic and the degree of spectral instability characterized by the stability index, the real-time adaptation factor is generated to simultaneously measure the degree of deviation of the rate of change from the historical normal and the degree of decline in spectral stability.

[0013] In one embodiment, generating the real-time adaptation factor based on the deviation of the instantaneous rate of change from the benchmark statistic and the spectral instability characterized by the stability index includes: The expression for calculating the real-time adaptation factor is:

[0014]

[0015]

[0016] in, For real-time adaptation factors; It is the instantaneous rate of change; and These are the arithmetic mean and standard deviation, respectively. As a stability index; Pearson correlation coefficient; Key band; The characteristic wavelength value; This represents the sampling time interval.

[0017] In one embodiment, the benchmark parameters include preset parameters, benchmark sampling frequency, and benchmark statistics; The step of matching the timer time corresponding to the current sampling period with the set of time windows, and configuring the target sampling frequency according to the matching result, the benchmark parameter, and the real-time adaptation factor, includes: If the timer time corresponding to the current sampling period falls into the time window set, a first frequency is determined based on the preset parameters, the benchmark statistics, and the real-time adaptation factor; otherwise, a second frequency is determined based on the preset parameters, the benchmark sampling frequency, and the real-time adaptation factor, wherein the first frequency is higher than the second frequency. The first frequency or the second frequency is used as the target sampling frequency.

[0018] In one embodiment, the preset parameters include a key frequency and a key gain; The first frequency is determined based on preset parameters, the benchmark statistics, and the real-time adaptation factor, and its expression is as follows:

[0019] in, For key frequencies; This is the key gain.

[0020] In one embodiment, the preset parameters include a minimum threshold frequency and a stationary period gain; The second frequency is determined based on the preset parameters, the reference sampling frequency, and the real-time adaptation factor, and its expression is as follows:

[0021] in, The reference sampling frequency; The lowest threshold frequency, This represents the gain during the plateau period.

[0022] Secondly, this application proposes a system for calculating the sampling frequency of a spectrometer, the system being installed within a semiconductor epitaxial growth detection device, the semiconductor epitaxial growth detection device including a spectrometer; the system comprising: The process interpreter is used to acquire process change signals issued in real time by the main controller of the epitaxial equipment, and in response to the process change signals, initialize the timer and determine the current process; based on the current process, it queries the preset historical knowledge base to obtain the corresponding baseline parameters and time window set of key events; A real-time sensing adapter is used to acquire spectral data sampled by the spectrometer based on the current sampling frequency for each sampling cycle during the current process duration, and to generate a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data. The decision-maker is used to match the timer time corresponding to the current sampling period with the set of time windows, and configure the target sampling frequency according to the matching result, the reference parameters and the real-time adaptation factor; the target sampling frequency is used as the sampling frequency of the spectrometer in the next sampling period, and a sampling instruction is generated. An actuator is used to generate a timer interrupt signal in response to the sampling command, so as to trigger the spectrometer to perform a data acquisition action.

[0023] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method steps in the first aspect.

[0024] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the method steps of the first aspect.

[0025] The above-mentioned method, system, equipment, and storage medium for calculating the sampling frequency of the spectrometer have at least the following advantages: This application retrieves the corresponding baseline parameters and key event time window sets from a pre-set historical knowledge base based on the current process, providing a priori basis for configuring the subsequent sampling frequency. During each sampling cycle of the current process, a real-time adaptation factor characterizing the degree of spectral anomalies is generated based on the collected spectral data, enabling sampling decisions to reflect immediate changes in the current operating conditions. Furthermore, the timer time and the key event time window sets are matched, and the target sampling frequency is configured by combining the matching results, baseline parameters, and real-time adaptation factors. The sampling frequency can be adaptively configured based on key events, achieving on-demand allocation of sampling resources. Finally, the target sampling frequency is used for the next sampling cycle, forming a continuous closed-loop update mechanism. This dynamically generates the optimal sampling strategy without continuous manual intervention, ultimately significantly improving the overall efficiency of data acquisition while ensuring reliable monitoring of the core quality formation process. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the method for calculating the sampling frequency of a spectrometer in one embodiment; Figure 2 This is a flowchart illustrating the steps involved in building a historical knowledge base in one embodiment. Figure 3 This is a flowchart illustrating the steps for generating real-time adaptation factors in one embodiment. Figure 4 This is a block diagram of the system for calculating the sampling frequency of a spectrometer in one embodiment; Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0027] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0028] Some exemplary embodiments of this application have been described for illustrative purposes. It should be understood that this application may be implemented in other ways not specifically shown in the accompanying drawings.

[0029] Please see Figure 1 In one embodiment, this application provides a method for calculating the sampling frequency of a spectrometer. This method is applicable to a semiconductor epitaxial growth detection device, which includes a spectrometer. Specifically, it includes the following steps: Step 102: In response to the process change signal, initialize the timer and determine the current process.

[0030] Specifically, the epitaxial growth process typically consists of multiple sequential process steps, each corresponding to different control conditions and exhibiting different spectral variation characteristics. A process change signal is an event signal output by the main control system of the semiconductor epitaxial equipment when a process step is switched during recipe execution, indicating that the epitaxial growth process has entered a new process stage. Typically, this process change signal carries information such as the step code and step ID.

[0031] The timer is used to keep track of the current process execution time. Upon detecting a process change signal, the timer is reset to zero or the current time is used as the start time to obtain the starting reference time of the process step. Thus, the relative time within the process can be obtained during the process duration.

[0032] Step 104: Based on the current process, retrieve the corresponding baseline parameters and time window set of key events from the preset historical knowledge base.

[0033] Specifically, the historical knowledge base is used to store prior information on the spectral behavior corresponding to each process step in the epitaxial growth process, so that the baseline parameters and time window sets of key events can be quickly retrieved in the online stage based on the current process. In this embodiment, the historical knowledge base is constructed based on historical detection data, which includes at least one of the following: raw reflectance spectral time series data, process time series logs, wafer endpoint detection data (EPD), and device electrical test data.

[0034] The aforementioned historical detection data is packaged into structured data packets in batches, and the timestamps are strictly aligned.

[0035] Step 106: For each sampling period during the current process, acquire the spectral data obtained by the spectrometer based on the current sampling frequency, and generate a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data.

[0036] Specifically, a process step typically lasts from several seconds to several minutes. In this embodiment, multiple sampling periods are set for each process step, with each sampling period corresponding to one sampling frequency update and at least one spectral acquisition. At the start of the current process, the sampling frequency of the previous process can be continued, or a preset sampling frequency can be used. The spectrometer analyzes the spectral data based on this frequency, extracts features from it, and obtains feature results characterizing the current process state. Based on these feature results, the degree of deviation of the current spectral dynamics from historical norms is comprehensively evaluated, thereby generating a real-time adaptation factor.

[0037] Step 107: Match the timer time and time window set corresponding to the current sampling period, and configure the target sampling frequency according to the matching result, reference parameters and real-time adaptation factor; use the target sampling frequency as the sampling frequency of the spectrometer in the next sampling period.

[0038] Specifically, during the process, at each sampling cycle, the current value of the timer is read as the relative time within the corresponding step. This relative time is then matched against a set of time windows for key events. If the relative time falls within the start and end range of any key event time window, the current sampling cycle is considered to be in a key event phase; otherwise, it is considered to be in a stable phase. Based on the phase of the current sampling cycle, a corresponding target sampling frequency is then configured.

[0039] The above-mentioned method for calculating the spectrometer sampling frequency is based on retrieving the corresponding benchmark parameters and key event time window sets from a preset historical knowledge base for the current process, providing a priori basis for configuring the subsequent sampling frequency. During each sampling cycle in the current process, a real-time adaptation factor characterizing the degree of spectral anomalies is generated based on the collected spectral data, enabling sampling decisions to reflect the immediate changes in the current operating conditions. Furthermore, the timer time and the key event time window sets are matched, and the target sampling frequency is configured by combining the matching results, benchmark parameters, and real-time adaptation factors. The sampling frequency can be adaptively configured based on key events, realizing on-demand allocation of sampling resources. Finally, the target sampling frequency is used for the next sampling cycle, forming a continuous closed-loop update mechanism. This allows for the dynamic generation of the optimal sampling strategy without continuous manual intervention, ultimately achieving a significant improvement in the overall efficiency of data acquisition while ensuring reliable monitoring of the core quality formation process.

[0040] Please see Figure 2 Optionally, when the benchmark parameters include benchmark statistics and benchmark sampling frequency, the methods for pre-setting the historical knowledge base include: Step 202: Obtain historical detection data.

[0041] Step 204: Extract key data from historical testing data. Key data includes process sequences that meet the screening criteria and their corresponding spectral characteristic curves.

[0042] Step 206: Identify key events in the key data and obtain the time window set and stationary time interval of the key events.

[0043] Step 208: Calculate the sample set of the rate of change of the spectral characteristic curve over time for each process series, and perform statistics on the sample set of the rate of change to obtain the baseline statistics.

[0044] Step 210: Statistically calculate the historical sampling frequency within the stable time interval to obtain the baseline sampling frequency.

[0045] Step 212: Store the baseline sampling frequency, baseline statistics, and time window set of key events according to the preset data structure to obtain the historical knowledge base.

[0046] Specifically, by setting filtering criteria, data packets that meet the criteria are selected from historical inspection data for analysis, thereby identifying key event segments in each process step and summarizing them into one or more time windows of key events. Simultaneously, in non-critical event segments, sampling behavior and spectral change characteristics are statistically analyzed to learn baseline parameters for online scheduling. The filtering criteria include meeting at least one of the following: a preset yield threshold, passing EPD inspection, and meeting key electrical indicators.

[0047] Furthermore, the spectral characteristic curve is a time-varying characteristic sequence extracted from the original reflectance spectrum, which typically includes stopband edge characteristic wavelengths, peak and valley positions, or fusion characteristic wavelengths.

[0048] Optionally, key events in the key data are identified to obtain a set of time windows and stationary time intervals for the key events, including: Perform time-series alignment and pattern clustering operations on key data to obtain multiple change templates corresponding to each process sequence; parse each change template to identify key events and obtain the time window set of key events corresponding to each change template; the time interval outside the time window set is denoted as the stationary time interval.

[0049] Specifically, due to slight differences in the actual growth rate and response delay among different batches under the same process, the occurrence time of the same physical process may be earlier or later in different batches, resulting in a stretching effect on the time axis of the spectral characteristic curves of each batch. Based on this, this application performs time alignment on the aforementioned key data, so that curve segments with the same change pattern correspond to a consistent relative position as much as possible after alignment.

[0050] The purpose of clustering is to track different variations that may exist under the same process, thereby obtaining the trajectory of changes caused by differences in equipment status, material batches, or formula fine-tuning.

[0051] Optionally, time-series alignment and pattern clustering operations are performed on the key data to obtain multiple variation templates corresponding to each process sequence, including: Spectral characteristic curves of different batches under the same process sequence are obtained; a dynamic time warping algorithm is used to perform nonlinear time axis alignment on each spectral characteristic curve to obtain a set of characteristic curves.

[0052] Cluster analysis is performed on the feature curve set to divide the feature curve set into a preset number of pattern categories; for each pattern category, the feature curves belonging to that pattern category are extracted representatively to generate corresponding typical change pattern templates, thereby obtaining multiple change templates.

[0053] Optionally, each change template is parsed, key events are identified, and a set of time windows for key events corresponding to each change template is obtained, including: Numerical differentiation is performed on each variation template to obtain the rate of change sequence and acceleration sequence representing the variation of the variation template with time in the process.

[0054] Local extrema are detected in the rate of change sequence to determine the critical moment when the change in the change template reaches its fastest or slowest point.

[0055] In the changing acceleration sequence, the turning point where the sign changes is detected is used to determine the key moment when the trend of change in the changing template changes.

[0056] The key moments corresponding to local extreme points and key moments corresponding to turning points are taken as key event points; and with each key event point as the center, the corresponding key event time window is formed by expanding in a preset direction, thereby obtaining the set of key event time windows corresponding to each change template.

[0057] Specifically, this application analyzes each variation template in the variation template set corresponding to each process step to identify key events and forms a time window set of key events corresponding to the variation template for subsequent online sampling frequency configuration.

[0058] The first derivative is calculated for each change template to obtain a rate of change sequence, which characterizes how quickly the change template changes over time. The second derivative is calculated for each change template to obtain a change acceleration sequence, which characterizes the increasing or decreasing trend of the rate of change over time, thereby reflecting whether the change process is in an accelerating, decelerating, or trend-reversing state.

[0059] Furthermore, to improve robustness, this application can also smooth the rate of change sequence before calculating the changing acceleration.

[0060] Furthermore, the preset direction includes forward, backward, or both directions. For example, expanding according to the preset direction involves extending forward and backward for a preset duration centered on the critical event point, thereby forming a time window covering the impact range of the critical event. In another embodiment, expanding according to the preset direction may only extend forward or only backward to cover the warning segment before the critical event or the response segment after the critical event. Accordingly, the time outside the critical event time window within a process step is the stable time interval.

[0061] Optionally, if the baseline statistic includes the arithmetic mean and standard deviation, a sample set of the rate of change of the spectral characteristic curves over time for each process series is calculated, and the sample set of the rate of change is statistically analyzed to obtain the baseline statistic, including: Obtain the spectral characteristic curves and their time information of different batches under the same process sequence.

[0062] Based on the spectral feature difference between adjacent sampling points and the sampling time interval, the rate of change of spectral features at each sampling point is calculated to form a sample set of the rate of change under this process sequence.

[0063] Statistical analysis was performed on the sample set of change rates to calculate the arithmetic mean used to characterize the normal degree of change in the process sequence, and the standard deviation used to characterize the normal fluctuation range of the process sequence.

[0064] Specifically, this application first determines the identification information of the target process sequence. Based on this identification information, it filters multiple historical batch data from a historical database and extracts the spectral characteristic curves and time information of each batch within the process sequence. The spectral characteristic curves are a sequence of features extracted from the original reflectance spectrum that change over time. The time information includes the timestamp or relative time within each step corresponding to each sampling point, used to characterize the time interval between adjacent sampling points.

[0065] For each historical batch, the spectral characteristic curves of that batch within the target process sequence are traversed in the order of sampling time, and adjacent sampling points are recorded as a group to calculate the rate of change sample, until all sampling points within that batch are completed, resulting in a set of rate of change samples.

[0066] Furthermore, the cumulative sum is obtained by summing all the rate of change samples in the rate of change sample set; then, the cumulative sum is divided by the total number of samples to obtain the arithmetic mean.

[0067] Further, calculate the deviation between each rate of change sample and the arithmetic mean; square each deviation and sum them up to obtain the sum of squares of deviations; then divide the sum of squares of deviations by the total number of samples to obtain the variance; finally, take the square root of the variance to obtain the standard deviation.

[0068] The baseline sampling frequency is a recommended sampling frequency learned autonomously from historical detection data that meets the screening criteria. Specifically, the effective sampling frequency within the stable time interval is statistically or optimized to obtain a frequency value that can both track normal process changes and reduce redundant data in non-critical stages. This frequency is then stored in the historical knowledge base as the baseline sampling frequency for this step, and can be used as a benchmark for sampling frequency configuration in the online stage.

[0069] Please see Figure 3Optionally, if the benchmark parameters also include benchmark statistics, spectral data sampled by the spectrometer based on the current sampling frequency is acquired, and a real-time adaptation factor characterizing the degree of spectral anomaly is generated based on the spectral data, including: Step 302: Obtain spectral data obtained by the spectrometer from two consecutive samplings based on the current sampling frequency.

[0070] Step 304: Extract the characteristic wavelength values ​​of the two continuous spectral data and their corresponding sampling time intervals, and calculate the instantaneous rate of change based on the characteristic wavelength values ​​and the sampling time intervals.

[0071] Step 306: Based on two continuous spectral data, calculate the correlation between the two in a preset key band to obtain a stability index characterizing the stability of the spectral shape; wherein the preset key band includes the stop band edge band.

[0072] Step 308: Based on the degree of deviation of the instantaneous rate of change from the benchmark statistic and the degree of spectral instability characterized by the stability index, a real-time adaptation factor is generated to simultaneously measure the degree of deviation of the rate of change from the historical normal and the degree of decline in spectral stability.

[0073] Optionally, based on the degree of deviation of the instantaneous rate of change from the benchmark statistic and the degree of spectral instability characterized by the stability index, a real-time adaptation factor is generated, including: The expression for calculating the real-time adaptation factor is:

[0074]

[0075]

[0076] in, For real-time adaptation factors; It is the instantaneous rate of change; and These are the arithmetic mean and standard deviation, respectively. As a stability index; Pearson correlation coefficient; Key band; The characteristic wavelength value; This represents the sampling time interval.

[0077] Optionally, when the benchmark parameters include preset parameters, benchmark sampling frequency, and benchmark statistics, the timer time and time window set corresponding to the current sampling period are matched, and the target sampling frequency is configured according to the matching result, benchmark parameters, and real-time adaptation factor, including: If the timer time corresponding to the current sampling period falls within the time window set, the first frequency is determined based on preset parameters, benchmark statistics, and real-time adaptation factors; otherwise, the second frequency is determined based on preset parameters, benchmark sampling frequency, and real-time adaptation factors, with the first frequency being higher than the second frequency; the first frequency or the second frequency is used as the target sampling frequency.

[0078] Optionally, if the preset parameters include a key frequency and a key gain, the first frequency is determined based on the preset parameters, benchmark statistics, and real-time adaptation factor, and its expression is:

[0079] in, For key frequencies; This is the key gain.

[0080] Optionally, if the preset parameters include a minimum threshold frequency and a stationary period gain, the second frequency is determined based on the preset parameters, a reference sampling frequency, and a real-time adaptation factor, and its expression is:

[0081] in, The reference sampling frequency; The lowest threshold frequency, This represents the gain during the plateau period.

[0082] The above-mentioned method for calculating the spectrometer sampling frequency is based on retrieving the corresponding benchmark parameters and key event time window sets from a preset historical knowledge base for the current process, providing a priori basis for configuring the subsequent sampling frequency. During each sampling cycle in the current process, a real-time adaptation factor characterizing the degree of spectral anomalies is generated based on the collected spectral data, enabling sampling decisions to reflect the immediate changes in the current operating conditions. Furthermore, the timer time and the key event time window sets are matched, and the target sampling frequency is configured by combining the matching results, benchmark parameters, and real-time adaptation factors. The sampling frequency can be adaptively configured based on key events, realizing on-demand allocation of sampling resources. Finally, the target sampling frequency is used for the next sampling cycle, forming a continuous closed-loop update mechanism. This allows for the dynamic generation of the optimal sampling strategy without continuous manual intervention, ultimately achieving a significant improvement in the overall efficiency of data acquisition while ensuring reliable monitoring of the core quality formation process.

[0083] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0084] Based on the same inventive concept, this application also provides a system for calculating the sampling frequency of a spectrometer. This system is applicable to the above-described method for calculating the sampling frequency of a spectrometer. The solution provided by this system is similar to the solution described in the above-described method. Therefore, the specific limitations in one or more device embodiments provided below can be found in the limitations of the method described above, and will not be repeated here.

[0085] Please see Figure 4 In one embodiment, the system for calculating the sampling frequency of the spectrometer is located within a semiconductor epitaxial growth detection device, which includes a spectrometer. The system for calculating the sampling frequency of the spectrometer includes a process interpreter, a real-time sensing adapter, a decision-maker, and an actuator.

[0086] The process interpreter is used to acquire process change signals released in real time by the main controller of the epitaxial equipment. In response to the process change signal, it initializes the timer and determines the current process. Based on the current process, it queries the preset historical knowledge base to obtain the corresponding baseline parameters and time window set of key events.

[0087] The real-time sensing adapter is used to acquire spectral data sampled by the spectrometer based on the current sampling frequency for each sampling cycle during the current process duration, and to generate a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data.

[0088] The decision-maker is used to match the timer time and time window set corresponding to the current sampling period, and configure the target sampling frequency according to the matching result, the reference parameters and the real-time adaptation factor; the target sampling frequency is used as the sampling frequency of the spectrometer in the next sampling period, and a sampling command is generated.

[0089] The actuator is used to generate a timer interrupt signal in response to the sampling command, so as to trigger the spectrometer to perform the acquisition action.

[0090] Optionally, the above-mentioned system for calculating the sampling frequency of the spectrometer also includes a database construction module.

[0091] The database building module is used to build a pre-defined historical knowledge base.

[0092] Optionally, when the benchmark parameters include benchmark statistics and benchmark sampling frequency, the database construction module constructs a preset historical knowledge base in the following ways: acquiring historical testing data; extracting key data from the historical testing data, including process sequences that meet the screening criteria and their corresponding spectral characteristic curves; identifying key events in the key data to obtain a set of time windows and a stable time interval for the key events; calculating a sample set of the rate of change of the spectral characteristic curves over time for each process series, and statistically analyzing the sample set of the rate of change to obtain benchmark statistics; statistically analyzing the historical sampling frequency within the stable time interval to obtain the benchmark sampling frequency; and storing the benchmark sampling frequency, benchmark statistics, and the set of time windows for key events according to a preset data structure to obtain the historical knowledge base.

[0093] Optionally, the database construction module identifies key events in the key data and obtains a set of time windows and a stable time interval for the key events, including: performing time sequence alignment and pattern clustering operations on the key data to obtain multiple change templates corresponding to each process sequence; parsing each change template, identifying the key events within it, and obtaining a set of time windows for the key events corresponding to each change template; the time interval outside the set of time windows is denoted as the stable time interval.

[0094] Optionally, the database construction module performs time-series alignment and pattern clustering operations on key data to obtain multiple variation templates corresponding to each process sequence. This includes: obtaining spectral characteristic curves of different batches under the same process sequence; using a dynamic time warping algorithm to perform nonlinear time axis alignment on each spectral characteristic curve to obtain a set of characteristic curves; performing clustering analysis on the set of characteristic curves to divide the set of characteristic curves into a preset number of pattern categories; and for each pattern category, extracting representative characteristic curves belonging to that pattern category to generate corresponding typical variation pattern templates, thereby obtaining multiple variation templates.

[0095] Optionally, the database construction module parses each change template, identifies key events within them, and obtains a set of time windows for key events corresponding to each change template. This includes: performing numerical differentiation on each change template to obtain a change rate sequence and a change acceleration sequence representing the change template's variation with time within the process; detecting local extreme points in the change rate sequence to determine the critical moments when the change in the change template reaches its fastest or slowest point; detecting inflection points where the signs change in the change acceleration sequence to determine the critical moments when the change trend in the change template reverses; taking the critical moments corresponding to the local extreme points and the critical moments corresponding to the inflection points as key event points; and expanding the corresponding key event time windows in a preset direction around each key event point to obtain a set of time windows for key events corresponding to each change template.

[0096] Optionally, the database construction module, assuming the baseline statistics include the arithmetic mean and standard deviation, calculates a sample set of the rate of change of the spectral characteristic curves over time for each process series, and performs statistical analysis on the rate of change sample set to obtain the baseline statistics. This includes: acquiring the spectral characteristic curves and their time information for different batches under the same process sequence; calculating the rate of change of the spectral characteristics at each sampling point based on the difference in spectral characteristics between adjacent sampling points and the sampling time interval, forming a rate of change sample set for the process sequence; and performing statistical analysis on the rate of change sample set to calculate the arithmetic mean used to characterize the normal degree of change in the process sequence, and the standard deviation used to characterize the normal fluctuation range of the process sequence.

[0097] Optionally, the real-time sensing adapter acquires spectral data sampled by the spectrometer at the current sampling frequency, and generates a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data. This includes: acquiring spectral data obtained by the spectrometer from two consecutive samplings at the current sampling frequency; extracting the characteristic wavelength values ​​of the two continuous spectral data and their corresponding sampling time intervals, and calculating the instantaneous rate of change based on the characteristic wavelength values ​​and the sampling time intervals; calculating the correlation between the two continuous spectral data on preset key bands to obtain a stability index characterizing spectral stability; wherein the preset key bands include stopband edge bands; and generating a real-time adaptation factor based on the degree of deviation of the instantaneous rate of change from the benchmark statistic and the degree of spectral instability characterized by the stability index, so as to simultaneously measure the degree of deviation of the rate of change from the historical normal and the degree of decrease in spectral stability.

[0098] Optionally, the real-time sensing adapter generates a real-time adaptation factor based on the degree of deviation of the instantaneous rate of change from the benchmark statistic and the degree of spectral instability characterized by the stability index, including: the expression for calculating the real-time adaptation factor is:

[0099]

[0100]

[0101] in, For real-time adaptation factors; It is the instantaneous rate of change; and These are the arithmetic mean and standard deviation, respectively. As a stability index; Pearson correlation coefficient; Key band; The characteristic wavelength value; This represents the sampling time interval.

[0102] Optionally, when the benchmark parameters include preset parameters, benchmark sampling frequency, and benchmark statistics, the decision-maker matches the timer time and time window set corresponding to the current sampling period, and configures the target sampling frequency according to the matching result, benchmark parameters, and real-time adaptation factor. This includes: if the timer time corresponding to the current sampling period falls into the time window set, determining a first frequency based on preset parameters, benchmark statistics, and real-time adaptation factor; otherwise, determining a second frequency based on preset parameters, benchmark sampling frequency, and real-time adaptation factor, wherein the first frequency is higher than the second frequency; and using the first frequency or the second frequency as the target sampling frequency.

[0103] Optionally, when the preset parameters include key frequencies and key gains, the decision-maker determines the first frequency based on the preset parameters, benchmark statistics, and real-time adaptation factors, expressed as:

[0104] in, For key frequencies; This is the key gain.

[0105] Optionally, if the preset parameters include a minimum threshold frequency and a stationary period gain, the decision-maker determines the second frequency based on the preset parameters, the reference sampling frequency, and the real-time adaptation factor, as expressed in the following expression:

[0106] in, The reference sampling frequency; The lowest threshold frequency, This represents the gain during the plateau period.

[0107] The aforementioned spectrometer sampling frequency calculation system retrieves corresponding benchmark parameters and key event time window sets from a pre-set historical knowledge base based on the current process, providing a priori basis for subsequent sampling frequency configuration. During each sampling cycle of the current process, a real-time adaptation factor characterizing the degree of spectral anomalies is generated based on the acquired spectral data, enabling sampling decisions to reflect immediate changes in the current operating conditions. Furthermore, the timer time and key event time window sets are matched, and the target sampling frequency is configured by combining the matching results, benchmark parameters, and real-time adaptation factors. This allows for adaptive configuration of the sampling frequency based on key events, achieving on-demand allocation of sampling resources. Finally, the target sampling frequency is used for the next sampling cycle, forming a continuous closed-loop update mechanism. This dynamically generates the optimal sampling strategy without continuous manual intervention, ultimately significantly improving overall data acquisition efficiency while ensuring reliable monitoring of the core quality formation process.

[0108] The modules in the above-mentioned spectrometer sampling frequency calculation system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0109] Optionally, in one embodiment, this application also provides a semiconductor epitaxial growth detection device, including: a system for calculating the spectrometer sampling frequency and a spectrometer provided in the above embodiments.

[0110] Specifically, the structure and working principle of the spectrometer sampling frequency calculation system in this embodiment are the same as those in the above embodiments, and will not be repeated here to save space.

[0111] The spectrometer is used to perform data acquisition by using the timer interrupt signal output by the calculation system based on the spectrometer's sampling frequency.

[0112] Using the above scheme, the semiconductor epitaxial growth detection device calculates the real-time adaptation factor from the real-time acquired spectral data in each sampling cycle during the process, and dynamically determines the sampling frequency of the next sampling cycle by combining the reference parameters of the current step and the time window of key events. This enables the spectrometer to automatically increase the sampling density during the key event stage to fully capture rapidly changing details, and automatically decrease the sampling density during the stable stage to reduce redundant data. Ultimately, this significantly improves the overall efficiency of data acquisition while ensuring reliable monitoring of the core quality formation process.

[0113] In one feasible embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements the method for calculating the sampling frequency of the spectrometer. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0114] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0115] In one feasible embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method steps in the above-described method for calculating the sampling frequency of a spectrometer.

[0116] In one feasible embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method steps in the above-described method for calculating the sampling frequency of a spectrometer.

[0117] In one feasible embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the method steps in the above-described method for calculating the sampling frequency of a spectrometer.

[0118] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0119] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for calculating the sampling frequency of a spectrometer, characterized in that, The method is applicable to a semiconductor epitaxial growth detection apparatus, wherein the semiconductor epitaxial growth detection apparatus includes a spectrometer; the method includes: In response to a process change signal, initialize the timer and determine the current process; Based on the current process, the corresponding baseline parameters and time window set of key events are obtained by querying the preset historical knowledge base; For each sampling period during the current process, acquire the spectral data sampled by the spectrometer based on the current sampling frequency, and generate a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data; The timer time corresponding to the current sampling period is matched with the set of time windows, and the target sampling frequency is configured according to the matching result, the benchmark parameter and the real-time adaptation factor. The target sampling frequency is used as the sampling frequency for the next sampling period of the spectrometer.

2. The method according to claim 1, characterized in that, The benchmark parameters include benchmark statistics and benchmark sampling frequency; Methods for pre-setting a historical knowledge base include: Acquire historical testing data, which includes at least one of the following: raw reflectance spectral time-series data, process time-series log, wafer endpoint testing data, and device electrical test data; Key data is extracted from the historical detection data, including process sequences that meet the screening criteria and their corresponding spectral characteristic curves; Identify key events in the key data to obtain the time window set and stable time interval of the key events; Calculate a sample set of the rate of change of the spectral characteristic curves over time for each process series, and perform statistics on the sample set of the rate of change to obtain a baseline statistic; The baseline sampling frequency is obtained by statistically analyzing the historical sampling frequencies within the aforementioned stable time interval. The baseline sampling frequency, baseline statistics, and time window set of key events are stored according to a preset data structure to obtain a historical knowledge base.

3. The method according to claim 2, characterized in that, The process of identifying key events in the key data and obtaining a set of time windows and stable time intervals for the key events includes: Perform time-series alignment and pattern clustering operations on the key data to obtain multiple variation templates corresponding to each process sequence; Each change template is analyzed to identify key events, resulting in a time window set for the key events corresponding to each change template; the time interval outside the time window set is denoted as the stationary time interval.

4. The method according to claim 3, characterized in that, The step of performing time-series alignment and pattern clustering operations on the key data yields multiple variation templates corresponding to each process sequence, including: Obtain the spectral characteristic curves of different batches under the same process sequence; A dynamic time warping algorithm is used to perform nonlinear time axis alignment on each of the spectral feature curves to obtain a set of feature curves. Cluster analysis is performed on the feature curve set to divide the feature curve set into a preset number of pattern categories; for each pattern category, the feature curves belonging to that pattern category are extracted representatively to generate corresponding typical change pattern templates, thereby obtaining multiple change templates.

5. The method according to claim 3, characterized in that, The process involves parsing each change template, identifying key events within them, and obtaining a set of time windows for the key events corresponding to each change template, including: Numerical differentiation is performed on each of the aforementioned change templates to obtain a sequence of the rate of change and a sequence of the acceleration of change that characterize the change template as a function of time within the process. Local extreme points are detected in the rate of change sequence to determine the critical moment when the change in the change template reaches its fastest or slowest point. In the changing acceleration sequence, the turning point where the sign changes is detected is used to determine the key moment when the changing trend in the changing template changes; The key moments corresponding to the local extreme points and the key moments corresponding to the turning points are taken as key event points; and with each key event point as the center, the corresponding key event time window is formed by expanding in a preset direction, thereby obtaining the time window set of key events corresponding to each change template.

6. The method according to claim 2, characterized in that, The baseline statistics include the arithmetic mean and standard deviation; The calculation of the rate of change of the spectral characteristic curve over time for each process series, and the statistical analysis of the rate of change sample set to obtain a baseline statistic, includes: Obtain the spectral characteristic curves and their time information of different batches under the same process sequence; Based on the spectral feature difference between adjacent sampling points and the sampling time interval, the rate of change of spectral features at each sampling point is calculated to form a sample set of the rate of change under the process sequence. Statistical analysis is performed on the sample set of change rates to calculate the arithmetic mean that characterizes the normal degree of change in the process sequence and the standard deviation that characterizes the normal fluctuation range of the process sequence.

7. The method according to claim 1, characterized in that, The benchmark parameters include benchmark statistics; The step of acquiring spectral data sampled by the spectrometer based on the current sampling frequency, and generating a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data, includes: Acquire the spectral data obtained by the spectrometer from two consecutive samplings based on the current sampling frequency; Extract the characteristic wavelength values ​​and their corresponding sampling time intervals from two continuous spectral data, and calculate the instantaneous rate of change based on the characteristic wavelength values ​​and the sampling time interval; Based on two continuous spectral data, the correlation between the two is calculated on a preset key band to obtain a stability index characterizing the stability of the spectral shape; wherein the preset key band includes the stop band edge band. Based on the degree of deviation of the instantaneous rate of change from the benchmark statistic and the degree of spectral instability characterized by the stability index, the real-time adaptation factor is generated to simultaneously measure the degree of deviation of the rate of change from the historical normal and the degree of decline in spectral stability.

8. The method according to claim 7, characterized in that, The real-time adaptation factor is generated based on the deviation of the instantaneous rate of change from the benchmark statistic and the degree of spectral instability characterized by the stability index, including: The expression for calculating the real-time adaptation factor is: in, For real-time adaptation factors; It is the instantaneous rate of change; and These are the arithmetic mean and standard deviation, respectively. As a stability index; Pearson correlation coefficient; Key band; The characteristic wavelength value; This represents the sampling time interval.

9. The method according to claim 1, characterized in that, The benchmark parameters include preset parameters, benchmark sampling frequency, and benchmark statistics; The step of matching the timer time corresponding to the current sampling period with the set of time windows, and configuring the target sampling frequency according to the matching result, the benchmark parameter, and the real-time adaptation factor, includes: If the timer time corresponding to the current sampling period falls into the time window set, a first frequency is determined based on the preset parameters, the benchmark statistics, and the real-time adaptation factor; otherwise, a second frequency is determined based on the preset parameters, the benchmark sampling frequency, and the real-time adaptation factor, wherein the first frequency is higher than the second frequency. The first frequency or the second frequency is used as the target sampling frequency.

10. The method according to claim 9, characterized in that, The preset parameters include key frequencies and key gains; The first frequency is determined based on preset parameters, the benchmark statistics, and the real-time adaptation factor, and its expression is as follows: in, For key frequencies; This is the key gain.

11. The method according to claim 9, characterized in that, The preset parameters include the minimum threshold frequency and the stationary period gain; The second frequency is determined based on the preset parameters, the reference sampling frequency, and the real-time adaptation factor, and its expression is as follows: in, The reference sampling frequency; The lowest threshold frequency, This represents the gain during the plateau period.

12. A system for calculating the sampling frequency of a spectrometer, characterized in that, The system is installed within a semiconductor epitaxial growth detection device, which includes a spectrometer; the system includes: The process interpreter is used to acquire process change signals issued in real time by the main controller of the epitaxial equipment, and in response to the process change signals, initialize the timer and determine the current process; based on the current process, it queries the preset historical knowledge base to obtain the corresponding baseline parameters and time window set of key events; A real-time sensing adapter is used to acquire spectral data sampled by the spectrometer based on the current sampling frequency for each sampling cycle during the current process duration, and to generate a real-time adaptation factor characterizing the degree of spectral anomaly based on the spectral data. The decision-maker is used to match the timer time corresponding to the current sampling period with the set of time windows, and configure the target sampling frequency according to the matching result, the reference parameters and the real-time adaptation factor; the target sampling frequency is used as the sampling frequency of the spectrometer in the next sampling period, and a sampling instruction is generated. An actuator is used to generate a timer interrupt signal in response to the sampling command, so as to trigger the spectrometer to perform a data acquisition action.

13. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-11.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-11.

Images