Remote plasma sheath characteristics regulation and etching uniformity optimization method

Abstract

The present application relates to the technical field of plasma etching, and discloses a remote plasma sheath characteristic regulation and etching uniformity optimization method.The method steps are as follows: collecting historical stable parameters to establish a benchmark parameter library, collecting real-time monitoring sets and determining the regulation and optimization period according to the data volume;collecting all plasma sheath parameters in the period and obtaining the average value as the regulation and optimization threshold, and combining the benchmark parameter library to analyze and judge whether there is suspected non-uniform etching and mark;when there is suspected non-uniform etching, the sheath thickness and distribution characteristics are obtained to determine whether it is non-uniform etching, and if it is non-uniform etching, the initial regulation coefficient is determined according to the frequency;extracting non-uniform etching and identifying the sheath potential, ion energy and boundary profile, classifying based on a classification algorithm to determine whether there is a fluctuating sheath;if there is, the initial regulation coefficient is corrected according to the fluctuating sheath characteristics to correct the initial regulation coefficient, and the next regulation and optimization period is adjusted.The method can dynamically adapt to the sheath change.

Description

Technical Field

[0001] This invention relates to the field of plasma etching technology, specifically to a method for remotely controlling plasma sheath characteristics and optimizing etching uniformity. Background Technology

[0002] In semiconductor manufacturing and micro / nanoelectronic device fabrication, plasma etching technology, with its high material selectivity and precision processing capabilities, has become one of the core processes for pattern transfer. The plasma sheath, as a crucial transition region between the plasma and the substrate surface to be etched, directly influences the energy, angle, and density distribution of ion bombardment, thus playing a decisive role in the consistency and uniformity of the etching profile. Inhomogeneous sheath thickness, fluctuations in potential distribution, and deviations in ion energy can all lead to localized over-etching or under-etching in the etched area, ultimately affecting the electrical performance of the device and the overall yield.

[0003] In existing technologies, monitoring of the plasma sheath mostly employs offline detection or fixed-period sampling, making it difficult to capture dynamic changes in sheath characteristics in real time. While some solutions attempt to introduce real-time monitoring, they lack effective correlation with historical stable parameters, failing to quickly define the boundary between normal and abnormal sheath states, resulting in delayed early warnings for non-uniform etching. Regarding the identification of non-uniform etching, traditional methods often rely on threshold judgments for a single parameter, such as judging solely by deviations in sheath thickness, ignoring the influence of sheath distribution characteristics. This easily leads to misjudging normal parameter fluctuations as non-uniform etching, or failing to promptly identify actual non-uniform etching.

[0004] Existing control strategies often employ fixed control coefficients, failing to consider the frequency variations of non-uniform etching. When non-uniform etching occurs frequently, fixed coefficients are insufficient to meet practical control requirements; conversely, for occasional non-uniform etching, applying strong control may introduce new disturbances, compromising the stability of the sheath. Furthermore, plasma sheaths often exhibit fluctuations due to factors such as gas source stability, RF power fluctuations, and cavity pressure changes, exhibiting characteristics significantly different from static non-uniform etching. Current technologies have limited ability to identify fluctuating sheaths, often treating them the same as conventional non-uniform etching, leading to control directions deviating from actual needs and further exacerbating etching non-uniformity.

[0005] Traditional methods for setting the control period often use a fixed period, which cannot be dynamically adjusted according to the drastic changes in sheath characteristics. When the sheath fluctuates violently, a fixed period may lead to monitoring lag and missing the optimal control opportunity; while when the sheath is in a stable state, an excessively short period will increase unnecessary computational load and reduce etching efficiency. These problems together result in insufficient control precision of sheath characteristics during plasma etching, and the etching uniformity is difficult to meet the requirements of high-precision manufacturing, thus restricting the further development of related technologies. Summary of the Invention

[0006] The purpose of this invention is to provide a method for remotely controlling the properties of a plasma sheath and optimizing the etching uniformity, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a method for remote plasma sheath characteristic modulation and etching uniformity optimization, the method comprising:

[0008] Historical stable parameters are collected to establish a benchmark parameter library, and a real-time monitoring set is collected. The control and optimization cycle is determined based on the amount of data in the real-time monitoring set.

[0009] All plasma sheath parameters are collected within the control and optimization cycle, and the average value of the sheath parameters is obtained. The average value of the sheath parameters is used as the control and optimization threshold. The plasma sheath parameters are analyzed according to the control and optimization threshold and the benchmark parameter library to determine whether there is any suspected non-uniform etching and mark it.

[0010] When there is a suspected non-uniform etching, the sheath thickness and distribution characteristics of each suspected non-uniform etching are obtained. Based on the sheath thickness and distribution characteristics, it is determined whether the suspected non-uniform etching is non-uniform etching. When the suspected non-uniform etching is determined to be non-uniform etching, the initial control coefficient is determined based on the non-uniform etching frequency.

[0011] Extract the non-uniform etching and identify the sheath potential, ion energy and boundary contour. Use the boundary contour as a feature parameter. Classify all the non-uniform etchings based on a classification algorithm. Determine whether there is a fluctuating sheath based on the classification results.

[0012] When a wavering sheath is determined to exist, a correction coefficient is determined based on the characteristics of the wavering sheath to correct the initial control coefficient, and the next control optimization cycle is adjusted.

[0013] Preferably, when determining the control optimization cycle based on the data volume of the real-time monitoring set, the following steps are included:

[0014] The data volume is compared with a first preset data volume and a second preset data volume, and the adjustment and optimization cycle is determined based on the comparison results; the first preset data volume is less than the second preset data volume.

[0015] When the amount of data is less than or equal to a first preset amount of data, the adjustment and optimization cycle is determined to be a first cycle; when the amount of data is greater than the first preset amount of data and less than or equal to a second preset amount of data, the adjustment and optimization cycle is determined to be a second cycle; when the amount of data is greater than the second preset amount of data, the adjustment and optimization cycle is determined to be a third cycle; the first cycle is greater than the second cycle, and the second cycle is greater than the third cycle.

[0016] Preferably, when analyzing the plasma sheath parameters based on the control optimization threshold and the benchmark parameter library to determine and mark the presence of suspected non-uniform etching, the process includes:

[0017] The sheath potential of all plasma sheath parameters is compared with the reference parameter library, and the sheath density of each plasma sheath parameter is compared with the average value of the sheath parameter. Based on the comparison results, it is determined whether there is any suspected non-uniform etching and it is marked.

[0018] When the sheath density of the plasma sheath parameter is greater than a preset multiple of the average value of the sheath parameter, the plasma sheath parameter is determined to be a suspected non-uniform etching and is marked.

[0019] When the sheath potential of the plasma sheath parameter is not in the reference parameter library, the plasma sheath parameter is determined to be a suspected non-uniform etching and is marked.

[0020] Preferably, when determining whether the suspected non-uniform etching is non-uniform etching based on the sheath thickness and distribution characteristics, the method includes:

[0021] When the thickness of the sheath layer suspected of being non-uniformly etched is different from the conventional thickness in the reference parameter library, the suspected non-uniform etching is determined to be non-uniform etching.

[0022] The distribution characteristics include the distribution pattern morphology. When the distribution pattern morphology of the suspected non-uniform etching shows discrete distribution peaks, the suspected non-uniform etching is determined to be non-uniform etching.

[0023] Preferably, when determining the initial control coefficient based on the non-uniform etching frequency, the following steps are included:

[0024] The difference between the non-uniform etching frequency and the conventional frequency in the reference parameter library is statistically analyzed. Combined with the deviation of the sheath thickness of each non-uniform etching layer from the thickness closest to the reference parameter library, the initial control coefficient is calculated.

[0025] Preferably, when classifying all the non-uniform etchings based on a classification algorithm, and determining whether a fluctuating sheath layer exists based on the classification results, the following steps are included:

[0026] The boundary profile includes the number of valleys in the profile curve and the gradient rate of the profile curve.

[0027] The initial classification intervals were determined using a distance analysis plot, and the classification cardinality was set to 2.

[0028] Each non-uniform etching feature parameter is treated as an object. All objects are traversed, and the object with a number of objects in the classification interval that is greater than or equal to the classification cardinality is identified as the core object.

[0029] Starting with each core object, examine the objects within its category range; if an object in the range is a core object, continue expanding the category; if an object in the range is a boundary object, add it to the current category; if an object is not within the category range of any core object and cannot form a category with other objects, it is marked as an independent object.

[0030] When there exists a class that includes at least two non-uniform etchings and the sheath potential of the non-uniform etchings is at least two different potentials, the non-uniform etchings in that class are determined to be wavering sheaths.

[0031] Preferably, when correcting the initial control coefficient based on the characteristics of the wave sheath, the correction coefficient is determined by:

[0032] The characteristics of the wave sheath include the total number of wave sheaths and the ion energy of each wave sheath.

[0033] The characteristics of the wave sheath are compared with historical correction schemes, and correction coefficients are determined based on the comparison results to correct the initial control coefficients. The historical correction schemes include the characteristics of several historical wave sheaths and several historical correction coefficients, and each characteristic of a historical wave sheath corresponds to a historical correction coefficient.

[0034] Calculate the similarity between the characteristics of the wave sheath and the characteristics of each historical wave sheath;

[0035] When there is data in the historical wave sheath features that has a similarity greater than a similarity threshold with the features of the wave sheath, the historical correction coefficient is determined based on the historical wave sheath features corresponding to the maximum similarity and is used to correct the initial control coefficient.

[0036] When the similarity between the characteristics of the historical wave sheath and the characteristics of the wave sheath is less than or equal to the similarity threshold, the initial control coefficient is corrected based on the total number of wave sheaths.

[0037] Preferably, when correcting the initial control coefficient based on the total number of wave sheaths, the correction includes:

[0038] The correction coefficient is proportional to the total number of wave sheaths, and the value of the correction coefficient is greater than 1 and less than or equal to 1.5.

[0039] Preferably, when adjusting the next regulation and optimization cycle, the following are included:

[0040] Obtain the corrected control coefficient, determine the periodic correction coefficient based on the control coefficient, and adjust the next control optimization cycle accordingly. The periodic correction coefficient is inversely proportional to the corrected control coefficient, and the value range of the periodic correction coefficient is greater than or equal to 0.5 and less than 1.

[0041] Preferably, the real-time monitoring data collection includes:

[0042] The potential, density, and thickness parameters of the plasma sheath are collected synchronously by multi-channel sensors, and the parameters are arranged in a time series to form a real-time monitoring set.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] This remote plasma sheath characteristic control and etching uniformity optimization method establishes a benchmark parameter library by collecting historical stable parameters and determines the control optimization cycle by combining it with real-time monitoring data. This allows the cycle setting to no longer be limited to a fixed value but to adapt to the actual changing trends of sheath characteristics. When sheath parameter fluctuations are small, the cycle can be appropriately extended to reduce unnecessary monitoring and control operations; when parameter changes are drastic, the cycle automatically shortens to ensure timely detection of abnormal states. This dynamic cycle adjustment method avoids monitoring lag under a fixed cycle and reduces resource consumption caused by ineffective operations.

[0045] In the identification of suspected non-uniform etching, this method uses the average value of sheath parameters within the optimization cycle as the optimization threshold and compares it with a benchmark parameter library. This ensures that the suspected marker no longer relies on the absolute value of a single parameter, but rather combines the parameter distribution characteristics under historical stable conditions, thereby reducing misjudgments caused by single parameter fluctuations. Based on this, by obtaining the sheath thickness and distribution characteristics of suspected non-uniform etching, it further determines whether it is indeed non-uniform etching, distinguishing between normal parameter fluctuations and actual etching anomalies, making the identification of non-uniform etching more targeted.

[0046] For the control of non-uniform etching, this method determines the initial control coefficient based on the frequency of non-uniform etching, ensuring that the control intensity matches the frequency of etching anomalies. For frequently occurring non-uniform etching, the initial control coefficient can be adjusted accordingly to enhance the control effect; for occasional anomalies, the control intensity is relatively mild to avoid over-controlling and causing new fluctuations. Furthermore, by extracting the sheath potential, ion energy, and boundary contour of non-uniform etching and classifying them using a classification algorithm, the differences between fluctuating sheaths and conventional non-uniform etching can be effectively identified, providing direction for subsequent precise control.

[0047] When a fluctuating sheath is detected, a correction coefficient is determined based on its characteristics to adjust the initial control coefficient. This ensures that the control strategy not only considers the frequency of non-uniform etching but also adapts to the dynamic characteristics of the fluctuating sheath. This correction mechanism compensates for the shortcomings of traditional fixed control coefficients, allowing the control measures to better align with the actual changes in the sheath. Consequently, it more effectively addresses various sheath anomalies during etching, reduces etching deviations caused by unstable sheath characteristics, improves the contour consistency of the etched area, and meets the demands of high-precision manufacturing scenarios. Attached Figure Description

[0048] Figure 1 This is a schematic diagram illustrating the working principle of the remote plasma sheath characteristic regulation and etching uniformity optimization method described in this invention.

[0049] Figure 2 A flowchart for determining the control optimization cycle based on the amount of data in the real-time monitoring set;

[0050] Figure 3 A flowchart for identifying and marking suspected non-uniform etching;

[0051] Figure 4 A flowchart for adjusting the next regulatory optimization cycle. Detailed Implementation

[0052] 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, and 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.

[0053] Please see Figures 1-4 This invention provides a method for remotely controlling plasma sheath properties and optimizing etching uniformity, the method comprising:

[0054] Collect historical stable parameters to establish a benchmark parameter library, collect real-time monitoring sets, and determine the control and optimization cycle based on the amount of data in the real-time monitoring sets.

[0055] All plasma sheath parameters were collected during the control and optimization cycle, and the average value of the sheath parameters was obtained. The average value of the sheath parameters was used as the control and optimization threshold. The plasma sheath parameters were analyzed based on the control and optimization threshold and the benchmark parameter library to determine whether there was any suspected non-uniform etching and to mark it.

[0056] When there is suspected non-uniform etching, the sheath thickness and distribution characteristics of each suspected non-uniform etching are obtained. Based on the sheath thickness and distribution characteristics, it is determined whether the suspected non-uniform etching is non-uniform etching. When the suspected non-uniform etching is determined to be non-uniform etching, the initial control coefficient is determined based on the non-uniform etching frequency.

[0057] Non-uniform etching is extracted and sheath potential, ion energy and boundary contour are identified. The boundary contour is used as a feature parameter. All non-uniform etching is classified based on a classification algorithm. The existence of a wavering sheath is determined based on the classification results.

[0058] When a wavering sheath is determined, the initial control coefficient is corrected based on the characteristics of the wavering sheath, and the next control optimization cycle is adjusted.

[0059] Example 1:

[0060] When determining the control optimization cycle based on the data volume of the real-time monitoring set, the data volume is compared with a first preset data volume and a second preset data volume, respectively. The control optimization cycle is determined based on the comparison results. The first preset data volume is less than the second preset data volume. When the data volume is less than or equal to the first preset data volume, the control optimization cycle is determined as the first cycle. When the data volume is greater than the first preset data volume and less than or equal to the second preset data volume, the control optimization cycle is determined as the second cycle. When the data volume is greater than the second preset data volume, the control optimization cycle is determined as the third cycle. The first cycle is greater than the second cycle, and the second cycle is greater than the third cycle. When collecting the real-time monitoring set, the potential, density, and thickness parameters of the plasma sheath are simultaneously collected by multi-channel sensors, and the parameters are arranged in a time series to form the real-time monitoring set.

[0061] The multi-channel sensor array needs to cover different regions of the plasma sheath to ensure simultaneous acquisition of potential, density, and thickness parameters at different spatial locations. The sensor sampling frequency is set based on the rate of change of sheath parameters in the actual application scenario to ensure that the acquired data accurately reflects the dynamic changes of the sheath. The time series is arranged with the acquisition time as the axis, categorizing parameters at different locations at the same moment to form a spatiotemporally combined real-time monitoring set, allowing subsequent data analysis to consider both temporal and spatial characteristics.

[0062] The setting of the first and second preset data amounts needs to be combined with the stability requirements of the plasma etching process and the computing power of the data processing system. In scenarios where the etching process is relatively stable and the sheath parameters change slowly, the first and second preset data amounts can be set to relatively large values ​​to reduce the frequency of adjustment and optimization. However, in the early stages of the etching process or during the process parameter adjustment phase, the sheath parameters may change frequently. In this case, the first and second preset data amounts can be set to smaller values ​​to improve the timeliness of adjustment and optimization.

[0063] The specific durations of the first, second, and third cycles need to be determined based on actual process requirements. The first cycle, being the longest adjustment and optimization cycle, is suitable for situations with small amounts of data. In this case, the sheath parameters change relatively smoothly, and the longer cycle reduces the number of data processing and adjustment operations, lowering system resource consumption. The second cycle, with a duration between the first and third cycles, is suitable for scenarios with moderate data volumes, ensuring timely adjustment while avoiding excessively frequent operations. The third cycle, being the shortest cycle, is suitable for situations with large amounts of data. In this case, the sheath parameters may be changing rapidly, and the shorter cycle allows for timely capture and optimization of these changes to maintain the stability of the etching process.

[0064] In practical applications, when the amount of data in the real-time monitoring set changes dynamically, the control and optimization cycle will switch accordingly between the first, second, and third cycles. For example, when the etching process transitions from a stable to an unstable state, the amount of data monitored in real-time may gradually increase from less than a first preset amount to more than a second preset amount. In this case, the control and optimization cycle will switch from the first cycle to the second and then the third cycle to adapt to the changes in data volume and the dynamic characteristics of the sheath parameters. Conversely, when the etching process gradually stabilizes from an unstable state, the amount of data decreases, and the control and optimization cycle will switch from the third cycle to the second and then the first cycle.

[0065] By dynamically adjusting the control and optimization cycle based on data volume, the entire control and optimization process can be matched with the actual changes in plasma sheath parameters. This allows for timely control through short cycles when parameters change drastically, and reduces unnecessary operations through long cycles when parameters are stable, thereby achieving a balance between reasonable resource allocation and control efficiency.

[0066] Example 2:

[0067] Based on the control optimization threshold and the reference parameter library, the plasma sheath parameters are analyzed to determine whether there is any suspected non-uniform etching and to mark it. The sheath potential of all plasma sheath parameters is compared with the reference parameter library, and the sheath density of each plasma sheath parameter is compared with the average value of the sheath parameter. Based on the comparison results, it is determined whether there is any suspected non-uniform etching and to mark it. When the sheath density of a plasma sheath parameter is greater than a preset multiple of the average value of the sheath parameter, the plasma sheath parameter is determined to be suspected non-uniform etching and to be marked. When the sheath potential of a plasma sheath parameter is not in the reference parameter library, the plasma sheath parameter is determined to be suspected non-uniform etching and to be marked.

[0068] The benchmark parameter library is constructed based on historically stable plasma etching processes, containing sheath potential data under different process conditions. This data, accumulated over a long period, covers the potential range that may occur during normal etching. During construction, historical data was filtered to remove outliers and data from unstable states to ensure the reliability of the benchmark parameter library. The sheath potentials in the benchmark parameter library are stored in interval format, with each interval corresponding to a specific combination of process parameters, facilitating rapid location during comparison.

[0069] The calculation of the average sheath parameter must cover the sheath density of all plasma sheath parameters collected within the control and optimization cycle. The calculation method is the sum of all sheath densities divided by the number of parameters. No data is excluded during the calculation to fully reflect the overall level of sheath density within that cycle. The average sheath parameter serves as a control and optimization threshold, reflecting the concentration trend of sheath density within the current cycle and providing a reference standard for judging whether the density of a single sheath is abnormal.

[0070] The determination of the preset multiplier needs to comprehensively consider the normal fluctuation range of the plasma sheath density, which is obtained by analyzing the standard deviation and range of historical stable data. The value of the preset multiplier should cover the upper limit of normal fluctuation. When a certain sheath density exceeds the product of the multiplier and the average value of the sheath parameters, it indicates that it deviates from the overall level and may be a characteristic of non-uniform etching.

[0071] During the comparison operation, the sheath potential of each plasma sheath parameter is first matched with the potential range in the reference parameter library. If a sheath potential falls within any range, it is determined to be within the normal range; if it cannot fall within any range, it is directly marked as suspected non-uniform etching. At the same time, the sheath density of each plasma sheath parameter is compared with the average value of the sheath parameter, and the ratio between the two is calculated. When the ratio is greater than a preset multiple, it is marked as suspected non-uniform etching.

[0072] The marking process employs digital identification, assigning a unique identifier to each suspected non-uniform etching. This identifier includes the time and location of parameter acquisition, as well as the reason for the suspected non-uniformity (abnormal sheath potential or abnormal sheath density). The identification information is stored along with the corresponding sheath parameters for easy traceability and further analysis.

[0073] In practice, the two comparison methods are performed independently and do not interfere with each other. The same plasma sheath parameter may simultaneously meet the conditions that the sheath potential is not in the reference parameter library and the sheath density is greater than a preset multiple of the average value of the sheath parameter. In this case, it is still only marked as a suspected non-uniform etching, but the identification information will record both reasons at the same time.

[0074] This dual comparison method allows for the screening of potential non-uniform etching patterns from both potential and density dimensions. The sheath potential reflects the electric field characteristics of the plasma sheath, and anomalies in this potential often indicate a disordered electric field distribution. The sheath density is related to ion concentration, and abnormal deviations in this density can lead to non-uniform etching rates. Combining these two methods improves the comprehensiveness of identifying suspected non-uniform etching, reduces the probability of missed or false positives, and provides reliable screening results for subsequent accurate assessment.

[0075] In different control and optimization cycles, the average value of the sheath parameters changes with the data within the cycle. The preset multiple remains unchanged, but the threshold used for comparison (the product of the preset multiple and the average value of the sheath parameters) is adjusted accordingly to adapt to the overall level in different cycles. This dynamically adjusted threshold allows the judgment criteria to match the actual situation of the current cycle, further improving the accuracy of suspected non-uniform etching marks.

[0076] Example 3:

[0077] When determining whether suspected non-uniform etching is indeed non-uniform based on sheath thickness and distribution characteristics, if the sheath thickness of the suspected non-uniform etching differs from the conventional thickness in the reference parameter library, it is classified as non-uniform etching. Distribution characteristics include the distribution pattern morphology; if the distribution pattern of the suspected non-uniform etching exhibits discrete distribution peaks, it is classified as non-uniform etching. When determining the initial control coefficient based on the frequency of non-uniform etching, the difference between the frequency of non-uniform etching and the conventional frequency in the reference parameter library is statistically analyzed. This, combined with the deviation of the sheath thickness of each non-uniform etching from the closest thickness in the reference parameter library, is used to comprehensively calculate the initial control coefficient.

[0078] The standard thickness in the benchmark parameter library is stored as a range. This range is determined by collecting a large amount of sheath thickness data from historical stable etching processes, covering the normal thickness range under different process conditions. Determining the standard thickness range requires excluding outliers from historical data, retaining only thickness values ​​that are stable. The mean and standard deviation of these thickness values ​​are calculated, and the range of the mean plus or minus three times the standard deviation is taken as the standard thickness range. When the sheath thickness of suspected non-uniform etching exceeds this range, it is determined to be different from the standard thickness.

[0079] The distribution pattern morphology presents the spatial distribution of sheath parameters as an image, with the horizontal axis representing spatial location and the vertical axis representing parameter values. Distribution patterns corresponding to normal etching typically exhibit a continuous and smooth curve, with concentrated and uniformly distributed peaks. Discrete distribution peaks refer to multiple independent, clearly spaced peaks in the pattern, with distinct troughs between these peaks, indicating significant differences in the spatial distribution of sheath parameters, consistent with the characteristics of non-uniform etching.

[0080] Non-uniform etching frequency refers to the number of times non-uniform etching is identified within the current adjustment and optimization cycle, while the normal frequency is the average number of non-uniform etchings within the same cycle length in the benchmark parameter library. When calculating the difference between the two, the difference between the current frequency and the normal frequency is calculated. A positive difference indicates that the current frequency of non-uniform etching is higher than the normal level.

[0081] The deviation between the sheath thickness of each non-uniform etching and the closest thickness in the reference parameter library is calculated as follows: subtract the closest normal thickness value from the reference parameter library from the non-uniform etching sheath thickness, and take the absolute value as the deviation value. The larger the deviation value, the greater the degree of deviation between the non-uniform etching and the normal state.

[0082] The formula for calculating the initial control coefficient is:

[0083] Where K1 represents the initial control coefficient, a and b are weighting coefficients, a+b=1, F1 is the non-uniform etching frequency in the current control optimization cycle, F0 is the normal frequency in the benchmark parameter library, Dᵢ is the deviation between the sheath thickness of the i-th non-uniform etching and the thickness closest to the one in the benchmark parameter library, n is the total number of non-uniform etchings in the current control optimization cycle, and ΣDᵢ is the sum of all deviation values.

[0084] The values ​​of the weighting coefficients a and b are determined based on the degree of influence of frequency difference and thickness deviation on etching uniformity in the actual process. If the influence of frequency difference is greater, then a takes a larger value, and vice versa. Represents the relative value of frequency difference.

[0085] The average deviation value is represented by the two values, which are combined through a weighting coefficient to form the initial control coefficient.

[0086] In practice, when a suspected non-uniform etching simultaneously satisfies both abnormal sheath thickness and abnormal distribution pattern morphology, it is still classified as a single non-uniform etching and not counted repeatedly. When calculating frequency, duplicate classifications must be excluded to ensure that each non-uniform etching is counted only once. When calculating the deviation value, if multiple close conventional thickness values ​​exist in the benchmark parameter library, the value with the smallest difference from the non-uniform etching sheath thickness is selected as the closest thickness.

[0087] The calculation results of the initial control coefficient need to be limited to a range to ensure that they are within a reasonable range, avoiding excessively large or small control coefficients due to extreme values. When the calculation result exceeds the upper limit, the upper limit value is taken as the initial control coefficient; when it is below the lower limit, the lower limit value is taken.

[0088] Example 4:

[0089] All non-uniform etchings are classified using a classification algorithm. When determining the presence of a fluctuating sheath based on the classification results, the boundary contour includes the number of valleys in the contour curve and the gradient rate of change of the contour curve. An initial classification interval is determined using a distance analysis map, with a classification cardinality of 2. Each non-uniform etching feature parameter is treated as an object, and all objects are traversed to identify those with a number of objects within their classification interval greater than or equal to the classification cardinality as core objects. Starting from each core object, objects within its classification interval are examined. If an object within the interval is a core object, the category is expanded. If an object within the interval is a boundary object, it is added to the current category. If an object is not within the classification interval of any core object and cannot form a category with other objects, it is marked as an independent object. When a category contains at least two non-uniform etchings with at least two different sheath potentials, the non-uniform etchings in that category are determined to be fluctuating sheaths.

[0090] The number of troughs in the contour curve of the sheath boundary refers to the number of troughs appearing on the contour curve. This is determined by differentiating the contour curve and finding the points where the slope changes from negative to positive; these are the trough points. The number of these points is then counted to obtain the total number of troughs. The gradient change rate of the contour curve is calculated by the change in slope between two adjacent points. This is obtained by calculating the ratio of the difference in slope between two adjacent points to the distance between them. This value reflects the degree of drastic change in the contour curve.

[0091] The distance analysis plot uses feature parameters as coordinate axes, drawing all non-uniformly etched feature parameters as points on the graph. The Euclidean distance between any two points is calculated, and the initial classification interval is determined based on the distance distribution. The size of the initial classification interval is determined by analyzing the distance distribution characteristics, ensuring that the interval contains a sufficient number of objects while avoiding excessively large intervals that would result in objects of different categories being included in the same interval.

[0092] Setting the classification cardinality to 2 means that a valid category must contain at least two objects. This setting avoids overly detailed classifications caused by a single object forming a category, ensuring the representativeness of the categories. During the traversal of all objects, the number of other objects within the initial classification range for each object is checked one by one. When the number is greater than or equal to 2, the object is identified as a core object.

[0093] When expanding categories starting from a core object, all objects within the core object's category range are first added to the initial category. Then, it is checked whether these objects are core objects. If so, objects within their category ranges are added to the new category, and this process is repeated until no new core objects can be added. For boundary objects within a range (i.e., objects whose number of objects within a category range is less than the category cardinality), they are directly added to the current category without further expansion.

[0094] The labeling of independent objects applies to objects that are neither within the classification interval of any core object nor can they form an interval containing at least two other objects with other objects. These objects are labeled separately in the classification results and are not included in any category.

[0095] When identifying wavering sheaths, sheath potential analysis is performed on non-uniform etching in each category, and the number of different types of sheath potentials in that category is counted. When a category contains at least two non-uniform etchings and the number of different types of sheath potentials is not less than two, it indicates that there is a change in the sheath potential in that category, which meets the characteristics of a wavering sheath. Therefore, the non-uniform etching in that category is identified as a wavering sheath.

[0096] In practice, the extraction of feature parameters must be accurate, and the counting of troughs must exclude false troughs caused by measurement noise. This can be achieved by setting a minimum trough depth threshold; only troughs with a depth exceeding the threshold are counted. The calculation of the gradient change rate of the contour curve requires smoothing the original curve to reduce the impact of noise on the calculation results.

[0097] Drawing a distance analysis plot requires selecting an appropriate combination of feature parameters. Typically, the number of troughs and the gradient rate of change of the contour curve are chosen as coordinate axes to ensure that different categories of objects can be clearly distinguished in the plot. The initial classification intervals can be optimized based on the actual classification results. If the classification results show too many or too few categories, the interval size can be adjusted appropriately for reclassification.

[0098] During category expansion, the expansion path needs to be recorded to trace the category formation process in subsequent analysis. When category overlap or intersection occurs, objects are assigned to the closer category based on their distance from the core object. If the number of independent objects is too large, it indicates that the initial classification interval setting may be unreasonable, and the distance distribution characteristics need to be re-analyzed, the interval size adjusted, and the classification process repeated.

[0099] Example 5:

[0100] When correcting the initial control coefficient based on the characteristics of the wave sheath, the characteristics of the wave sheath include the total number of wave sheaths and the ion energy of each wave sheath. The characteristics of the wave sheaths are compared with historical correction schemes, and the correction coefficient is determined based on the comparison results to correct the initial control coefficient. The historical correction scheme includes the characteristics of several historical wave sheaths and several historical correction coefficients, with each historical wave sheath characteristic corresponding to a historical correction coefficient. The similarity between the characteristics of the wave sheaths and the characteristics of each historical wave sheath is calculated. When there is data in the historical wave sheath characteristics whose similarity to the characteristics of the wave sheath is greater than a similarity threshold, the historical correction coefficient corresponding to the historical wave sheath with the highest similarity is determined as the correction coefficient to correct the initial control coefficient. When the similarity between the historical wave sheath characteristics and the characteristics of the wave sheath is all less than or equal to the similarity threshold, the correction coefficient is determined based on the total number of wave sheaths to correct the initial control coefficient. The correction coefficient is directly proportional to the total number of wave sheaths, and the value of the correction coefficient ranges from greater than 1 to less than or equal to 1.5. When adjusting the next regulation optimization cycle, the corrected regulation coefficient is obtained, and the cycle correction coefficient is determined based on the regulation coefficient to adjust the next regulation optimization cycle. The cycle correction coefficient is inversely proportional to the corrected regulation coefficient, and the value range of the cycle correction coefficient is greater than or equal to 0.5 and less than 1.

[0101] The total number of wave sheaths is the total number of non-uniform etchings identified as wave sheaths within the current control and optimization cycle. Duplicate counts must be excluded during the statistical analysis; each wave sheath is counted only once. The ion energy of each wave sheath is collected using a dedicated energy analyzer. The collected ion energy is stored in numerical form, retaining two decimal places to ensure data accuracy meets comparison requirements.

[0102] The historical correction scheme is constructed based on past cases of processing wave sheaths. Each case includes the total number of wave sheaths at that time, the ion energy of each wave sheath, and the corresponding correction coefficient. These cases are stored in chronological order, and an index is created for fast querying. The index uses the total number of wave sheaths and the range of ion energies as keywords to improve comparison efficiency.

[0103] The similarity is calculated from two dimensions: the difference in the total number of wave sheaths and the difference in the overall distribution of ion energy. The difference in total number is calculated as the absolute value of the difference between the current total number and the historical total number; the smaller the difference, the higher the similarity in that dimension. The difference in ion energy distribution is assessed by calculating the difference between the current average ion energy and the average ion energy in historical cases, as well as the difference in their standard deviations, to comprehensively evaluate the closeness of the energy distribution. The similarities from the two dimensions are combined in a certain proportion to obtain the overall similarity; the higher this value, the closer the current wave sheath characteristics are to historical cases.

[0104] The similarity threshold is determined based on the effectiveness of historical correction schemes. By analyzing the actual application effects of correction coefficients in historical cases, the similarity between cases that can achieve effective control and the current features is used as a reference to set the threshold. When the similarity of a historical case exceeds this threshold, it indicates that its correction coefficient is applicable to the current situation.

[0105] When the similarity of multiple historical cases exceeds a threshold, the correction coefficient corresponding to the case with the highest similarity is selected as the current correction coefficient. If there are multiple cases with the highest similarity, the correction coefficient corresponding to the most recent case is selected to adapt to the latest changes in the process.

[0106] When the similarity of all historical cases does not exceed the threshold, the correction coefficient is determined based on the total number of wave sheaths. The larger the total number, the larger the correction coefficient. For example, the correction coefficient can be 1.1 when the total number is 2, 1.3 when the total number is 5, and 1.5 when the total number is 8, ensuring that the correction coefficient increases linearly between 1 and 1.5.

[0107] The corrected control coefficient is the product of the initial control coefficient and the corrected coefficient, and this value serves as the final basis for control. The determination of the periodic correction coefficient is based on the corrected control coefficient; the larger the corrected control coefficient, the smaller the periodic correction coefficient. For example, when the corrected control coefficient is 1.2, the periodic correction coefficient can be 0.8; when the corrected control coefficient is 1.5, the periodic correction coefficient can be 0.5. During calculation, it must be ensured that the periodic correction coefficient is between 0.5 and 1. If the result calculated according to the inverse relationship exceeds the range, the extreme value within the range should be taken.

[0108] The duration of the next adjustment and optimization cycle is calculated by multiplying the current cycle duration by the cycle correction coefficient, with the result rounded to the nearest whole minute to ensure the reasonableness of the cycle duration. For example, if the current cycle is 10 minutes and the cycle correction coefficient is 0.7, then the next cycle will be 7 minutes; if the current cycle is 8 minutes and the cycle correction coefficient is 0.6, then the next cycle will be 5 minutes.

[0109] In practice, the calculation process of the correction coefficient and the periodic correction coefficient needs to be recorded in a log. The log includes historical case information on which the calculation was based, the specific values ​​of each parameter, and the final result, which facilitates subsequent traceability and process optimization analysis. At the same time, historical correction schemes are updated regularly, and new processing cases are added to continuously enrich the case library and improve the accuracy of similarity comparison.

[0110] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0111] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for remotely controlling plasma sheath properties and optimizing etching uniformity, characterized in that, include: Historical stable parameters are collected to establish a benchmark parameter library, and a real-time monitoring set is collected. The control and optimization cycle is determined based on the amount of data in the real-time monitoring set. All plasma sheath parameters are collected within the control and optimization cycle, and the average value of the sheath parameters is obtained. The average value of the sheath parameters is used as the control and optimization threshold. The plasma sheath parameters are analyzed according to the control and optimization threshold and the benchmark parameter library to determine whether there is any suspected non-uniform etching and mark it. When there is a suspected non-uniform etching, the sheath thickness and distribution characteristics of each suspected non-uniform etching are obtained. Based on the sheath thickness and distribution characteristics, it is determined whether the suspected non-uniform etching is non-uniform etching. When the suspected non-uniform etching is determined to be non-uniform etching, the initial control coefficient is determined based on the non-uniform etching frequency. Extract the non-uniform etching and identify the sheath potential, ion energy and boundary contour. Use the boundary contour as a feature parameter. Classify all the non-uniform etchings based on a classification algorithm. Determine whether there is a fluctuating sheath based on the classification results. When a wavering sheath is determined, a correction coefficient is determined based on the characteristics of the wavering sheath to correct the initial control coefficient, and the next control optimization cycle is adjusted.

2. The method for remote plasma sheath characteristic regulation and etching uniformity optimization according to claim 1, characterized in that, When determining the control optimization cycle based on the amount of data in the real-time monitoring set, the following are included: The data volume is compared with a first preset data volume and a second preset data volume, and the adjustment and optimization cycle is determined based on the comparison results; the first preset data volume is less than the second preset data volume. When the amount of data is less than or equal to a first preset amount of data, the adjustment and optimization cycle is determined to be a first cycle; when the amount of data is greater than the first preset amount of data and less than or equal to a second preset amount of data, the adjustment and optimization cycle is determined to be a second cycle; when the amount of data is greater than the second preset amount of data, the adjustment and optimization cycle is determined to be a third cycle; the first cycle is greater than the second cycle, and the second cycle is greater than the third cycle.

3. The method for remote plasma sheath characteristic regulation and etching uniformity optimization according to claim 2, characterized in that, When analyzing the plasma sheath parameters based on the aforementioned control optimization threshold and benchmark parameter library, and determining and marking the presence of suspected non-uniform etching, the following steps are taken: The sheath potential of all plasma sheath parameters is compared with the reference parameter library, and the sheath density of each plasma sheath parameter is compared with the average value of the sheath parameter. Based on the comparison results, it is determined whether there is any suspected non-uniform etching and it is marked. When the sheath density of the plasma sheath parameter is greater than a preset multiple of the average value of the sheath parameter, the plasma sheath parameter is determined to be a suspected non-uniform etching and is marked. When the sheath potential of the plasma sheath parameter is not in the reference parameter library, the plasma sheath parameter is determined to be a suspected non-uniform etching and is marked.

4. The method for remote plasma sheath characteristic regulation and etching uniformity optimization according to claim 3, characterized in that, When determining whether the suspected non-uniform etching is non-uniform etching based on the sheath thickness and distribution characteristics, the following methods are included: When the thickness of the sheath layer suspected of being non-uniformly etched is different from the conventional thickness in the reference parameter library, the suspected non-uniform etching is determined to be non-uniform etching. The distribution characteristics include the distribution pattern morphology. When the distribution pattern morphology of the suspected non-uniform etching shows discrete distribution peaks, the suspected non-uniform etching is determined to be non-uniform etching.

5. The method for remote plasma sheath characteristic regulation and etching uniformity optimization according to claim 4, characterized in that, When determining the initial control coefficient based on the non-uniform etching frequency, the following are included: The difference between the non-uniform etching frequency and the conventional frequency in the reference parameter library is statistically analyzed. Combined with the deviation of the sheath thickness of each non-uniform etching layer from the thickness closest to the reference parameter library, the initial control coefficient is calculated.

6. The method for remote plasma sheath characteristic regulation and etching uniformity optimization according to claim 5, characterized in that, When classifying all the non-uniform etchings based on a classification algorithm, and determining whether a fluctuating sheath layer exists based on the classification results, the following steps are taken: The boundary profile includes the number of valleys in the profile curve and the gradient rate of the profile curve. The initial classification intervals were determined using a distance analysis plot, and the classification cardinality was set to 2. Each non-uniform etching feature parameter is treated as an object. All objects are traversed, and the object with a number of objects in the classification interval that is greater than or equal to the classification cardinality is identified as the core object. Starting with each core object, examine the objects within its category range; if an object in the range is a core object, continue expanding the category; if an object in the range is a boundary object, add it to the current category; if an object is not within the category range of any core object and cannot form a category with other objects, it is marked as an independent object. When there exists a class that includes at least two non-uniform etchings and the sheath potential of the non-uniform etchings is at least two different potentials, the non-uniform etchings in that class are determined to be wavering sheaths.

7. The method for remote plasma sheath characteristic regulation and etching uniformity optimization according to claim 6, characterized in that, When correcting the initial control coefficient based on the characteristics of the wave sheath, the correction coefficient is determined, including: The characteristics of the wave sheath include the total number of wave sheaths and the ion energy of each wave sheath. The characteristics of the wave sheath are compared with historical correction schemes, and correction coefficients are determined based on the comparison results to correct the initial control coefficients. The historical correction schemes include the characteristics of several historical wave sheaths and several historical correction coefficients, and each characteristic of a historical wave sheath corresponds to a historical correction coefficient. Calculate the similarity between the characteristics of the wave sheath and the characteristics of each historical wave sheath; When there is data in the historical wave sheath features that has a similarity greater than a similarity threshold with the features of the wave sheath, the historical correction coefficient is determined based on the historical wave sheath features corresponding to the maximum similarity and is used to correct the initial control coefficient. When the similarity between the characteristics of the historical wave sheath and the characteristics of the wave sheath is less than or equal to the similarity threshold, the initial control coefficient is corrected based on the total number of wave sheaths.

8. The method for remote plasma sheath characteristic regulation and etching uniformity optimization according to claim 7, characterized in that, When correcting the initial control coefficient based on the total number of wave sheaths, the correction coefficient includes: The correction coefficient is proportional to the total number of wave sheaths, and the value of the correction coefficient is greater than 1 and less than or equal to 1.

5.

9. The method for remote plasma sheath characteristic regulation and etching uniformity optimization according to claim 8, characterized in that, When adjusting the next regulatory optimization cycle, the following should be included: Obtain the corrected control coefficient, determine the periodic correction coefficient based on the control coefficient, and adjust the next control optimization cycle accordingly. The periodic correction coefficient is inversely proportional to the corrected control coefficient, and the value range of the periodic correction coefficient is greater than or equal to 0.5 and less than 1.

10. The method for remote plasma sheath characteristic modulation and etching uniformity optimization according to claim 1, characterized in that, When collecting real-time monitoring data, it includes: The potential, density, and thickness parameters of the plasma sheath are collected synchronously by multi-channel sensors, and the parameters are arranged in a time series to form a real-time monitoring set.

Images