An optical path cleanliness detection device, monitoring method and processing system

Abstract

The present invention provides an optical path cleanliness detection device, monitoring method, and processing system, relating to the field of chip manufacturing technology. The present invention utilizes a digital micromirror device (DMD) to generate a pure white image, captures the reflected light image of the optical path lens using an area array CCD, and then uses a CCD controller to collect the reflected light image of the area array CCD. All pixels are accumulated and processed, and the accumulated value is used as the lens cleanliness data. By monitoring the optical path cleanliness in real time and performing on-demand maintenance based on the monitoring results, the cleanliness of the optical path lens can be effectively maintained, improving the imaging quality and lithography accuracy of the lithography machine, thereby improving the chip yield rate, avoiding the problems of excessive and insufficient maintenance, reducing the number of maintenance times and maintenance time, and reducing maintenance costs. In addition, the present invention can also predict defective product batches and analyze and predict pollution sources based on relevant data, further mining and utilizing the collected data to achieve a comprehensive assessment of affected defective product batches and pollution sources.

Description

Technical Field

[0001] The present invention relates to the technical field of chip manufacturing, and in particular to an optical path cleanliness detection device, a monitoring method and a processing system. Background Art

[0002] As a key piece of equipment in semiconductor manufacturing, the performance and long-term stability of photolithography machines are crucial to chip quality. The cleanliness of the lenses in the optical path is a key factor influencing their performance. Despite stringent air cleanliness requirements for the machine's installation environment, dust and other contaminants inevitably accumulate on the lenses during long-term operation, leading to reduced light transmittance and image quality. This, in turn, impacts the long-term stability of the machine and increases the risk of defects during production.

[0003] Currently, the optical path cleanliness maintenance of photolithography machines mainly relies on regular manual cleaning, such as monthly lens wiping. This maintenance method has the following limitations:

[0004] Fixed maintenance cycles lack flexibility: Different lithography machines vary in frequency of use, environmental conditions, and other factors, so a fixed maintenance cycle may not meet actual needs. For lithography machines with high usage and poor environmental conditions, monthly maintenance may not be sufficient. However, for lithography machines with low usage and good environmental conditions, monthly maintenance may lead to excessive maintenance, increasing maintenance costs.

[0005] Manual operation can lead to errors: When manually cleaning lenses, factors such as the operator's skill level and operating procedures can affect the cleaning effect. Differences in operation between different operators can lead to inconsistent maintenance of optical path cleanliness, which in turn affects the performance of the lithography machine.

[0006] Unable to monitor the cleanliness of the optical path in real time: Between two maintenance sessions, the changes in the cleanliness of the optical path cannot be monitored in real time. Once the contaminants accumulate to a certain level, they may have a serious impact on the photolithography process, but it is impossible to detect and take measures in time.

[0007] Unable to assess the impact of contamination: When contamination occurs but is not discovered in time, it will lead to a decline in the quality of subsequent wafer production and even cause serious defects in the chip. Traditional regular cleaning methods cannot detect and locate problematic product batches in time, which makes it impossible to conduct subsequent contamination impact assessments.

[0008] Therefore, it is necessary to provide an optical path cleanliness detection device, monitoring method and processing system to solve the technical problems that traditional optical path cleanliness maintenance relies on regular manual cleaning, cannot monitor the optical path cleanliness in real time, and cannot evaluate the impact of contamination. Summary of the Invention

[0009] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide an easy-to-operate optical path cleanliness detection device, monitoring method and processing system. It solves the technical problems that traditional optical path cleanliness maintenance relies on regular manual cleaning, cannot monitor the optical path cleanliness in real time, and cannot assess the impact of contamination. It monitors the optical path cleanliness in real time, assists users in dynamic time cleaning, and comprehensively evaluates the affected defective product batches and contamination sources.

[0010] To achieve the above-mentioned purpose, the present application proposes an optical path cleanliness detection device for detecting the cleanliness of lenses in the optical path of a photolithography machine, comprising an area array CCD, a digital micromirror device (DMD), a CCD controller, a data acquisition module, a data analysis module, and an alarm prompt module; wherein,

[0011] Area array CCD: A high-resolution area array CCD is installed at the edge of the lithography machine stage and away from the wafer stage. The area array CCD is used to capture the image of reflected light from the optical path lens.

[0012] Digital micromirror device (DMD): used to generate a pure white image and project it onto the target surface of the area array CCD with a 100% reflected light path;

[0013] CCD controller: responsible for collecting the reflected light image of the area array CCD, accumulating all pixels, and using the accumulated value as the lens cleanliness data;

[0014] Data acquisition module: used to continuously collect and store clean big data, including comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data and reflected light images;

[0015] Data analysis module: used to perform data analysis and prediction based on clean big data, including maintenance cycle prediction, defective product batch prediction and pollution source analysis and prediction;

[0016] Alarm prompt module: When the lens cleanliness data is lower than the set lens cleanliness threshold, a cleanliness alarm signal is issued and a prompt is given that cleaning maintenance is required.

[0017] As a further solution, perform optical path cleanliness monitoring through the following steps:

[0018] Step 1: Control the digital micromirror device (DMD) to generate a pure white image and project it onto the target surface of the area array CCD with a 100% reflected light path;

[0019] Step 2: Capture the reflected light image of the optical path lens through the area array CCD;

[0020] Step 3: The reflected light image of the area array CCD is collected by the CCD controller, and all pixels are accumulated and processed, and the accumulated value is used as the lens cleanliness data;

[0021] Step 4: Continuously collect and store clean big data through the data acquisition module; this includes comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data, and reflected light images;

[0022] Step 5: Use the data analysis module to perform data analysis and prediction based on the clean big data, including maintenance cycle prediction, defective product batch prediction, and pollution source analysis and prediction;

[0023] Step 6: When the lens cleanliness data is lower than the set lens cleanliness threshold, the alarm prompt module is activated to issue a cleanliness alarm signal and prompt that cleaning maintenance is required.

[0024] On the other hand, the present invention provides a light path pollution treatment system, comprising a light path cleanliness detection device, a maintenance cycle prediction unit, a defective product batch prediction unit, and a pollution source analysis and prediction unit; wherein,

[0025] Optical path cleanliness detection device: used to collect cleaning big data and lens cleanliness data, and when the lens cleanliness data is lower than the set lens cleanliness threshold, it will issue a cleanliness alarm signal and prompt that cleaning maintenance is required;

[0026] Maintenance cycle prediction unit: Based on lens cleanliness data and cleaning big data, a cleaning data prediction model and a cleaning data comprehensive scoring model are constructed to set the maintenance cycle.

[0027] Defective batch prediction unit: Based on lens cleanliness data, cleaning big data and cleaning data comprehensive scoring model, it performs forward prediction and / or backward prediction of defective batches to obtain future and / or past predicted defective batches;

[0028] Pollution source analysis and prediction unit: By comparing the similarity between the pollution source database and the clean big data and reflected light image at the time point corresponding to the issuance of the cleanliness alarm, the pollution source information whose similarity reaches the similarity threshold is used as the pollution source analysis prediction output; which includes image similarity and big data similarity.

[0029] As a further solution, the maintenance cycle prediction unit performs maintenance cycle prediction through the following steps:

[0030] Collecting cleaning big data, including comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data, and historical cleaning data;

[0031] Extract cleaning time nodes and intervals from historical cleaning data, and perform time alignment on lithography machine comprehensive operation data, chip tape-out quality data, cleanliness inspection data, and historical cleaning data;

[0032] The comprehensive operation data of the lithography machine, chip flow quality data and lens cleanliness data are divided into time by cleaning time nodes to obtain large data sets before cleaning and large data sets after cleaning;

[0033] Build a clean data prediction model, use the pre-cleaning large dataset and the corresponding pre- and post-cleaning intervals as input, and the post-cleaning large dataset as output for model training.

[0034] Construct a comprehensive scoring model for clean data and set scoring thresholds; the comprehensive scoring model for clean data is a data calculation model, including factor weights corresponding to operation scoring factors, quality scoring factors, cleanliness scoring factors, and time interval factors;

[0035] Continuously collect cleaning big data from the most recent cleaning time node to the current cleaning time node, and input it into the cleaning data prediction model to perform cleaning data prediction and obtain cleaning prediction data;

[0036] Substitute the clean prediction data into the clean data comprehensive scoring model, substitute the scoring threshold and use the time interval factor as the solution to obtain the predicted interval time before and after reaching the scoring threshold;

[0037] The predicted time between the front and back intervals is set as the maintenance cycle, and when the predicted time between the front and back intervals returns to zero, a notification is issued to clean the optical path.

[0038] As a further solution, the comprehensive operation data of the lithography machine includes overlay accuracy, illumination uniformity and pre-alignment accuracy; the chip flow quality data includes functional test data, reliability test data and parameter test data, the operation scoring factor comprehensively includes the overlay accuracy score, illumination uniformity score and pre-alignment accuracy score, and the quality scoring factor comprehensively includes the functional test score, reliability test score and parameter test score.

[0039] As a further solution, the defective batch prediction unit performs forward prediction of defective batches through the following steps:

[0040] Set defective product quality threshold;

[0041] Obtain the manufacturing plan and retrieve the corresponding comprehensive operation data of the lithography machine according to the manufacturing plan;

[0042] By fitting the historical lens cleanliness data, a normal lens cleanliness variation function is obtained;

[0043] Perform forward iteration on the interval time before and after;

[0044] Obtaining the comprehensive operation data of the lithography machine corresponding to the interval time between the front and back times as the comprehensive operation prediction data of the lithography machine, substituting the interval time between the front and back times into the lens cleanliness change function to obtain the lens cleanliness prediction data;

[0045] Substitute the comprehensive operation prediction data of the lithography machine, the lens cleanliness prediction data, the scoring threshold, and the interval time into the comprehensive scoring model of the cleaning data, and use the quality scoring factor as the solution to obtain the forward prediction value of the chip tape-out quality;

[0046] Compare the defective product quality threshold and the chip tape-out quality forward prediction value; if it is judged to be a defective product, output the current interval time; otherwise, return to forward iteration of the interval time;

[0047] The output interval time is used as the defective batch generation time, and the batch products corresponding to the defective batch generation time in the future are set as defective predicted batch products.

[0048] As a further solution, the defective batch prediction unit performs backward prediction of defective batches through the following steps:

[0049] Set defective product quality threshold;

[0050] Obtain the comprehensive operation history data of the lithography machine before the current time, the lens cleanliness history data, and the comprehensive cleaning data score history data;

[0051] Iterate backwards on the interval time before and after;

[0052] Substitute the comprehensive operation history data of the lithography machine, the historical data of the lens cleanliness, the historical data of the comprehensive cleaning data score, and the interval time into the comprehensive cleaning data score model, and use the quality score factor as the solution to obtain the backward prediction value of the chip tape-out quality;

[0053] Compare the defective product quality threshold and the backward prediction value of chip tape-out quality; if it is judged to be a defective product, output the current interval time; otherwise, return to the backward iteration of the interval time;

[0054] The output interval time is used as the defective batch generation time, and the batch products corresponding to the defective batch generation time in the past are set as the defective predicted batch products.

[0055] As a further solution, the pollution source database is constructed through the following steps:

[0056] Setting a lens cleanliness threshold and obtaining lens cleanliness data through an optical path cleanliness detection device;

[0057] When the lens cleanliness data exceeds the lens cleanliness threshold, the current time point is marked and a cleanliness alarm is issued;

[0058] Obtain clean big data and reflected light images at marked time points and set up pollution source tracing tasks;

[0059] Professionals conduct retrospective analysis on pollution source tracing tasks and locate pollution source information;

[0060] Clean big data, reflected light images and pollution source information are added to the pollution source analysis library as pollution source data.

[0061] Compared with related technologies, the optical path cleanliness detection device, monitoring method, and processing system provided by the present invention have the following advantages:

[0062] The present invention utilizes a digital micromirror device (DMD) to generate a pure white image, uses an area array CCD to capture the reflected light image of the optical path lens, and then uses a CCD controller to collect the reflected light image of the area array CCD. The image is then accumulated for all pixels, and the accumulated value is used as the lens cleanliness data. By monitoring the optical path cleanliness in real time and performing on-demand maintenance based on the monitoring results, the cleanliness of the optical path lens can be effectively maintained, improving the imaging quality and lithography accuracy of the lithography machine, thereby increasing the yield rate of the chip, avoiding the problems of excessive and insufficient maintenance, reducing the number of maintenance times and time, and lowering maintenance costs. Furthermore, the present invention can predict defective product batches and analyze pollution sources based on relevant data, further mining and utilizing the collected data to achieve a comprehensive assessment of affected defective product batches and pollution sources. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0065] Figure 1 A schematic structural diagram of an optical path cleanliness detection device provided by the present invention;

[0066] Figure 2 A schematic diagram of the maintenance cycle prediction steps provided by the present invention;

[0067] Figure 3 Schematic diagram of the steps for forward prediction of defective product batches provided by the present invention;

[0068] Figure 4 Schematic diagram of the backward prediction steps for defective product batches provided by the present invention.

[0069] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0071] See also Figure 1 The embodiment of the present application provides an optical path cleanliness detection device for detecting the cleanliness of lenses in the optical path of a lithography machine, comprising an area array CCD, a digital micromirror device (DMD), a CCD controller, a data acquisition module, a data analysis module, and an alarm prompt module; wherein,

[0072] Area array CCD: A high-resolution area array CCD is installed at the edge of the lithography machine stage and away from the wafer stage. The area array CCD is used to capture the image of reflected light from the optical path lens.

[0073] Digital micromirror device (DMD): used to generate a pure white image and project it onto the target surface of the area array CCD with a 100% reflected light path;

[0074] CCD controller: responsible for collecting the reflected light image of the area array CCD, accumulating all pixels, and using the accumulated value as the lens cleanliness data;

[0075] Data acquisition module: used to continuously collect and store clean big data, including comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data and reflected light images;

[0076] Data analysis module: used to perform data analysis and prediction based on clean big data, including maintenance cycle prediction, defective product batch prediction and pollution source analysis and prediction;

[0077] Alarm prompt module: When the lens cleanliness data is lower than the set lens cleanliness threshold, a cleanliness alarm signal is issued and a prompt is given that cleaning maintenance is required.

[0078] It should be noted that this embodiment utilizes a digital micromirror device (DMD) to generate a pure white image, captures the reflected light image of the optical path lens through an area array CCD, and then uses a CCD controller to collect the reflected light image of the area array CCD. All pixels are accumulated and processed, and the accumulated value is used as the lens cleanliness data. By monitoring the cleanliness of the optical path in real time and performing on-demand maintenance based on the monitoring results, the cleanliness of the optical path lens can be effectively maintained, the imaging quality and lithography accuracy of the lithography machine can be improved, thereby improving the yield rate of the chip, avoiding the problems of excessive and insufficient maintenance, reducing the number of maintenance times and maintenance time, and reducing maintenance costs. In addition, the present invention can also predict defective product batches and analyze pollution sources based on relevant data, further mining and utilizing the collected data to achieve a comprehensive assessment of affected defective product batches and pollution sources.

[0079] The present invention provides an optical path pollution treatment system, comprising an optical path cleanliness detection device, a maintenance cycle prediction unit, a defective product batch prediction unit, and a pollution source analysis and prediction unit; wherein,

[0080] Optical path cleanliness detection device: used to collect cleaning big data and lens cleanliness data, and when the lens cleanliness data is lower than the set lens cleanliness threshold, it will issue a cleanliness alarm signal and prompt that cleaning maintenance is required;

[0081] Maintenance cycle prediction unit: Based on lens cleanliness data and cleaning big data, a cleaning data prediction model and a cleaning data comprehensive scoring model are constructed to set the maintenance cycle.

[0082] Defective batch prediction unit: Based on lens cleanliness data, cleaning big data and cleaning data comprehensive scoring model, it performs forward prediction and / or backward prediction of defective batches to obtain future and / or past predicted defective batches;

[0083] Pollution source analysis and prediction unit: By comparing the similarity between the pollution source database and the clean big data and reflected light image at the time point corresponding to the issuance of the cleanliness alarm, the pollution source information whose similarity reaches the similarity threshold is used as the pollution source analysis prediction output; which includes image similarity and big data similarity.

[0084] like Figure 2 As shown, the maintenance cycle prediction unit performs maintenance cycle prediction through the following steps:

[0085] Collecting cleaning big data, including comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data, and historical cleaning data;

[0086] Extract cleaning time nodes and intervals from historical cleaning data, and perform time alignment on lithography machine comprehensive operation data, chip tape-out quality data, cleanliness inspection data, and historical cleaning data;

[0087] The comprehensive operation data of the lithography machine, chip flow quality data and lens cleanliness data are divided into time by cleaning time nodes to obtain large data sets before cleaning and large data sets after cleaning;

[0088] Build a clean data prediction model, use the pre-cleaning large dataset and the corresponding pre- and post-cleaning intervals as input, and the post-cleaning large dataset as output for model training.

[0089] Construct a comprehensive scoring model for clean data and set scoring thresholds; the comprehensive scoring model for clean data is a data calculation model, including factor weights corresponding to operation scoring factors, quality scoring factors, cleanliness scoring factors, and time interval factors;

[0090] Continuously collect cleaning big data from the most recent cleaning time node to the current cleaning time node, and input it into the cleaning data prediction model to perform cleaning data prediction and obtain cleaning prediction data;

[0091] Substitute the clean prediction data into the clean data comprehensive scoring model, substitute the scoring threshold and use the time interval factor as the solution to obtain the predicted interval time before and after reaching the scoring threshold;

[0092] The predicted time between the front and back intervals is set as the maintenance cycle, and when the predicted time between the front and back intervals returns to zero, a notification is issued to clean the optical path.

[0093] Operation score factor (comprehensive operation data of lithography machine):

[0094] Overlay Accuracy: The overlay accuracy score can be calculated based on its deviation from the standard value. For example, if the standard value is ±1μm, a scoring function can be set up to give full score when the actual value is within ±0.5μm, and 0 score when it exceeds ±1.5μm, with linear interpolation for intermediate values.

[0095] Illumination uniformity: The score for illumination uniformity can be calculated based on its deviation from the standard value. For example, if the standard value is ±3%, a scoring function can be set up to give full marks when the actual value is within ±1%, 0 marks when it exceeds ±5%, and linear interpolation for intermediate values.

[0096] Pre-alignment accuracy: The pre-alignment accuracy score can be calculated based on its deviation from the standard value. For example, if the standard value is ±15μm, a scoring function can be set up to give full score when the actual value is within ±10μm, 0 score when it exceeds ±20μm, and linear interpolation for intermediate values.

[0097] Quality Scoring Factor (Chip Tape Quality Data): The functional test score can be calculated based on the pass rate. For example, if all functional tests pass, full marks are awarded; if some functional tests fail, points are deducted based on the percentage of tests that fail.

[0098] Reliability testing: Reliability test scores can be calculated based on the stability of the test results. For example, if the performance is stable during long-term operation and high and low temperature tests, full marks will be awarded; if there are intermittent failures, points will be deducted based on the frequency of failures.

[0099] Parameter Testing: Parameter testing scores can be calculated based on the deviation of parameter values from the design specifications. For example, if all parameters are within the design specifications, full marks are awarded; if some parameters are outside the range, points are deducted based on the proportion of the deviation.

[0100] Cleanliness Scoring Factor (Lens Cleanliness Data): The scoring of mirror cleanliness can be calculated based on the cleanliness grading standard. For example, if the mirror cleanliness is "excellent", it will be scored full points; if it is "good", it will be scored 80 points; if it is "fair", it will be scored 60 points; if it is "poor", it will be scored 0 points.

[0101] Cleaning effect: The cleaning effect score can be calculated based on the cleaning effect evaluation criteria. For example, if the cleaning effect is "effective", full marks will be awarded; if it is "effective", 80 points will be awarded; if it is "ineffective", 0 points will be awarded.

[0102] Interval Factor: The interval factor can be calculated based on the rationality of the cleaning cycle and the equipment's operating status. For example, if the equipment is operating well and the cleaning cycle meets the recommended values, full marks are awarded. If the cleaning cycle is too long or too short, points are deducted based on the degree of deviation from the recommended values. A scoring function can be set up to award full marks when the actual cleaning cycle is within ±10% of the recommended value, 0 points when it exceeds ±20%, and linearly interpolate between values.

[0103] The comprehensive scoring model can be expressed as:

[0104] S=w1*F1+w2*F2+w3*F3+w4*F4;

[0105] Among them: S is the comprehensive score, F1 is the operation score factor, F2 is the quality score factor, F3 is the cleanliness score factor, F4 is the time interval factor, w1, w2, w3, w4 are factor weights, which can be adjusted according to actual conditions.

[0106] Through the above-mentioned scoring factors and comprehensive scoring model, a comprehensive evaluation can be made on the operating status of the lithography machine, chip flow quality, lens cleanliness and cleaning cycle to determine whether optical path cleaning is required.

[0107] Based on the above comprehensive scoring model, we can perform forward and backward predictions of defective batches. Among them, the forward prediction of defective batches mainly aims to make early predictions of defective batches that may appear in the future due to substandard cleanliness, so as to conduct targeted and focused screening of the batch of products to eliminate potential defective products. The backward prediction of defective batches mainly focuses on situations where there are no monitoring alarms. It can trace back to the defective batches that may have appeared in the past due to substandard cleanliness, so as to preliminarily determine the scope of the problem, thereby reducing the investigation work and facilitating subsequent maintenance.

[0108] like Figure 3 As shown in the figure, the good batch prediction unit performs forward prediction of defective batches through the following steps:

[0109] Set defective product quality threshold;

[0110] Obtain the manufacturing plan and retrieve the corresponding comprehensive operation data of the lithography machine according to the manufacturing plan;

[0111] By fitting the historical lens cleanliness data, a normal lens cleanliness variation function is obtained;

[0112] Perform forward iteration on the interval time before and after;

[0113] Obtaining the comprehensive operation data of the lithography machine corresponding to the interval time between the front and back times as the comprehensive operation prediction data of the lithography machine, substituting the interval time between the front and back times into the lens cleanliness change function to obtain the lens cleanliness prediction data;

[0114] Substitute the comprehensive operation prediction data of the lithography machine, the lens cleanliness prediction data, the scoring threshold, and the interval time into the comprehensive scoring model of the cleaning data, and use the quality scoring factor as the solution to obtain the forward prediction value of the chip tape-out quality;

[0115] Compare the defective product quality threshold and the chip tape-out quality forward prediction value; if it is judged to be a defective product, output the current interval time; otherwise, return to forward iteration of the interval time;

[0116] The output interval time is used as the defective batch generation time, and the batch products corresponding to the defective batch generation time in the future are set as defective predicted batch products.

[0117] like Figure 4 As shown in the figure, the defective batch prediction unit performs backward prediction of defective batches through the following steps:

[0118] Set defective product quality threshold;

[0119] Obtain the comprehensive operation history data of the lithography machine before the current time, the lens cleanliness history data, and the comprehensive cleaning data score history data;

[0120] Iterate backwards on the interval time before and after;

[0121] Substitute the comprehensive operation history data of the lithography machine, the historical data of the lens cleanliness, the historical data of the comprehensive cleaning data score, and the interval time into the comprehensive cleaning data score model, and use the quality score factor as the solution to obtain the backward prediction value of the chip tape-out quality;

[0122] Compare the defective product quality threshold and the backward prediction value of chip tape-out quality; if it is judged to be a defective product, output the current interval time; otherwise, return to the backward iteration of the interval time;

[0123] The output interval time is used as the defective batch generation time, and the batch products corresponding to the defective batch generation time in the past are set as the defective predicted batch products.

[0124] When building the contamination source database, we use an optical path cleanliness detection device to obtain lens cleanliness data and set a lens cleanliness threshold. For example, the threshold can be set to 80% cleanliness. When the lens cleanliness data exceeds the set lens cleanliness threshold, the current time point is marked and a cleanliness alarm is issued. For example, if the cleanliness falls below 80%, the system automatically marks the current time point and issues an alarm signal. At this time, the clean data and reflected light image of the marked time point are obtained, and a contamination source tracing task is set. The clean data includes comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data, and reflected light images.

[0125] Professionals conduct retrospective analysis of pollution source tracing tasks and locate pollution source information. This can be done by comparing historical and current data to analyze the characteristics and origins of pollution sources. Clean big data, reflected light images, and pollution source information are added to the pollution source analysis database as pollution source data. This helps to establish a comprehensive pollution source database, providing data support for future pollution source analysis and prevention.

[0126] The above are only some embodiments of the present application and are not intended to limit the patent scope of the present application. All equivalent structural transformations made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A light path cleanliness detection device for detecting the cleanliness of lenses in the light path of a photolithography machine, characterized in that: It includes area array CCD, digital micromirror device DMD, CCD controller, data acquisition module, data analysis module and alarm prompt module; among them, Area array CCD: A high-resolution area array CCD is installed at the edge of the lithography machine stage and away from the wafer stage. The area array CCD is used to capture the image of reflected light from the optical path lens. Digital micromirror device (DMD): used to generate a pure white image and project it onto the target surface of the area array CCD with a 100% reflected light path; CCD controller: responsible for collecting the reflected light image of the area array CCD, accumulating all pixels, and using the accumulated value as the lens cleanliness data; Data acquisition module: used to continuously collect and store clean big data, including comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data and reflected light images; Data analysis module: used to perform data analysis and prediction based on clean big data, including maintenance cycle prediction, defective product batch prediction and pollution source analysis and prediction; Alarm prompt module: When the lens cleanliness data is lower than the set lens cleanliness threshold, a cleanliness alarm signal is issued and a prompt is given that cleaning maintenance is required.

2. A method for monitoring optical path cleanliness, applied to an optical path cleanliness detection device as claimed in claim 1, characterized in that: Perform optical path cleanliness monitoring through the following steps: Step 1: Control the digital micromirror device (DMD) to generate a pure white image and project it onto the target surface of the area array CCD with a 100% reflected light path; Step 2: Capture the reflected light image of the optical path lens through the area array CCD; Step 3: The reflected light image of the area array CCD is collected by the CCD controller, and all pixels are accumulated and processed, and the accumulated value is used as the lens cleanliness data; Step 4: Continuously collect and store clean big data through the data acquisition module; this includes comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data, and reflected light images; Step 5: Use the data analysis module to perform data analysis and prediction based on the clean big data, including maintenance cycle prediction, defective product batch prediction, and pollution source analysis and prediction; Step 6: When the lens cleanliness data is lower than the set lens cleanliness threshold, the alarm prompt module is activated to issue a cleanliness alarm signal and prompt that cleaning maintenance is required.

3. A light path pollution treatment system, characterized in that: It includes an optical path cleanliness detection device, a maintenance cycle prediction unit, a defective product batch prediction unit and a pollution source analysis and prediction unit; among which, Optical path cleanliness detection device: used to collect cleaning big data and lens cleanliness data, and when the lens cleanliness data is lower than the set lens cleanliness threshold, it will issue a cleanliness alarm signal and prompt that cleaning maintenance is required; Maintenance cycle prediction unit: Based on lens cleanliness data and cleaning big data, a cleaning data prediction model and a cleaning data comprehensive scoring model are constructed to set the maintenance cycle. Defective batch prediction unit: Based on lens cleanliness data, cleaning big data and cleaning data comprehensive scoring model, it performs forward prediction and / or backward prediction of defective batches to obtain future and / or past predicted defective batches; Pollution source analysis and prediction unit: By comparing the similarity between the pollution source database and the clean big data and reflected light image at the time point corresponding to the issuance of the cleanliness alarm, the pollution source information whose similarity reaches the similarity threshold is used as the pollution source analysis prediction output; which includes image similarity and big data similarity.

4. The light path pollution treatment system according to claim 3, characterized in that: The maintenance cycle prediction unit performs maintenance cycle prediction through the following steps: Collecting cleaning big data, including comprehensive lithography machine operation data, chip flow quality data, lens cleanliness data, and historical cleaning data; Extract cleaning time nodes and intervals from historical cleaning data, and perform time alignment on lithography machine comprehensive operation data, chip tape-out quality data, cleanliness inspection data, and historical cleaning data; The comprehensive operation data of the lithography machine, chip flow quality data and lens cleanliness data are divided into time by cleaning time nodes to obtain large data sets before cleaning and large data sets after cleaning; Build a clean data prediction model, use the pre-cleaning large dataset and the corresponding pre- and post-cleaning intervals as input, and the post-cleaning large dataset as output for model training. Construct a comprehensive scoring model for clean data and set scoring thresholds; the comprehensive scoring model for clean data is a data calculation model, including factor weights corresponding to operation scoring factors, quality scoring factors, cleanliness scoring factors, and time interval factors; Continuously collect cleaning big data from the most recent cleaning time node to the current cleaning time node, and input it into the cleaning data prediction model to perform cleaning data prediction and obtain cleaning prediction data; Import the clean prediction data into the clean data comprehensive scoring model, substitute the scoring threshold and use the time interval factor as the solution to obtain the predicted interval time before and after reaching the scoring threshold; The predicted time between the front and back intervals is set as the maintenance cycle, and when the predicted time between the front and back intervals returns to zero, a notification is issued to clean the optical path.

5. The light path pollution treatment system according to claim 4, characterized in that: The comprehensive operation data of the lithography machine includes overlay accuracy, illumination uniformity and pre-alignment accuracy; the chip flow quality data includes functional test data, reliability test data and parameter test data, the operation scoring factor comprehensively integrates the overlay accuracy score, illumination uniformity score and pre-alignment accuracy score, and the quality scoring factor comprehensively integrates the functional test score, reliability test score and parameter test score.

6. The light path pollution treatment system according to claim 3, characterized in that: The defective product batch prediction unit performs forward prediction of defective product batches through the following steps: Set defective product quality threshold; Obtain the manufacturing plan and retrieve the corresponding comprehensive operation data of the lithography machine according to the manufacturing plan; By fitting the historical lens cleanliness data, a normal lens cleanliness variation function is obtained; Perform forward iteration on the interval time before and after; Obtaining the comprehensive operation data of the lithography machine corresponding to the interval time between the front and back times as the comprehensive operation prediction data of the lithography machine, substituting the interval time between the front and back times into the lens cleanliness change function to obtain the lens cleanliness prediction data; Substitute the comprehensive operation prediction data of the lithography machine, the lens cleanliness prediction data, the scoring threshold, and the interval time into the comprehensive scoring model of the cleaning data, and use the quality scoring factor as the solution to obtain the forward prediction value of the chip tape-out quality; Compare the defective product quality threshold and the chip tape-out quality forward prediction value; if it is judged to be a defective product, output the current interval time; Otherwise, return to the forward iteration of the interval time before and after; The output interval time is used as the defective batch generation time, and the batch products corresponding to the defective batch generation time in the future are set as defective predicted batch products.

7. The light path pollution treatment system according to claim 3, characterized in that: The defective product batch prediction unit performs backward prediction of defective product batches through the following steps: Set defective product quality threshold; Obtain the comprehensive operation history data of the lithography machine before the current time, the lens cleanliness history data, and the comprehensive cleaning data score history data; Iterate backwards on the interval time before and after; Substitute the comprehensive operation history data of the lithography machine, the historical data of the lens cleanliness, the historical data of the comprehensive score of the cleaning data, and the interval time into the comprehensive score model of the cleaning data, and use the quality score factor as the solution to obtain the backward prediction value of the chip tape-out quality; Compare the defective product quality threshold and the backward prediction value of chip tape-out quality; if it is judged to be a defective product, output the current interval time; Otherwise, return to iterate backwards over the intervals before and after; The output interval time is used as the defective batch generation time, and the batch products corresponding to the defective batch generation time in the past are set as the defective predicted batch products.

8. The light path pollution treatment system according to claim 3, characterized in that: The pollution source database is constructed through the following steps: Setting a lens cleanliness threshold and obtaining lens cleanliness data through an optical path cleanliness detection device; When the lens cleanliness data exceeds the lens cleanliness threshold, the current time point is marked and a cleanliness alarm is issued; Obtain clean big data and reflected light images at marked time points and set up pollution source tracing tasks; Professionals conduct retrospective analysis on pollution source tracing tasks and locate pollution source information; Clean big data, reflected light images and pollution source information are added to the pollution source analysis library as pollution source data.

Images