CN113158610B - Overlay measurement method and system in integrated circuit chip manufacturing process - Google Patents
Overlay measurement method and system in integrated circuit chip manufacturing process Download PDFInfo
- Publication number
- CN113158610B CN113158610B CN202110390067.0A CN202110390067A CN113158610B CN 113158610 B CN113158610 B CN 113158610B CN 202110390067 A CN202110390067 A CN 202110390067A CN 113158610 B CN113158610 B CN 113158610B
- Authority
- CN
- China
- Prior art keywords
- overlay
- data
- overlay measurement
- model
- measurement data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/39—Circuit design at the physical level
- G06F30/392—Floor-planning or layout, e.g. partitioning or placement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/39—Circuit design at the physical level
- G06F30/398—Design verification or optimisation, e.g. using design rule check [DRC], layout versus schematics [LVS] or finite element methods [FEM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Architecture (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Testing Or Measuring Of Semiconductors Or The Like (AREA)
- Exposure And Positioning Against Photoresist Photosensitive Materials (AREA)
Abstract
The invention discloses an overlay measuring method in the manufacturing process of an integrated circuit chip, which comprises the following steps: acquiring the data of each overlay mark image on the wafer or acquiring overlay measurement data in a partial area on the wafer aiming at each process level of the chip manufacturing process; respectively establishing a machine learning model by utilizing the collected overlapping marked image data or the overlapping measurement data: an overlay mark image processing model and an overlay measurement data prediction model; during the overlay measurement, the overlay mark image is processed through the established overlay mark image processing model, or the overlay measurement data in other unmeasured areas on the same wafer is predicted through the established overlay measurement data prediction model and the overlay measurement data collected in the upper area of the wafer. The invention processes or predicts the data of the overlay measurement through the machine learning model, thereby improving the stability, the precision and the measurement efficiency of the overlay measurement.
Description
Technical Field
The invention relates to the technical field of pattern overlay of integrated circuit chips, in particular to an overlay measuring method and system in the manufacturing process of the integrated circuit chips.
Background
The fabrication process of Integrated Circuit (Integrated Circuit) chips involves multiple process levels (layers). The process layers form a pattern of a photoetching material on the wafer through a photoetching process, and the pattern is used as a template to form a circuit device and an interconnection structure which meet the design specification by utilizing various process means such as etching, deposition, chemical mechanical polishing and the like. The Overlay (Overlay) of process patterns between different process levels is critical. If the overlay error is outside the allowable range, the circuit devices and the interconnect structure will not work properly.
In order to ensure accurate pattern alignment between process levels, Overlay measurements are performed using a specific Overlay Mark (Overlay Mark) after the photolithography process is completed. One possible overlay mark design is shown in fig. 1. The overlay marks 110 and 210 are generated during two different process levels, and each of the overlay marks includes 4 bar marks, forming a block structure. By measuring the distances X1, X2, Y1, Y2 and other parameters between the corresponding bar marks, the pattern overlay error between two process levels can be calculated. If the overlay error is outside the allowable range, Rework of the photolithography process step (Rework) is required. Besides monitoring the overlay error, the result of the overlay measurement is also used for establishing an overlay compensation model, and the process parameters are adjusted through the overlay compensation model, so that the overlay error between the process layers is reduced.
Several situations that may occur in the overlay measurement are shown in fig. 2. As shown in fig. 2(a), overlay marks 121, 122, 123, 124 are generated during the first process level, and are located at different positions on the wafer. Overlay marks 221, 222, 223, 224 are generated during the second process level, and are located on the wafer at positions corresponding to 121, 122, 123, 124, respectively. The centers of the overlay marks 121, 122, 123, 124 coincide with the centers of the overlay marks 221, 222, 223, 224, respectively, indicating that the pattern alignment between the first process level and the second process level is good. As shown in fig. 2(b), the overlay marks 131, 132, 133, 134 are generated during the first process level, and are located at different positions of the wafer. Overlay marks 231, 232, 233, 234 are generated during the second process level, and are located on the wafer at positions corresponding to 131, 132, 133, 134, respectively. The overlay marks 131, 132, 133, 134 are horizontally offset from the overlay marks 231, 232, 233, 234, respectively, indicating that there is an overlay error in the horizontal direction in the pattern alignment between the first process level and the second process level. As shown in fig. 2(c), the overlay marks 141, 142, 143, 144 are generated during the first process level, and are located at different positions on the wafer. Overlay marks 241, 242, 243, and 244 are generated during the second process level, and are located on the wafer at positions corresponding to 141, 142, 143, and 144, respectively. The overlay marks 141, 142, 143, 144 are vertically offset from the overlay marks 241, 242, 243, 244, respectively, indicating that there is a vertical overlay error in the pattern alignment between the first process level and the second process level. As shown in fig. 2(d), overlay marks 151, 152, 153, 154 are generated during the first process level, and are located at different positions on the wafer. Overlay marks 251, 252, 253, 254 are generated during the second process level, corresponding to locations 151, 152, 153, 154 on the wafer, respectively. The overlay marks 151, 152, 153, 154 and the overlay marks 251, 252, 253, 254 have relative offsets caused by rotation, indicating that overlay errors caused by rotation exist in the alignment of the patterns between the first process level and the second process level. In addition to the above-mentioned overlay errors, other factors, such as image magnification, lens distortion, wafer stage wobble, etc., may also cause overlay errors between different process levels.
Various defects may be created during chip fabrication that may affect the stability and accuracy of the measurement process. In addition, due to the limitation of the equipment productivity and the measurement time, only the overlay marks of partial areas on the wafer can be measured, and overlay data of other areas cannot be directly obtained. In view of the above, there is a need for an improved method and system for overlay measurement, which can enhance the stability and accuracy of the process and effectively obtain overlay data of each region of the wafer.
Disclosure of Invention
The invention provides an overlay measurement method and system in the manufacturing process of an integrated circuit chip to overcome the defects of the technology.
In the chip manufacturing process, overlay measurement is needed to monitor the pattern alignment condition between different process levels, ensure that the pattern alignment condition meets the specification requirements of design and process, and provide a compensation model for the subsequent process. Overlay measurement calculates overlay error by measuring the relative position between corresponding overlay mark patterns on each process level. The overlay mark pattern is affected by subsequent process steps after it is created. For example, subsequent thin film deposition steps may cause blurring of the image of the overlay mark, subsequent chemical mechanical polishing steps may cause the overlay mark to be partially missing due to over-polishing, and so on. In addition, particles generated during the process may also affect the overlay measurement. As shown in fig. 3, overlay marks 160 and 260 are generated at two different process levels, and each of them includes four bar marks, which form a block structure. In overlay mark 160, bar mark 1601 is partially missing, and the edge of bar mark 1603 is unclear. In addition, particles 1605 generated during the process are located near overlay marks 160 and 260, which generate interference signals, making overlay measurements impossible.
In the chip manufacturing process, due to the limitation of the equipment capacity and the test time, the overlay measurement can be performed only on a partial area on the wafer. Fig. 4 shows a set of regions 10 on a wafer, each region having a set of overlay marks, each overlay mark comprising a plurality of overlay marks. By using overlay marks between different process levels in a set of areas 10 on the wafer for overlay measurement, the overlay error between different process levels in a set of areas 10 on the wafer can be obtained, but the overlay error between different process levels in other areas on the wafer cannot be obtained.
The overlay measurement method and system of the present invention utilizes machine learning to overcome the above-described deficiencies. Machine learning is a subset of the field of artificial intelligence. Different from the traditional computer programming method based on the known rules, the machine learning utilizes the existing data for training, automatically learns the rules according to the machine learning algorithm, constructs a corresponding machine learning model, and utilizes the model to calculate the new input data to obtain an output result. Common machine learning models include, but are not limited to, linear regression, logistic regression, K-nearest neighbors, naive bayes, decision trees, clustering, support vector machines, random forests, boosted trees, and various neural networks, and the like. Machine learning has been widely used in many fields such as prediction, classification, image processing, and natural language processing in recent years, and has achieved excellent effects.
The technical scheme adopted by the invention for overcoming the technical problems is as follows:
an overlay measurement method in the manufacturing process of an integrated circuit chip comprises the following steps:
acquiring the data of each overlay mark image on the wafer or acquiring overlay measurement data in a partial area on the wafer aiming at each process level of the chip manufacturing process;
respectively establishing a machine learning model by utilizing the collected overlapping marked image data or the overlapping measurement data: an overlay mark image processing model and an overlay measurement data prediction model;
during the overlay measurement, the overlay mark image is processed through the established overlay mark image processing model, or the overlay measurement data in other unmeasured areas on the same wafer is predicted through the established overlay measurement data prediction model and the overlay measurement data collected in the upper area of the wafer.
Further, the step of establishing the machine learning model comprises the following steps:
collecting or generating input data required for establishing a machine learning model;
preprocessing the input data;
the preprocessed data is divided into two parts: one part is model training and verification data, and the other part is test data;
performing model training and verification through model training and verification data, and performing model evaluation through test data;
judging the evaluation result of the model: if the evaluation is qualified, model deployment is carried out to deliver the model for use; otherwise, returning to the previous step until the evaluation is qualified.
Preferably, the step of processing the overlay mark image data by the overlay mark image processing model comprises the following steps:
collecting input data of the overlay measurement;
preprocessing the input data;
inputting the preprocessed data into an overlay marked image processing model, wherein the overlay marked image processing model processes an overlay marked image by a machine learning method, and the processing comprises at least one of reconstructing a missing graph, reconstructing a blurred graph and eliminating particle influence;
the processed overlay mark image is used for overlay measurement data calculation to obtain final overlay measurement data.
Preferably, the step of processing the overlay mark image data by the overlay mark image processing model comprises the following steps:
collecting input data of the overlay measurement;
preprocessing the input data;
inputting the preprocessed data into an overlay marked image processing model, wherein the overlay marked image processing model processes an overlay marked image by a machine learning method, and the processing comprises at least one of reconstructing a missing graph, reconstructing a blurred graph and eliminating particle influence;
the processed overlay mark image is used for calculating overlay measurement data to obtain new overlay measurement data;
after new preprocessed data and new overlay measurement data are obtained, part or all of the new preprocessed data and part or all of the new overlay measurement data are used for on-line model training, and a new model generated after the on-line model training is used for updating an overlay labeled image processing model through model deployment, so that the overlay labeled image processing model is continuously updated according to the new data.
Preferably, the step of predicting the overlay measurement data by the overlay measurement data prediction model comprises the following steps:
collecting input data of overlapping measurement in partial area on the wafer;
preprocessing the input data;
the preprocessed data are input into an overlay measurement data prediction model, the overlay measurement data prediction model predicts overlay measurement data in other unmeasured areas on the same wafer by a machine learning method by using the preprocessed overlay measurement data obtained in the upper area of the wafer, and final overlay measurement data are obtained.
Preferably, the step of predicting the overlay measurement data by the overlay measurement data prediction model comprises the following steps:
collecting input data of overlapping measurement in partial area on the wafer;
preprocessing the input data;
inputting the preprocessed data into an overlay measurement data prediction model, wherein the overlay measurement data prediction model predicts overlay measurement data in other unmeasured areas on the same wafer by using the preprocessed overlay measurement data obtained in the upper area of the wafer through a machine learning method to obtain new overlay measurement data;
after new preprocessed data and new overlay measurement data are obtained, part or all of the new preprocessed data and part or all of the new overlay measurement data are used for on-line model training, and a new model generated after the on-line model training is used for updating an overlay measurement data prediction model through model deployment, so that the overlay measurement data prediction model is continuously updated according to the new data.
The invention also provides an overlay measurement system in the manufacturing process of integrated circuit chips, which at least comprises:
the mechanical transmission unit is at least used for transmitting and positioning the wafer;
the data acquisition unit is used for acquiring and receiving data required by the overlay measurement;
the input end of the overlay measuring unit is connected with the output end of the data acquisition unit and is used for processing or predicting overlay measuring data by a machine learning method; and the number of the first and second groups,
and the input end of the data output unit is connected with the output end of the overlay measurement unit and is used for outputting the overlay measurement result.
Preferably, the overlay measurement unit at least includes an overlay mark image processing module, and the overlay mark image processing module processes the overlay mark image by a machine learning method, where the processing includes at least one of reconstructing a missing pattern, reconstructing a blurred pattern, and eliminating a particle effect.
As another preferred scheme, the overlay measurement unit at least includes an overlay marker image processing module and an online model training module, the overlay marker image processing module processes the overlay marker image by a machine learning method, and the online model training module trains and updates an overlay marker image processing model for overlay marker image processing by using newly-added data, and deploys the updated model to the overlay marker image processing module.
Preferably, the overlay measurement unit at least includes an overlay measurement data prediction module, and the overlay measurement data prediction module predicts overlay measurement data of other unmeasured areas on the same wafer by using a machine learning method and the overlay measurement data in the area above the wafer.
Preferably, the overlay measurement unit at least includes an overlay measurement data prediction module and an online model training module, the overlay measurement data prediction module predicts overlay measurement data of other unmeasured areas on the same wafer through a machine learning method and overlay measurement data in an upper area of the wafer, and the online model training module trains and updates an overlay measurement data prediction model for the overlay measurement data prediction by using newly added data and deploys the updated model to the overlay measurement data prediction module.
The invention has the beneficial effects that:
the invention processes or predicts the data of the overlay measurement through the machine learning model, thereby improving the stability, the precision and the measurement efficiency of the overlay measurement.
1. The invention can process the overlay mark image data through the overlay mark image processing model, reduces the influence of factors such as process defects, image blurring or image deletion on the overlay measurement result, and improves the stability and the precision of overlay measurement.
2. The invention predicts the overlay measurement data in other unmeasured areas on the same wafer through the overlay measurement data prediction model and the overlay measurement data collected in the upper area of the wafer, and can obtain the overlay error in each area on the wafer in shorter test time, thereby greatly shortening the measurement time and improving the measurement efficiency.
Drawings
FIG. 1 is a schematic diagram of one possible overlay mark design.
Fig. 2 is a schematic diagram of four possible overlay mark alignment situations, and fig. 2 includes fig. 2(a), fig. 2(b), fig. 2(c) and fig. 2 (d).
FIG. 3 is a schematic diagram of one possible overlay mark image.
FIG. 4 is a schematic diagram of a possible overlay measurement distribution of several regions on a wafer.
FIG. 5 is a diagram illustrating the results of one possible overlay mark image processing by the overlay measurement method of the present invention.
FIG. 6 is a schematic view of overlay measurement distribution of two sets of regions on a wafer according to the present invention.
Fig. 7 is a schematic flowchart of establishing a machine learning model according to embodiment 1 of the present invention.
Fig. 8 is a flowchart illustrating an overlay mark image processing model according to embodiment 2 of the present invention processing overlay mark image data.
Fig. 9 is a flowchart illustrating another overlay mark image processing model according to embodiment 3 of the present invention for processing overlay mark image data.
Fig. 10 is a flowchart illustrating an iterative measurement data prediction model for predicting iterative measurement data according to embodiment 4 of the present invention.
Fig. 11 is a flowchart illustrating another iterative measurement data prediction model for predicting iterative measurement data according to embodiment 5 of the present invention.
Fig. 12 is a schematic diagram of an overlay measurement system according to embodiment 6 of the present invention.
Fig. 13 is a schematic diagram of a first overlay measurement system according to embodiment 7 of the present invention.
Fig. 14 is a schematic diagram of a second overlay measurement system according to embodiment 8 of the present invention.
Fig. 15 is a schematic diagram of a third overlay measurement system according to embodiment 9 of the present invention.
Fig. 16 is a schematic diagram of a fourth overlay measurement system according to embodiment 10 of the present invention.
Detailed Description
In order to facilitate a better understanding of the invention for those skilled in the art, the invention will be described in further detail with reference to the accompanying drawings and specific examples, which are given by way of illustration only and do not limit the scope of the invention.
Examples 1,
The invention relates to an overlay measurement method in the manufacturing process of an integrated circuit chip, which comprises the following steps:
and acquiring overlay mark image data on the wafer or acquiring overlay measurement data in partial areas on the wafer aiming at each process level of the chip manufacturing process.
Respectively establishing a machine learning model by utilizing the collected overlay mark image data or the overlay measurement data, specifically: and establishing an overlay mark image processing model by using the acquired overlay mark image data, and establishing an overlay measurement data prediction model by using the acquired overlay measurement data.
As shown in fig. 7, a process of establishing a machine learning model is shown. Firstly, collecting or generating input data 1000 required for establishing a machine learning model by a process test or simulation method, wherein the input data at least comprises names and process parameter settings of related process steps, process monitoring data, types and name of overlay measuring equipment, equipment parameter settings, types and position coordinates of overlay marks, horizontal direction deviation values and vertical direction deviation values between different overlay marks, image data of the overlay marks, types and parameter settings of overlay compensation models, residual values after overlay compensation, overlay error values and the like; secondly, preprocessing the input data 1000, wherein the preprocessing at least comprises cleaning, standardizing and feature extraction on the input data 1000, and the preprocessed data is recorded as 1100; then, the preprocessed data 1100 is divided into two parts: a portion of model training and validation data 1200 for model training and validation and a portion of test data 1300; model training and verification are performed through model training and verification data 1400, and after the model training and verification are completed, model evaluation is performed through test data 1300 1500; and finally, judging the evaluation result of the model: if the evaluation is qualified, model deployment is carried out 1600, and the model is delivered for use; otherwise, returning to the previous step, and performing training, verification and evaluation again until the evaluation is qualified.
During the overlay measurement, the overlay mark image is processed through the established overlay mark image processing model, or the overlay measurement data in other unmeasured areas on the same wafer is predicted through the established overlay measurement data prediction model and the overlay measurement data collected in the upper area of the wafer, so that the stability, the precision and the measurement efficiency of the overlay measurement can be improved.
Examples 2,
As shown in fig. 5, overlay marks 170 and 270 are generated at two different process levels, and each of the overlay marks includes four bar marks, which form a box structure. In the overlay mark 170, the bar mark 1701 is partially missing; the edges of the bar 1703 are not clear; in addition, particles 1705 generated during the process are located near the overlay marks 170 and 270, which generate interference signals that affect the proper performance of the overlay measurements.
In this embodiment, the overlay mark image processing model described in embodiment 1 is used to process the overlay mark image data, so as to solve the problem shown in fig. 5. As shown in fig. 8, the steps include the following:
for each process level, collecting each overlay mark image on the wafer as input data 2000 for overlay measurement, wherein the input data described in this embodiment is the same as the input data of embodiment 1;
the input data 2000 is preprocessed, and the preprocessing in this embodiment is the same as that in embodiment 1;
the preprocessed data 2100 is input to an overlay marker image processing model 2200, and the overlay marker image processing model 2200 processes the overlay marker image by a machine learning method, as shown in fig. 5, and the deployed overlay marker image processing model 2200 can reconstruct a bar marker pattern 1702 and a blur pattern 1704 and eliminate the pattern of particles 1705 from the image (shown by dotted lines).
The processed overlay mark images are used for overlay measurement data calculation 2300 to obtain final overlay measurement data 2400. Because the overlapped pair of marked images after being processed by the machine learning model restores the missing marked images, the unclear marked edges are reconstructed, the influence of particles is eliminated, and better measurement stability and precision can be obtained.
Examples 3,
As a more preferable scheme than embodiment 2, as shown in fig. 9, the present embodiment is different from embodiment 2 in that an online model training function is added, specifically:
for each process level, collecting each overlay mark image on the wafer as input data 3000 of overlay measurement, wherein the input data described in the embodiment is the same as the input data of embodiment 1;
preprocessing the input data 3000, which is the same as the preprocessing of embodiment 1;
the preprocessed data 3100 is input to an overlay marker image processing model 3200, and the overlay marker image processing model 3200 processes the overlay marker image by a machine learning method, as shown in fig. 5, and the overlay marker image processing model 3200 reconstructs a bar-shaped marker pattern 1702 and a blur pattern 1704, and removes the pattern of particles 1705 (shown by dotted lines in the figure) from the image.
The processed overlay mark images are used for overlay measurement data calculation 3300, resulting in new overlay measurement data 3400.
After new preprocessed data 3100 and new overlay measurement data 3400 are obtained, part or all of the new preprocessed data 3100 and part or all of the new overlay measurement data 3400 are used for on-line model training 3500, a new model generated after the on-line model training is updated through the model deployment 3600 overlay marker image processing model 3200, and therefore the overlay marker image processing model 3200 is updated continuously according to the new data, and the method can guarantee the effectiveness of the overlay marker image processing model under the condition that process drift exists.
Examples 4,
Fig. 6 shows a set of regions 11 on a wafer, each region having a set of overlay marks, each overlay mark comprising a plurality of overlay marks. By using overlay marks between different process levels in one set of areas 11 on the wafer for overlay measurement, overlay errors between different process levels in one set of areas 11 on the wafer can be obtained, but overlay errors between different process levels in another set of areas 21 on the wafer cannot be obtained.
In this embodiment, the iterative measurement data prediction model described in embodiment 1 is used to predict iterative measurement data, so as to solve the problem shown in fig. 6. As shown in fig. 10, the steps include the following:
for each process level, collecting overlay measurement data in each region on the wafer as input data 4000 for overlay measurement, wherein the input data in the embodiment is the same as the input data in embodiment 1;
preprocessing the input data, wherein the preprocessing is the same as that of the embodiment 1;
the preprocessed data 4100 is input into an overlay measurement data prediction model 4200, and the overlay measurement data prediction model 4200 predicts overlay measurement data in other unmeasured regions on the same wafer by a machine learning method using the preprocessed overlay measurement data obtained in the upper region of the wafer, so as to obtain final overlay measurement data 4300. Due to the fact that overlapping measurement data of the partial area on the wafer and other unmeasured areas can be obtained only by overlapping measurement of the partial area on the wafer, measurement time is greatly shortened, and measurement efficiency is improved.
Examples 5,
As a more preferable scheme than embodiment 4, as shown in fig. 11, the present embodiment is different from embodiment 4 in that an online model training function is added, specifically:
for each process level, collecting overlay measurement data in each region on the wafer as input data 5000 of overlay measurement, wherein the input data described in the embodiment is the same as the input data of embodiment 1;
preprocessing the input data, wherein the preprocessing is the same as that of the embodiment 1;
the preprocessed data 5100 is input into an overlay measurement data prediction model 5200, and the overlay measurement data prediction model 5200 predicts overlay measurement data in other unmeasured areas on the same wafer by a machine learning method by using the preprocessed overlay measurement data obtained in the upper area of the wafer to obtain new overlay measurement data 5300;
after new preprocessed data 5100 and new overlay measurement data 5300 are obtained, part or all of the new preprocessed data 5100 and part or all of the new overlay measurement data 5300 are used for online model training 5400, the new model generated after the online model training updates the overlay measurement data prediction model 5200 through model deployment 5500, and therefore the overlay measurement data prediction model 5200 is continuously updated according to the new data.
Examples 6,
The embodiment discloses an overlay measurement system 6000 in the manufacturing process of an integrated circuit chip, as shown in fig. 12, comprising at least a mechanical transmission unit 6100, a data acquisition unit 6200, an overlay measurement unit 6300, a data output unit 6400, and other units 6500. The mechanical transmission unit 6100 is at least used for wafer transportation and positioning; the data acquisition unit 6200 is configured to acquire and receive various types of data required for overlay measurement; the input end of the overlay measurement unit 6300 is connected to the output end of the data acquisition unit 6200, and is configured to process or predict overlay measurement data by a machine learning method; the input end of the data output unit 6400 is connected to the output end of the overlay measurement unit 6300, and is configured to output an overlay measurement result to a related hardware device and a software system; the other unit 6500 is responsible for other functions of the system, such as communication, data storage, etc. The overlay measurement unit 6300 at least includes a machine learning module 6310 and other measurement modules 6320, where the machine learning module 6310 processes or predicts overlay measurement data by using a machine learning model, so as to improve stability, accuracy and measurement efficiency of overlay measurement, and the other measurement modules 6320 are used to complete other overlay measurement functions.
Example 7,
The overlay measurement system in the integrated circuit chip manufacturing process described in this embodiment is used for processing overlay measurement data, as shown in fig. 13, specifically:
the overlay measurement system 7000 at least includes a mechanical transmission unit 7100, a data acquisition unit 7200, an overlay measurement unit 7300, a data output unit 7400 and other units 7500. The mechanical transmission unit 7100 is at least used for transmitting and positioning wafers; the data acquisition unit 7200 is configured to acquire and receive various types of data required for overlay measurement; the input end of the overlay measurement unit 7300 is connected to the output end of the data acquisition unit 7200, and is configured to process overlay measurement data by a machine learning method; the input end of the data output unit 7400 is connected to the output end of the overlay measurement unit 7300, and is used for outputting the overlay measurement result to the relevant hardware device and software system; the other unit 7500 is responsible for other functions of the system, such as communication, data storage, etc. In this embodiment, the overlay measurement unit 7300 at least includes an overlay mark image processing module 7310, another machine learning module 7320, and another measurement module 7330, where the overlay mark image processing module 7310 processes the overlay mark image by a machine learning method, reconstructs a missing pattern, reconstructs a blurred pattern, and eliminates particle influence, thereby improving stability and accuracy of overlay measurement; other machine learning module 7320 is used to perform other machine learning tasks, such as processing, classification, prediction, etc. of data; other measurement modules 7330 are used to perform other overlay measurement functions.
Example 8,
As a more preferable scheme than embodiment 7, as shown in fig. 14, the present embodiment is different from embodiment 7 in that an online model training module is added to the overlay measurement unit, specifically:
the overlay measurement system 8000 includes at least a mechanical transmission unit 8100, a data acquisition unit 8200, an overlay measurement unit 8300, a data output unit 8400, and other units 8500. The mechanical transmission unit 8100 is at least used for conveying and positioning wafers; the data acquisition unit 8200 is used for acquiring and receiving various types of data required by overlay measurement; the input end of the overlay measurement unit 8300 is connected with the output end of the data acquisition unit 8200 and is used for processing or predicting overlay measurement data by a machine learning method; the input end of the data output unit 8400 is connected with the output end of the overlay measurement unit 8300 and is used for outputting the overlay measurement result to related hardware equipment and a software system; the other unit 8500 is responsible for other functions of the system, such as communication, data storage, etc. In this embodiment, the overlay measurement unit 8300 at least includes an overlay mark image processing module 8310, an online model training module 8320, another machine learning module 8330, and another measurement module 8340, where the overlay mark image processing module 8310 processes the overlay mark image by a machine learning method, reconstructs a missing pattern, reconstructs a blurred pattern, and eliminates particle influence, thereby improving stability and precision of overlay measurement; the online model training module 8320 trains and updates the overlay mark image processing model by using the newly added data, and deploys the updated model to the overlay mark image processing module 8310, so that the effectiveness of the overlay measurement data prediction can be ensured under the condition of process drift; the other machine learning module 8330 is configured to perform other machine learning tasks, such as processing, classifying, predicting, etc., of data; the other measurement module 8340 is configured to perform other overlay measurement functions.
Examples 9,
The overlay measurement system in the integrated circuit chip manufacturing process according to this embodiment is used for predicting overlay measurement data, as shown in fig. 15, and specifically includes:
the overlay measurement system 9000 comprises at least a mechanical transmission unit 9100, a data acquisition unit 9200, an overlay measurement unit 9300, a data output unit 9400 and other units 9500. The mechanical transmission unit 9100 is at least used for conveying and positioning wafers; the data acquisition unit 9200 is used for acquiring and receiving various types of data required by overlay measurement; the input end of the iterative measurement unit 9300 is connected with the output end of the data acquisition unit 9200 and is used for predicting iterative measurement data by a machine learning method; the input end of the data output unit 9400 is connected with the output end of the overlay measurement unit 9300, and is used for outputting the overlay measurement result to related hardware equipment and software systems; the other unit 9500 is responsible for other functions of the system, such as communication, data storage, etc. In this embodiment, the overlay measurement unit 9300 at least includes an overlay measurement data prediction module 9310, another machine learning module 9320, and another measurement module 9330, and the overlay measurement data prediction module 9310 predicts overlay measurement data of other unmeasured areas on the same wafer by using a machine learning method and overlay measurement data in an upper partial area of the wafer, so as to obtain overlay measurement data in the upper partial area of the wafer and the other unmeasured areas, thereby greatly shortening the measurement time and improving the measurement efficiency; the other machine learning module 9320 is used to perform other machine learning tasks, such as data processing, classification, prediction, etc.; the other measurement module 9330 is used to perform other overlay measurement functions.
Examples 10,
As a more preferable scheme than embodiment 9, as shown in fig. 16, the difference between this embodiment and embodiment 9 is that an online model training module is added to the overlay measurement unit, specifically:
the overlay measurement system 10000 at least comprises a mechanical transmission unit 10100, a data acquisition unit 10200, an overlay measurement unit 10300, a data output unit 10400 and other units 10500. The mechanical transmission unit 10100 is at least used for conveying and positioning wafers; the data acquisition unit 10200 is used for collecting and receiving various types of data required for overlay measurement; the input end of the overlay measurement unit 10300 is connected to the output end of the data acquisition unit 10200, and is configured to predict overlay measurement data by a machine learning method; the input end of the data output unit 10400 is connected to the output end of the overlay measurement unit 10300, and is configured to output an overlay measurement result to a related hardware device and a software system; the other unit 10500 is responsible for other functions of the system, such as communication, data storage, etc. In this embodiment, the overlay measurement unit 10300 at least includes an overlay measurement data prediction module 10310, an online model training module 10320, another machine learning module 10330, and another measurement module 10340, where the overlay measurement data prediction module 10310 predicts overlay measurement data of other unmeasured regions on the same wafer by using a machine learning method and the overlay measurement data in the upper partial region of the wafer, so as to obtain overlay measurement data in the upper partial region of the wafer and the other unmeasured regions, thereby greatly shortening the measurement time and improving the measurement efficiency; the online model training module 10320 trains and updates the iterative measured data prediction model by using the newly added data, and deploys the updated model to the iterative measured data prediction module 10310, so that the effectiveness of iterative measured data prediction can be ensured under the condition of process drift; other machine learning modules 10330 are used to perform other machine learning tasks, such as processing, classification, prediction, etc. of data; other measurement modules 10340 are used to perform other overlay measurement functions.
The foregoing merely illustrates the principles and preferred embodiments of the invention and many variations and modifications may be made by those skilled in the art in light of the foregoing description, which are within the scope of the invention.
Claims (7)
1. An overlay measurement method in the manufacturing process of an integrated circuit chip, comprising the steps of:
acquiring the data of each overlay mark image on the wafer or acquiring overlay measurement data in a partial area on the wafer aiming at each process level of the chip manufacturing process;
respectively establishing a machine learning model by utilizing the collected overlapping marked image data or the overlapping measurement data: an overlay mark image processing model and an overlay measurement data prediction model;
during the overlay measurement, processing the overlay mark image through the established overlay mark image processing model, or predicting the overlay measurement data in other unmeasured areas on the same wafer through the established overlay measurement data prediction model and the overlay measurement data collected in the upper area of the wafer;
wherein, the step of processing the overlay mark image data by the overlay mark image processing model comprises the following steps:
collecting input data of the overlay measurement;
preprocessing the input data;
inputting the preprocessed data into an overlay marked image processing model, wherein the overlay marked image processing model processes an overlay marked image by a machine learning method, and the processing comprises at least one of reconstructing a missing graph, reconstructing a blurred graph and eliminating particle influence;
the processed overlay mark image is used for calculating overlay measurement data to obtain final overlay measurement data;
the step of predicting the overlay measurement data by the overlay measurement data prediction model comprises the following steps:
collecting input data of overlapping measurement in partial area on the wafer;
preprocessing the input data;
the preprocessed data are input into an overlay measurement data prediction model, the overlay measurement data prediction model predicts overlay measurement data in other unmeasured areas on the same wafer by a machine learning method by using the preprocessed overlay measurement data obtained in the upper area of the wafer, and final overlay measurement data are obtained.
2. The method of claim 1, wherein the step of building a machine learning model comprises the steps of:
collecting or generating input data required for establishing a machine learning model;
preprocessing the input data;
the preprocessed data is divided into two parts: one part is model training and verification data, and the other part is test data;
performing model training and verification through model training and verification data, and performing model evaluation through test data;
judging the evaluation result of the model: if the evaluation is qualified, model deployment is carried out to deliver the model for use; otherwise, returning to the previous step until the evaluation is qualified.
3. The method of claim 2, wherein the step of processing the overlay marker image data by the overlay marker image processing model further comprises:
collecting input data of the overlay measurement;
preprocessing the input data;
inputting the preprocessed data into an overlay marked image processing model, wherein the overlay marked image processing model processes an overlay marked image by a machine learning method, and the processing comprises at least one of reconstructing a missing graph, reconstructing a blurred graph and eliminating particle influence;
the processed overlay mark image is used for calculating overlay measurement data to obtain new overlay measurement data;
after new preprocessed data and new overlay measurement data are obtained, part or all of the new preprocessed data and part or all of the new overlay measurement data are used for on-line model training, and a new model generated after the on-line model training is used for updating an overlay labeled image processing model through model deployment, so that the overlay labeled image processing model is continuously updated according to the new data.
4. The method of claim 2, wherein the step of predicting the overlay measurement data by the overlay measurement data prediction model further comprises the steps of:
collecting input data of overlapping measurement in partial area on the wafer;
preprocessing the input data;
inputting the preprocessed data into an overlay measurement data prediction model, wherein the overlay measurement data prediction model predicts overlay measurement data in other unmeasured areas on the same wafer by using the preprocessed overlay measurement data obtained in the upper area of the wafer through a machine learning method to obtain new overlay measurement data;
after new preprocessed data and new overlay measurement data are obtained, part or all of the new preprocessed data and part or all of the new overlay measurement data are used for on-line model training, and a new model generated after the on-line model training is used for updating an overlay measurement data prediction model through model deployment, so that the overlay measurement data prediction model is continuously updated according to the new data.
5. An overlay measurement system in the manufacture of integrated circuit chips, comprising:
the mechanical transmission unit is at least used for transmitting and positioning the wafer;
the data acquisition unit is used for acquiring and receiving data required by the overlay measurement;
the input end of the overlay measuring unit is connected with the output end of the data acquisition unit and is used for processing or predicting overlay measuring data by a machine learning method; and the number of the first and second groups,
the input end of the data output unit is connected with the output end of the overlay measurement unit and used for outputting the overlay measurement result;
the overlay measurement unit at least comprises an overlay mark image processing module, the overlay mark image processing module processes an overlay mark image by a machine learning method, and the processing comprises at least one of reconstructing a missing graph, reconstructing a blurred graph and eliminating particle influence; or the like, or, alternatively,
the overlay measurement unit at least comprises an overlay measurement data prediction module which predicts overlay measurement data of other unmeasured areas on the same wafer through a machine learning method and the overlay measurement data in the area above the wafer.
6. The system of claim 5, wherein the overlay measurement unit comprises at least an overlay marker image processing module that processes overlay marker images by a machine learning method and an online model training module that trains and updates an overlay marker image processing model for overlay marker image processing with the newly added data and deploys the updated model to the overlay marker image processing module.
7. The system of claim 5, wherein the overlay measurement unit comprises at least an overlay measurement data prediction module and an online model training module, the overlay measurement data prediction module predicts overlay measurement data of other unmeasured areas on the same wafer through a machine learning method and overlay measurement data in an upper area of the wafer, the online model training module trains and updates an overlay measurement data prediction model for the overlay measurement data prediction by using newly added data, and deploys the updated model to the overlay measurement data prediction module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110390067.0A CN113158610B (en) | 2021-04-12 | 2021-04-12 | Overlay measurement method and system in integrated circuit chip manufacturing process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110390067.0A CN113158610B (en) | 2021-04-12 | 2021-04-12 | Overlay measurement method and system in integrated circuit chip manufacturing process |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113158610A CN113158610A (en) | 2021-07-23 |
| CN113158610B true CN113158610B (en) | 2022-04-01 |
Family
ID=76889990
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110390067.0A Active CN113158610B (en) | 2021-04-12 | 2021-04-12 | Overlay measurement method and system in integrated circuit chip manufacturing process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113158610B (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100790826B1 (en) * | 2006-06-30 | 2008-01-02 | 삼성전자주식회사 | Method for measuring overlay and management system of semiconductor manufacturing equipment at the same |
| CN101369548B (en) * | 2007-08-14 | 2010-09-29 | 中芯国际集成电路制造(上海)有限公司 | Method for detecting whether present layer aligning with anterior layer of chip |
| CN104183573B (en) * | 2013-05-24 | 2018-11-30 | 华邦电子股份有限公司 | Repeatedly to label and its manufacturing method |
| US10430719B2 (en) * | 2014-11-25 | 2019-10-01 | Stream Mosaic, Inc. | Process control techniques for semiconductor manufacturing processes |
| CA3042819A1 (en) * | 2018-05-09 | 2019-11-09 | Postureco, Inc. | Method and system for postural analysis and measuring anatomical dimensions from a digital image using machine learning |
-
2021
- 2021-04-12 CN CN202110390067.0A patent/CN113158610B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113158610A (en) | 2021-07-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11520238B2 (en) | Optimizing an apparatus for multi-stage processing of product units | |
| CN107004060B (en) | Improved process control techniques for semiconductor manufacturing processes | |
| US10234401B2 (en) | Method of manufacturing semiconductor devices by using sampling plans | |
| US6535774B1 (en) | Incorporation of critical dimension measurements as disturbances to lithography overlay run to run controller | |
| CN110648305B (en) | Industrial image detection method, system and computer readable recording medium | |
| CN108475351A (en) | The acceleration training of the model based on machine learning for semiconductor application | |
| US11094057B2 (en) | Semiconductor wafer measurement method and system | |
| TW201828335A (en) | Method and apparatus for image analysis | |
| KR20100135846A (en) | Method for classifying defects, computer storage medium, and device for classifying defects | |
| CN108369915A (en) | Reduce the noise examined in the bare die caused near registration and design | |
| KR20220019717A (en) | Board inspection apparatus, board inspection system and board inspection method | |
| KR20220126761A (en) | Determination technique of lithography matching performance | |
| CN113168116A (en) | Method for determining root causes affecting yield in semiconductor manufacturing processes | |
| US11144702B2 (en) | Methods and systems for wafer image generation | |
| KR20210135416A (en) | Automatic selection of algorithmic modules for examination of a specimen | |
| CN113158610B (en) | Overlay measurement method and system in integrated circuit chip manufacturing process | |
| US7342643B2 (en) | Method of aligning wafer using database constructed of alignment data in a photolithography process | |
| JPH07504978A (en) | Product inspection method | |
| CN117518747B (en) | Method, device, equipment and storage medium for generating photoetching measurement intensity | |
| CN110741466B (en) | Broadband plasma verification based on a perturbed dot pattern | |
| KR20230075369A (en) | Mask inspection for semiconductor specimen fabrication | |
| KR20230149223A (en) | Lithography method, article manufacturing method, information processing method, program, and information processing apparatus | |
| CN117529803A (en) | Manufacturing fingerprints for active yield management | |
| CN116057682A (en) | System and method for determining measurement position in semiconductor wafer metrology | |
| CN111052326A (en) | Inspection-guided critical site selection for critical dimension measurement |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |