CN111276411A - Method for identifying wafer particles, electronic device and computer readable recording medium - Google Patents
Method for identifying wafer particles, electronic device and computer readable recording medium Download PDFInfo
- Publication number
- CN111276411A CN111276411A CN201811472662.3A CN201811472662A CN111276411A CN 111276411 A CN111276411 A CN 111276411A CN 201811472662 A CN201811472662 A CN 201811472662A CN 111276411 A CN111276411 A CN 111276411A
- Authority
- CN
- China
- Prior art keywords
- particle
- suspected
- wafer
- information
- determined
- 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.)
- Granted
Links
- 239000002245 particle Substances 0.000 title claims abstract description 299
- 238000000034 method Methods 0.000 title claims abstract description 71
- 239000004065 semiconductor Substances 0.000 claims abstract description 35
- 238000013136 deep learning model Methods 0.000 claims abstract description 21
- 239000000463 material Substances 0.000 claims abstract description 17
- 238000004519 manufacturing process Methods 0.000 claims abstract description 15
- 238000004458 analytical method Methods 0.000 claims description 40
- 230000007547 defect Effects 0.000 claims description 30
- 238000000724 energy-dispersive X-ray spectrum Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 4
- 238000000921 elemental analysis Methods 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 2
- 238000001000 micrograph Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 description 13
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 239000000203 mixture Substances 0.000 description 4
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 3
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 description 3
- 229910052799 carbon Inorganic materials 0.000 description 3
- 238000001878 scanning electron micrograph Methods 0.000 description 3
- 229910052710 silicon Inorganic materials 0.000 description 3
- 239000010703 silicon Substances 0.000 description 3
- 229910052742 iron Inorganic materials 0.000 description 2
- 238000012827 research and development Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
- H01L22/00—Testing or measuring during manufacture or treatment; Reliability measurements, i.e. testing of parts without further processing to modify the parts as such; Structural arrangements therefor
- H01L22/10—Measuring as part of the manufacturing process
- H01L22/12—Measuring as part of the manufacturing process for structural parameters, e.g. thickness, line width, refractive index, temperature, warp, bond strength, defects, optical inspection, electrical measurement of structural dimensions, metallurgic measurement of diffusions
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
- H01L22/00—Testing or measuring during manufacture or treatment; Reliability measurements, i.e. testing of parts without further processing to modify the parts as such; Structural arrangements therefor
- H01L22/20—Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps
Abstract
The present disclosure provides a method, an electronic device and a computer-readable recording medium for identifying wafer particles. The method comprises the following steps: obtaining a plurality of suspected particle information on a wafer after a semiconductor process, wherein the suspected particle information respectively corresponds to a plurality of suspected particles on the wafer; identifying whether the suspected particles corresponding to the suspected particle information are at least one judged particle on the wafer according to a deep learning model; classifying the material of the at least one determined particle according to a plurality of particle information corresponding to the at least one determined particle, and generating a classification result corresponding to the at least one determined particle; and presenting the classification result of the at least one determined particle to improve the semiconductor manufacturing process.
Description
Technical Field
Embodiments of the present disclosure relate to semiconductor processing, and more particularly, to a method, an electronic device and a computer-readable recording medium for identifying wafer particles.
Background
During a semiconductor process using a tool, particles (particles) may fall off from the tool onto a wafer, or the wafer may be contaminated by a certain process during the semiconductor process, thereby affecting the yield of the wafer. Therefore, developers typically monitor the amount of particles on a wafer after semiconductor processing. When the amount of particles is detected to be excessive and exceed the Out of Control (OOC) limit, in order to know which program the source of the particles is contaminated by or the related parts of which program are dropped, a developer performs the related program of the semiconductor manufacturing process by using a Control Wafer (CW), and captures and analyzes the defects (e.g., particles, pits, etc.) on the processed CW to obtain a high-resolution image of the defects corresponding to the element components. Then, the researchers use manpower to read the causes of these defects one by one to adjust the related procedures in the semiconductor process, so as to reduce the number of particles on the processed wafer.
However, the noise of the high power microscope or the many depressions on the monitor wafer itself can be considered as a defect on the monitor wafer. These drawbacks are not related to the particle cause required by the developers, but consume the workload of the developers and take the developers' working hours. On the other hand, the judgment of these defects is performed empirically by subjective judgment of developers, and thus, there is no judgment standard for consistency, and it is difficult to perform normalization through the number of historical events.
Disclosure of Invention
The method for identifying wafer particles in the embodiment of the disclosure comprises the following steps: obtaining a plurality of suspected particle information on a wafer after a semiconductor process, wherein the suspected particle information respectively corresponds to a plurality of suspected particles on the wafer; identifying whether the suspected particles corresponding to the suspected particle information are at least one determined particle on the wafer according to a deep learning model; classifying the material of the at least one determined particle according to a plurality of particle information corresponding to the at least one determined particle, and generating a classification result corresponding to the at least one determined particle; and presenting the classification result of the at least one determined particle to improve the semiconductor manufacturing process.
The electronic device of the embodiment of the disclosure comprises a sensor, a processor and a display. The sensor is used for sensing defects on a wafer after a semiconductor manufacturing process so as to obtain a plurality of suspected particle information. A processor is coupled to the sensor. And a processor obtains the suspected particle information on the wafer, wherein the suspected particle information respectively corresponds to a plurality of suspected particles on the wafer. The processor identifies whether the suspected particles corresponding to the suspected particle information are at least one determined particle on the wafer according to a deep learning model, classifies the material of the at least one determined particle according to a plurality of particle information corresponding to the at least one determined particle, and generates a classification result corresponding to the at least one determined particle. A display is coupled to the processor. A display is used to display the classification result of the at least one determined particle.
The non-volatile computer readable recording medium of the embodiments of the present disclosure is used to record a computer program, which executes the method for identifying wafer particles according to the embodiments of the present disclosure.
Drawings
The accompanying drawings are included to provide a further understanding of the disclosure, and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the disclosure and together with the description serve to explain the principles of the disclosure.
FIG. 1 is a block diagram of an electronic device for implementing a method for identifying wafer grains according to an embodiment of the present disclosure;
FIG. 2 is a flowchart illustrating a method for identifying wafer particles according to an embodiment of the present disclosure;
FIG. 3 is a diagram illustrating the amount of information for the EDX spectrum reduction in step S250 of FIG. 2;
fig. 4 is a detailed flowchart of step S250 in fig. 2.
Description of the reference numerals
100: an electronic device;
110: a sensor;
120: a processor;
130: a display;
s210 to S270, S410 to S450: a step of;
CW: a control wafer;
PL: wafer particles;
c: an elemental carbon;
fe: elemental iron;
si: elemental silicon;
LB 1: a baseline;
PK 1-PK 3: wave crest;
RF 1-RF 3: reference points.
Detailed Description
Reference will now be made in detail to the exemplary embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings and the description to refer to the same or like parts.
Fig. 1 is a block diagram of an electronic device 100 for implementing a method for identifying wafer grains according to an embodiment of the disclosure. The electronic device 100 generally includes a sensor 110, a processor 120, and a display 130. The developer can utilize the monitor wafer CW to perform related procedures of the semiconductor manufacturing process, and capture and analyze the defects (e.g., wafer particles PL, monitor wafer pits, etc.) on the processed monitor wafer by the electronic device 100 to obtain high-resolution images of the defects corresponding to the elemental compositions. The present embodiment can be applied to semiconductor processes of various processes (e.g., 20 nm, 16 nm, 10 nm, etc.), and therefore, the application range of the semiconductor process is not limited.
In detail, the sensor 110 may include a Scanning Electron Microscope (SEM) and a particle element analysis apparatus. The scanning electron microscope captures the image of the defect on the control wafer to obtain an SEM image, and the particle element analysis equipment can analyze the components of the defect to obtain an energy-dispersive X-ray (EDX) spectrum, so as to obtain the element components corresponding to the defects. The sensor 110 may be implemented by at least one dedicated particle analyzer station.
After obtaining the suspected particle information of each defect (the suspected particle information includes the SEM image and the EDX spectrum of each defect), the processor 120 may perform the method for identifying wafer particles PL according to the embodiment of the disclosure, and display or display the identification result of these defects in the display 130 for the researchers to refer to. However, there are many types of defects, common being: the noise present in the SEM itself causes image blurring defects, pits in the wafer, and wafer particles. Many of the above-mentioned defects (e.g., noise from SEM and dishing in the wafer) are not associated with wafer particles. However, defects that are not related to wafer particles are difficult to distinguish through the processor 120 itself, so that the developer needs to take time to subjectively judge the defects in a manual manner, and the accuracy of such judgment is very related to the experience of the developer, so that a consistent judgment standard is lacked. In addition, the developer often ignores the historical adjustment factors, processing recipes, and related information of the tools in the semiconductor manufacturing process, resulting in poor efficiency of the developer in processing the wafer particles.
Embodiments of the present disclosure utilize a combination of deep learning and particle classification (clustering approach) to more easily identify noise, wafer dishing and true wafer particles in these defects. In some embodiments, the processor 120 may also retrieve historical processing information from the defect recognition results and display the historical processing information on the display 130, or may be authorized by a researcher to allow the processor 120 to directly readjust parameters of a part of the semiconductor manufacturing process, or replace a part of the semiconductor manufacturing process that may cause wafer particles.
Fig. 2 is a flowchart illustrating a method for identifying wafer particles according to an embodiment of the disclosure. In step S210, the processor 120 obtains a plurality of suspected particle information on a wafer (e.g., a wafer control wafer) after a semiconductor process through the sensor 110. The suspected particle information respectively corresponds to a plurality of suspected particles on the wafer. The suspected particles are defects sensed from the wafer (wafer CW) by the sensor 110. The particle image information (e.g., SEM image) and particle elemental analysis information (e.g., EDX spectrum) are included in the corresponding particle information for each particle.
In step S220, the processor 120 identifies whether the suspected particle corresponding to the suspected particle information is at least one determined particle on the wafer according to the deep learning model. The deep learning model of this embodiment is implemented by a RestNet convolutional layer neural network (CNN) having 12 layers. Historical particle information and corresponding historical recognition results (e.g., noise patterns) collected in the past can be used as a basis for building a deep learning model. In other words, the deep learning model of the present embodiment is mainly used to identify the SEM noise and wafer recess that do not need to be distinguished from the defects on the wafer, and the wafer particles required by the method. The deep learning model mainly identifies wafer particles in the defect through particle image information in the suspected particle information. The deep learning model of the embodiment also generates a plurality of corresponding residual layers (residual layers) according to the requirement of the model design. These residual layers can prevent over-adaptation (over-fitting), shorten model setup time, and make the learning model deeper.
In this embodiment, before the step S220, the boundary characteristics of the particle image information in the suspected particle information may be enhanced. For example, the processor may perform image processing on the grain image information using a Sobel operator (Sobel operator), so that the deep learning model may more easily detect the boundary of the grain image information. Specifically, the particle image information may be subjected to Sobel-vertical (Sobel-vertical) processing and Sobel-vertical (Sobel-horizontal) processing, respectively, and the image subjected to the Sobel-vertical processing, the original image, and the image subjected to the Sobel-vertical processing may be integrated to obtain the boundary-enhanced particle image information. The deep learning model judges whether the suspected particles are judged particles or not by utilizing the image information of the particles after the boundary strengthening.
When the processor 120 identifies that the suspected particle corresponding to the suspected particle information is not the determined particle on the wafer, it indicates that the suspected particle may be the noise of the SEM itself or a recess on the control wafer. Therefore, from step S220 to step S230, the processor determines whether the suspected particle is the last suspected particle. If the suspected particle is not the last suspected particle, the process proceeds from step S230 to step S240, and the processor 120 obtains the next suspected particle and the image information of the particle corresponding to the next suspected particle to perform step S220. On the contrary, if the suspected particle is the last suspected particle, the method proceeds to step S250 to continue the method.
When the processor 120 identifies that the suspected particle corresponding to the suspected particle information is a determined particle on the wafer, the process proceeds from step S220 to step S250, and the processor reduces the information amount of the particle element analysis information in the particle information corresponding to the determined particle by using an asymmetric least squares (asymmetric mean square) smoothing technique and peak detection based on continuous wavelet transform. The particle element analysis information in the present embodiment is presented by the EDX spectrum, but the EDX spectrum is generally presented by a curve and has an excessive amount of information, but such an excessive amount of data is not necessary when classifying the material of the determined particle. Fig. 3 is a diagram illustrating the amount of information for EDX spectrum reduction in step S250 in fig. 2. As shown in fig. 3, part (a) of fig. 3 is the original EDX spectrum. From the EDX spectrum, it is found that the elemental composition of the particles may be a mixture of carbon (C), iron (Fe), and silicon (Si). The processor will find the baseline LB1 in the EDX spectrum and subtract the baseline LB1 from the EDX spectrum to become the EDX spectrum presented in part (b) of fig. 3 to remove the effects of noise. The processor then finds peaks, e.g., three peaks PK 1-PK 3, in the EDX spectrum of section (b) of FIG. 3, as shown in section (c) of FIG. 3. Each peak PK 1-PK 3 corresponds to different element components, for example, peak PK1 corresponds to element carbon, peak PK2 corresponds to element iron, and peak PK3 corresponds to element silicon. Then, the processor takes a predetermined number (e.g., 10) of points in the vicinity of each peak in the EDX spectrum as a plurality of reference points RF1 to RF3, as shown in part (d) of fig. 3, and uses these reference points RF1 to RF3 as data-simplified particle element analysis information corresponding to the determined particles.
In step S260, the processor classifies the material of the determined particles according to the particle element analysis information (e.g., the particle element analysis information simplified by the data in step S250) in the particle information corresponding to the determined particles, and generates a classification result corresponding to the determined particles. The processor of the present embodiment performs classification of the determined particles using the particle element analysis information of the determined particles as a basis. This embodiment takes a hybrid nested classification (hybrid clustering) as an example of how to classify these judged particles, and the hybrid nested classification is detailed in the detailed flow of fig. 4.
Fig. 4 is a detailed flowchart of step S250 in fig. 2. In step S410, the processor randomly selects two of the determined grains as two initial groups, and sets group profiles (profiles) corresponding to the two initial groups as grain element analysis information corresponding to the randomly selected two determined grains. In step S420, the processor randomly selects one of the unselected determined particles as a selected particle. In step S430, the processor compares the particle element analysis information corresponding to the selected particle with the group profiles corresponding to the plurality of groups (including the initial group) to determine whether the particle element analysis information corresponding to the selected particle is close to one of the group profiles.
When the particle element analysis information corresponding to the selected particle is determined to be close to one of the group profiles, proceeding from step S430 to step S440, the processor sets the selected particle in the group corresponding to the one of the close group profiles. From step S440, the processor then returns to step S420 to continue the classification process for the other determined particles. In contrast, when it is determined that the particle element analysis information corresponding to the selected particle is not close to one of the group profiles, the process proceeds from step S430 to step S450, the processor sets up another group, sets the selected particle in the newly set up another group, and sets the group profile corresponding to the another group as the particle element analysis information corresponding to the selected particle. From step S450, the processor then returns to step S420 to continue the classification process for the other determined particles.
After all the determined particles have been classified in steps S410 to S450 of fig. 4, the processor presents the classification result of the determined particles in step S270 to improve the semiconductor manufacturing process. In detail, the processor can integrate the groups and the corresponding determined particles to form diversified graphs and data as the classification results of the determined particles, and the classification results are displayed on the screen through the display, so that the research and development personnel can intuitively use the classification results to judge the factors in the semiconductor process to directly or indirectly cause the generation of the wafer particles, thereby improving part of the process in the semiconductor process. For example, when most defects do not belong to wafer particles, developers may improve wafer preparation procedures or modify monitoring techniques during an off-line semiconductor process. Alternatively, the classification result of the wafer particles is used to compare the replacement history of tools or parts used in the machine during the off-line semiconductor process, and the tools or parts are replaced to prevent more wafer particles from being generated.
From the perspective of research and development personnel, the method according to the embodiment of the disclosure can eliminate SEM noise and depression of the monitor wafer among defects on the monitor wafer by using the deep learning module, thereby reducing the workload by about 15% compared with the workload of conventional wafer particle recognition, and easily obtaining the material of the wafer particles from the classification result. For example, it can be determined whether the material of the wafer particles is similar to that of the control wafer, and further, whether the wafer particles are particles generated by peeling off the parts in the semiconductor process. In summary, the method of the embodiment of the present disclosure can reduce the workload of the developer in wafer grain recognition by about 10% to 30%. The method for identifying wafer particles in the embodiment of the disclosure belongs to one of big data analysis, and carries out automatic interpretation of particle events in a semiconductor process, thereby reducing the idle time of the semiconductor process under an on-line state. The embodiment of the present disclosure may also present the computer program of the method for identifying wafer particles and record the computer program in a non-volatile computer readable recording medium.
In summary, the method, the electronic device and the non-volatile computer readable recording medium for identifying wafer particles in the embodiments of the disclosure utilize a combination of a deep learning module and a particle classification method to identify defects on a control wafer. In other words, the embodiments of the present disclosure identify the noise, the wafer dishing and the wafer grain in the defects through the deep learning module, and perform the elemental composition analysis on the identified wafer grain to classify the grain, thereby reducing the workload of the developers and shortening the standby time of the related equipment using the semiconductor process.
The embodiment of the disclosure discloses a method for identifying wafer particles, which includes: obtaining a plurality of suspected particle information on a wafer after a semiconductor process, wherein the suspected particle information respectively corresponds to a plurality of suspected particles on the wafer; identifying whether the suspected particles corresponding to the suspected particle information are at least one determined particle on the wafer according to a deep learning model; classifying the material of the at least one determined particle according to a plurality of particle information corresponding to the at least one determined particle, and generating a classification result corresponding to the at least one determined particle; and presenting the classification result of the at least one determined particle to improve the semiconductor manufacturing process.
In some embodiments, the suspected particle information or the particle information corresponding to the at least one determined particle includes particle image information and particle element analysis information.
In some embodiments, the particle image information is a scanning electron microscope image. The particle elemental analysis information is an energy dispersive X-ray spectrum. The deep learning model is realized by a convolution layer neural network.
In some embodiments, before the step of identifying whether the suspected particle corresponding to the suspected particle information is at least one determined particle on the wafer, the method further includes: and enhancing the boundary characteristics of the particle image information.
In some embodiments, the step of classifying the material of the at least one determined particle further comprises the following steps: the information amount of the grain element analysis information is reduced using an asymmetric least squares smoothing technique and peak detection based on continuous wavelet transform.
In some embodiments, the step of classifying the material of the at least one determined particle comprises: randomly selecting one of the at least one decided particle that is not selected as a selected particle; comparing the particle element analysis information corresponding to the selected particle with a plurality of group profiles (profiles) corresponding to a plurality of groups to determine whether the particle element analysis information corresponding to the selected particle is close to one of the group profiles; when the particle element analysis information corresponding to the selected particle is determined to be close to one of the group summaries, setting the selected particle in the group corresponding to the close one of the group summaries; and when the particle element analysis information corresponding to the selected particle is judged not to be close to one of the group summaries, setting up another group, setting the selected particle in the other group, and setting the group summary corresponding to the other group as the particle element analysis information corresponding to the selected particle.
In some embodiments, the step of classifying the material of the at least one determined particle further comprises: randomly selecting two of the at least one judged particle as the initial group, and setting the group summary corresponding to the initial group as the particle element analysis information corresponding to the randomly selected at least two judged particles.
In some embodiments, the method further comprises: collecting a plurality of historical particle information and a plurality of corresponding historical recognition results to build the deep learning model.
In some embodiments, the method further comprises: and presenting historical processing information related to the classification result according to the classification result.
In some embodiments, the method further comprises: and readjusting part of the procedures in the semiconductor manufacturing process according to the classification result, or replacing part of the parts in the semiconductor manufacturing process according to the classification result.
An embodiment of the present disclosure discloses an electronic device, which includes a sensor, a processor, and a display. The sensor is used for sensing defects on a wafer after a semiconductor manufacturing process so as to obtain a plurality of suspected particle information. A processor is coupled to the sensor. And a processor obtains the suspected particle information on the wafer, wherein the suspected particle information respectively corresponds to a plurality of suspected particles on the wafer. The processor identifies whether the suspected particles corresponding to the suspected particle information are at least one determined particle on the wafer according to a deep learning model, classifies the material of the at least one determined particle according to a plurality of particle information corresponding to the at least one determined particle, and generates a classification result corresponding to the at least one determined particle. A display is coupled to the processor. A display is used to display the classification result of the at least one determined particle.
In some embodiments, the suspected particle information or the particle information corresponding to the at least one determined particle includes particle image information and particle element analysis information.
In some embodiments, the particle image information is a sem image, the particle element analysis information is an energy dispersive X-ray spectrum, and the deep learning model is implemented using a convolutional layer neural network.
In some embodiments, the processor reduces the amount of information in the grain element analysis information using asymmetric least squares smoothing techniques and continuous wavelet transform based peak detection.
In some embodiments, the processor randomly selects one of the at least one determined particle that is not selected as a selected particle, compares the particle element analysis information corresponding to the selected particle with a plurality of group profiles corresponding to a plurality of groups to determine whether the particle element analysis information corresponding to the selected particle is close to one of the group profiles, sets the selected particle in the group corresponding to the close one of the group profiles when the particle element analysis information corresponding to the selected particle is close to one of the group profiles, and sets another group when the particle element analysis information corresponding to the selected particle is not close to one of the group profiles, setting the selected particle to be located in the other group, and setting a group summary corresponding to the other group as the particle element analysis information corresponding to the selected particle.
In some embodiments, the processor randomly selects two of the at least two determined particles as the initial group, and sets the group summary corresponding to the initial group as the particle element analysis information corresponding to the randomly selected at least two determined particles.
In some embodiments, the processor builds the deep learning model according to the collected historical particle information and the corresponding historical recognition results.
In some embodiments, the processor further controls the display to display historical processing information associated with the classification result according to the classification result.
In some embodiments, the processor further enhances boundary features of the particle image information before identifying whether the suspected particles corresponding to the suspected particle information are at least one determined particle on the wafer.
The present disclosure discloses a non-volatile computer readable recording medium, which records a computer program, wherein the computer program executes the method for identifying wafer particles as described above.
Finally, it should be noted that: the above embodiments are only used for illustrating the technical solutions of the present disclosure, and not for limiting the same; while the present disclosure has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the present disclosure.
Claims (10)
1. A method for identifying wafer particles, comprising:
obtaining a plurality of suspected particle information on a wafer after a semiconductor process, wherein the suspected particle information respectively corresponds to a plurality of suspected particles on the wafer;
identifying whether the suspected particles corresponding to the suspected particle information are at least one judged particle on the wafer according to a deep learning model;
classifying the material of the at least one determined particle according to a plurality of particle information corresponding to the at least one determined particle, and generating a classification result corresponding to the at least one determined particle; and
presenting the classification result of the at least one determined particle to improve the semiconductor manufacturing process.
2. The method of claim 1, wherein the particle information corresponding to the suspected particle information or the at least one determined particle comprises particle image information and particle element analysis information.
3. The method of claim 2, wherein the particle image information is scanning electron microscope images, the particle elemental analysis information is energy dispersive X-ray spectra, and the deep learning model is implemented with a convolutional layer-like neural network.
4. The method of claim 2, further comprising, before the step of identifying whether the suspected particle corresponding to the suspected particle information is at least one determined particle on the wafer:
and enhancing the boundary characteristics of the particle image information.
5. The method of claim 2, wherein the step of classifying the material of the at least one determined particle further comprises the steps of:
the information amount of the grain element analysis information is reduced using an asymmetric least squares smoothing technique and peak detection based on continuous wavelet transform.
6. The method of claim 2, the step of classifying the material of the at least one determined particle comprising:
randomly selecting one of the at least one judged particle that is not selected as a selected particle;
comparing the particle element analysis information corresponding to the selected particle with a plurality of group summaries corresponding to a plurality of groups to determine whether the particle element analysis information corresponding to the selected particle is close to one of the group summaries;
when the particle element analysis information corresponding to the selected particle is judged to be close to one of the group summaries, setting the selected particle in the group corresponding to the close one of the group summaries; and
when it is determined that the particle element analysis information corresponding to the selected particle is not close to one of the group profiles, another group is established, the selected particle is set to be located in the other group, and the group profile corresponding to the other group is set as the particle element analysis information corresponding to the selected particle.
7. The method of claim 6, the step of classifying the material of the at least one determined particle further comprising:
randomly selecting two of the at least one judged particle as the initial group, and setting the group summary corresponding to the initial group as the particle element analysis information corresponding to the randomly selected at least two judged particles.
8. The method of claim 1, further comprising:
collecting a plurality of historical particle information and a plurality of corresponding historical recognition results to build the deep learning model.
9. An electronic device, comprising:
the sensor is used for sensing the defects on the wafer after the semiconductor manufacturing process so as to obtain a plurality of suspected particle information;
a processor coupled to the sensor, wherein the processor obtains the suspected particle information on the wafer, the suspected particle information respectively corresponds to a plurality of suspected particles on the wafer,
the processor identifies whether the suspected particles corresponding to the suspected particle information are at least one determined particle on the wafer according to a deep learning model, classifies the material of the at least one determined particle according to a plurality of particle information corresponding to the at least one determined particle, and generates a classification result corresponding to the at least one determined particle; and
a display, coupled to the processor, for displaying the classification result of the at least one determined particle.
10. A non-volatile computer readable recording medium recording a computer program, wherein the computer program performs the method for identifying wafer particles according to claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811472662.3A CN111276411B (en) | 2018-12-04 | 2018-12-04 | Method for identifying wafer particles, electronic device and computer readable recording medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811472662.3A CN111276411B (en) | 2018-12-04 | 2018-12-04 | Method for identifying wafer particles, electronic device and computer readable recording medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111276411A true CN111276411A (en) | 2020-06-12 |
| CN111276411B CN111276411B (en) | 2024-01-26 |
Family
ID=71002865
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811472662.3A Active CN111276411B (en) | 2018-12-04 | 2018-12-04 | Method for identifying wafer particles, electronic device and computer readable recording medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111276411B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114077164A (en) * | 2020-08-21 | 2022-02-22 | 长鑫存储技术有限公司 | Semiconductor machine cleaning system and semiconductor machine cleaning method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030058436A1 (en) * | 2001-09-26 | 2003-03-27 | Makoto Ono | Inspection data analysis program, defect inspection apparatus, defect inspection system and method for semiconductor device |
| CN102759533A (en) * | 2011-04-27 | 2012-10-31 | 中国科学院微电子研究所 | Wafer detecting method and wafer detecting device |
| CN108108745A (en) * | 2016-11-23 | 2018-06-01 | 财团法人工业技术研究院 | Classification method, classification module and computer program product |
| CN108765389A (en) * | 2018-05-18 | 2018-11-06 | 浙江大学 | A kind of microcosmic wafer surface defects image detecting method |
-
2018
- 2018-12-04 CN CN201811472662.3A patent/CN111276411B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030058436A1 (en) * | 2001-09-26 | 2003-03-27 | Makoto Ono | Inspection data analysis program, defect inspection apparatus, defect inspection system and method for semiconductor device |
| CN102759533A (en) * | 2011-04-27 | 2012-10-31 | 中国科学院微电子研究所 | Wafer detecting method and wafer detecting device |
| CN108108745A (en) * | 2016-11-23 | 2018-06-01 | 财团法人工业技术研究院 | Classification method, classification module and computer program product |
| CN108765389A (en) * | 2018-05-18 | 2018-11-06 | 浙江大学 | A kind of microcosmic wafer surface defects image detecting method |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114077164A (en) * | 2020-08-21 | 2022-02-22 | 长鑫存储技术有限公司 | Semiconductor machine cleaning system and semiconductor machine cleaning method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111276411B (en) | 2024-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP1238367B1 (en) | Power assisted automatic supervised classifier creation tool for semiconductor defects | |
| JP7174117B2 (en) | Closed loop automated defect inspection and classification | |
| US7904195B2 (en) | Method for prognostic maintenance in semiconductor manufacturing equipments | |
| JP5030582B2 (en) | Apparatus and method for data analysis | |
| US6377898B1 (en) | Automatic defect classification comparator die selection system | |
| US7356177B2 (en) | Defect image classifying method and apparatus and a semiconductor device manufacturing process based on the method and apparatus | |
| JP6391083B2 (en) | Preparation and maintenance of classifiers in automatic defect classification | |
| CN102473660B (en) | Automatic fault detection and classification in a plasma processing system and methods thereof | |
| TWI776945B (en) | Method of classifying defects in a semiconductor specimen and system thereof | |
| US20080240544A1 (en) | System for creating an inspection recipe, system for reviewing defects, method for creating an inspection recipe and method for reviewing defects | |
| JP4443270B2 (en) | Defect classification method | |
| US20080298670A1 (en) | Method and its apparatus for reviewing defects | |
| US7646476B2 (en) | Process excursion detection | |
| JP5707291B2 (en) | Charged particle beam system that supports image classification | |
| JP4950946B2 (en) | Defect analysis apparatus and defect analysis method | |
| US6973390B2 (en) | Method and system for analyzing wafer yield against uses of a semiconductor tool | |
| CN111276411B (en) | Method for identifying wafer particles, electronic device and computer readable recording medium | |
| US6424881B1 (en) | Computer generated recipe selector utilizing defect file information | |
| TWI689722B (en) | Method for identifying dust particles on wafer, electronic device and computer readable recording media | |
| US6512842B1 (en) | Composition based association engine for image archival systems | |
| US6423557B1 (en) | ADC based in-situ destructive analysis selection and methodology therefor | |
| US8126257B2 (en) | Alignment of semiconductor wafer patterns by corresponding edge groups | |
| US6272393B1 (en) | Efficient tool utilization using previous scan data | |
| JP2005123566A (en) | Method of controlling defect | |
| Matsushita et al. | EES data analysis of lithography trouble detected by MapSSA |
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 |