TMBS device manufacturing method for real-time defect detection

Abstract

The invention relates to the technical field of semiconductor manufacturing, and provides a TMBS device manufacturing method for real-time defect detection, and the method comprises the steps: collecting temperature and humidity data in photoetching and etching processes in a TMBS substrate, and outputting a first real-time monitoring parameter value; acquiring a TMBS substrate surface image in real time by using a sensor, and outputting a second real-time monitoring parameter value; in the process of forming a semiconductor grid electrode on the TMBS substrate, a sensor is used for collecting a temperature distribution diagram in real time, and a third real-time monitoring parameter value is output; setting a process window for each step and constructing parameter tolerance, and when real-time monitoring parameter values exceed the corresponding parameter tolerance, immediately alarming and adjusting the parameters of the current step; and packaging the TMBS device. According to the invention, temperature, humidity and surface image data are collected in real time and are converted into images with orientation features, and real-time monitoring and parameter adjustment are carried out, so that the effect of real-time defect detection is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of semiconductor manufacturing, and in particular relates to a method for manufacturing a TMBS device for real-time defect detection. Background Art

[0002] The manufacturing process of semiconductor devices often includes multiple steps. A slight error in any process step may cause the failure of the semiconductor device. Therefore, during the preparation process, it should be ensured that each process step meets the process requirements.

[0003] Although sensor technology and real-time monitoring have been applied to a certain extent in the traditional TMBS device manufacturing process, they are often used to monitor and adjust a single parameter. The existing technology may be more sufficient in considering the parameter impact of a single step, but it cannot comprehensively consider the parameter impact of multiple steps.

[0004] For example, the temperature and humidity parameters of the lithography and etching processes affect the temperature distribution during semiconductor gate formation, which in turn affects the final device performance. Therefore, monitoring and adjusting parameters for each step individually cannot fully account for the complex impact of the entire manufacturing process. Summary of the Invention

[0005] In view of the above-mentioned defects of the prior art, the present invention proposes a method for manufacturing a TMBS device for real-time defect detection. The technical solution designed by the present invention includes the following steps:

[0006] S10: integrating a sensor in the TMBS substrate to collect temperature and humidity data during the photolithography and etching process, converting the temperature and humidity data into a pseudo-color image of the TMBS substrate with orientation features based on Cartesian coordinates and computer vision, and using the pseudo-color image of the TMBS substrate with orientation features as a first real-time monitoring parameter value;

[0007] S20: After photolithography and etching, using a sensor to collect a TMBS substrate surface image in real time, and outputting the TMBS substrate surface image with orientation features based on Cartesian coordinates and computer vision, wherein the TMBS substrate surface image with orientation features serves as a second real-time monitoring parameter value;

[0008] S30: During the process of forming the semiconductor gate on the TMBS substrate, a sensor is used to collect a temperature distribution map in real time, and a temperature distribution map with azimuthal features is output based on Cartesian coordinates and computer vision, wherein the temperature distribution map with azimuthal features serves as a third real-time monitoring parameter value;

[0009] S40: Set a process window for each step from S10 to S30. Based on historical data and performance targets, define the optimal parameter value for the real-time monitoring parameter value corresponding to each process window, and set upper and lower limits to build parameter tolerance. When the real-time monitoring parameter value exceeds the corresponding parameter tolerance, an alarm is immediately issued and the parameters of the current step are adjusted.

[0010] S50: packaging the TMBS device.

[0011] Preferably, the outputting of the TMBS base pseudo-color image with orientation features in S10 includes:

[0012] Establish Cartesian coordinates and define the positions of the temperature sensor, humidity sensor, and TMBS substrate. Convert the real-time collected temperature and humidity data into corresponding pseudo-color values. Use computer vision to map them to the Cartesian coordinates of the TMBS substrate, extract and annotate similar annotation points of color values, and output a TMBS substrate pseudo-color image with orientation features.

[0013] Preferably, the step of outputting the TMBS substrate surface image with orientation features in S20 includes:

[0014] A Cartesian coordinate system is established and the positions of the high-definition camera and the TMBS substrate are defined. Computer vision is used to extract similar annotation points in the real-time TMBS substrate surface image. The similar annotation points are annotated on the TMBS substrate surface image and combined with the camera position information. The coordinate information of the camera shooting position is superimposed on the TMBS substrate surface image, and a TMBS substrate surface image with orientation features is output.

[0015] Preferably, outputting the temperature distribution map with azimuthal characteristics in S30 includes:

[0016] A Cartesian coordinate system was established to define the positions of the temperature sensor and the TMBS substrate. The real-time collected temperature data was converted into corresponding pseudo-color values. Computer vision was used to map the data to the Cartesian coordinates of the TMBS substrate, and similar annotation points of the color values ​​were extracted and annotated to generate a temperature distribution map with azimuthal characteristics.

[0017] Preferably, the parameter tolerance in S40 is constructed as follows:

[0018]

[0019] Where, is the parameter tolerance, is the number of similar identification points, is the mth identification point of the real-time parameter, is the mth identification point of the optimal parameter value, is the sensor position weight, is the real-time sensor coordinate information, is the historical sensor coordinate information corresponding to the optimal parameter value.

[0020] Preferably, adjusting the parameters of the current step in S40 includes:

[0021] Define the target parameter value, and gradually approach the target parameter value and the optimal parameter value by controlling the real-time monitoring parameter value until the real-time monitoring parameter value is within the parameter tolerance range;

[0022] The formula for the real-time monitoring parameter value to gradually approach the target parameter value and the optimal parameter value is as follows:

[0023]

[0024] Where, To monitor the approach speed of parameter values ​​in real time, is the weight of the impact of the previous speed change on the current change, is the target parameter value, To monitor parameter values ​​in real time, is the optimal parameter value, for Towards The acceleration weight of the boost, for Towards The acceleration weight of the boost, is a real-time parameter, is the optimal parameter value, and For the value in A random number, Move the weight coefficient for the parameters of the current step;

[0025] The formula for the real-time monitoring parameter value to be within the parameter tolerance range is as follows:

[0026]

[0027]

[0028] Where, Real-time monitoring of parameter values ​​after gradually approaching the target parameter values ​​and optimal parameter values.

[0029] Preferably, the parameter shift weight coefficient of the current step includes the following steps:

[0030] S4001: Determine the numerical limit, discreteness, and variation of the real-time monitoring parameter value of the current step, and map the parameter of the current step to the numerical space;

[0031] S4002: Randomly select a value in the numerical space as a baseline sample and add a variation to generate a variation data set;

[0032] S4003: Calculate the impact of various parameters under the current benchmark sample on the output results of the benchmark sample and the changed data set;

[0033] S4004: Calculate the absolute mean and standard deviation of the impact, and take the maximum of the absolute mean and standard deviation as the weight coefficient of the current step .

[0034] Preferably, the numerical space of S4001 is as follows:

[0035]

[0036] Where, is the degree of discreteness, is the change amount, For dimension.

[0037] Preferably, the formula for the change data set in S4002 is as follows:

[0038]

[0039] Preferably, the influence of the output result is expressed as follows:

[0040]

[0041] Where, is the output result of the changing data set, The output result of the benchmark sample.

[0042] Beneficial effects: The present invention achieves the effect of real-time defect detection by collecting temperature, humidity and surface image data in real time and converting them into images with orientation features, and using computer vision technology for real-time monitoring and parameter adjustment. By timely discovering and adjusting production parameters, the manufacturing quality of TMBS devices can be improved, the generation of defective products can be reduced, and production efficiency and product reliability can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a preferred embodiment of the present invention;

[0044] Figure 2 It is a flow chart of calculating parameter movement weight coefficients according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0045] The embodiments of the present invention are described in detail below. The following embodiments are implemented based on the technical solutions of the present invention, and provide detailed implementation methods and specific operating procedures. However, the protection scope of the present invention is not limited to the following embodiments.

[0046] The present invention designs a TMBS device manufacturing method for real-time defect detection, and the technical solution includes the following steps: Figure 1 As shown, specifically including:

[0047] S10: integrating a sensor in the TMBS substrate to collect temperature and humidity data during the photolithography and etching process, converting the temperature and humidity data into a pseudo-color image of the TMBS substrate with orientation features based on Cartesian coordinates and computer vision, and using the pseudo-color image of the TMBS substrate with orientation features as a first real-time monitoring parameter value;

[0048] S20: After photolithography and etching, using a sensor to collect a TMBS substrate surface image in real time, outputting the TMBS substrate surface image with orientation features based on Cartesian coordinates and computer vision, and using the TMBS substrate surface image with orientation features as a second real-time monitoring parameter value;

[0049] S30: During the process of forming the semiconductor gate on the TMBS substrate, a sensor is used to collect a temperature distribution map in real time, and a temperature distribution map with azimuthal features is output based on Cartesian coordinates and computer vision. The temperature distribution map with azimuthal features is used as a third real-time monitoring parameter value;

[0050] S40: Set a process window for each step from S10 to S30. Based on historical data and performance targets, define the optimal parameter value for the real-time monitoring parameter value corresponding to each process window, and set upper and lower limits to build parameter tolerance. When the real-time monitoring parameter value exceeds the corresponding parameter tolerance, an alarm is immediately issued and the parameters of the current step are adjusted.

[0051] S50: packaging the TMBS device.

[0052] Preferably, outputting the TMBS base pseudo-color image with orientation features in S10 includes:

[0053] Establish Cartesian coordinates and define the positions of the temperature sensor, humidity sensor, and TMBS substrate. Convert the real-time collected temperature and humidity data into corresponding pseudo-color values. Use computer vision to map them to the Cartesian coordinates of the TMBS substrate, extract and annotate similar annotation points of color values, and output a TMBS substrate pseudo-color image with orientation features.

[0054] Preferably, outputting the TMBS substrate surface image with orientation features in S20 includes:

[0055] A Cartesian coordinate system is established and the positions of the high-definition camera and the TMBS substrate are defined. Computer vision is used to extract similar annotation points in the real-time TMBS substrate surface image. The similar annotation points are annotated on the TMBS substrate surface image and combined with the camera position information. The coordinate information of the camera shooting position is superimposed on the TMBS substrate surface image, and a TMBS substrate surface image with orientation features is output.

[0056] Preferably, outputting the temperature distribution map with azimuth characteristics in S30 includes:

[0057] A Cartesian coordinate system was established to define the positions of the temperature sensor and the TMBS substrate. The real-time collected temperature data was converted into corresponding pseudo-color values. Computer vision was used to map the data to the Cartesian coordinates of the TMBS substrate, and similar annotation points of the color values ​​were extracted and annotated to generate a temperature distribution map with azimuthal characteristics.

[0058] Preferably, the parameter tolerance in S40 is constructed as follows:

[0059]

[0060] Where, is the parameter tolerance, is the number of similar identification points, is the mth identification point of the real-time parameter, is the mth identification point of the optimal parameter value, is the sensor position weight, is the real-time sensor coordinate information, is the historical sensor coordinate information corresponding to the optimal parameter value.

[0061] Specifically, the formula for parameter tolerance is used to monitor the difference between the real-time parameter value and the optimal parameter value. The first part of the formula is to calculate the average value of the cosine similarity between the real-time parameter value and the optimal parameter value. This part can help us measure the similarity between the real-time parameter value and the optimal parameter value. If the average value of the cosine similarity is small, it means that the real-time parameter value is more different from the optimal parameter value; the second part of the formula is to calculate the Euclidean distance between the real-time parameter value and the optimal parameter value. This part takes into account the absolute difference in the parameter value, not just the difference in direction, which helps to more comprehensively evaluate the difference between the real-time parameter value and the optimal parameter value. When the direction is opposite, the cosine value of the angle is negative, and the formula handles these negative numbers by taking the absolute value and then dividing it by 2M. This operation ensures that when the real-time parameter value is in the opposite direction to the optimal parameter value, some information is still retained instead of being completely ignored; the formula also introduces weights The system comprehensively considers directional and absolute differences in parameter values, as well as differences in sensor coordinate information, to promptly detect parameter values ​​outside the tolerance range during real-time monitoring, triggering alarms and parameter adjustments. This helps ensure that the system always maintains optimal parameters during operation, improving operational efficiency and safety.

[0062] Preferably, adjusting the parameters of the current step in S40 includes:

[0063] Define the target parameter value, and gradually approach the target parameter value and the optimal parameter value by controlling the real-time monitoring parameter value until the real-time monitoring parameter value is within the parameter tolerance range;

[0064] The formula for real-time monitoring parameter values ​​gradually approaching the target parameter values ​​and optimal parameter values ​​is as follows:

[0065]

[0066] Where, To monitor the approach speed of parameter values ​​in real time, is the weight of the impact of the previous speed change on the current change, is the target parameter value, To monitor parameter values ​​in real time, is the optimal parameter value, for Towards The acceleration weight of the advance, for Towards The acceleration weight of the advance, is a real-time parameter, is the optimal parameter value, and For the value in A random number, Move the weight coefficient for the parameters of the current step;

[0067] The formula for real-time monitoring parameter values ​​within the parameter tolerance range is as follows:

[0068]

[0069]

[0070] Where, Real-time monitoring of parameter values ​​after gradually approaching the target parameter values ​​and optimal parameter values.

[0071] Specifically, in the traditional parameter optimization algorithm, each parameter is assigned the same random value during the speed update process. and , which means that the importance of all process parameters is considered equal. However, in semiconductor parameter adjustment, special attention should be paid to the parameters of key steps, that is, by assigning different weight coefficients to the parameters of each step. To achieve weighted optimization.

[0072] Preferably, the parameter shift weight coefficient of the current step is, such as Figure 2 As shown, the following steps are included:

[0073] S4001: Determine the numerical limit, discreteness, and variation of the real-time monitoring parameter value of the current step, and map the parameter of the current step to the numerical space;

[0074] S4002: Randomly select a value in the numerical space as a baseline sample and add a variation to generate a variation data set;

[0075] S4003: Calculate the impact of various parameters under the current benchmark sample on the output results of the benchmark sample and the changed data set;

[0076] S4004: Calculate the absolute mean and standard deviation of the impact, and take the maximum of the absolute mean and standard deviation as the weight coefficient of the current step .

[0077] Specifically, S4001 is to limit and adjust the parameters to ensure that they change within an acceptable range. At the same time, the parameters are mapped to the numerical space to prepare for the subsequent steps; S4002 is to generate a variable data set by randomly selecting a benchmark sample in the numerical space and adding a variation, so that different parameter value combinations can be obtained to simulate the parameter changes and provide input data for the subsequent steps. S4003 is to change the parameters and observe the changes in the output results to determine the degree of influence of each parameter on the system output. This evaluation can help us understand the importance and sensitivity of the parameters; S4004 is to select the maximum value as the weight coefficient; S4001-S4004 steps work together in the calculation process of the parameter movement weight coefficient. By evaluating and calculating the parameter range, variation, and impact, the weight coefficient is obtained to guide the parameter adjustment and optimization in the parameter tolerance step, which can help better understand the importance of each parameter, so as to make targeted adjustments and optimizations, thereby achieving the effect of improving performance and stability.

[0078] Preferably, the numerical space of S4001 is as follows:

[0079]

[0080] Where, is the degree of discreteness, is the change amount, For dimension.

[0081] Preferably, the formula for the change data set of S4002 is as follows:

[0082]

[0083] Preferably, the output result is expressed as follows:

[0084]

[0085] Where, is the output result of the changing data set, The output result of the benchmark sample.

[0086] Specifically, for and In the field of TMBS device manufacturing, process simulation software can be used as a model to simulate the design parameters and process parameters of TMBS devices. The trained model inputs the variable data set and the benchmark sample output results.

[0087] In addition, S4004 finally takes the absolute mean and the maximum standard deviation as the weight coefficient of the current step The formula is as follows:

[0088]

[0089]

[0090]

[0091] Where, is the absolute average of the impact, is the standard deviation of the effect.

[0092] The above describes in detail the preferred embodiments of the present invention. It should be understood that numerous modifications and variations based on the concepts of the present invention are possible by those skilled in the art without inventive effort. Therefore, any technical solution that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A method for manufacturing a TMBS device for real-time defect detection, characterized in that: include: S10: integrating a sensor in the TMBS substrate, collecting temperature and humidity data during the photolithography and etching process, converting the temperature and humidity data into a TMBS substrate pseudo-color image with orientation features based on Cartesian coordinates and computer vision, and outputting the TMBS substrate pseudo-color image with orientation features as a first real-time monitoring parameter value; S20: After photolithography and etching, using a sensor to collect a TMBS substrate surface image in real time, and outputting a TMBS substrate surface image with orientation features based on Cartesian coordinates and computer vision, wherein the TMBS substrate surface image with orientation features is used as a second real-time monitoring parameter value; S30: In the process of forming a semiconductor gate on the TMBS substrate, a sensor is used to collect a temperature distribution map in real time, and a temperature distribution map with azimuthal features is output based on Cartesian coordinates and computer vision, and the temperature distribution map with azimuthal features is used as a third real-time monitoring parameter value; S40: Set a process window for each step of S10-S30, define the optimal parameter value for the real-time monitoring parameter value corresponding to each process window based on historical data and performance targets, and set upper and lower limits to build parameter tolerance. When the real-time monitoring parameter value exceeds the corresponding parameter tolerance, an alarm is immediately issued and the parameters of the current step are adjusted; S50: packaging the TMBS device.

2. A method for manufacturing a TMBS device for real-time defect detection according to claim 1, characterized in that: The output of the TMBS base pseudo-color image with orientation features in S10 includes: Cartesian coordinates are established and the positions of temperature sensors, humidity sensors and TMBS substrates are defined. The real-time collected temperature and humidity data are converted into corresponding pseudo-color values, which are mapped to the Cartesian coordinates of the TMBS substrate using computer vision, and similar annotation points of color values ​​are extracted for annotation, and a TMBS substrate pseudo-color image with orientation features is output.

3. The method for manufacturing a TMBS device for real-time defect detection according to claim 1, characterized in that: The output of the TMBS substrate surface image with orientation features in S20 includes: A Cartesian coordinate system is established and the positions of the high-definition camera and the TMBS substrate are defined. The similar annotation points in the real-time TMBS substrate surface image are extracted using computer vision. The similar annotation points are annotated on the TMBS substrate surface image and combined with the camera position information. The coordinate information of the camera shooting position is superimposed on the TMBS substrate surface image, and a TMBS substrate surface image with orientation features is output.

4. The method for manufacturing a TMBS device for real-time defect detection according to claim 1, characterized in that: The output of the temperature distribution map with azimuth characteristics in S30 includes: A Cartesian coordinate system is established and the positions of the temperature sensor and the TMBS substrate are defined. The real-time collected temperature data are converted into corresponding pseudo-color values, which are mapped to the Cartesian coordinates of the TMBS substrate using computer vision, and similar annotation points of color values ​​are extracted for annotation to generate a temperature distribution map with azimuthal characteristics.

5. A method for manufacturing a TMBS device for real-time defect detection according to any one of claims 1 to 4, characterized in that: The construction parameter tolerance in S40 is as follows: ; In the formula, is the parameter tolerance, is the number of similar identification points, is the mth identification point of the real-time parameter, is the mth identification point of the optimal parameter value, is the sensor position weight, is the real-time sensor coordinate information, is the historical sensor coordinate information corresponding to the optimal parameter value.

6. The method for manufacturing a TMBS device for real-time defect detection according to claim 1, characterized in that: The parameters of the current step in S40 include: Define the target parameter value, and gradually approach the target parameter value and the optimal parameter value by controlling the real-time monitoring parameter value until the real-time monitoring parameter value is within the parameter tolerance range; The formula for the real-time monitoring parameter value to gradually approach the target parameter value and the optimal parameter value is as follows: ; In the formula, To monitor the approach speed of parameter values ​​in real time, is the weight of the impact of the previous speed change on the current change, is the target parameter value, To monitor parameter values ​​in real time, is the optimal parameter value, for Towards The acceleration weight of the boost, for Towards The acceleration weight of the boost, is a real-time parameter, is the optimal parameter value, and For the value in A random number, Move the weight coefficient for the parameters of the current step; The formula for the real-time monitoring parameter value to be within the parameter tolerance range is as follows: ; ; In the formula, To monitor parameter values ​​in real time after gradually approaching the target parameter values ​​and optimal parameter values.

7. A method for manufacturing a TMBS device for real-time defect detection according to claim 6, characterized in that: The parameter movement weight coefficient of the current step includes the following steps: S4001: Determine the numerical limit, discreteness and variation of the real-time monitoring parameter value of the current step, and map the parameter of the current step to the numerical space; S4002: randomly select a value in the numerical space as a benchmark sample and add a variation to generate a variation data set; S4003: Calculate the impact of various parameters under the current benchmark sample on the output results of the benchmark sample and the changed data set; S4004: Calculate the absolute mean value and standard deviation of the impact, and take the maximum value of the absolute mean value and standard deviation as the weight coefficient of the current step .

8. The method for manufacturing a TMBS device for real-time defect detection according to claim 7, characterized in that: The numerical space of S4001 is as follows: ; In the formula, is the degree of discreteness, is the change amount, For dimension.

9. The method for manufacturing a TMBS device for real-time defect detection according to claim 8, characterized in that: The formula for the change data set of S4002 is as follows:

10. The method for manufacturing a TMBS device for real-time defect detection according to claim 9, characterized in that: The impact of the output result is as follows: ; In the formula, is the output result of changing the data set, The output result of the benchmark sample.

Images