Method for evaluating ridge opening degree of czochralski silicon shouldering ridge

Abstract

The invention discloses a method for evaluating the ridge opening degree of a czochralski silicon shouldering ridge, and belongs to the field of czochralski silicon growth. The method comprises the following steps: (1) acquiring a shouldering image in the single crystal furnace in real time through an image acquisition system; (2) processing the image, determining the edge of the crystal bar, and respectively calculating the diameter Dmax of the maximum part of the shouldered crystal bar representing the arc and the diameter Dmin of the minimum part of the shouldered crystal bar representing the ridge after the ridge is opened; (3) calculating evaluation parameters, wherein the evaluation parameters at least comprise an edge opening degree delta D and an edge opening trend degree K; wherein the edge opening degree delta D is equal to Dmax-Dmin; the edge opening trend degree K is the ratio of the change rate of Dmax to the change rate of Dmin; and (4) comparing the edge opening degree delta D and the edge opening trend degree K with a preset threshold range, evaluating the state of the shouldering process based on a comparison result, and outputting evaluation information. By means of the method, edge breaking and line breaking can be effectively avoided, the shouldering success rate and production efficiency are improved, and dependence on experience of operators is reduced.

Description

Technical Field

[0001] This invention relates to a method for evaluating the degree of shoulder opening of Czochralski single-crystal silicon, belonging to the field of Czochralski single-crystal silicon growth. Background Technology

[0002] The Czochralski method, also known as the Czochralski method, involves placing high-purity polycrystalline silicon in a quartz crucible, heating and melting it in a single-crystal furnace, and then pulling it into a crystal rod using a seed crystal with a specific crystal orientation. It is the most important and widely used method for producing silicon single crystals in the semiconductor industry today, accounting for over 90% of all silicon single crystal production. This method, through precise control of temperature, pulling speed, and rotation, can grow large-diameter, high-purity, defect-free silicon single crystals, perfectly meeting the stringent requirements of large-scale integrated circuit manufacturing for substrate materials. The process mainly includes the following steps: loading, melting, stabilization, crystal pulling, shoulder formation, equal diameter growth, and finishing. Shoulder formation is the most critical process step affecting crystal formation. Its main purpose is to smoothly and controllably enlarge the diameter of the defect-free fine crystals (after dislocation removal) to the target crystal rod diameter, laying the foundation for subsequent equal-diameter growth.

[0003] Currently, the majority of mainstream 8-inch, 12-inch, and larger monocrystalline silicon products are <100> oriented monocrystalline silicon, exhibiting four raised ridges, referred to as four edges. During the shoulder formation process in <100> oriented silicon monocrystalline silicon production, if the shoulder formation is rapid, the (111) crystal plane family will gradually open from the edge state, later growing in a planar or stepped shape as the diameter continuously expands, thus forming four "open edges" or "wide faces," such as... Figure 1 As shown. In traditional crystal pulling experience, the degree of opening of the edge, that is, the expansion speed of the "open edge" and "wide face", should not be too fast. If it is too fast, the line will easily break and grow into polycrystalline. According to the crystal plane elimination law and the periodic bond chain theory, when the silicon single crystal growth has a large degree of undercooling, the (111) crystal plane can be continuously expanded to the entire shoulder shape. The top view of the shoulder shape is close to a square. Therefore, it is necessary to detect and evaluate the expansion speed of the diameter at the "wide face" and match it with the expansion speed of the diameter at the arc during the shoulder process. The matching relationship should be related to the crystal growth principle and the silicon single crystal structure. If the growth speed is not matched, the diameter increases too fast in the shoulder stage, and the (111) crystal plane will not have time to grow the step surface, which will cause the single crystal to break easily and have to be melted back and start over. If the diameter increases too slowly, a cone shoulder will appear, such as Figure 2 As shown, wasting raw materials increases production costs.

[0004] In the current mainstream shoulder-forming process, the process settings such as crystal rise rate and temperature change in the process slope table of the shoulder-forming stage are mainly used to make the thinner single crystal rod at the end of the crystal pulling stage grow to the desired size through the shoulder-forming stage. However, it does not have the ability to detect and evaluate whether the shoulder-forming is too fast or too slow. Some single crystal furnace equipment can provide data on the shoulder-forming angle, but none of them have the ability to detect and evaluate the speed of edge opening.

[0005] As production continues, the furnace thermal field gradually ages, often resulting in repeated shoulder breakage during a stable crystal pulling process. This is because parameters such as temperature and pulling speed in the process slope table no longer match the current environment, causing the shoulder breakage edge to not open according to the ideal shape, severely impacting production efficiency. Current solutions involve engineers and operators monitoring the shoulder breakage process on-site or remotely using a centralized control system, assessing the degree of edge opening in real time to manually adjust process parameters. The advantage of this method is its ability to adjust the process at any time and a high success rate. However, its disadvantages are: firstly, it requires significant labor costs and has poor repeatability; secondly, it relies heavily on extensive manual experience, as new employees lack the necessary skills.

[0006] In summary, this invention proposes a method for evaluating the degree of edge opening of the shoulder line in Czochralski single-crystal silicon, so that engineers can adjust the shoulder shape and optimize process parameters more quickly. Summary of the Invention

[0007] To address the aforementioned problems in existing technologies, the present invention aims to propose a method for evaluating the degree of edge opening of shoulder-grown monocrystalline silicon. By detecting and evaluating the degree of edge opening, engineers can easily adjust the process to control the edge opening degree, effectively avoiding edge breakage and improving the success rate and production efficiency of shoulder growth.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: A method for evaluating the degree of edge opening of the shoulder line in Czochralski single-crystal silicon includes the following steps: (1) Obtain the shoulder-forming image inside the single crystal furnace in real time through the image acquisition system; (2) Process the image to determine the edge of the crystal rod, and calculate the maximum diameter D at the arc representing the shoulder of the crystal rod. max And the minimum diameter D after the edge is cut. min ; (3) Calculate the evaluation parameters, which include at least the opening degree ΔD and the opening trend degree K; wherein, the opening degree ΔD = D max -D min The degree of edge opening tendency K is D max The rate of change of D min The ratio of the rates of change; (4) Compare the opening degree ΔD and the opening trend degree K with the preset threshold range, evaluate the state of the shoulder release process based on the comparison results, and output the evaluation information.

[0009] Preferably, in step (2), the image is converted to grayscale, edge detection is performed, and contour fitting is performed to determine the edge of the crystal rod. The edge detection algorithm can be either the Sobel operator, the Canny edge detection algorithm, or the Scharr operator.

[0010] Preferably, in step (3), the method for calculating the ridge tendency degree K is as follows: for D max and D min Differentiate the curves of change with time or crystal length respectively to obtain the rate of change V. max and V min Then calculate K=V max / V min .

[0011] Preferably, in step (4), the evaluation information includes: when the ridge opening trend K is 1.0, it is evaluated that the ridge line has not been opened, or the diameter at the maximum and minimum points after ridge opening is parallel enlargement; when the ridge opening trend K is between 1.0 and 1.4, it is evaluated that the single crystal is accelerating the ridge opening; when the ridge opening trend K is greater than 1.4, it is evaluated that the shoulder is expanding too fast, and there is a risk of ridge breakage or line breakage.

[0012] Preferably, the evaluation information output in step (4) is used to prompt adjustments to the crystal pulling process parameters, including pulling speed and heating power. If the edge opening trend value increases and the edge opening degree also increases, the pulling speed is increased or the temperature control closed loop is opened to raise the temperature. After adjusting the pulling speed or temperature control, the edge opening trend value and edge opening degree are monitored in real time. If the edge opening trend value shows a decreasing trend, the temperature control closed loop is closed. If the edge opening trend value decreases rapidly, the pulling speed is reduced.

[0013] Preferably, steps (1) to (4) are executed in a cycle of 10-60 seconds to achieve real-time monitoring and evaluation of the shoulder placement process.

[0014] The beneficial effects of this invention are: The advantage of this invention over conventional shoulder placement methods is that it provides a method for numerically evaluating the speed of shoulder placement, which facilitates simple adjustments and judgments for less experienced operators in actual production, reducing reliance on operator experience.

[0015] The advantage of this invention compared to the previous fixed-angle shoulder setting is that the measurement data of the diameter of the fixed-angle shoulder setting is smoothed, which makes it impossible to effectively monitor the diameter of the shoulder arc and the diameter of the ridge separately. Therefore, it is impossible to adjust the process in time when the ridge is abnormal. In contrast, this invention processes the diameter of the shoulder arc and the diameter of the ridge separately. By defining the ridge trend degree and the ridge degree, it cyclically monitors the changes in the shoulder shape and then adjusts the process in real time.

[0016] Compared with previous methods for controlling the shoulder formation and edge opening of single crystals, the present invention has the advantage that it does not require the prior collection and screening of data on the good shape of the single crystal shoulder formation. It only requires the pre-written cooling curve and basic pulling speed of the basic matching shoulder formation process. The changes in the values ​​can remind the operator to adjust the process parameters in a timely manner until the crystal is successfully formed. Attached Figure Description

[0017] Figure 1 A top view of the shoulder shape of a single-crystal silicon wafer.

[0018] Figure 2 This is a schematic diagram of the shoulder shape of a single-crystal silicon cone.

[0019] Figure 3 The fitted single-crystal aperture ellipse at the end of the crystal pulling process.

[0020] Figure 4 The curves showing the changes in the maximum and minimum diameters of the present invention are shown, with the difference representing the open edge.

[0021] Figure 5 The ratio of the slope of the maximum diameter curve to the minimum diameter curve in this invention is the curve representing the change in the ridge tendency.

[0022] Figure 6 This is an evaluation curve for the degree and trend of edge opening of a certain un-edged object.

[0023] Figure 7 The evaluation curves for the opening degree and opening trend of normal opening are shown.

[0024] Figure 8 The evaluation curves for the degree and trend of excessively rapid edge opening are shown.

[0025] Figure 9 This is a schematic diagram of the shoulder shape without controlling for new variables.

[0026] Figure 10 This is a schematic diagram of the shoulder shape after controlling the new variables on the same furnace platform. Detailed Implementation

[0027] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments, but this does not mean that the scope of protection of the present invention is limited.

[0028] This invention addresses the problem of low production efficiency caused by the reliance on manual experience and the inability to quantify the speed of shoulder formation in existing technologies. The invention uses an image acquisition system to acquire real-time images of the shoulder formation process within the single crystal furnace. The images are processed to identify and fit the crystal rod edges, calculating the minimum diameter representing the edge after it has been opened and the maximum diameter representing the arc. "Edge opening degree" is defined as the difference between the maximum and minimum diameters, and "edge opening trend degree" is defined as the ratio of the rate of change of the maximum and minimum diameters. These two parameters are continuously calculated. By cyclically monitoring these two parameters, the shoulder formation status is quantified. When the parameters exceed a preset reasonable range, timely prompts are made to adjust process parameters (such as pulling speed and temperature), thereby effectively avoiding edge breakage and wire breakage, improving the success rate of shoulder formation and production efficiency, and reducing reliance on operator experience.

[0029] In the implementation of this invention, the single crystal furnace needs to be connected to a centralized control server and a vision server provided by the equipment manufacturer. The image acquisition and analysis system of the single crystal furnace can be used to acquire high-definition images in BMP or other formats in real time and transmit them to the centralized control vision server, thereby meeting the requirements for real-time image acquisition and processing. The image acquisition and analysis system of the single crystal furnace includes a monocular CCD camera and a CCD vision centralized control image transmission system. The monocular CCD camera is installed on the observation window on the side of the furnace cover of the single crystal furnace. The images acquired by the camera are transmitted to the vision server for storage via optical fiber. The vision server can save the crystal pulling process video in real time or save the crystal pulling process images at a sampling rate exceeding 15 FPS. The required software environment is deployed on the centralized control vision server to save the shoulder-forming images to the hard disk in real time. Then, the processing of the shoulder-forming images is performed in a regular cycle, and the measured diameter, calculated edge opening degree, edge opening trend degree, and other data and curves are output to the hard disk.

[0030] As a specific embodiment of the present invention, the method for evaluating the degree of edge opening of the shoulder line of Czochralski single crystal silicon provided by the present invention includes the following steps: S1: The image acquisition system can acquire images of the shoulder formation inside the single crystal furnace in real time. When acquiring images, the crystal pulling process data can be collected simultaneously and correspond to the images in time series. This allows the images to be processed in conjunction with the crystal pulling process data, increasing the value of data analysis.

[0031] S2: Convert the image to grayscale and transform it into a numerical matrix. By calculating the grayscale gradient in the central region of the image in the horizontal and vertical directions, fit the single-crystal aperture ellipse at the end of crystal extraction, such as... Figure 3 As shown, the center coordinates are calculated and provided to the region of interest (ROI) for subsequent processing.

[0032] S3: Using an edge detection algorithm for grayscale images, calculate the average grayscale gradient of the single crystal diameter within the ROI region during the shoulder-laying stage, highlighting the edge areas, obtaining pixel size data, and converting it to the actual millimeter scale. In this step, the image edge detection algorithm can not only use the Sobel operator with a 3×3 convolution kernel, but also, depending on the actual effect, can use Canny edge detection, Scharr operator, etc.; at the same time, the gradient magnitude can also be flexibly selected according to the actual image effect or automatically selected by the algorithm design.

[0033] S4: Use a higher-order polynomial to fit the upper and lower envelopes, i.e., the diameter D at the maximum point. max and the diameter D at the minimum point min The diameter D at its maximum point max D represents the diameter at the arc, and the diameter at the minimum point. min Represents the diameter at the cleavage point, such as Figure 1 As shown; simultaneously calculate the result of subtracting the lower envelope from the upper envelope, i.e., the opening angle ΔD = D. max -D min ,like Figure 4 As shown.

[0034] S5: Use the maximum diameter D max and the diameter D at the minimum point min By taking the derivative of the crystal length data, we can obtain its rate of change V. max and V min Then calculate the edge opening trend degree K=V max / V min ,like Figure 5 As shown. The diameter D at the maximum point can also be fitted. max and the minimum diameter D min Curve showing how it changes over time.

[0035] S6: Set different image sampling rates and different calculation cycle intervals to output the parameters of edge opening trend and edge opening in real time. Based on production experience, it is necessary to make timely adjustments during the shoulder setting stage. Even when using different detection algorithms and equipment with different computing power, the time required for one cycle should not exceed 60 seconds, preferably 10-30 seconds. Otherwise, the data lag will be too long, and the evaluation will lose its meaning.

[0036] The evaluation parameters of edge opening degree and edge opening trend degree described in this invention can be used to explain whether a single crystal is accelerating edge opening. According to the crystal structure of silicon single crystal, an edge opening trend degree of 1.0 indicates that the edge is not yet open, or that the diameter at the maximum and minimum points is parallel and enlarged after edge opening; an edge opening trend degree of 1.0-1.4 indicates that the single crystal is accelerating edge opening, and a value greater than 1.4 indicates that the critical value of edge breaking has been reached.

[0037] Based on the edge-opening tendency, when the value is within a reasonable range, continue shoulder formation according to the preset crystal pulling speed and temperature changes in the process slope table. For 8-inch monocrystalline products, the pulling speed during the shoulder formation stage is usually set between 50-70 mm / h, and the overall temperature drop during the shoulder formation stage is 2.5-3%. For other <100> crystal orientation products with different thermal fields or different sizes, the pulling speed and temperature changes may be slightly different, but the same applies.

[0038] If the edge opening trend value increases and the edge opening degree also continues to increase, exceeding the reasonable upper limit set by the system, it is evaluated as too fast shoulder release. The engineer needs to be reminded to increase the drawing speed or open the temperature control closed loop to raise the temperature when necessary.

[0039] After adjusting the drawing speed or temperature control, continue to monitor the opening trend and opening degree in real time. If there is a decreasing trend, close the temperature control closed loop. If the value drops rapidly, the drawing speed also needs to be reduced until the shoulder is successfully formed and the shoulder rotation begins.

[0040] Example The production equipment is a TDR110B single crystal furnace manufactured by Zhejiang Jingsheng Electromechanical Co., Ltd., equipped with a 24-inch hot zone, a feed rate of 170kg, and produces 8-inch single crystal products. The shoulder-forming stage process is set with a constant pulling speed of 60mm / h, and the temperature drops according to the length, with a cumulative temperature drop of 2.5-3%, which allows the single crystal to be shouldered to the required size.

[0041] The single crystal furnace is connected to the centralized control server and vision server provided by Zhejiang Jingsheng Electromechanical Co., Ltd. The required software environment is deployed on the vision server of the centralized control system. The shoulder image is saved to the hard disk in real time. Then, it is processed in a 30-second cycle. Subsequently, the measured diameter, calculated ridge opening degree, ridge opening trend degree and other data and curves are output to the hard disk.

[0042] Based on the output data such as edge opening degree and edge opening trend degree, the following judgments are made: If the ridge opening value is between 0-4mm and there is no abnormal increase, and the ridge opening trend is between 1.0-1.1, it is judged as either not yet ridged or about to be ridged. Figure 6 As shown, the normal shoulder-forming process is then carried out according to the preset process until the target diameter is successfully reached.

[0043] The ridge opening value is between 0-4mm, with no abnormal increase, and the ridge opening trend is between 1.1-1.2, indicating that ridge opening has begun in a single crystal, but the diameter expansion at the ridge opening is normal. Figure 7 As shown, the normal shoulder-forming process continues according to the preset process until the target diameter is successfully reached.

[0044] If the edge opening value is greater than 2mm and increases rapidly, and the edge opening trend is between 1.2 and 1.4, it is judged that the single crystal edge opening is relatively fast. When it approaches 1.4, there is a risk of edge breakage or wire breakage at any time. Figure 8 As shown, the diameter expansion rate at the ridge is slightly slower than that at the arc, indicating a large degree of melt undercooling. This suggests that the temperature or casting speed curve in the preset process is not suitable. In this case, the process parameters should be adjusted in time, such as increasing the casting speed by 2-5 mm / h or opening the closed loop of the temperature control. After adjusting the process parameters, continue to monitor the ridge expansion trend. If there is a decreasing trend, the casting speed should be reduced by 1-3 mm / h to keep the ridge expansion trend around 1-1.1 until the target diameter is successfully reached.

[0045] This invention provides a method for accurately and in real-time monitoring and evaluating the speed of the shoulder formation stage in Czochralski single-crystal silicon. Combining this method with data such as edge opening degree and edge opening trend, it can reduce single-crystal breakage caused by abnormal shoulder opening, effectively solving the problem of low shoulder formation efficiency and controlling shoulder shape consistency in large-scale single-crystal silicon production. Figure 9 and Figure 10 As shown, the shoulder shape consistency after controlling the new variables (i.e., ridge opening degree and ridge opening trend degree) on the same furnace platform is significantly better than the shoulder shape consistency without controlling the new variables. Therefore, the method of the present invention can significantly reduce the possibility of shoulder turning failure and improve efficiency.

Claims

1. A method for evaluating the degree of edge opening of the shoulder line in Czochralski-grown single-crystal silicon, characterized in that, Includes the following steps: (1) Obtain the shoulder-forming image inside the single crystal furnace in real time through the image acquisition system; (2) Process the image to determine the edge of the crystal rod, and calculate the maximum diameter D at the arc representing the shoulder of the crystal rod. max And the minimum diameter D after the edge is cut. min ; (3) Calculate the evaluation parameters, which include at least the opening degree ΔD and the opening trend degree K; wherein, the opening degree ΔD = D max -D min The degree of edge opening tendency K is D max The rate of change of D min The ratio of the rates of change; (4) Compare the opening degree ΔD and the opening trend degree K with the preset threshold range, evaluate the state of the shoulder release process based on the comparison results, and output the evaluation information.

2. The method for evaluating the degree of edge opening of the shoulder line in Czochralski single crystal according to claim 1, characterized in that, In step (2), the image is grayscaled, edge detected, and contour fitted to determine the crystal rod edge.

3. The method for evaluating the degree of edge opening of the shoulder line in Czochralski single crystal according to claim 2, characterized in that, The edge detection algorithm used is either the Sobel operator, the Canny edge detection algorithm, or the Scharr operator.

4. The method for evaluating the degree of edge opening of the shoulder line in Czochralski single crystal according to claim 1, characterized in that, In step (3), the method for calculating the ridge tendency degree K is as follows: for D max and D min Differentiate the curves of change with time or crystal length respectively to obtain the rate of change V. max and V min Then calculate K=V max / V min .

5. The method for evaluating the degree of edge opening of the shoulder line in Czochralski single crystal according to claim 1, characterized in that, In step (4), the preset threshold range of the angularity ΔD is 0-4mm, and the preset threshold range of the angularity tendency K is 1.0-1.

4.

6. The method for evaluating the degree of edge opening of the shoulder line in Czochralski single crystal according to claim 5, characterized in that, In step (4), the evaluation information includes: when the ridge opening trend K is 1.0, it is evaluated that the ridge line has not been opened, or the diameter at the maximum and minimum points after ridge opening is parallel enlargement; when the ridge opening trend K is between 1.0 and 1.4, it is evaluated that the single crystal is accelerating the ridge opening; when the ridge opening trend K is greater than 1.4, it is evaluated that the shoulder is expanding too fast, and there is a risk of ridge breakage or line breakage.

7. The method for evaluating the degree of edge opening of the shoulder line in Czochralski single crystal according to claim 1, characterized in that, The evaluation information output in step (4) is used to prompt adjustments to the crystal pulling process parameters, including pulling speed and heating power.

8. The method for evaluating the degree of edge opening of the shoulder line in Czochralski single crystal according to claim 7, characterized in that, If the opening trend value increases and the opening degree also continues to increase, increase the drawing speed or open the temperature control closed loop to raise the temperature; after adjusting the drawing speed or temperature control, continue to monitor the opening trend value and opening degree in real time. If the opening trend value shows a decreasing trend, close the temperature control closed loop. If the opening trend value decreases rapidly, reduce the drawing speed.

9. The method for evaluating the degree of edge opening of the shoulder line in Czochralski single crystal according to claim 1, characterized in that, Steps (1) to (4) are executed in a cycle of 10-60 seconds to achieve real-time monitoring and evaluation of the shoulder placement process.

10. A Czochralski-grown single-crystal silicon, characterized in that, During the shoulder formation stage of Czochralski single-crystal silicon, the degree of edge opening of the shoulder formation edge of Czochralski single-crystal silicon is evaluated using the method described in any one of claims 1 to 9.

Images