Process optimization method for thickness uniformity of silicon epitaxial growth film

Abstract

The invention provides a process optimization method for thickness uniformity of a silicon epitaxial growth film, which comprises the steps of determining basic conditions, designing an experimental scheme, calculating sensitivity, solving nonlinear programming and performing experimental verification in sequence, finally obtaining a process optimization group, and optimizing an actual process. According to the process optimization method, the DOE experimental design thought is combined, the number of experimental test groups can be optimized by determining the process variables and respectively designing i groups (i is larger than or equal to 3) of different experimental process parameters, wafer and machine consumption of experiments is effectively reduced, the number of times of the experiments is effectively reduced, and prediction of the optimal uniformity is achieved.

Description

Technical Field

[0001] The invention relates to the technical field of silicon epitaxial growth, and in particular to a process optimization method for uniformity of silicon epitaxial growth film thickness. Background Art

[0002] Silicon epitaxial growth is usually carried out in a silicon epitaxial furnace, and the precursor of the reaction is usually SiH 4 Or dichlorosilane (DCS) etc. 2 After being transported to the reaction chamber, the wafer and gas are heated by infrared lamps or other heating devices, and the gas is decomposed and adsorbed on the surface of the wafer to form an epitaxial layer. For silicon epitaxy, its thickness uniformity is an important indicator for considering the quality of its epitaxial growth. In the daily process debugging, one of the main tasks of process engineers is to achieve uniform epitaxial growth by adjusting various parameters.

[0003] The growth rate of epitaxial silicon is closely related to parameters such as temperature, pressure, process gas flow rate and process time. Taking temperature as an example, both the growth rate and temperature follow the Arrhenius curve. Under the condition of a given flow rate, the growth rate of epitaxial silicon on the surface of the wafer is mainly related to the temperature at the corresponding position. To optimize the film thickness of the epitaxial silicon layer on the surface of the wafer, it is mainly achieved by adjusting the power of the heating element. For the epitaxial reaction chamber, there are often many types of heating elements, such as strip lamps or round lamps. It is difficult to explore and optimize the best lamp power ratio by adjusting the power of a single lamp, which requires a large number of experiments. Moreover, it is often impossible to obtain the optimal uniformity by adjusting only a single lamp, and a combination of multiple lamps is required to obtain better uniformity.

[0004] In actual operation, the DOE method is usually used for experimental design, and then the linear response of each group of parameters is calculated, and good uniformity is obtained through the linear combination of these parameter group responses. This method can simplify the number of experimental groups required for single variable adjustment to a certain extent during experimental design. However, the solution to the optimal lamp ratio still requires many linear combination experiments.

[0005] Therefore, it is necessary to develop a process optimization method for silicon epitaxial growth film thickness uniformity in order to reduce the need for actual experiments and achieve the purpose of rapid debugging. Summary of the invention

[0006] In view of the problems existing in the prior art, the present invention provides a process optimization method for the uniformity of silicon epitaxial growth film thickness, which combines the experimental design idea of ​​DOE. By determining the process variables and designing i groups (i≥3) of different experimental process parameters respectively, the number of experimental test groups can be optimized, which effectively reduces the experimental wafer and machine consumption, effectively reduces the number of experiments, and realizes the prediction of optimal uniformity.

[0007] To achieve this object, the present invention adopts the following technical solutions:

[0008] The object of the present invention is to provide a process optimization method for silicon epitaxial growth film thickness uniformity, the process optimization method comprising the following steps:

[0009] (1) Determine the basic conditions: prepare i identical substrates, and grow Si layers on the substrates respectively using the same basic process parameters based on the chemical vapor deposition method;

[0010] (2) Designing an experimental plan: Determine the process variables, design i groups of different experimental process parameters, i ≥ 3, perform silicon epitaxial growth on the i groups of Si layers obtained in step (1), and measure the silicon epitaxial growth film thickness curves Ai respectively;

[0011] (3) Calculating sensitivity: normalizing the silicon epitaxial growth film thickness curve Ai obtained in step (2), and calculating the sensitivity of the silicon epitaxial growth film thickness to i groups of different experimental process parameters, and obtaining i sensitivity curves;

[0012] (4) Nonlinear programming solution: Based on the i sensitivity curves obtained in step (3), a set of parameters (α1, α2, to αi) is set as the prediction parameter set, and the two are linearly combined to obtain the ideal film thickness curve and solve its ideal uniformity characterization value R 理想 ;

[0013] (5) Experimental verification: Integrate the predicted parameter group described in step (4) with the basic process parameters described in step (1), conduct experimental verification on the obtained verification process parameters, obtain the film thickness curve of the verification group and solve its verification uniformity characterization value R 验证 ; Calculate the contrast difference Q =│100%-R 验证 / R 理想 │; If the comparison difference Q meets the requirements and is consistent with the prediction, the process parameters are verified as the process optimization group; otherwise, it does not meet the prediction and returns to re-establish the optimization model.

[0014] The process optimization method of the present invention combines the experimental design idea of ​​DOE. By determining the process variables and designing i groups (i≥3) of different experimental process parameters, the wafer and machine consumption of the experiment is effectively reduced. After the first round of experiments, the optimal parameter group is obtained by using a python script to calculate the nonlinear programming, which effectively reduces the number of experiments and realizes the prediction of the optimal uniformity. Through this prediction script, the optimal uniformity of the Si epitaxial layer under the corresponding parameter changes can be quickly known, and effective adjustments can be made according to the parameter group. If the Si uniformity obtained by the current process variables cannot meet the target process requirements, the process variables can be changed or added according to the results and predictions, and nonlinear programming can be re-performed for prediction. This prediction model establishes virtual measurement of the epitaxial machine, effectively reduces the demand for actual experiments, and provides a solution for rapid debugging.

[0015] Regarding step (5), the contrast difference Q =│100%-R 验证 / R 理想 │ is the absolute value. If the comparison difference Q meets the requirements and is consistent with the prediction, it proves that the optimization model is highly feasible. According to the prediction parameters given by the optimization model, the optimal uniformity under the corresponding variable conditions can be obtained; otherwise, it does not meet the prediction; if the deviation is large, it proves that the sensitivity of the parameter variables in the optimization model is not feasible, and it is necessary to return to step (2) and re-collect sensitivity data under the predicted optimal uniformity conditions and re-establish the optimization model; if the deviation is small, it can return to step (4) to adjust the value of the prediction parameter group.

[0016] As a preferred technical solution of the present invention, the process variables in step (2) include any one or a combination of at least two of temperature, pressure, process gas flow rate and process time.

[0017] As a preferred technical solution of the present invention, the process variable in step (2) is temperature, and i groups of different experimental process parameters are designed by grouping the heating elements and adjusting the power ratio of the heating elements relative to the rated power; the grouping of the heating elements is often based on the positional relationship of the heating elements, especially the symmetrical relationship.

[0018] As a preferred technical solution of the present invention, the normalization processing in step (3) includes: dividing the silicon epitaxial growth film thickness curve Ai obtained in step (2) by the respective average film thickness values.

[0019] As a preferred technical solution of the present invention, the nonlinear programming solution in step (4) includes: using the nonlinear programming solver in the sicpy program package, defining the upper and lower bounds of the variables and initializing them, calling the minimize function, performing optimization, solving the minimum value and outputting the optimization result, and providing parameter conditions.

[0020] As a preferred technical solution of the present invention, the ideal uniformity characterization value R 理想 , Verify the uniformity characterization value R 验证 They are the maximum differences obtained by subtracting the maximum and minimum values ​​of the corresponding film thickness curves.

[0021] As a preferred technical solution of the present invention, in step (5), if the comparison difference Q≤10%, it is in line with the prediction and a process optimization group is obtained.

[0022] As a preferred technical solution of the present invention, in step (5), if 10% < contrast difference Q < 20% is satisfied, it does not meet the prediction, and the process returns to step (4) to adjust the specific values ​​of the prediction parameter group.

[0023] As a preferred technical solution of the present invention, in step (5), if the comparison difference Q ≥ 20%, which does not meet the prediction, return to step (2), redefine the process variables or increase the process variables, and redesign i groups of different experimental process parameters.

[0024] As a preferred technical solution of the present invention, in step (5), the process variables are re-determined, i groups of different experimental process parameters are redesigned, and the obtained comparison difference satisfies 10%<comparison difference Q<20%, ​​and the process returns to step (4) to adjust the specific values ​​of the prediction parameter group.

[0025] Compared with the prior art solutions, the present invention has at least the following beneficial effects:

[0026] The present invention provides a process optimization method for uniformity of silicon epitaxial growth film thickness, including determining basic conditions, designing experimental schemes, calculating sensitivity, solving nonlinear programming, and experimental verification in sequence, and finally obtaining a process optimization group to optimize the actual process. The process optimization method of the present invention combines the experimental design idea of ​​DOE, and by determining process variables and designing i groups (i≥3) of different experimental process parameters respectively, the number of experimental test groups can be optimized, effectively reducing the consumption of experimental wafers and machines, effectively reducing the number of experiments, and realizing the prediction of optimal uniformity. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of an epitaxial reaction chamber according to a specific embodiment of the present invention.

[0028] Figure 2 yes Figure 1 Schematic diagram of the lamp arrangement in the middle.

[0029] Figure 3 It is a schematic diagram of sequential deposition on a substrate according to a specific embodiment of the present invention.

[0030] Figure 4It is a film thickness distribution map of the basic group Si layer of a specific embodiment of the present invention.

[0031] Figure 5 This is a graph showing the distribution of the Si layer thickness on the X-axis of a basic group according to a specific embodiment of the present invention.

[0032] Figure 6 It is a sensitivity curve corresponding to adjustment group 1 according to a specific embodiment of the present invention.

[0033] Figure 7 It is a sensitivity curve corresponding to adjustment group 2 according to a specific embodiment of the present invention.

[0034] Figure 8 It is a sensitivity curve corresponding to adjustment group 3 according to a specific embodiment of the present invention.

[0035] Fig. 9 It is a sensitivity curve corresponding to adjustment group 4 according to a specific embodiment of the present invention.

[0036] Fig.10 It is a three-line comparison diagram of the normalized basic group, prediction parameter group and validation group in a specific implementation manner of the present invention.

[0037] Fig.11 It is a logic execution block diagram of the process optimization method for silicon epitaxial growth film thickness uniformity of the present invention. DETAILED DESCRIPTION

[0038] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0039] To better illustrate the present invention and facilitate understanding of the technical solution of the present invention, typical but non-limiting embodiments of the present invention are as follows:

[0040] The present invention provides a process optimization method for uniformity of silicon epitaxial growth film thickness, such as Fig.11 As shown, the process optimization method comprises the following steps:

[0041] (1) Determine the basic conditions: prepare i identical substrates, and grow Si layers on the substrates respectively using the same basic process parameters based on the chemical vapor deposition method;

[0042] (2) Designing an experimental plan: Determine the process variables, design i groups of different experimental process parameters, i ≥ 3, perform silicon epitaxial growth on the i groups of Si layers obtained in step (1), and measure the silicon epitaxial growth film thickness curves Ai respectively;

[0043] (3) Calculating sensitivity: normalizing the silicon epitaxial growth film thickness curve Ai obtained in step (2), and calculating the sensitivity of the silicon epitaxial growth film thickness to i groups of different experimental process parameters, and obtaining i sensitivity curves;

[0044] (4) Nonlinear programming solution: Based on the i sensitivity curves obtained in step (3), a set of parameters (α1, α2, to αi) is set as the prediction parameter set, and the two are linearly combined to obtain the ideal film thickness curve and solve its ideal uniformity characterization value R 理想 ;

[0045] (5) Experimental verification: Integrate the predicted parameter group described in step (4) with the basic process parameters described in step (1), conduct experimental verification on the obtained verification process parameters, obtain the film thickness curve of the verification group and solve its verification uniformity characterization value R 验证 ; Calculate the contrast difference Q =│100%-R 验证 / R 理想 │; If the comparison difference Q meets the requirements and meets the prediction, the process parameters are verified as the process optimization group; otherwise, it does not meet the prediction, and the optimization model is re-established; specifically, if the comparison difference Q≤10%, it meets the prediction, and the process optimization group is obtained; if it satisfies 10%<comparison difference Q<20%, ​​it does not meet the prediction, and the specific values ​​of the prediction parameter group are adjusted; if the comparison difference Q≥20%, it does not meet the prediction, and the process parameters are re-determined, and the process parameters of i groups are re-designed. Specifically, returning to step (2) includes but is not limited to re-determining the process variables, or re-determining the range (upper and lower limits) of the process variables, adding process variables, etc.

[0046] The process variable selection of the present invention includes any one or a combination of at least two of temperature, pressure, process gas flow rate and process time.

[0047] A specific embodiment of the present invention provides a process optimization method for uniformity of silicon epitaxial growth film thickness, wherein the process variable is determined to be temperature, and the process optimization method includes the following contents:

[0048] The growth is carried out using a silicon epitaxial furnace, such as Figure 1 As shown, the heating of the epitaxial reaction chamber is usually carried out by heating the heating element (infrared lamp) outside the quartz chamber. The energy is radiated into the gas and wafer in the quartz chamber through the reflective gold plate above the infrared lamp. The outside of the chamber is cooled by water and air to ensure that the parts will not be damaged by high temperature. During the heating process, the actual temperature in the chamber is measured by a thermometer and compared with the set value. PID temperature control is performed through the program, and the program outputs a PID calculation value, which is multiplied by the power ratio of the infrared lamp to obtain the actual output power of each infrared lamp.

[0049] Since the infrared lamp and heat dissipation structure are relatively complex, in order to ensure the uniformity of the temperature field on the surface of the quartz cavity wafer, it is necessary to adjust the power ratio of each infrared lamp to obtain a relatively uniform temperature field. Figure 2 As shown in the figure, there are a total of 22 linear infrared lamps in the actual silicon epitaxial furnace and 4 spot lamps located in the middle of the quartz cavity wafer ( Figure 2 It is relatively complicated to adjust the power ratio of the infrared lamps individually to obtain uniform temperature field. The power ratio change of each infrared lamp will cause the power change of other infrared lamps and the temperature change at every point on the wafer surface.

[0050] During the silicon epitaxial growth process, due to SiH 4 The growth rate in the low temperature section is mainly determined by the temperature, and the uniformity of the epitaxial silicon film thickness is mainly adjusted by optimizing the uniformity of the temperature field. In the actual process debugging, in order to effectively measure the thickness of the epitaxial Si layer, a SiGe layer is first grown on the substrate, and then a Si layer is grown. Since the n and k coefficients of SiGe and Si are different, both SiGe and Si layers can be measured and fitted by ellipsometer.

[0051] Determine the basic conditions: In the process of determining the basic conditions, prepare 4 identical substrates, and set the power of the 22 linear infrared lamps and the 4 spot lights at the bottom according to the power ratio of the basic group compared to the rated power in Table 1; it should be noted that the power ratio of the 4 spot lights changes uniformly. Taking the basic group as an example, the power ratio of the spot light No. 23 is 50%, which means that the power of the 4 spot lights is set to 50% of the rated power; Figure 3 As shown, chemical vapor deposition is used to deposit a SiGe layer and a Si layer on a Si substrate in sequence, and the basic Si layer film thickness curve is obtained by measuring with an ellipsometer. Subsequently, silicon epitaxial growth is performed on the Si layer, and four Si layer film thickness curves are obtained by measuring with an ellipsometer. The four Si layer film thickness curves are respectively subtracted from the Si layer film thickness curve corresponding to the basic group, thereby obtaining four silicon epitaxial growth film thickness curves Ai. Although the basic group is performed four times to obtain four deposited Si layers, using exactly the same process conditions, one of the deposited Si layers can be selected to test the distribution of the deposited film thickness. Figure 4 The basic group Si layer film thickness distribution map is shown. Figure 5 The X-axis Si layer thickness distribution curve of the basic group is shown. Figure 5 The horizontal axis is the distance from the origin of the coordinate system (unit: mm), and the vertical axis is the thickness of the Si layer on the X-axis (unit: angstrom). A total of 49 test points were selected;

[0052] Design of experimental plan: Experimental design was carried out through DOE, and the positions of the heating elements were divided, mainly according to the symmetry relationship. Four groups of infrared lamps or lights were selected, and the power ratio of each group was adjusted. The specific adjustment is shown in Table 1. Compared with the basic group, the adjustment group 1 adjusted the central infrared lamp, specifically increasing the power ratio of the three infrared lamps at the center of the upper layer (numbered 3, 4, and 5) and the three infrared lamps at the center of the lower layer (numbered 10, 11, and 12). Compared with the basic group, the adjustment group 2 adjusted the inner The adjustment group 3 adjusts the outer circle infrared lamps, specifically, the power ratio of the two infrared lamps at the upper layer (numbered 1, 2, 6, 7) and the two infrared lamps at the lower layer (numbered 8, 9, 13, 14) is increased respectively; compared with the basic group, the adjustment group 3 adjusts the outer circle infrared lamps, specifically, the power ratio of the two infrared lamps at the upper layer (numbered 17, 18, 21, 22) and the two infrared lamps at the lower layer (numbered 15, 16, 19, 20) is increased respectively; compared with the basic group, the adjustment group 4 adjusts the lighting, specifically, the power ratio of the four lightings numbered 23 is increased;

[0053] Table 1

[0054]

[0055] Note: The percentages in the table refer to the ratio of the actual power of the infrared lamp or lighting to the rated power.

[0056] It should be noted that during the recipe execution, the power ratio distribution of the lamp tubes of the SiGe layer is still based on the basic power ratio. After the SiGe deposition is completed, the power ratio of the lamp tubes is switched to a new power ratio, and the Si epitaxial deposition is performed after the temperature stabilizes. The grown wafer is measured again by the ellipsometer to obtain the film thickness distribution of the Si layer after the lamp tube power ratio is changed.

[0057] Calculate the sensitivity: subtract the four Si layer thickness curves from the Si layer thickness curve corresponding to the basic group, and then obtain four silicon epitaxial growth film thickness curves Ai. Perform normalization on each silicon epitaxial growth film thickness curve Ai, that is, divide each silicon epitaxial growth film thickness curve Ai by its own average film thickness. Then calculate the sensitivity of the silicon epitaxial growth film thickness to the power change of the above four groups of infrared lamps or lighting, and obtain four sensitivity curves, as shown in FIG. Figure 6-Figure 9 As shown;

[0058] Nonlinear programming solution: Based on the above four sensitivity curves, the parameter groups (α1, α2, α3, α4) corresponding to adjustment group 1, adjustment group 2, adjustment group 3, and adjustment group 4 are set to (2, 1, 0.5, -2) as the prediction parameter group. The upper and lower bounds are specified according to the process conditions. The linear combination of the prediction parameter group and the corresponding sensitivity curve is calculated to obtain an ideal film thickness curve under the condition of adjusting the prediction parameter group. The maximum difference corresponding to the film thickness curve is 18.97 angstroms, which is taken as the ideal uniformity characterization value R. 理想 ;

[0059] Experimental verification: According to the integration of the predicted parameter group and the power ratio of the basic group in Table 1, the power ratio corresponding to the verification group in Table 1 is obtained, and the film thickness curve of the verification group is obtained by experimental verification. The maximum difference corresponding to the film thickness curve is 20.23 angstroms, which is used as the verification uniformity characterization value R 验证 ; Calculate the contrast difference Q =│100%-R 验证 / R 理想 │=6.64%, the comparison difference Q≤10%, which is in line with the prediction, and the power ratio corresponding to the verification group is used as the process optimization group.

[0060] For intuitive comparison, the X-axis Si layer film thickness distribution curve is normalized, that is, each silicon epitaxial growth film thickness curve is divided by its own average film thickness, and the normalized curves of the basic group, prediction parameter group and verification group are integrated in Fig.10 It can be seen from the three-line comparison diagram in that the prediction parameter group and the verification group basically overlap, indicating that the process optimization method of the present invention is highly feasible.

[0061] It should be noted that all parameters in the DOE design experiment are measured and calculated as above to obtain the sensitivity data corresponding to each parameter. After obtaining these data, program writing is performed. For each parameter, a numerical variable is set, a reasonable upper and lower limit is determined according to the actual needs of the site, and then a target thickness is calculated. The specific calculation method is the normalized thickness of the basic film thickness plus the linear combination of these parameters and sensitivity data. Then the minimum value of the range of the target thickness is solved by nonlinear programming in the sicpy package of python. The specific logical steps include: defining the upper and lower bounds of the variables and initializing them, calling the minimize function in the SciPy library, performing optimization, outputting the optimization results, and giving the parameter conditions. Through script calculation, the target film thickness and corresponding parameter values ​​of the best uniformity under the current parameter groups can be obtained. According to the parameter values ​​obtained by calculation, the corresponding parameters can be directly modified and re-experimentally verified to see whether the actual film thickness and the calculated film thickness under this condition are consistent. If it meets the prediction, it proves that the optimization model is highly feasible. According to the best parameters given by the model, the best uniformity under the corresponding variable conditions can be obtained. If it does not meet expectations, it means that the sensitivity of these parameter variables is not feasible, and it is necessary to re-collect sensitivity data based on the predicted parameter group and re-establish the optimization model. In addition, the optimal uniformity under the current parameter group conditions can be judged based on the calculated film thickness. If the current uniformity still cannot meet the requirements, it is necessary to introduce other parameter groups for sensitivity testing. Perform nonlinear programming again to obtain the film thickness morphology with the best uniformity. Through continuous iteration and optimization, better uniformity can be obtained. This method can effectively and quickly give the corresponding experimental conditions and predict the best experimental results. As an important auxiliary means of virtual measurement, it greatly improves the actual debugging efficiency.

[0062] The present invention provides a process optimization method for uniformity of silicon epitaxial growth film thickness, including determining basic conditions, designing experimental schemes, calculating sensitivity, solving nonlinear programming, and experimental verification in sequence, and finally obtaining a process optimization group to optimize the actual process. The process optimization method of the present invention combines the experimental design idea of ​​DOE, and by determining process variables and designing i groups (i≥3) of different experimental process parameters respectively, the number of experimental test groups can be optimized, effectively reducing the consumption of experimental wafers and machines, effectively reducing the number of experiments, and realizing the prediction of optimal uniformity.

[0063] The present invention illustrates the detailed structural features of the present invention through the above embodiments, but the present invention is not limited to the above detailed structural features, that is, it does not mean that the present invention must rely on the above detailed structural features to be implemented. Those skilled in the art should understand that any improvement of the present invention, equivalent replacement of the components selected by the present invention, addition of auxiliary components, selection of specific methods, etc., all fall within the protection scope and disclosure scope of the present invention.

[0064] The preferred embodiments of the present invention are described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, a variety of simple modifications can be made to the technical solution of the present invention, and these simple modifications all belong to the protection scope of the present invention.

[0065] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

[0066] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A process optimization method for uniformity of silicon epitaxial growth film thickness, characterized in that: The process optimization method comprises the following steps: (1) Determine the basic conditions: prepare i identical substrates, and grow Si layers on the substrates respectively using the same basic process parameters based on the chemical vapor deposition method; (2) Designing an experimental plan: Determine the process variables, design i groups of different experimental process parameters, i ≥ 3, perform silicon epitaxial growth on the i groups of Si layers obtained in step (1), and measure the silicon epitaxial growth film thickness curves Ai respectively; (3) Calculating sensitivity: normalizing the silicon epitaxial growth film thickness curve Ai obtained in step (2), and calculating the sensitivity of the silicon epitaxial growth film thickness to i groups of different experimental process parameters, and obtaining i sensitivity curves; (4) Nonlinear programming solution: Based on the i sensitivity curves obtained in step (3), a set of parameters (α1, α2, to αi) is set as the prediction parameter set, and the two are linearly combined to obtain the ideal film thickness curve and solve its ideal uniformity characterization value R 理想 ; (5) Experimental verification: Integrate the predicted parameter group described in step (4) with the basic process parameters described in step (1), conduct experimental verification on the obtained verification process parameters, obtain the film thickness curve of the verification group and solve its verification uniformity characterization value R 验证 ; Calculate the contrast difference Q =│100%-R 验证 / R 理想 │; If the comparison difference Q meets the requirements and conforms to the prediction, the process parameters are verified as the process optimization group; otherwise, it does not conform to the prediction and returns to re-establish the optimization model.

2. The process optimization method according to claim 1, characterized in that: The process variables in step (2) include any one of temperature, pressure, process gas flow rate and process time, or a combination of at least two of them.

3. The process optimization method according to claim 2, characterized in that: In step (2), the process variable is temperature. By grouping the heating elements and adjusting the power ratio of the heating elements relative to the rated power, i groups of different experimental process parameters are designed.

4. The process optimization method according to claim 1, characterized in that: The normalization process in step (3) includes: dividing the silicon epitaxial growth film thickness curve Ai obtained in step (2) by the respective average film thickness values.

5. The process optimization method according to claim 1, characterized in that: The nonlinear programming solution in step (4) includes: defining upper and lower bounds of variables and initializing them, calling the minimize function, performing optimization, solving the minimum value and outputting the optimization result, and providing parameter conditions.

6. The process optimization method according to claim 1, characterized in that: Ideal uniformity characterization value R 理想 , Verify the uniformity characterization value R 验证 They are the maximum differences obtained by subtracting the maximum and minimum values ​​of the corresponding film thickness curves.

7. The process optimization method according to claim 1, characterized in that: In step (5), if the comparison difference Q ≤ 10%, it is consistent with the prediction and the process optimization group is obtained.

8. The process optimization method according to claim 1, characterized in that: In step (5), if 10% < contrast difference Q < 20% is satisfied, the prediction is not met, and the process returns to step (4) to adjust the specific values ​​of the prediction parameter group.

9. The process optimization method according to claim 1, characterized in that: In step (5), if the comparison difference Q ≥ 20%, which does not meet the prediction, return to step (2), redefine the process variables or increase the process variables, and redesign i groups of different experimental process parameters.

10. The process optimization method according to claim 9, characterized in that: In step (5), the process variables are re-determined, and i groups of different experimental process parameters are redesigned. The obtained comparison difference satisfies 10% < comparison difference Q < 20%. Return to step (4) to adjust the specific values ​​of the prediction parameter group or add process variables.

Images