A method for predicting etching rate during electronic-grade polysilicon cleaning process

Abstract

This invention discloses an etching rate prediction method for the cleaning process of electronic-grade polycrystalline silicon. By incorporating key process parameters such as the concentration of hydrofluoric acid, nitric acid, temperature, and total concentration of metallic impurities in the etching solution into a comprehensive prediction model, the method corrects the influence of multiple factors on the etching rate and solves the problem of prediction distortion caused by neglecting the aging of the etching solution in traditional methods. Based on the prediction method proposed in this invention, the optimal etching time can be dynamically calculated for the aging degree of different batches, fundamentally avoiding over-etching and material waste caused by overestimating the rate in early batches, and under-etching and impurity residue caused by underestimating the rate in later batches. This significantly improves the consistency and yield of product quality between batches. At the same time, this method can guide the maximization and online control of the etching solution's lifespan, extending the service life of expensive ultra-high purity acid solutions and significantly reducing production costs.

Description

Technical Field

[0001] This invention relates to the field of polycrystalline silicon technology, and more specifically to a method for predicting the etching rate during the cleaning process of electronic-grade polycrystalline silicon. Background Technology

[0002] After the initial processes of pulling, cutting, and grinding, electronic-grade polycrystalline silicon inevitably acquires a damage layer on its surface, along with metallic impurities from equipment, tools, and the environment. These impurities severely affect the electrical performance and yield of subsequent devices. Therefore, before proceeding to later processes, the silicon core surface must be cleaned and etched to thoroughly remove the damage layer and metallic contaminants. Currently, wet chemical etching is one of the most mainstream and effective cleaning methods. It typically uses a mixture of hydrofluoric acid and nitric acid, utilizing their strong oxidizing and dissolving properties to uniformly etch away a certain thickness of silicon layer from the silicon core surface, thereby exposing a fresh, clean crystal surface.

[0003] During the cleaning process, as the etching reaction continues, reactants are constantly consumed, and the acid concentration gradually decreases, leading to a weakening of the driving force of the chemical reaction and a corresponding decrease in the etching rate. Furthermore, this reaction is exothermic and extremely sensitive to temperature; temperature changes significantly alter the reaction rate constant, typically following Arrhenius's law—the reaction rate doubles for every approximately 10°C increase in temperature. In addition, metal ions dissolved from the silicon core surface into the acid solution accumulate continuously. These metal ions occupy active reaction sites on the silicon surface or form complexes with acid radicals, producing significant catalytic poisoning or inhibitory effects, further reducing the effective etching rate. Therefore, in actual cleaning, the etching rate is not constant but dynamically decreases with batch production, time, and bath aging. The etching rate directly determines the etching time. If the etching rate cannot be accurately predicted, the etching time cannot be precisely controlled, easily leading to under-etching or over-etching. Under-etching leaves behind a damaged layer and impurities, failing to meet the cleanliness requirements of electronic-grade polysilicon and affecting the performance and yield of subsequent devices; over-etching results in unnecessary waste of silicon material, increasing production costs. Therefore, accurately predicting the etching rate is a core prerequisite for ensuring qualified etching depth and consistent product quality.

[0004] However, in industrial production, to meet the ultra-high cleanliness requirements of electronic-grade polycrystalline silicon, the nitric acid and hydrofluoric acid used in the cleaning solution are both ultra-high purity reagents of UP-S or higher grade, which are extremely expensive. To spread the cost of acid replacement as much as possible, the cleaning solution usually needs to be reused to process multiple batches of silicon cores continuously until its cleaning capacity is exhausted before replacement. This reuse method causes the etching solution to age continuously with each batch, the acid concentration to decrease continuously, the metal impurities to accumulate continuously, and accompanied by temperature fluctuations, resulting in a significant downward trend in the etching rate. Currently, before each cleaning, due to the lack of an effective means to assess the actual etching rate of the current etching solution in real time online, operators can only rely on experience to set a fixed etching time according to a fixed etching rate. This crude control method completely ignores the dynamic impact of acid concentration decay, temperature fluctuations, and metal impurity accumulation on the etching rate. In early batches with high cleaning solution activity, the actual etching rate is far higher than the set value, leading to over-etching and material waste. In late batches where the cleaning solution has aged, the etching rate has significantly decreased, and the fixed etching time cannot achieve the required etching depth, resulting in under-etching, damaged layers, and metal impurity residue. This leads to the problem of over-etching in early batches and under-etching in later batches, resulting in large fluctuations and poor consistency in product quality between batches, severely restricting the production efficiency and final yield of electronic-grade polysilicon. Summary of the Invention

[0005] To address the aforementioned problems, the present invention aims to provide an etching rate prediction method during the cleaning process of electronic-grade polysilicon, which can comprehensively consider the effects of acid concentration, temperature, and impurity accumulation on the etching rate, thereby accurately predicting the etching rate of the current batch.

[0006] This invention is implemented by the following technical solution: A method for predicting the etching rate during the cleaning process of electronic-grade polysilicon includes the following steps: S1: Obtain the process parameters of the current etching solution, including the concentration of hydrofluoric acid, the concentration of nitric acid, the temperature of the etching solution, and the total concentration of metal impurities in the etching solution; S2: Substitute the process parameters into the etching rate prediction formula to calculate the predicted etching rate. The etching rate prediction formula is as follows:

[0007] In the formula: R is the predicted etching rate; R0 is the pre-exponential factor; E a It is the apparent activation energy; T represents the current etching solution temperature; [HF] represents the concentration of hydrofluoric acid in the current etching solution; [HNO3] represents the current concentration of nitric acid in the etching solution; α and β are the reaction orders of HF and HNO3, respectively; k is the impurity suppression coefficient; C imp This represents the total concentration of metallic impurities in the current etching solution.

[0008] Furthermore, in step S2, the pre-exponential factor R0 and the apparent activation energy E... a The methods for determining the reaction order α of HF, the reaction order β of HNO3, and the impurity inhibition coefficient k are as follows: S21. Record the etching thickness data of silicon samples under multiple sets of different experimental conditions, including hydrofluoric acid concentration, nitric acid concentration, temperature of etching solution and total concentration of metal impurities in etching solution. S22: Calculate the actual etching rate R under each experimental condition based on the etching thickness data of the silicon sample and the corresponding etching time; S23: Using the actual etching rate R as the training set, substitute the data from multiple sets of experimental conditions and the corresponding silicon sample etching thickness data into the etching rate prediction formula, and obtain the pre-exponential factor R0 and the apparent activation energy E through regression analysis. a The values ​​of the reaction order α of HF, the reaction order β of HNO3, and the impurity inhibition coefficient k.

[0009] Furthermore, in step S21, the formula for calculating the etching thickness of the silicon sample is: H=(M2-M1)÷ρ÷(6×L 2 ) Where M1 and M2 are the masses of the sample before and after etching, respectively, ρ is the density of silicon, and L is the side length of the sample.

[0010] Furthermore, the total concentration of metallic impurities in the etching solution is expressed in ppm as iron equivalent.

[0011] Furthermore, the pre-exponential factor R0 = 2.47 × 10 6 μm / min; Apparent activation energy E a =42.3kJ / mol; The reaction order of HF is α = 0.68; The reaction order of HNO3 is β = 0.42; Impurity suppression coefficient k = 0.0025 ppm -1 .

[0012] Advantages of this invention: This invention corrects the influence of multiple factors on the etching rate by incorporating key process parameters such as the concentration of hydrofluoric acid, nitric acid, temperature, and total concentration of metallic impurities in the etching solution into a comprehensive prediction model. This solves the problem of prediction distortion caused by neglecting the aging of the etching solution in traditional methods. Based on the prediction method proposed in this invention, the optimal etching time can be dynamically calculated for the aging degree of different batches. This fundamentally avoids over-etching and material waste caused by overestimating the rate in early batches, and under-etching and impurity residue caused by underestimating the rate in later batches, significantly improving the consistency and yield of product quality between batches. At the same time, this method can guide the maximization and online control of the etching solution's lifespan, extending the service life of expensive ultra-high purity acid solutions and significantly reducing production costs. Detailed Implementation

[0013] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example 1 A method for predicting the etching rate during the cleaning process of electronic-grade polysilicon includes the following steps: S1: Obtain the current concentrations of hydrofluoric acid, nitric acid, and temperature of the etching solution, as well as the total concentration of metallic impurities in the etching solution; S2: Substitute the process parameters into the etching rate prediction formula to calculate the predicted etching rate. The etching rate prediction formula is as follows:

[0015] In the formula: R is the predicted etching rate; R0 is the pre-exponential factor, R0 = 2.47 × 10 6 μm / min; E a E is the apparent activation energy. a =42.3kJ / mol; T represents the current etching solution temperature; [HF] represents the concentration of hydrofluoric acid in the current etching solution; [HNO3] represents the current concentration of nitric acid in the etching solution; α and β are the reaction orders of HF and HNO3, respectively, with α = 0.68 and β = 0.42. k is the impurity suppression coefficient, k = 0.0025 ppm -1 ; C imp This represents the total concentration of metallic impurities in the current etching solution.

[0016] Application Example 1 The dynamic control method for an electronic-grade polysilicon cleaning process provided in Example 1, used to control a cleaning batch in a normally functioning cleaning tank, includes the following steps: S1: Obtain the process parameters of the etching solution in the current cleaning tank. Specifically, the concentration of hydrofluoric acid [HF] is 3.8 mol / L, the concentration of nitric acid [HNO3] is 7.2 mol / L, the temperature of the etching solution is T = 298 K, and the total concentration of metallic impurities in the etching solution is C. imp =65ppb; S2: Substituting the above process parameters into the etching rate prediction formula, the predicted etching rate is calculated to be R = 0.463 μm / min; S3: Calculate the etched height of the sample by dividing the weight reduction after cleaning (during the cleaning process) by the density, and then by the surface area of ​​the sample. This height is 6.0 μm, which is the target etching depth. Based on the target etching depth D... target =6.0μm and the predicted etching rate, calculate the required baseline etching time t base =778 seconds; S4: Due to C imp =65ppb, which is lower than the preset first metal impurity concentration threshold of 155ppb, and does not meet the dynamic control conditions. Therefore, the system determines that the etching solution in the current cleaning tank is in a normal state and does not trigger the dynamic control strategy. It then uses the etching solution in the current cleaning tank according to the calculated baseline etching time t. base The cleaning process takes 778 seconds. The PLC-controlled robotic arm immerses the silicon core in the etching solution in the cleaning tank, and removes it after 778 seconds.

[0017] After cleaning, the actual etching depth was measured to be 5.95 μm, with a relative error of 0.83% compared to the target etching depth of 6.0 μm, and the surface metallic impurities Fe < 5 × 10⁻⁶. 9 atoms / cm 2 It fully meets electronic-grade requirements. The system did not trigger dynamic control strategies, and the cleaning task was efficiently completed using waste acid without the need to replenish fresh high-purity mixed acid after replacement.

[0018] Application Example 2 The method for dynamically controlling an electronic-grade polysilicon cleaning process, as provided in Example 1, is used to regulate a cleaning tank with a high accumulation of impurities. The method includes the following steps: S1: Obtain the process parameters of the etching solution in the current cleaning tank. Specifically, the concentration of hydrofluoric acid [HF] = 4.5 mol / L, the concentration of nitric acid [HNO3] = 8.6 mol / L, the temperature of the etching solution T = 298 K, and the total concentration of metallic impurities C in the etching solution. imp=185ppb; S2: Substituting the above process parameters into the etching rate prediction formula, the predicted etching rate is calculated to be R = 0.349 μm / min; S3: Based on the target etching depth D target =6.0μm and the predicted etching rate, calculate the required baseline etching time t base =1031 seconds; S4: Due to C imp =185ppb, which is higher than the preset first metal impurity concentration threshold of 155ppb, and meets the dynamic control conditions. Therefore, the system judges that the current state of the etching solution in the cleaning tank is abnormal and triggers the dynamic control strategy. S5: Due to C imp =185ppb, between the preset first metal impurity concentration threshold of 155ppb and the preset second metal impurity concentration threshold of 200ppb. If the current etching solution is used for cleaning, simply extending the etching time may cause surface roughness and uncontrollable defects. Therefore, an online doping strategy is implemented: fresh high-purity mixed acid is added to the cleaning tank at a volume ratio of 1:4 to dilute the metal impurity concentration and increase the acid concentration. The concentration of HF in the fresh high-purity mixed acid is 6.0mol / L, and the concentration of HNO3 is 11.0mol / L. After doping, the parameters of the new etching solution measured by the sensor are approximately [HF]=4.8mol / L, [HNO3]=9.08mol / L, C imp ≈150 ppb, lower than the preset first metal impurity concentration threshold of 155 ppb. R was recalculated. new The etching rate is approximately 0.437 μm / min. Based on the new etching rate, the new etching time is calculated to be 824 seconds. Therefore, the modified etching solution is used for cleaning according to the calculated new etching time of 824 seconds. The PLC-controlled robot arm immerses the silicon core in the etching solution in the cleaning tank and removes it after 824 seconds.

[0019] After cleaning, the actual etching depth was measured to be 6.08 μm, with a relative error of 1.3% compared to the target etching depth of 6.0 μm, and the surface metallic impurities Fe < 6 × 10⁻⁶. 9 atoms / cm 2 It fully meets electronic-grade requirements. By triggering a dynamic control strategy, the system replenishes a certain amount of fresh, high-purity mixed acid and then uses the mixed etching solution to complete the cleaning task, saving a batch of acid that was about to be scrapped and preventing product spoilage.

[0020] Comparative Example 1 The traditional control methods are used to control the cleaning process in use case 2, specifically as follows: The process parameters of the etching solution in the cleaning tank are: hydrofluoric acid concentration [HF] = 4.5 mol / L, nitric acid concentration [HNO3] = 8.6 mol / L, etching solution temperature T = 298 K, and total concentration of metallic impurities in the etching solution C. imp =185ppb; Using traditional methods, only the acid concentration is monitored to determine if it is within acceptable limits. The operator sets the etching time to 554 seconds based on an empirical formula.

[0021] After cleaning, testing revealed that the etching depth was only 3.91 μm, far below the target value of 6.0 μm, and the surface Fe impurities reached as high as 3.2 × 10¹. 0 atoms / cm 2 The product is scrapped.

[0022] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the etching rate during the cleaning process of electronic-grade polysilicon, characterized in that, Includes the following steps: S1: Obtain the process parameters of the etching solution in the current cleaning tank. The process parameters include the concentration of hydrofluoric acid, the concentration of nitric acid, the temperature of the etching solution, and the total concentration of metal impurities in the etching solution. S2: Substitute the process parameters into the etching rate prediction formula to calculate the predicted etching rate. The etching rate prediction formula is as follows: In the formula: R is the predicted etching rate; R0 is the pre-exponential factor; E a It is the apparent activation energy; T represents the current etching solution temperature, expressed in Kelvin. [HF] represents the current concentration of hydrofluoric acid in the etching solution; [HNO3] represents the current concentration of nitric acid in the etching solution; α and β are the reaction orders of HF and HNO3, respectively; k is the impurity suppression coefficient; C imp This represents the total concentration of metallic impurities in the current etching solution; In step S2, the pre-exponential factor R0 and the apparent activation energy E are... a The methods for determining the reaction order α of HF, the reaction order β of HNO3, and the impurity inhibition coefficient k are as follows: S21. Record the etching thickness data of silicon samples under multiple sets of different experimental conditions, including hydrofluoric acid concentration, nitric acid concentration, temperature of etching solution and total concentration of metal impurities in etching solution. S22: Calculate the actual etching rate R under each experimental condition based on the etching thickness data of the silicon sample and the corresponding etching time; S23: Using the actual etching rate R as the training set, substitute the data from multiple sets of experimental conditions and the corresponding silicon sample etching thickness data into the etching rate prediction formula, and obtain the pre-exponential factor R0 and the apparent activation energy E through regression analysis. a The values ​​of the reaction order α of HF, the reaction order β of HNO3, and the impurity inhibition coefficient k.

2. The etching rate prediction method in the electronic-grade polysilicon cleaning process according to claim 1, characterized in that, In step S21, the formula for calculating the etching thickness of the silicon sample is: H=(M2-M1)÷ρ÷(6×L 2 ) Where M1 and M2 are the masses of the sample before and after etching, respectively, ρ is the density of silicon, and L is the side length of the sample.

3. The etching rate prediction method in the electronic-grade polysilicon cleaning process according to claim 1, characterized in that, The total concentration of metallic impurities in the etching solution is expressed in ppm as iron equivalent.

4. The etching rate prediction method in the electronic-grade polysilicon cleaning process according to claim 1, characterized in that, Pre-exponential factor R0 = 2.47 × 10 6 μm / min; Apparent activation energy E a =42.3kJ / mol; The reaction order of HF is α = 0.68; The reaction order of HNO3 is β = 0.42; Impurity suppression coefficient k = 0.0025 ppm -1 .

Images