JP6760243B2 - A method for determining the quality of a silicon single crystal, a method for manufacturing a silicon single crystal, and a method for manufacturing a silicon wafer. - Google Patents

Description

The present invention relates to a method for determining the quality of a silicon single crystal, a method for producing a silicon single crystal, and a method for producing a silicon wafer.

In recent years, when setting operating conditions in a plant and identifying operation abnormality detection, a method using a pattern recognition technique called MT method (Mahalanobis Taguchi Method) has been used.
For example, Patent Document 1 describes MD (Mahalanobis Distance) based on a command value used for plant operation, a measured value measured in the plant, and group data composed of a plurality of variables representing the state of the plant. A technique for calculating a value and determining a plant abnormality based on the calculated MD value is disclosed.

Further, Patent Document 2 describes the maximum value and the residual of the power supplied to the heater wire in determining whether or not there is a sign of disconnection of the heater wire used in the heat treatment device in the semiconductor heat treatment device. A technique for calculating an MD value based on the sum of squares and determining whether or not there is a sign of disconnection of the heater wire is disclosed.

Japanese Unexamined Patent Publication No. 2013-41492 JP-A-2009-140933

However, in the techniques described in Patent Document 1 and Patent Document 2, since the MT method is applied as it is to the actual command values and measured values used in the plant and the heat treatment apparatus, variations between data and noise There is a problem that the discrimination accuracy is lowered because it may be affected by.

An object of the present invention is to provide a method for determining the quality of a silicon single crystal, a method for producing a silicon single crystal, and a method for producing a silicon wafer, which can determine the quality of a silicon single crystal with high accuracy based on the pulling control data of the silicon single crystal. There is.

The method for determining the quality of a silicon single crystal of the present invention is
Based on the pull-up control data of the silicon single crystal, the quality determination of the pulled-up silicon single crystal is a method for determining the quality of the silicon single crystal.
A pull-up control data extraction step for extracting pull-up control data from the start of pulling up the silicon single crystal to shoulder formation, and pull-up control data for straight body formation after shoulder formation, and
The pulling control data up to the shoulder formation and the pulling control data for forming the straight body are linked, and the standardization processing of each pulling control data is performed based on the pulling control data at the start of pulling the silicon single crystal. Standardization processing steps to be performed and
A moving average processing step that performs moving average processing of the standardized pull-up control data, and
A plurality of sample lines are set for the moving average processed pull-up control data, the number of the pull-up control data intersecting each sample line is quantified as a change amount, and the width of the mountain of the pull-up control data is calculated. A waveform processing step that digitizes the abundance,
A unit space calculation step for calculating the unit space of the pulling control data of the silicon single crystal, and
An MD value calculation step for calculating an MD value for the unit space calculated based on the amount of change and the abundance of the digitized pull-up control data, and
Based on the calculated MD value, the quality determination step for determining the quality of the interstitial oxygen concentration of the silicon single crystal pulled up by the pulling control data, and the quality determination step.
It is characterized by carrying out.

According to the present invention, by performing the standardization process, it is possible to eliminate the variation between batches of the raised silicon single crystal, determine the base point, and perform MT analysis of the amount of change.
Further, by performing the moving average processing, it is possible to remove the noise of the pull-up control data, which is the actual result, and perform the MT analysis.
Further, since the pull-up control data can be quantified by the waveform processing step, MT analysis can be performed based on the quantified amount of change and abundance.
Then, by the unit space calculation step, the MD value calculation step, and the quality determination step, the quality determination of the silicon single crystal can be performed with high accuracy based on the MT analysis.
Therefore, before performing MT analysis, noise of the actual pull-up control data can be removed and quantified appropriately, so that highly accurate MT analysis can be performed and the quality of the silicon single crystal can be determined. It can be done with high accuracy.

In the present invention, when the accuracy of determining the quality of the interstitial oxygen concentration of the silicon single crystal by the quality determination step is poor, a plurality of sample lines are reset, and the waveform is based on the preset plurality of sample lines. It is conceivable to carry out the re-waveform processing step for performing the processing, and then carry out the MD value calculation step and the pass / fail determination step.
According to the present invention, by carrying out the re-waveform processing step, the sample line can be reset centering on the portion that affects the interstitial oxygen concentration of the silicon single crystal, and therefore the sample line can be reset by the MD value calculation step. The MD value can be calculated with higher accuracy, and the accuracy of the quality determination of the silicon single crystal by the quality determination step is improved.

The method for producing a silicon single crystal of the present invention is
A method for producing a silicon single crystal that pulls up the silicon single crystal based on the pulling control data of the silicon single crystal, the step of carrying out the above-mentioned method for determining the quality of the silicon single crystal, and the quality of the silicon single crystal. Based on the implementation result of the discrimination method, in the pulling control data, while monitoring the step of identifying the control factor that affects the fluctuation of the interstitial oxygen concentration of the silicon single crystal and the fluctuation amount of the specified control factor. It is characterized in that the step of pulling up the silicon single crystal is carried out.

According to the present invention, the control factor that influences the fluctuation of the interstitial oxygen concentration of the silicon single crystal is specified in advance by carrying out the above-mentioned method for determining the quality of the silicon single crystal while pulling up the silicon single crystal. Can be done. Therefore, by pulling up the silicon single crystal while monitoring the fluctuation amount of the specified control factor, it is possible to obtain a silicon single crystal with less fluctuation in the interstitial oxygen concentration, and the silicon single crystal that fails can be obtained. The probability of sending to the next process can be reduced.

The method for manufacturing a silicon wafer of the present invention is a method for manufacturing a silicon wafer by cutting out a silicon wafer from the silicon single crystal obtained by the above-mentioned method for manufacturing a silicon single crystal to manufacture a silicon wafer. An inspection sample is cut out from the part of the silicon single crystal whose quality is expected to be out of specification by carrying out the process of pulling up, and the out-of-specification part is confirmed by performing quality evaluation, and the out-of-specification part is used in the wafer processing process. It is characterized by not flowing.

According to the present invention, an inspection sample is cut out from a site of a silicon single crystal whose quality is expected to be out of specification, and quality evaluation is performed. Therefore, since the non-standard portion is not flowed to the wafer processing process, the defective silicon wafer is not processed in the next process such as the grinding and polishing forming process, and waste in the next process can be reliably prevented.

The schematic diagram which shows the silicon single crystal pulling apparatus which concerns on embodiment of this invention. Schematic and graphs for explaining the background of the embodiment. The graph for demonstrating the background of the said embodiment. The flowchart for demonstrating the quality determination method of the silicon single crystal in the said embodiment. The schematic diagram for demonstrating the method of waveform processing in said embodiment. The graph which shows the variation of the oxygen concentration between lattices by the batch of the raised silicon single crystal in the said embodiment. The graph which confirmed the discrimination accuracy of each control factor by the MT method in the said embodiment. The graph which compares the discrimination accuracy of the good or bad judgment of the interstitial oxygen concentration of a silicon single crystal when the normalization process in the said embodiment is not performed, and when it is performed. The graph which shows the setting of the sample line in the waveform processing in the said embodiment, and the discrimination accuracy of the quality judgment of the interstitial oxygen concentration of a silicon single crystal. The graph which shows the determination accuracy of the resetting of a sample line in the waveform processing in the said embodiment, and the judgment of the quality of the interstitial oxygen concentration of a silicon single crystal. The graph which shows the result of the moving averaging process in the said embodiment, and the discrimination accuracy of the quality judgment of the interstitial oxygen concentration of a silicon single crystal. The flowchart which shows the manufacturing method of the silicon single crystal in the said embodiment. The flowchart which shows the manufacturing method of the silicon wafer in the said embodiment. The graph which shows the heater power intensity before improvement which concerns on the comparative example of this invention. The graph which shows the heater power strength after improvement which concerns on Example of this invention. The graph which shows the average value and variation of the shoulder length in an Example and a comparative example of this invention. The graph which shows the average value and variation of the interstitial oxygen concentration [Oi] in the Example and the comparative example of this invention.

[1] Structure of Silicon Single Crystal Pulling Device 1 FIG. 1 shows a schematic diagram showing an example of the structure of a silicon single crystal pulling device 1 to which the method for producing a silicon single crystal according to the embodiment of the present invention can be applied. Has been done. The pulling device 1 is a device for pulling the silicon single crystal 10 by the Czochralski method, and includes a chamber 2 forming an outer shell and a crucible 3 arranged in the center of the chamber 2.
The crucible 3 has a double structure composed of an inner quartz crucible 3A and an outer graphite crucible 3B, and is fixed to the upper end of a support shaft 4 capable of rotating and raising and lowering.

On the outside of the crucible 3, resistance heating type two-stage heaters 5A and 5B surrounding the crucible 3 are provided, and on the outside thereof, a heat insulating material 6 is provided along the inner surface of the chamber 2.
Above the crucible 3, a pull-up shaft 7 such as a wire that rotates coaxially with the support shaft 4 in the opposite direction or the same direction at a predetermined speed is provided. A seed crystal 8 is attached to the lower end of the pulling shaft 7.

A tubular heat shield plate 12 is arranged in the chamber 2.
The heat shield plate 12 blocks high-temperature radiant heat from the silicon melt 9 in the crucible 3, heaters 5A and 5B, and the side wall of the crucible 3 with respect to the growing silicon single crystal 10, and is a crystal growth interface. In the vicinity of the solid-liquid interface, it suppresses the diffusion of heat to the outside and plays a role of controlling the temperature gradient in the pulling axial direction of the central portion of the single crystal and the outer peripheral portion of the single crystal.

A gas introduction port 13 for introducing an inert gas such as Ar gas into the chamber 2 is provided in the upper part of the chamber 2. An exhaust port 14 is provided in the lower part of the chamber 2 to suck and discharge the gas in the chamber 2 by driving a vacuum pump (not shown).
The inert gas introduced into the chamber 2 from the gas introduction port 13 descends between the growing silicon single crystal 10 and the heat shield plate 12, and the lower end of the heat shield plate 12 and the liquid level of the silicon melt 9. After passing through the gap between the gas and the gas, the gas flows toward the outside of the heat shield plate 12 and further toward the outside of the crucible 3, then descends the outside of the crucible 3 and is discharged from the exhaust port 14.

When growing the silicon single crystal 10 using such a pulling device 1, the heater 5A uses a solid raw material such as polycrystalline silicon filled in the crucible 3 while maintaining the inside of the chamber 2 in an inert gas atmosphere under reduced pressure. The silicon melt 9 is formed by melting by heating 5B. When the silicon melt 9 is formed in the crucible 3, the pulling shaft 7 is lowered to immerse the seed crystal 8 in the silicon melt 9, and the crucible 3 and the pulling shaft 7 are rotated in a predetermined direction while pulling the shaft. 7 is gradually pulled up, thereby growing a silicon single crystal 10 connected to the seed crystal 8.

[2] Background to the Present Invention The silicon single crystal 10 pulled up by the pulling device 1 having the above-mentioned structure is cut into product blocks 101, 102 and the like as shown in FIG. 2 (A). At the time of cutting, sample wafers SW1, SW2, and SW3 are sampled at the ends of the product blocks 101 and 102, and the interstitial oxygen concentration [Oi] is measured for each of the sample wafers SW1, SW2, and SW3. If the interstitial oxygen concentration [Oi] of SW1 and SW2 is equal to or less than a predetermined threshold value, it is determined that all the interstitial oxygen concentrations [Oi] of the product block 101 are equal to or less than the threshold value, and the wafer is sent to the next step. ..

However, for the sample wafers SW1 and SW2 at the end of the product block 101, the sample wafer SW4 was sampled from the inside of the product block 101 for the product block 101 for which it was determined that the interstitial oxygen concentration [Oi] had no problem, and the interstitial oxygen concentration [Oi] was determined to be no problem. When the oxygen concentration [Oi] was measured, it was found that the interstitial oxygen concentration [Oi] of the sample wafer SW4 exceeded the threshold value as shown in FIG. 2 (B).

That is, based on the measurement results of the interstitial oxygen concentrations [Oi] of the sample wafers SW1, SW2, and SW3 collected from the end portions, it is determined that the interstitial oxygen concentrations [Oi] of the product blocks 101 and 102 are all equal to or less than the threshold value. Then, there is a possibility that a product having an interstitial oxygen concentration [Oi] exceeding the threshold value may be mixed inside the product blocks 101 and 102, which causes a problem that a defective portion may be sent to the next step.
There is also a method of measuring the interstitial oxygen concentration [Oi] by scanning the entire length of the product blocks 101 and 102, but there is a problem in processing capacity and it is impossible to measure all of them.

The present inventors have diligently investigated what factors cause the interstitial oxygen concentration [Oi] between the pulling devices 1 and the batches of the silicon single crystal 10 to vary. As a result, FIGS. 3 (A) and 3 (Fig. 3) As shown in B), it was found that the pulling control data of the silicon single crystal 10 affects the variation in the interstitial oxygen concentration [Oi] inside the silicon single crystal 10.
Therefore, the present inventors decided to analyze the pull-up control data using the MT method to understand the factors that affect the interstitial oxygen concentration [Oi].

[3] Method for determining the quality of a silicon single crystal FIG. 4 shows a flowchart showing a method for determining the quality of a silicon polycrystal according to the present embodiment.
First, the factors in the pull-up control data of the silicon single crystal 10 are extracted (step S1: pull-up control data extraction step). Specifically, the pulling control data includes the straight body diameter of the silicon single crystal 10, the heater power of the heaters 5A and 5B, the vertical heater power ratio of the heaters 5A and 5B, the heater temperature, and the pulling speed of the silicon single crystal 10 (seed). The ascending speed), the number of rotations of the crucible 3, the magnetic field strength, and the like, and the factors that affect the interstitial oxygen concentration [Oi] of the silicon single crystal 10 are extracted from these.

Next, regarding the extracted pull-up control data, the pull-up control data up to the shoulder formation of the silicon single crystal 10 and the pull-up control data of the straight body formation after the shoulder portion formation of the silicon single crystal 10 are linked and pulled up. The control data standardization process is performed (step S2: standardization process step).
Specifically, in the pulling control data up to the shoulder formation, the pulling control data of the straight body portion is converted and standardized based on the pulling control data at the start of pulling the shoulder portion of the silicon single crystal 10.
Moving average processing is performed on the standardized pull-up control data (step S3: moving average processing step). Specifically, the noise of the pull-up control data is removed by calculating the average value of the pull-up control data for 10 minutes and using this as the representative value of the pull-up control data.

Waveform processing is performed on the pull-up control data that has undergone moving average processing (step S4: waveform processing step).
Specifically, in the waveform processing, as shown in FIG. 5, a plurality of sample lines L1 to sample lines L4 are set for the pull-up control data L0, and each sample line L1 to the sample line L4 intersects with the sample line L4. The number of control data is quantified as the amount of change, and the width of the peak of the raised control data is quantified as the amount of existence. For example, in the case of the sample line L1, since the number of points where the pull-up control data L0 intersects is 4, the amount of change is 4, and the width of the mountain W1 + W2 is the existence amount. Further, in the case of the sample line L2, since the number of points where the pull-up control data L0 intersects is 6, the amount of change is 6, and the width of the mountain W3 + W4 + W5 is the abundance amount. The same applies to the sample line L3 and the sample line L4. The number of sample lines is not limited to four, and can be increased according to the number of pull-up control data of the accepted silicon single crystal 10 constituting the unit space described later.

Returning to FIG. 4, the unit space is calculated based on the pulling control data of the plurality of silicon single crystals 10 in which the interstitial oxygen concentration [Oi] is all good (passed product) (step S5: unit space calculation step). ). Specifically, the unit space is calculated from the change amount and the abundance amount of a plurality of pull-up control data of the known accepted silicon single crystal 10.

The MD value of the pull-up control data of the rejected product of the silicon single crystal 10 is calculated based on the amount of change and the abundance of the digitized pull-up control data (step S6: MD value calculation step). Specifically, a threshold value that can be discriminated from the MD value of the pull-up control data of the rejected product and the MD value of the unit space is obtained. Next, the pull-up control data of rejected products is standardized, a covariance matrix is created, and the inverse matrix of the covariance matrix is calculated. Next, an operation is performed by sandwiching both sides of the inverse matrix with data, the square value of the MD value is calculated, and the square root thereof is taken to calculate the MD value.

MT analysis of the calculated MD value is performed to determine the quality of the interstitial oxygen concentration [Oi] of the silicon single crystal 10 to be inspected (step S7: quality determination step).
It is determined whether or not the accuracy of the pass / fail judgment is sufficient (step S8), and if the accuracy of the pass / fail judgment is poor, the process returns to step S4, a plurality of sample lines are reset, and a plurality of reset sample lines are reset. Based on the above, the re-waveform processing step is carried out, and then steps S5 to S7 are carried out thereafter.
Hereinafter, each step will be described in detail with respect to the specific verification results performed by the present inventors.

[4] Extraction of factors in pull-up control data (step S1)
As shown in FIG. 6, the present inventors have defined 37 acceptable products (OK) in which the interstitial oxygen concentration [Oi] was equal to or lower than a predetermined threshold among the plurality of silicon single crystals 10 raised in the past. For 9 rejected products (NG) that exceeded the threshold of, the diameter of the straight body of the silicon single crystal 10, the power of the heaters 5A and 5B, the vertical power ratio of the heaters 5A and 5B, the heater temperature, and the seed rise rate. The rotation speed and magnetic field strength of the crucible 3 were compared as pull-up control data.
However, a clear difference between the accepted product and the rejected product was not found only by comparing the individual control data of the passed product and the rejected product. Therefore, the analysis was performed by the MT method using the diameter of the straight body of the silicon single crystal 10, the heater power of the heaters 5A and 5B, the vertical power ratio of the heaters 5A and 5B, the heater temperature, and the seed rise rate as factors.

FIG. 7 shows the results of calculating the MD values of the pull-up control data for 37 accepted products and 9 rejected products of the silicon single crystal 10.
First, FIG. 7A is a graph showing the results of calculating the MD values of the accepted product (OK) and the rejected product (NG) for the diameter of the straight body portion of the silicon single crystal 10. ing.
In the case of FIG. 7A, only 22% (2) of the 9 rejected products could be determined to be rejected products, and the silicon single crystal could be determined. It was found that the diameter of the straight body portion of No. 10 is inappropriate as the pulling control data for determining the quality of the silicon single crystal 10.

Similarly, as shown in FIG. 7B, with respect to the upper and lower heater power ratios, it is not possible to determine that all nine rejected products are rejected products, and this is used as pull-up control data for determining the quality of the silicon single crystal 10. It turned out to be inappropriate.
Further, as shown in FIG. 7 (F), regarding the pulling speed, only 11% (1) of the 9 rejected products can be discriminated, and the pulling control for determining the quality of the silicon single crystal 10 is controlled. It turned out to be inappropriate as data.

On the other hand, as shown in FIG. 7C, regarding the heater power of the heaters 5A and 5B, 67% (6) of the 9 rejected products can be determined to be rejected products, and silicon. It was confirmed that there is a possibility that it can be used as pull-up control data for determining the quality of the single crystal 10.
Further, as shown in FIG. 7 (D), 67% (6) of the 9 rejected products can be determined to be rejected with respect to the heater temperature, and the quality of the silicon single crystal 10 can be determined. It was confirmed that there is a possibility that it can be used as pull-up control data.

Further, as shown in FIG. 7 (E), regarding the heater power of the heaters 5A and 5B that have undergone the standardization processing described later, 78% (7) of the 9 rejected products are rejected products. It was confirmed that there is a possibility that it can be used as pull-up control data for determining the quality of the silicon single crystal 10.
Based on the above, it was decided to extract the heater power and heater temperature as control data, and determine whether the interstitial oxygen defect concentration [Oi] of the silicon single crystal 10 was good or bad.

[5] Normalization process of pull-up control data (step S2)
The number of passed products of the silicon single crystal 10 was increased from 37 to 52 in order to increase the number of sample lines to improve the discrimination accuracy, and the MD values were calculated for the passed products and the rejected products of the silicon single crystal 10. FIG. 8A shows an MD value calculated based on the actual value. In FIG. 8B, the pulling control data of the straight body portion of the silicon single crystal 10 is linked to the pulling control data up to the shoulder formation, and standardized based on the control data at the start of pulling the silicon single crystal 10. It was done.
In FIG. 8 (A) based on the actual values, only 44% (4) of the 9 rejected products could be determined to be rejected products.

On the other hand, in FIG. 8B, which has undergone standardization processing, 67% (6) of the 9 rejected products could be determined to be rejected products.
Therefore, it has been confirmed that the discrimination accuracy is improved because the variation between batches due to the replacement of parts of the pulling device 1 and the change with time can be canceled by performing the standardization process of the pulling control data.

[6] Waveform processing of pull-up control data (step S4)
As shown in FIG. 9 (A), 6 lines at the shoulder forming position of the silicon single crystal 10 and 7 lines at the straight body forming position of the silicon single crystal 10 are uniform with respect to the entire change of the pulling control data. Sample lines were set at various intervals. The number of heater power raising control data intersecting each sample line was quantified as the amount of change, and the width of the peak of the raising control data was quantified as the abundance amount to calculate the MD value.
When the sample lines are evenly arranged on the shoulder and the straight body of the silicon single crystal 10, it can be determined that the nine rejected products are rejected, as shown in FIG. 9B. It was 67% (6 bottles).

Next, as shown in FIG. 10 (A), 6 at the shoulder forming position of the silicon single crystal 10, 7 at the straight body forming position of the silicon single crystal 10, and the shoulder forming position of the silicon single crystal 10. From, the sample line was set in a dense position at the position where it shifts to the straight body. The number of heater power raising control data intersecting each sample line was quantified as the amount of change, and the width of the peak of the raising control data was quantified as the abundance amount to calculate the MD value.
When the sample lines are densely arranged on the shoulder and the straight body of the silicon single crystal 10, it is determined that the nine rejected products are rejected as shown in FIG. 10 (B). Only 78% (7) were produced.
Therefore, in the waveform processing, it is not possible to arrange the sample lines densely at the position where the silicon single crystal 10 shifts from the shoulder forming position to the straight body portion with respect to the pulling control data of the silicon single crystal 10. It was confirmed that the discrimination accuracy of accepted products was improved.

[7] Moving average processing of pull-up control data (step S3)
When the pull-up control data is subjected to the moving average processing for 10 minutes, as shown in FIG. 11 (A), the waveform of the pull-up control data becomes clear with noise removed. After the moving average treatment, 89% (8 pieces) of the 9 rejected products of the silicon single crystal 10 could be determined to be rejected products as shown in FIG. 11 (B).
When the moving averaging process is performed, the discrimination accuracy of the rejected products increases to 89% (8 pieces), while the discrimination accuracy of 78% in the pull-up control data without the moving average process shown in FIG. 10 (A). ing.

[8] Summary According to this embodiment, the following items were found.
By analyzing the heater powers of the heaters 5A and 5B by the MT method as the raising control data, it is possible to discriminate the variation in the interstitial oxygen concentration [Oi] of the raised silicon single crystal 10 with a discrimination accuracy of 89%. I was able to confirm.
When performing analysis by the MT method, it is possible to improve the discrimination accuracy of the passed product and the rejected product of the silicon single crystal 10 by performing standardization processing, moving average processing, and waveform processing of the pulling control data in advance. It could be confirmed.

When performing waveform processing, by concentrating the sample lines from the shoulder forming position of the silicon single crystal 10 to the position where it shifts to the straight body portion, the accuracy of distinguishing the accepted product and the rejected product of the silicon single crystal 10 can be determined. I was able to confirm that it could be increased.
When pulling up the silicon single crystal 10, it is preferable to strengthen the monitoring of the heater power of the heaters 5A and 5B as the pulling control data, and in particular, strengthen the monitoring in the portion transitioning from the shoulder formation position to the straight body formation position. By doing so, it is predicted that the silicon single crystal 10 having little variation in the interstitial oxygen concentration [Oi] can be pulled up.

[9] Method for Producing Silicon Single Crystal 10 Next, a flowchart shown in FIG. 12 regarding a method for producing the silicon single crystal 10 for pulling up the silicon single crystal 10 based on the above-mentioned method for determining the quality of the silicon single crystal 10. The explanation will be based on.
Based on the pull-up control data of the silicon single crystal 10, the quality determination method of the silicon single crystal 10 is carried out (steps S1-S8).

Based on the execution result of the quality determination method for the silicon single crystal 10, the control factor in the pull-up control data is specified (step S9).
While monitoring the fluctuation amount of the specified control factor, the silicon single crystal 10 is pulled up to produce the silicon single crystal 10 (step S10).

[10] Method for Manufacturing a Silicon Wafer Next, FIG. 13 shows a method for manufacturing a silicon wafer by cutting out a silicon wafer from the silicon single crystal 10 manufactured by the above-described method for manufacturing a silicon single crystal 10 to manufacture a silicon wafer. The explanation will be given based on the flowchart.
The silicon single crystal 10 is produced by the method for producing the silicon single crystal 10 (steps S1-S10).

The outer circumference of the silicon single crystal 10 is ground (step S11).
The silicon single crystal 10 is cut into a plurality of ingot blocks having a predetermined length (step S12).
Along with cutting each ingot block, a plurality of test samples are collected from the end of the ingot block (step S13).
A quality inspection of the obtained plurality of inspection samples is carried out (step S14).
It is determined whether the inspection sample subjected to the quality inspection is out of specification (step S15).

If it is determined that the inspection sample is out of specification, the out-of-specification part is excluded as a defective product (step S16), and the out-of-specification part is not flowed to the wafer processing process.
When it is determined that the inspection sample is not within the standard range, the site is dispensed to the wafer processing step (step S17).
In the wafer processing step, the silicon wafer is sliced and processed such as grinding or wrapping, chamfering, etching, and mirror polishing (step S18).

[11] Actions and Effects of the Embodiment According to the present embodiment, the silicon single crystal 10 is previously subjected to the above-mentioned method for determining the quality of the silicon single crystal 10 while the silicon single crystal 10 is being pulled up. The control factors that influence the fluctuation of interstitial oxygen concentration can be identified. Therefore, by pulling up the silicon single crystal 10 while monitoring the fluctuation amount of the specified control factor, it is possible to obtain the silicon single crystal 10 having less fluctuation in the interstitial oxygen concentration, and the silicon single crystal is rejected. The probability of sending the crystal to the next process can be reduced.

In addition, the inspection sample is cut out from the site of the silicon single crystal whose quality is expected to be out of specification, and the quality is evaluated. Therefore, since the non-standard portion is not flowed to the wafer processing process, the defective silicon wafer is not processed in the next process such as the grinding and polishing processes, and waste in the next process can be reliably prevented.

Next, examples of the present invention will be described. The present invention is not limited to the following examples.
In pulling up the silicon single crystal 10 of the IGBT (Insulated Gate Bipolar Transistor) crystal (low oxygen MCZ: 200 mm product) in a certain factory, the variation of the interstitial oxygen concentration [Oi] in the Top between furnaces and batches became a problem. ..
Therefore, as a result of the analysis by the MT method described above, it was clarified that the power behavior of the heaters 5A and 5B affects the interstitial oxygen concentration [Oi].

Crystals for n-type 200 mm wafers with a resistivity target value of 60 Ω cm were pulled up by the Czochralski method in which a transverse magnetic field was applied using a multi-stage heater, and the interstitial oxygen concentration [Oi] before and after the improvement was compared.
In the comparative example before the improvement, the sudden shape change of the latter half of the shoulder portion (especially before the shoulder change) of the silicon single crystal 10 induces the heater power fluctuation (higher power), and the first half portion (100 mm) of the straight body portion Body. It is affected by diameter and power fluctuations up to (near). Specifically, as shown in FIG. 14, the power behaviors of the heaters 5A and 5B fluctuate, and in the latter half of the shoulder portion, the heater power behavior gradually shifts to the straight body portion Body.

In the improved example, the analysis results by the MT method are used to suppress fluctuations in the heaters 5A and 5B in the latter half of the shoulder (especially before changing the shoulder) so that the process can smoothly shift to the straight body body forming process. did. Specifically, temperature profile tuning and shoulder scanning line adjustment of the control camera were carried out aiming at the shoulder shape of the silicon single crystal 10. As a result, as shown in FIG. 15, fluctuations in the power of the heaters 5A and 5B in the latter half of the shoulder portion were suppressed, and the transition to the straight body portion Body could be performed smoothly.

Comparing the shoulder lengths before and after the improvement, as shown in FIG. 16, the average value of the shoulder lengths of the examples after the improvement decreased as compared with the shoulder lengths of the comparative examples before the improvement. , It was confirmed that the variation was significantly improved.
Comparing the interstitial oxygen concentration [Oi] in the silicon single crystal 10 before and after the improvement, as shown in FIG. 17, the interstitial oxygen concentration [Oi] of the comparative example before the improvement was compared with the average value. the average value of interstitial oxygen concentration is after improvement embodiment [Oi] ^{is, 0.3 × 10 17 atoms / cm} 3 has decreased, in the embodiment, more than the standard value of interstitial oxygen concentration [Oi] No site of the silicon single crystal 10 was formed.
The interstitial oxygen concentration [Oi] is a value measured by FTIR (Fourier Transform Infrared Spectroscopy) specified in ASTM F-121 (1979).

Considering these results, it is speculated that an excessive increase in heater power to control abrupt shoulder shape changes promoted oxygen dissolution from the quartz crucible 3A and increased the level of interstitial oxygen concentration [Oi]. To. It is considered that by stabilizing the shoulder shape of the silicon single crystal 10, the heater power was also stabilized, and the average value level of the interstitial oxygen concentration [Oi] was lowered.
From the above, it is possible to obtain a silicon single crystal 10 having a low average value level of the interstitial oxygen concentration [Oi] by carrying out the method for determining the quality of a silicon single crystal of the present invention to produce a silicon single crystal. It was confirmed that it could be done.

1 ... Pulling device, 2 ... Chamber, 3 ... Crucible, 3A ... Quartz crucible, 3B ... Graphite crucible, 4 ... Support shaft, 5A, 5B ... Heater, 6 ... Insulation material, 7 ... Pulling shaft, 8 ... Seed crystal, 9 ... Silicon melt, 10 ... Silicon single crystal, 12 ... Heat shield plate, 13 ... Gas inlet, 14 ... Exhaust port, 101 ... Product block, 102 ... Product block, L0 ... Pulling control data, L1, L2, L3, L4 ... sample line, SW1 ... sample wafer, SW2 ... sample wafer, SW3 ... sample wafer, SW4 ... sample wafer.

Claims (4)

A method for determining the quality of a silicon single crystal, which evaluates the quality of the pulled-up silicon single crystal based on the pull-up control data of the silicon single crystal.
A pull-up control data extraction step for extracting pull-up control data from the start of pulling up the silicon single crystal to shoulder formation, and pull-up control data for straight body formation after shoulder formation, and
The pulling control data up to the shoulder formation and the pulling control data for forming the straight body are linked, and the standardization processing of each pulling control data is performed based on the pulling control data at the start of pulling the silicon single crystal. Standardization processing steps to be performed and
A moving average processing step that performs moving average processing of the standardized pull-up control data, and
A plurality of sample lines are set for the moving average processed pull-up control data, the number of the pull-up control data intersecting each sample line is quantified as a change amount, and the width of the mountain of the pull-up control data is calculated. A waveform processing step that digitizes the abundance,
A unit space calculation step for calculating the unit space of the pulling control data of the silicon single crystal, and
An MD value calculation step for calculating an MD (Mahalanobis Distance) value with respect to the unit space calculated based on the amount of change and the abundance of the digitized pull-up control data.
Based on the calculated MD value, the quality determination step for determining the quality of the interstitial oxygen concentration of the silicon single crystal pulled up by the pulling control data, and the quality determination step.
A method for determining the quality of a silicon single crystal, which comprises carrying out. In the method for determining the quality of a silicon single crystal according to claim 1,
When the accuracy of determining the quality of the interstitial oxygen concentration of the silicon single crystal by the quality determination step is poor, a plurality of sample lines are reset and waveform processing is performed based on the preset plurality of sample lines. A method for determining the quality of a silicon single crystal, which comprises performing a waveform processing step, performing the unit space calculation step, the MD value calculation step, and the quality determination step. A method for producing a silicon single crystal, which pulls up the silicon single crystal based on the pulling control data of the silicon single crystal.
The step of carrying out the method for determining the quality of a silicon single crystal according to claim 1 or 2.
Based on the implementation result of the method for determining the quality of the silicon single crystal, the step of identifying the control factor that affects the fluctuation of the interstitial oxygen concentration of the silicon single crystal in the pulling control data, and
The step of pulling up the silicon single crystal while monitoring the fluctuation amount of the identified control factor, and
A method for producing a silicon single crystal, which comprises carrying out. A method for manufacturing a silicon wafer, which is a method for manufacturing a silicon wafer by cutting out a silicon wafer from the silicon single crystal obtained by the method for manufacturing a silicon single crystal according to claim 3.
The step of pulling up the silicon single crystal is carried out, an inspection sample is cut out from the part of the silicon single crystal whose quality is expected to be out of specification, and the non-standard part is confirmed by performing quality evaluation, and the non-standard part is confirmed. A method for manufacturing a silicon wafer, which is characterized in that the wafer is not flowed into a wafer processing process.

Images