JP2018193270A - Evaluation method of silicon single crystal, and production method of silicon single crystal - Google Patents

Abstract

To provide an evaluation method of a silicon single crystal capable of estimating an occurrence state of LPD in an early stage after production of the silicon single crystal.SOLUTION: An evaluation method of a silicon single crystal includes an oxygen concentration measurement step for measuring an oxygen concentration in the silicon single crystal containing COP, and an LPD occurrence state estimation step for estimating the occurrence state of LPD of the silicon single crystal.SELECTED DRAWING: Figure 12

Description

  The present invention relates to a silicon single crystal evaluation method and a silicon single crystal manufacturing method.

A silicon wafer used as a substrate of a semiconductor device (hereinafter sometimes referred to as “wafer”) is generally cut out from a silicon single crystal grown by the Czochralski method (hereinafter sometimes referred to as “CZ method”). Then, it is manufactured through processes such as polishing and heat treatment.
FIG. 1 is a longitudinal sectional view of a pulled silicon single crystal, and schematically shows an example of the relationship between defect distribution and V / G. V is the pulling speed of the silicon single crystal, and G is the temperature gradient in the growth direction of the silicon single crystal immediately after the pulling. The temperature gradient G is considered to be substantially constant during the pulling of the silicon single crystal due to the thermal characteristics of the hot zone structure of the CZ furnace. For this reason, V / G can be controlled by adjusting the pulling speed V.

In FIG. 1, COP (Crystal Originated Particle) is an agglomeration of vacancies lacking atoms that should constitute a crystal lattice when a silicon single crystal is grown.
When an OSF (Oxidation induced Stacking Fault) region is thermally oxidized at a high temperature (generally 1000 ° C. to 1200 ° C.), OSF nuclei become apparent as OSF.
P _V region and the _{P I} area, COP also not include dislocation clusters, _{P V} region includes oxygen precipitation nuclei in as-grown state, when subjected to heat treatment, oxygen precipitate (BMD) occurs easy. As shown in FIG. 2, as V / G increases, the concentration of hole-type point defects in the _Pv region increases.
P _I area does not include the most oxygen precipitation nuclei in as-grown state, BMD be subjected to a heat treatment hardly occurs.

A method for evaluating the number of COPs in manufacturing such a silicon single crystal is known (see, for example, Patent Document 1).
Patent Document 1 discloses a method for evaluating the number of COPs after mirror-polishing a wafer cut from a silicon single crystal.

Japanese Patent Laying-Open No. 2005-281071

  However, in the method described in Patent Document 1, since it takes a long time to complete COP evaluation after mirror polishing after the production of a silicon single crystal, LPD (Light Point Defect) caused by COP ), It may take time to obtain an evaluation result.

  An object of the present invention is to provide a method for evaluating a silicon single crystal and a method for producing a silicon single crystal that can estimate the occurrence of LPD at an early stage after the production of the silicon single crystal.

  The method for evaluating a silicon single crystal according to the present invention includes an oxygen concentration measurement step for measuring the oxygen concentration of a silicon single crystal containing COP, and the LPD occurrence state of the silicon single crystal is estimated based on the measurement result of the oxygen concentration. And an LPD occurrence state estimating step.

The measurement of the oxygen concentration of a silicon single crystal is generally performed on a sample wafer sampled at the time of block division after pulling the silicon single crystal. For this reason, the measurement result of the oxygen concentration can be obtained at an earlier stage than the evaluation result of the mirror-polished wafer.
According to the present invention, it is possible to estimate the occurrence of LPD at an early stage after the production of a silicon single crystal based on the measurement result of the oxygen concentration.

  In the silicon single crystal evaluation method of the present invention, the oxygen concentration measurement step preferably measures the oxygen concentration in a region where the solidification rate in the straight body portion is 15% or less.

According to the present invention, it is possible to accurately estimate the LPD occurrence state by a simple method of selecting an oxygen concentration measurement position.
The solidification rate means the ratio of the solidified weight to the total weight of the silicon melt before pulling up the silicon single crystal.

  A method for producing a silicon single crystal according to the present invention includes a crucible, a heater that generates a silicon melt by heating the crucible, and a pulling unit that pulls up a seed crystal after contacting the silicon melt. A method for producing a silicon single crystal by the Czochralski method using a crystal pulling apparatus, the production process for producing a first silicon single crystal containing COP, and the above-described silicon single crystal evaluation method, An evaluation step for estimating the LPD occurrence state of the first silicon single crystal, and a manufacturing condition setting step for setting a manufacturing condition for the second silicon single crystal including the COP based on the estimation result in the evaluation step. It is characterized by.

In the case where the first silicon single crystal is a defective product in which a large amount of LDP is generated, all the silicon single crystals manufactured thereafter are defective products until the manufacturing conditions are changed.
According to the present invention, since the manufacturing condition of the second silicon single crystal is set based on the estimation result of the LPD occurrence state obtained at an early stage after the manufacturing of the first silicon single crystal, the number of defective products Can be reduced.

Explanatory drawing which shows an example of the relationship between defect distribution and V / G in a silicon single crystal. The graph which shows the relationship between V / G and a vacancy-type point defect density | concentration in a silicon single crystal. It is a schematic diagram which shows schematic structure of the single crystal pulling apparatus used in Experiment 1, 2 conducted in order to guide | lead this invention, and one Embodiment of this invention. It is a schematic diagram which shows schematic structure of the said single crystal pulling apparatus, and shows the state which pulling progressed rather than the state of FIG. The graph which shows the relationship between the heater batch number in the said experiment 1, and the moving average value of the oxygen concentration of a 1st area | region. The graph which shows the relationship between the number of heater batches in the said experiment 1, and the moving average value of LPD of the 1st field. The graph which shows the relationship between the moving average value of the oxygen concentration of the 1st area | region in the said experiment 1, and the moving average value of LPD. The graph which shows the relationship between the moving average value of the oxygen concentration of the 2nd area | region in the said experiment 2, and the moving average value of LPD. The graph which shows the relationship between the moving average value of the oxygen concentration of the 3rd area | region in the said experiment 2, and the moving average value of LPD. The graph which shows the relationship between the moving average value of the oxygen concentration of the 4th area | region in the said experiment 2, and the moving average value of LPD. The graph which shows the relationship between the solidification rate in the said experiment 2, and the absolute value of the correlation coefficient of the moving average value of oxygen concentration, and the moving average value of LPD. The flowchart which shows the manufacturing method of the silicon single crystal in the said one Embodiment. Explanatory drawing of the effect of the said one Embodiment.

[Background to Leading the Present Invention]
[Experiment 1: Investigation of relationship between oxygen concentration of silicon single crystal and LPD generation status]
First, the single crystal pulling apparatus used in this experiment will be described.
As shown in FIG. 3, the single crystal pulling apparatus 1 is an apparatus used for the CZ method (Czochralski method), and includes a pulling apparatus main body 2, a memory 3, and a control unit 4.
The lifting device main body 2 includes a chamber 21, a crucible 22 disposed at the center of the chamber 21, a heater 23 for heating the crucible 22, a lifting portion 24, and a heat shield 25.
A gas inlet 21 </ b> A for introducing an inert gas such as Ar gas into the chamber 21 is provided at the upper portion of the chamber 21. A gas exhaust port 21 </ b> B that exhausts the gas in the chamber 21 is provided in the lower portion of the chamber 21.
The crucible 22 melts polycrystalline silicon, which is a raw material of a wafer, to form a silicon melt M. The crucible 22 is supported by a support shaft 26 that can be rotated and moved up and down at a predetermined speed.
The heater 23 is disposed around the crucible 22 and melts silicon in the crucible 22.
The pulling unit 24 includes a pulling cable 241 to which the seed crystal SC is attached at one end, and a pulling drive unit 242 that moves the pulling cable 241 up and down and rotates.
The heat shield 25 blocks radiant heat radiated upward from the heater 23.
The memory 3 stores various information necessary for manufacturing the silicon single crystal SM, such as the gas flow rate in the chamber 21 and the furnace pressure, the electric power supplied to the heater 23, and the rotational speed of the crucible 22 and the silicon single crystal SM. .

A silicon single crystal SM having a diameter of 200 mm was manufactured using the single crystal pulling apparatus 1 described above. Specifically, a new heater 23 was attached to the chamber 21. Then, after the chamber 21 is maintained in an inert gas atmosphere under reduced pressure and the seed crystal SC is deposited on the silicon melt M, the value of V / G becomes larger than the value corresponding to A in FIG. By pulling up in this way, a silicon single crystal SM having a neck portion SM1, a shoulder portion SM2, a straight body portion SM3, and a tail portion (not shown) and including COP in the straight body portion SM3 was manufactured. During this pulling up, the crucible 22 rotates and the distance GP (gap GP) between the liquid surface of the silicon melt M and the lower end of the thermal shield 25 becomes substantially constant as shown in FIGS. Rose to.
Thereafter, a plurality of silicon single crystals SM were manufactured using the same single crystal pulling apparatus 1 without replacing the heater 23.

The oxygen concentration in the silicon single crystal SM produced as described above and the occurrence of LPD were evaluated.
The wafer for evaluation was obtained from an area where the solidification rate in the straight body part SM3 is 15% or less (hereinafter referred to as the first area). After performing mirror etching on this wafer, the oxygen concentration was measured by ASTM F-121 (1979) using FTIR (Fourier Transform Infrared Spectrometer). The oxygen concentration measurement result of one wafer was taken as one batch measurement result.
In the LPD measurement, the wafer obtained from the lower end side in the pulling direction from the evaluation wafer acquisition area in the first area is subjected to mirror polishing through chamfering, lapping, surface grinding, etching, mirror chamfering, primary polishing, etc. Using a surface inspection device (SP-1 manufactured by KLA-Tencor), the LPD per sheet observed on the surface was counted. LPD having a size of 120 nm or more was used as a measurement target, and an average value of LPD measurement results of about 100 to 200 wafers was taken as one batch of measurement results.

FIG. 5 shows the relationship between the number of batches (hereinafter referred to as “heater batch number”) and the moving average value of oxygen concentration. Further, FIG. 6 shows the relationship between the number of heater batches and the moving average value of LPD.
5 and 6, the moving average value of the Nth batch means the average value of the latest four batches including the Nth batch. Since the same single crystal pulling apparatus 1 was used to manufacture a silicon single crystal having a quality different from that of the subject of this experiment (a silicon single crystal including COP), the moving average value is different from that of the subject of this experiment. It is the average value of the last 4 batches that do not contain crystals.

As shown in FIG. 5 and FIG. 6, there was a tendency for the oxygen concentration to decrease and LPD to increase with the number of heater batches reaching 100 batches.
The inventor presumed this reason as follows.

During the manufacture of the silicon single crystal SM, SiO gas is generated from the silicon melt M, and this SiO gas passes between the crucible 22 and the heater 23 and is guided below the chamber 21. At this time, the SiO gas reacts with the carbon of the heater 23, and the carbon of the heater 23 is thinned. Since the SiO gas is guided from the upper side to the lower side, the above reaction occurs more significantly in the upper part of the heater 23 than in the lower part, and the upper part of the heater 23 is deteriorated more than the lower part. When the upper portion of the heater 23 deteriorates in this way, the resistivity of the upper portion becomes higher than that of the lower portion, the upper heating ratio increases, and the lower heating ratio decreases.
In addition, since the main supply source of oxygen to the silicon melt M is the bottom of the crucible 22, if the heating ratio at the lower part of the heater 23 is reduced, the amount of heating at the bottom of the crucible 22 is also reduced. Oxygen supply is also reduced. From the above, it was estimated that when the number of heater batches increased, deterioration of the upper portion of the heater 23 progressed, and as a result, the oxygen concentration of the silicon single crystal SM decreased.

  Moreover, when the heating ratio of the upper part of the heater 23 increases with the increase in the number of heater batches, the cooling of the silicon single crystal SM pulled up from the silicon melt M is suppressed, and compared with the case where the heating ratio of the upper part of the heater 23 does not increase. The value of V / G also increases. As a result, as shown in FIG. 2, it was estimated that the vacancy point defect concentration increased, the COP formed by aggregation of the vacancy point defects increased, and the LPD increased.

When the relationship between the moving average value of oxygen concentration obtained from the results of FIGS. 5 and 6 and the moving average value of LPD was examined, it was found that there was a correlation between the two as shown in FIG.
The equation of the approximate straight line LA1 based on the data in FIG. 7 is as follows when the moving average value of oxygen concentration is X (× 10 ¹⁷ atoms / cm ³ ) and the moving average value of LPD is Y (pieces / sheet). It is represented by (1).
Y = −103.72 × X + 1680.3 (1)

From the relationship shown in FIG. 7, in the single crystal pulling apparatus 1 used in this experiment, when the oxygen concentration of the silicon single crystal SM is 13.4 × 10 ¹⁷ atoms / cm ³ or less, it is obtained from the silicon single crystal SM. When the number of LPDs per wafer is 300 or more and exceeds 13.4 × 10 ¹⁷ atoms / cm ³ , it can be estimated that the number of LPDs is less than 300.
Therefore, it was confirmed that the LPD generation status can be estimated at an early stage after the production of the silicon single crystal SM based on the measurement result of the oxygen concentration. Then, when the silicon single crystal SM used for the oxygen concentration measurement is the first silicon single crystal, the second silicon single crystal manufactured after the first silicon single crystal based on the measurement result of the oxygen concentration. The number of defective products can be reduced by lowering the pulling speed.

[Experiment 2: Investigation of the relationship between the solidification rate of the straight body part and the correlation between oxygen concentration and LPD generation status]
The solidification rate in the silicon single crystal SM of Experiment 1 is 15% to 30% or less (hereinafter referred to as the second region), 30% to 50% or less (hereinafter referred to as the third region), 50% A wafer was obtained from a region exceeding this (hereinafter referred to as a fourth region). Then, the relationship between the number of heater batches and the moving average value of oxygen concentration in these wafers, and the relationship between the number of heater batches and the moving average value of LPD are obtained, and the moving average value of oxygen concentration and the moving average of LPD are obtained from these relationships. The relationship with the value was examined.
FIG. 8 shows the relationship of the second region, FIG. 9 shows the relationship of the third region, and FIG. 10 shows the relationship of the fourth region.
As shown in FIGS. 8 to 10, referring to the approximate lines LA <b> 2 to LA <b> 4 based on the data in each figure, the moving average value of the oxygen concentration increases in the second to fourth regions as in the first region. It was confirmed that the moving average value of LPD was small.

Further, the absolute value of the correlation coefficient between the moving average value of oxygen concentration and the moving average value of LPD in the first to fourth regions was calculated. The result is shown in FIG.
As shown in FIG. 11, the correlation in the first region was higher than in the other regions.
The inventor presumed this reason as follows.

  The lower the solidification rate, the greater the amount of silicon melt M. During the pulling of the silicon single crystal SM, the gap GP is substantially constant, so that the lower the solidification rate, the lower the position of the crucible 22 and the shorter the distance between the bottom of the crucible 22 and the lower portion of the heater 23. For this reason, the heating ratio of the lower part decreases with the deterioration of the upper part of the heater 23, and the influence when the heating amount of the bottom part of the crucible 22 decreases increases as the solidification rate decreases. From the above, it was speculated that by measuring the oxygen concentration in the region where the solidification rate was 15% or less, it was possible to estimate the occurrence of LPD more accurately than in the case of measuring in other regions.

  Although the results of Experiments 1 and 2 shown in FIGS. 5 to 11 differ depending on the size and arrangement position of each constituent member of the single crystal pulling apparatus, the correlation between the oxygen concentration and the LPD occurrence state is different for each constituent member. It is considered that the results of Experiments 1 and 2 are the same regardless of the size and arrangement position.

[Embodiment]
Next, as an embodiment of the present invention, a method for producing a silicon single crystal SM containing COP will be described with reference to the drawings.
In the present embodiment, a case where a p-type silicon single crystal SM having a resistivity of 8 Ω · cm or more and 12 Ω · cm or less and a diameter of the straight barrel portion SM3 after outer peripheral grinding is 200 mm is illustrated. Further, the case where the single crystal pulling apparatus 1 is used in which the relationship between the oxygen concentration in the first region of the silicon single crystal SM and the LPD generation state becomes the relationship shown in FIG. Further, the memory 3 stores information for estimating the occurrence of LPD indicating the relationship as shown in FIG.
Note that the diameter of the straight body portion SM3 after the outer periphery grinding may be other sizes such as 300 mm and 450 mm.

First, the control unit 4 of the single crystal pulling apparatus 1 sets the manufacturing conditions of the silicon single crystal SM, for example, heater power, Ar flow rate, furnace pressure, the number of revolutions of the crucible 22 and the silicon single crystal SM, and the like.
Next, the control unit 4 heats the crucible 22 to melt the polysilicon material (silicon raw material) and the dopant in the crucible 22, thereby generating a silicon melt M. Thereafter, as shown in FIG. 12, the control unit 4 maintains the inside of the chamber 21 in an inert atmosphere under reduced pressure, rotates the crucible 22, and raises the gap GP so as to be substantially constant. A first silicon single crystal SM containing COP in the part SM3 is manufactured (step S1: manufacturing process).

  Next, the operator measures the oxygen concentration of the straight body portion SM3 in the first silicon single crystal SM (step S2: oxygen concentration measurement step (evaluation step)). In step S2, the operator acquires an evaluation wafer from the straight body portion SM3 of the first silicon single crystal. The evaluation wafer is preferably obtained from a region where the solidification rate in the straight body portion SM3 is 15% or less. Then, as in Experiment 1, the operator measures the oxygen concentration of the evaluation wafer after mirror etching using FTIR. The worker may measure the resistivity and the carbon concentration in addition to the oxygen concentration.

Thereafter, when the oxygen concentration is input by the operator, the control unit 4 estimates the LPD occurrence status based on the oxygen concentration and LPD occurrence status estimation information as shown in FIG. (Step S3: LPD occurrence state estimation step (evaluation step)).
In step S3, the control unit 4 estimates that the LPD per wafer is equal to or higher than the quality reference value when the oxygen concentration is equal to or lower than the determination threshold value, and estimates that the LPD is lower than the quality reference value when the oxygen concentration is lower than the determination threshold value. To do. In the present embodiment, the determination threshold is 13.4 × 10 ¹⁷ atoms / cm ³ and the quality reference value is 300 pieces.

Then, when it is estimated that the LPD is equal to or higher than the quality reference value, the control unit 4 changes the manufacturing condition (step S4: manufacturing condition setting step), and includes the COP in the straight body portion SM3 under the changed manufacturing condition. A second silicon single crystal SM is manufactured (step S5). In step S4, the controller 4 slows the pulling rate in order to reduce the number of LPDs generated in the second silicon single crystal SM. In this case, the relationship between the COP concentration and the pulling rate is obtained in advance, and the pulling rate corresponding to the number of LPDs to be reduced may be obtained based on this relationship.
Here, when the silicon single crystal SM is manufactured during the process of step S3, the manufacturing condition change in step S4 is not performed on the silicon single crystal SM, but the second silicon in the next batch. This is done for single crystal SM. When the silicon single crystal SM is not manufactured during the process of step S3, the process is performed on the second silicon single crystal SM of the next batch.

  On the other hand, when it is estimated that the LPD is less than the quality reference value, the control unit 4 maintains the manufacturing conditions (step S6: manufacturing condition setting step) and performs the second manufacturing process under the same manufacturing conditions as the first silicon single crystal SM. A silicon single crystal SM is manufactured (step S5).

[Effects of Embodiment]
According to the said embodiment, there can exist the following effects.
Conventionally, as shown in FIG. 13, the manufacturing condition setting based on the LPD occurrence state is performed from the slice to the mirror surface with respect to the remaining portion of the straight body portion SM3 after performing the steps S1 and S2 of the present embodiment. A process for polishing is performed (step S11). Then, the LPD of the wafer after mirror polishing was measured (step S12), and the manufacturing conditions were changed or maintained based on whether the LPD is equal to or higher than the quality reference value (step S13). In such a method, since the process of step S11 takes time in particular, when the next batch of the production batch of the first silicon single crystal SM is set to the first batch, for example, the LPD measurement result is only in the middle of the seventh batch. The manufacturing conditions can be set only from the 8th batch. In this case, the silicon single crystals SM from the first batch to the seventh batch may be defective.
On the other hand, in this embodiment, without performing the processes of steps S11 and S12, the occurrence of LPD is estimated based on the oxygen concentration measurement result of the evaluation wafer (step S3), and the manufacturing is performed based on the estimation result. The conditions are changed or maintained (steps S4 and S6). Thus, since the time-consuming step S11 is not performed, for example, an LPD estimation result can be obtained in the middle of the second batch. In addition, based on the LPD estimation result, the manufacturing conditions of the third batch (second silicon single crystal SM) can be set. Accordingly, even if the first batch of silicon single crystal SM is defective, the second and subsequent batches of silicon single crystal SM can be made non-defective, and the number of defective products can be reduced.

[Modification]
Note that the present invention is not limited to the above embodiment, and various improvements and design changes can be made without departing from the scope of the present invention.
For example, the determination threshold value and the quality reference value are not limited to the above values, and may be other values according to needs.
In addition to the first area or in place of the first area, information for estimating the LPD occurrence status of at least one of the second to fourth areas may be stored in the memory 3. In this case, when the oxygen concentration and the acquisition region of the evaluation wafer are input by the operator, the control unit 4 generates the LPD based on the information for estimating the LPD occurrence state corresponding to the acquisition region and the oxygen concentration. The situation may be estimated.
An operator may perform the processing of steps S3 to S6.
Although the relationship between the oxygen concentration and the LPD generation state is obtained based on the oxygen concentration and the moving average value of LPD, the data of each batch may be obtained as it is.

  DESCRIPTION OF SYMBOLS 1 ... Single crystal pulling apparatus, 22 ... Crucible, 23 ... Heater, 24 ... Pulling part, SM ... Silicon single crystal, SM3 ... Straight body part.

Claims (3)

An oxygen concentration measurement step for measuring the oxygen concentration of the silicon single crystal containing COP;
And a LPD generation state estimation step of estimating an LPD generation state of the silicon single crystal based on the measurement result of the oxygen concentration. In the evaluation method of the silicon single crystal according to claim 1,
The method for evaluating a silicon single crystal, wherein the oxygen concentration measuring step measures the oxygen concentration in a region where the solidification rate in the straight body portion is 15% or less. Crucible,
A heater for generating a silicon melt by heating the crucible;
A method for producing a silicon single crystal by a Czochralski method using a single crystal pulling device including a pulling unit that pulls a seed crystal after contacting the silicon melt.
A manufacturing process for manufacturing a first silicon single crystal containing COP;
Using the silicon single crystal evaluation method according to claim 1 or 2, an evaluation step of estimating an LPD occurrence state of the first silicon single crystal;
A method for manufacturing a silicon single crystal, comprising: a manufacturing condition setting step for setting a manufacturing condition for a second silicon single crystal containing COP based on an estimation result in the evaluation step.

Images