JPH0416435B2 - - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a method for producing semiconductor silicon single crystals (hereinafter referred to as single crystals) by a pulling method, and in particular, relates to a method for producing semiconductor silicon single crystals (hereinafter referred to as single crystals) by a pulling method. After melting semiconductor silicon in a crucible without adding dopants, a thin silicon rod containing a high concentration of impurities is immersed and melted in the molten silicon liquid, thereby freely controlling the amount of doping into the single crystal and increasing the resistance value. The present invention relates to a method for manufacturing a semiconductor single crystal that makes it possible to obtain a single crystal with little variation or deviation from the target.

[Conventional technology and its problems] In recent years, when devices are manufactured using single crystals by the pulling method, as the devices become more sophisticated, the range of usable resistance values of single crystals becomes limited and narrows. There is a tendency to

When producing single crystals by conventional pulling methods,
Usually, a method is used in which the required amount of impurity is weighed according to the desired conductivity type and resistance value, mixed with raw material silicon, and simultaneously dissolved. When a single crystal is pulled from this, the resistance value changes in the length direction of the crystal at the beginning and end of solidification, depending on the segregation coefficient of impurities and the pulling conditions.

In particular, the commonly used N-type (phosphorous-doped)
From the segregation coefficient of phosphorus, the resistance value on the top side (beginning of solidification) of the single crystal is compared to the resistance value on the bottom side (end of solidification).
Because of the large changes that occur, only a portion of the single crystal may be used for devices, and other portions may not be made into products.

In this way, the conventional method of mixing and simultaneously dissolving raw silicon and dopant impurities before pulling a single crystal requires a long dissolution time (the time it takes for the raw silicon to completely dissolve in the crucible). Because the silicon liquid phase is short and long, there is a difference in the amount of impurity volatilized into the gas phase, and therefore, even if the same amount of dopant is doped, the impurity concentration in the silicon liquid phase differs from batch to batch. Ru. In other words, even if the raw material concentration at the time of preparation is the same, if you compare the finished single crystals one by one, not only will there be a difference in resistance value between each single crystal, but there will also be differences from the target resistance value. A shift occurs.

This situation will be shown as Reference Example 1 below.

From Reference Example 1, the variation in the predictive value of the resistance value for each single crystal is about 10%, and such a situation is not desirable in terms of device fabrication. A method for obtaining single crystals is highly desired.

When the cause of the variation was investigated in detail using the PL method (photoluminescence method), it was found that it was due to the variation in the amount of dopant phosphorus incorporated into the single crystal, but further investigation revealed that the first cause was It has become clear that the reason for this is that the melting time of the raw silicon cannot be made constant, and the second reason is that the resistance value of the dopant itself varies.

These are all problems that must be solved.

Another problem is that during single crystal pulling,
It is necessary to freely control only the resistance value without affecting the pulling process.

As a way to solve this problem, for example,
52-40966, JP-A-130195, JP-A-58-
No. 156,993 has been proposed, but all of these methods involve adding dopant impurities into the molten silicon liquid during the pulling process through a conduit that is constantly immersed.

In these methods, contaminants are eluted from the conduit that is continuously immersed in high-temperature silicon liquid, impairing the physical properties of the single crystal. There is a risk of introducing oxygen into the
When dopant is added, there is a great risk of disturbing the liquid level of the silicon solution being pulled and breaking the single crystal, and the dopant that has been added floats in the solution until it is completely dissolved in the silicon solution, which hits the single crystal being pulled. There are problems such as inhibiting single crystallization.

As a means to solve the above problems,
No. 190,292 provides a method using a gas as a dopant impurity. However, in this method, the ratio of the amount of impurities taken into the single crystal to the amount of dopant gas supplied is extremely low compared to the case of using a solid dopant, and it is not efficient, and the resistance of each single crystal is There is a problem in that the dispersion of the predictive value of the values is, in principle, quite large.

[Means for Solving the Problems] The present invention has been made to solve these problems, and its purpose is to increase the doping accuracy of single crystals.

The gist of the present invention is to provide a manufacturing method using a semiconductor single crystal pulling method, in which a dopant is not added;
After the raw silicon is completely melted in a quartz crucible, a certain amount of the thin silicon rod containing the dope material, which is suspended and held by a lifting device near the single crystal pulling area, is added to the molten silicon just before pulling the single crystal. After melting, the thin silicon rod is lifted up and separated from the melt surface using a lifting device, and the single crystal is then pulled up.

Incidentally, if a potential difference is created between the silicon thin rod containing the dopant and the pulled single crystal, and the current flowing between the silicon thin rod and the pulled single crystal is detected when the silicon thin rod comes into contact with molten silicon, The amount of doping can also be controlled extremely accurately from the relationship between the lowering speed of the silicon rod and the time during which the current flows.

Hereinafter, the present invention will be described in detail using drawings and based on examples.

FIG. 1 schematically shows a single crystal pulling apparatus used in carrying out the present invention.

In Figure 1, 1 is molten silicon, 2 is a single crystal, 3 is a seed crystal, 4 is a thin silicon rod for doping, 5
1 is a lifting device for lifting and lowering the thin silicon rod, 6 is a motor, 7 is a controller for controlling the lifting and lowering of the silicon thin rod, 8 is a single crystal pulling length measuring device (encoder), and a single crystal pulling mechanism (not shown) is connected to.

Next, a method for manufacturing a single crystal according to the present invention will be explained.

A quartz crucible 9 is filled with raw silicon obtained by crushing polycrystals without addition of dopants, and heated and melted using a conventional heating device. Next, the thin silicon rod 4 containing one type of dopant impurity among phosphorus, boron, arsenic, and antimony, whose concentration has been measured in advance, is lowered by the lifting device 5 and brought into contact with the liquid surface of the molten silicon 1. , the silicon rod is gently lowered and immersed for a preset length, a certain length of the silicon rod is doped with a certain amount of dopant, and then the thin rod is raised. Thereafter, the single crystal 2 is pulled and grown to the end by a normal pulling method.

FIG. 2 shows the relationship between the pulled length and resistivity of the single crystal pulled in this way.

In addition, in combination with the present invention described above, it is also possible to further dope by lowering the thin silicon rod again during the pulling of the single crystal. In this case, measure the resistivity distribution in the pulled length direction of the single crystal shown in Figure 2 in advance, and measure the length a of the product acceptable part (shaded area in the figure).
Know. Then, after inputting the pulse amount of the pulling length measuring device (encoder) 8 attached to the single crystal pulling mechanism in FIG. 1 to the controller 7 of the doping device, pulling of the single crystal is started in the same manner as described above. Pull up the single crystal 2. When the single crystal pulling length reaches the preset product acceptable length a, the elevating device 5 automatically melts a predetermined length of the silicon thin rod 4, that is, a predetermined doping amount into the molten silicon 1. Sink. This doping amount corresponds to another product standard, and is determined by testing in advance. Continue pulling to obtain a single crystal of a predetermined length.

FIG. 3 shows the relationship between the pulled length and resistivity of the single crystal pulled in this way.

In FIG. 3, the horizontal axis shows the pulling length and the vertical axis shows the resistivity distribution. In this manner, the resistance value of the portion of the single crystal b can be arbitrarily lowered by accurately doping any predetermined amount during the pulling of the single crystal. As a result, part b is added to the product acceptance section, and the acceptance rate is significantly improved.

Further, when a voltage is applied between the thin silicon rod 4 and the single crystal 2 and the thin silicon rod 4 comes into contact with the liquid surface of the molten silicon 1, a current flows through the thin silicon rod 4, the molten silicon 1, and the pulled single crystal 2. By doing so, the molten liquid surface can be detected with good reproducibility and high precision.

FIG. 4 shows a molten liquid level detection device in which a potential difference is provided between a thin silicon rod 4 and a single crystal 2.

Hereinafter, the present invention will be explained in detail with reference to reference examples and examples.

Reference Example 1 This reference example shows a reference example of a single crystal production method using a conventional doping method and the results thereof.

1000 to 3000Ω− in a quartz crucible with a diameter of 12″φ
20 kg of polycrystalline silicon of cm was loaded as raw silicon, and 827 mg of dopant of 2 × 10 ^-3 Ω-cm was put into it.
After evacuation to 0.01 Torr, replace with argon.
dissolve. Thereafter, a seed crystal was attached, a shoulder was made, and the material was pulled to a predetermined length.

FIG. 5 is a distribution diagram showing variations in resistance values of each of the top portions of 20 single crystals obtained according to this reference example. The horizontal axis shows the deviation of the resistance value of the top part from the target in percentage, and the vertical axis shows the frequency.

From this result, the deviation of the resistance value from the center value (target value) (portion c in the figure) χ was approximately 10%, and the standard deviation value σ of the variation was 6.1%. As mentioned above, these deviations (χ) and variations (σ) are due to the fact that when investigated using the PL method, the amount of phosphorus incorporated into the single crystal of the dapant varies.
The reason for this is that it is difficult to maintain a constant melting time until the raw material polycrystalline silicon is completely dissolved in the crucible, so the amount of phosphorus volatilized varies in proportion to time, and the amount of phosphorus volatilized varies in proportion to time. This is because the impurity content of the dopant itself varies.

Example 1 This example shows an example of the method for producing a single crystal according to the present invention and its results.

First, by using the floating zone method (FZ method), the resistivity is 3.
×10 ^-3 Ω−cm, resistance value variation standard deviation 2 ×
500 mm long, uniform along the length, which is 10 ^-5 Ω-cm
A thin silicon rod with a cross-sectional area of 1 cm ² is attached to the pulling device shown in Figure 1. 1000 in a quartz crucible with a diameter of 12″φ
20 kg of polycrystalline silicon having a diameter of 3,000 Ω-cm is loaded as a raw material silicon, and after evacuating to 0.01 Torr, the system is replaced with argon and melted. Thereafter, the thin silicon rod 4 was lowered by the lifting device 5, and after coming into contact with the silicon solution 1, a predetermined length, that is, a predetermined doping amount, was accurately dissolved. Thereafter, a seed crystal was attached, a shoulder was made, and the crystal was pulled to obtain a single crystal with a diameter of 100 mmφ and a length of 800 mm.

FIG. 6 is a distribution diagram showing variations in resistance values of each of the top portions of 20 single crystals obtained in this example.

The horizontal and vertical axes are the same as in FIG.

As a result, the deviation of the resistivity from the center order (target value) (portion d in FIG. 6) (χ) was 2%, and the standard deviation of variation (σ) was 0.8%. It can be seen that this is much improved over the value shown in FIG. 5, which is the result of the conventional method of Reference Example 1.

Furthermore, when the amount of impurities incorporated into the single crystal 2 was measured, it was found that more than 95% of the impurities supplied by the melting of the silicon thin rod 4 were incorporated.

This is a much higher value than the ratio of the amount taken in by methods using gas as an impurity doping method, which is around 20%.

Reference Example 2 This reference example shows a reference example that can be implemented in combination with the single crystal production method of the present invention and its results.

A rectangular thin silicon rod 4 with a resistance value of 5 x 10 ^-3 Ω-cm and a size of 6.5 mm x 6.4 mm was attached to the pulling device shown in FIG. 1000 to 3000Ω− in a quartz crucible with a diameter of 12″φ
Loaded with 20 kg of polycrystalline cm as raw silicon,
After evacuation to 0.01 Torr, the mixture was replaced with argon and dissolved. Thereafter, a predetermined dope amount of 27.4 mm was dissolved using the same doping method as in Example 1.

After that, seed crystal is attached, shoulder is made, single crystal 100
The length of mmφ was increased to 450mm. At this point, while continuing to pull up the silicone rod 4, a length of 9.9 mm was gently dissolved into the silicone solution.
Continue pulling, and finally the diameter is 100mmφ,
A single crystal with a length of 800 mm was obtained.

A wafer was cut out from the obtained single crystal, and the resistance value at the center of the wafer was measured. The results are shown in FIG. The vertical axis shows the resistance value (Ω-cm), and the horizontal axis shows the pulled length (mm) of the single crystal.

Line 12 is the theoretical resistance value, and line 13 is the resistance value of the single crystal obtained in this reference example.

Furthermore, FIG. 8 schematically represents this single crystal, and the shaded area indicates the part that passed the product.

As shown in FIG. 7, it can be seen that the theoretical values and the results of this reference example show good agreement.

As a result of this reference example, in addition to the single crystal part e in FIG. 8, by doping once more during the pulling process, a single crystal part f having the desired resistance was obtained. The length of the e part is 420mm, and the length of the f part is
It was 330mm.

Reference Example 3 This Reference Example was carried out in order to eliminate the phenomenon observed in Reference Example 2 that was expected to interfere with single crystallization. That is, in Reference Example 2, as shown in FIG.
In some cases, crystals 11 were observed to grow at the portions where the thin silicon rods 4 and 0 were in contact with each other. This is because heat escapes through the silicon thin rod 4.
If this becomes large, when pulling up the silicon thin rod,
This may disturb the silicon liquid level 10 and hinder the single crystal growth of the single crystal being pulled.

In order to prevent this, in this reference example, the thin rod 4
preheating, melting speed of the thin rod 4, and the thin rod 4.
We investigated the thickness of the

As a result, before melting the silicon thin rod 4, by preheating it for 10 to 20 minutes just above the molten silicon liquid level 10 and keeping the dipping speed at 0.03 mm/min. or less,
It was found that even if the cross-sectional area of the silicon thin rod was increased to 1.0 cm ² , it could be carried out without preventing single crystal growth.

However, this is a reference example using the doping mechanism used in the present invention, and there are various means for the preheating, such as using a laser.

However, if the thin silicon rod 4 is made too thick, the preheating method and equipment will become complicated, so 1.0
^cm2 or less is reasonable.

In addition, as mentioned above, the cross-sectional area of the silicon thin rod 4,
By appropriately selecting the doping rate and preheating time, it is also possible to prevent the molten silicon liquid level 10 from vibrating during doping and when the thin silicon rod 4 is raised after doping.

Reference Example 4 This reference example shows a different reference example that can be implemented in combination with the method for producing a single crystal using the doping method of the present invention.

In order to reduce variations in the resistance value of product single crystals, it is necessary to dope a certain amount of dopant into molten silicon, that is, to melt a thin silicon rod of a certain length with high precision.

Therefore, as shown in FIG. 4, if a voltage is applied between the silicon thin rod 4 and the single crystal 2, when the silicon thin rod descends and its tip contacts the liquid surface of the molten silicon 1, the molten silicon 1 is passed through the silicon thin rod 4. Current flows. As a result, the liquid level position was accurately detected with good reproducibility. Thereafter, a thin silicon rod of a predetermined length was dissolved.

If this liquid level detection method and device are used, for example, in Reference Example 2 or Reference Example 3, it is possible to obtain a single crystal with a high precision and small variations in resistance target value of the single crystal.

[Effects of the Invention] According to the present invention, since it is possible to dope a certain amount of dopant accurately, the deviation from the target resistance center value (χ) is small, and the variation in resistance value (σ) from one wire to another is small. Fewer single crystals can be obtained.

Furthermore, the rate of impurity incorporation into the single crystal reaches 95% of the supplied impurities, which is much more efficient than when using a doping device using gas.

Furthermore, since all doping operations can be performed from outside the pulling device, there is no need to entrain air when doping material is introduced, and there is no risk of undesirable impurities being mixed into the single crystal.

In addition, if the cross-sectional area of the thin silicon rod, doping speed, and preheating time are appropriately selected according to the reference example, it is possible to prevent liquid level vibration during doping and when raising the thin silicon rod after doping. , pulling can be continued without inhibiting single crystallization.

As described above, when the single crystal manufacturing method of the present invention is used, much superior effects can be obtained compared to the conventional methods.

[Brief explanation of drawings]

FIG. 1 is a sectional view schematically showing a single crystal pulling apparatus used in carrying out the present invention. FIG. 2 is a diagram showing the resistivity distribution of the pulled length of a single crystal produced by the method of doping immediately before pulling the single crystal according to the present invention. FIG. 3 is a diagram showing the resistivity distribution in the length direction of a single crystal produced by a method of doping just before and during pulling of the single crystal, which is carried out in combination with the present invention. FIG. 4 is a conceptual diagram of a liquid level detection device in which a potential difference is provided between a single crystal and a thin silicon rod. FIG. 5 is a distribution diagram showing variations in resistance values of the top portion of a single crystal pulled by a conventional method. FIG. 6 is a distribution diagram showing variations in the resistance value of the top portion of a single crystal according to an embodiment of the present invention. FIG. 7 is a diagram comparing the longitudinal resistivity with the theoretical resistivity of a single crystal produced by a method of doping immediately before and during the pulling of the single crystal, which is carried out in combination with the present invention. FIG. 8 is implemented in combination with the present invention,
Schematic diagram of a single crystal created by doping immediately before and during pulling of the single crystal. FIG. 9 is a diagram showing a state in which crystals partially grow at the portion where the molten silicon liquid surface and the silicon thin rod are in contact.

Claims (1)

[Claims] 1. In a method for producing a semiconductor silicon single crystal by the pulling method, after completely melting the raw material silicon in a quartz crucible without adding a dopant,
A certain amount of the thin silicon rod containing the dope material, which is near the single crystal pulling area and is held by a lifting device above the raw silicon melt, is lowered through the lifting device to perform single crystal pulling. 1. A method for producing a semiconductor silicon single crystal, comprising immediately before dissolving it in molten silicon, separating the thin silicon rod from the surface of the silicon melt after melting, and then pulling the single crystal.