JPH0380200A - Production of high-strength silicon wafer - Google Patents

Abstract

PURPOSE:To obtain the high-strength silicon wafer which is decreased in warpage, dislocation, deposit, etc., by specifically regulating the kind and amt. of impurities to be added, heat treating conditions, rapid cooling conditions, etc., treating a single crystal ingot of silicon in inert atmosphere and then cutting out the wafer. CONSTITUTION:The signal crystal ingot of the silicon contg. >=1X10<17>cm<-3> boron impurity and the impurity (e.g. Ga, Sn) of group IIIb or IVb elements of the periodic table at the ratio higher than the ratio of the silicon atoms corresponding to the ratio to compensate the distortions occurring in this boron impurity is prepd. This single crystal ingot of the silicon is then disposed in the inert atmosphere and is held for >=30 minutes at >=1200 deg.C; thereafter, the ingot is cooled at 20 to 60 deg.C/min cooling rate. The wafer is cut out of the cooled single crystal ingot of the silicon, by which the high-strength silicon wafer is obtd. The element having the good characteristics is produced at a good yield by using this silicon wafer.

Description

[Detailed Description of the Invention] [Summary] A method for producing a high-strength silicon substrate that exhibits very little warping even after being heat-treated at 700°C or higher for a long time (repeatedly) for silicon wafers that are semiconductor (single-crystal) substrates. Cl11-3 or higher boron (B) impurity Th8i holon impurity compensating for distortion caused by a considerable amount of group 3B elements of the periodic table (Al1.
GaIn) or group 4B elements (Ge, Sn,
A silicon single crystal ingot containing Pb) impurities was heated at 1200°C in an inert atmosphere (Ar, Nz).
Hold at the above temperature for 30 minutes or more, and then 20-60℃/
After cooling at a cooling rate of 10 minutes, a wafer is cut from the silicon single crystal ingot to form a high-strength silicon wafer.

(2) [Industrial Application Field] The present invention relates to a silicon wafer that is a semiconductor (single crystal) substrate, and more specifically to a method for manufacturing a high-strength silicon wafer.

[Prior Art] Silicon wafers used as substrates for semiconductor integrated circuits (super r-s r ) are repeatedly heat-treated at high temperatures for oxide film formation, impurity diffusion, annealing after ion implantation, and the like. At this time, it is difficult with current heat treatment technology to heat and cool the entire wafer surface while keeping it at the same temperature, so a temperature distribution occurs within the wafer surface, which causes thermal stress. Due to this thermal stress, the silicon wafer undergoes plastic deformation, which is observed as ``warp''.

If warpage occurs in the silicon wafer, it becomes difficult to uniformly process the entire surface using photolithography.

In addition, there is a 1O'cur2 inside the plastically deformed silicon wafer.
Because of the presence of more than one dislocation, bipolar transistors,
It deteriorates the electrical characteristics of elements such as MOSFET, and (3
) Decrease yield. On the other hand, super 1. ．． With the increase in the diameter of silicon wafers and the increase in the degree of integration and miniaturization of elements promoted by the development of S1, the allowable warpage of silicon wafers has become reasonably small.

Silicon wafers are generally manufactured by processing (cutting) dislocation-free silicon single crystals (ingots) grown by the Czochralski method (hereinafter referred to as the CZ method) or the floating thin method (hereinafter referred to as the FZ method). be done. Currently, about 90% of silicon single crystals for substrates are grown by the CZ method.

CZ method silicon single crystals are grown by immersing a seed crystal in raw material silicon melted in a quartz crucible and continuously raising the seed crystal. Since the quartz crucible reacts with and dissolves the silicon melt, the silicon melt inevitably has a high oxygen concentration. Therefore, the CZ method silicon single crystal also contains oxygen, and its concentration is usually 1017 to 1018 cI11-3. Oxygen in single crystals is known to suppress the occurrence of warpage mentioned above (for example, (4) Yasuo Tarui, ed., “Wafer Warpage and Deformation”, Super R, SR Technology, Ohmsha. (198]) 211 (Reference 1) and H
, Shimizu et al., Jpn, J., 8pp.
1. Phys, 25 (1986) 68 (Reference 2)
reference). For example, as is clear from the diagram in Figure 7, which shows the relationship between the oxygen concentration in silicon wafers and warpage measured by Tarui et al. It is overwhelmingly small, and the high oxygen concentration substrate has little warpage. however,
The data in Figure 7 is 1150 as the wafer heat treatment condition.
These are the measurement results after only 30 minutes at ℃. Generally, 1
It is known that when heat treated at around 000"C for several hours, the oxygen in the silicon wafer forms oxygen precipitates (SiO7, etc.), and these precipitates increase the warpage of the wafer (
Reference 2). Moreover, wafers with high oxygen concentration tend to cause oxygen to precipitate. In other words, with a heat treatment of just 30 minutes as shown in Figure 7, oxygen hardly forms as a precipitate, so warpage decreases as the oxygen concentration increases, but if heat treatment is performed for a long time and oxygen precipitates are formed, As shown in Figure 8 (Reference 2), as the density of precipitates (5) increases, the warpage increases. In this way CZ
Oxygen in silicon single crystals (wafers) is a double-edged sword, and in order to actually use it in the VLSI process, heat treatment at around 1000°C continues for a long time, so the oxygen concentration and heat treatment conditions must be carefully combined. This method has major drawbacks as a method for suppressing wafer warpage.

In addition, it has been found that wafers processed from nitrogen-doped FZ silicon single crystals experience less warpage due to heat treatment than wafers processed from regular CZ silicon.
tal,, Sem1conductor 5ili
Con 1981pp, 54-71 (Reference 3). In Fig. I3 of Document 3, a nitrogen-doped FZ method silicon wafer (nitrogen concentration 4.5 × l015c
The slip lines generated by heat-treating a wafer with many slip lines and a regular CZ silicon wafer for 10 minutes at 1150°C are shown. The warpage is larger than that of the FZ method silicon wafer. This was noticed as an indication of the effect of nitrogen doping. However, (6
) The diffusion coefficient of nitrogen in high temperature silicon wafers is very large, e.g. ~106 CII12 at 1200 °C.
SeC (T, 1toh and T, hebe pp1. Phys, 1.ett, 53 (1988) 3
9 (see Reference 4)). Diffusion coefficient: Dcm2sec Diffusion time: 1. sec, the distance traveled by diffusion is 2j■
Therefore, if heat treatment is performed at 1200°C for 30 minutes, the amount of nitrogen will be 27TfT - 2j - one book m nu lower F! 850
Spread the sign. Usually, the thickness of the wafer is 600-800μ...
Therefore, by heat treatment at 1200° C. for 130 minutes, all the nitrogen in the wafer escapes to the outside of the wafer, and the effect of suppressing warpage is lost. The data in Reference 3 shows the effect of nitrogen because the heat treatment is only 10 minutes, but in the actual VLSI manufacturing process, the heat treatment is longer (for example, CCD
Since the process involves heat treatment at 1200° C. for 3 hours, this nitrogen-doped FZ method silicon wafer has poor practicality.

(7) [Problem to be solved by the invention] We will explain what causes the macroscopic plastic deformation of warping of silicon wafers in terms of crystal engineering, and then explain how oxygen and nitrogen can be used to reduce warping as described above. Explain the mechanism.

The fundamental cause of plastic deformation in crystals is the movement of lattice defects called dislocations. It is known that dislocations move in a specific direction in a crystal under stress and their density increases (referred to as dislocation multiplication).

When a single dislocation reaches the crystal surface from inside the crystal, a one-atom fault appears on the surface. Naturally, this fault has strain energy.

Since dislocations multiply in huge numbers during movement, these faults also become large and can be observed using an optical microscope using the etching method. When such faults occur at various locations in the crystal, the crystal no longer remains in its initial shape and exhibits macroscopic deformation. Furthermore, since this deformation is caused by the movement of dislocations, it does not return to its original state even if the stress is removed; in other words, it is plastic deformation. When this phenomenon (8) occurs with a thin disk-shaped wafer, the balance between the strain accumulated on one side of the wafer and the strain accumulated on the other side is lost, causing the wafer to become unstable as a whole. It will warp. Therefore, when observing the surface of a warped wafer on an atomic scale, there are a huge number of "atomic fault lines".

As can be seen from the above explanation, the first approach to reducing wafer warpage is to suppress the occurrence of dislocations. However, this is not possible with current silicon crystal technology. For one thing, although single crystals (ingots) grown by the CZ method or the FZ method are said to be free of -6 dislocations, recent research has revealed that they contain defects called A defects (J , Chika prisoner a
et al.

Sem1coductor 5ilicon 1986
+ p, 61 (see reference 5)). It has been confirmed that this A defect is a micro dislocation loop, and even in the crystal immediately after growth, there are no "large (on an atomic scale) dislocation", but a large number of micro dislocation loops are formed. Second, is wafer processing technology so complete that it can create atomically flat processed surfaces?
) That means there is no. In particular, there is still no technology for finishing the rounded part (also called hevelling part) around the wafer into a mirror surface, and this is the part where the most processing strain is accumulated on the wafer. Therefore, when a wafer is subjected to thermal stress caused by taking it in and out of a furnace during heat treatment, there is a very high possibility that dislocations will be introduced from the wafer's peripheral area. I”
- It is impossible to completely eliminate the occurrence of dislocations for two reasons.

The next possible method for reducing warpage is to suppress the movement of existing dislocations as much as possible under thermal stress. By inhibiting movement, proliferation is also inhibited. We have already mentioned that oxygen and nitrogen reduce warpage, but from a crystal engineering perspective, oxygen and nitrogen have the effect of fixing dislocations at high temperatures and preventing their movement even when stress is applied. That is to say (1.
and K., Sumin.

J, 8pp1. Phys., 56 (1984) 234
6 (Reference 6), K. Sumin.

et at,, Jpn, J, Appl, Phys, +
19 (1980) L763 (Reference 7) 1. Yon
enaga and K, Sumino, Jpn, J
, Appl, Phys.

(10) 23 (1984) L590 (Reference 8) and S
Umino et al.

J. Appl, Phys., 54 (1983) 501
6 (see reference 9)).

Therefore, CZ method silicon wafers and nitrogen-doped FZ method silicon wafers have small warpage. However, when stress is applied to a CZ silicon wafer in which oxygen has been precipitated by long-term heat treatment, dislocation loops are punched out from the precipitates (
1, Yonenaga and K.

Sumino, Jpn, J., Appl, Phys.
21 (1982) 47 (Reference 10)), these dislocations multiply and cause plastic deformation of the wafer. Further, since the diffusion coefficient of nitrogen at high temperatures is extremely large, when a nitrogen-doped FZ wafer is subjected to high-temperature heat treatment, nitrogen escapes from the wafer before fixing dislocations. For this,
Nitrogen-doped FZ wafers also undergo long-term heat treatment,
Warpage appears.

The purpose of the present invention is to propose a method for manufacturing a high-strength silicon substrate that exhibits extremely little warping even after repeated heat treatments at temperatures of 700°C or higher for long periods of time, thereby facilitating the photolithography process in the VLSI process and preventing dislocations. The purpose is to prevent the deterioration of the electrical characteristics of the device due to the occurrence of oxidation (11) and to improve the yield.

[Means for Solving the Problems] The above-mentioned purpose is to compensate for the boron (B) impurity of IX1017cm-'2 and a corresponding amount of silicon atoms larger than 3B of the periodic table to compensate for the distortion caused by the boron impurity. Group elements (Af
f, Ga, In) or group 4B elements (Ge, 5
A silicon single crystal ingot containing nPb) impurities was heated for 1200 min in an inert atmosphere (Ar, N2).
Hold at a temperature above ℃ for more than 30 minutes, and then 20-60℃
This is achieved by a method for producing a high-strength silicon wafer, which is characterized in that the wafer is cut from the silicon single crystal ingot after cooling at a cooling rate of 1/2 min.

In addition, I
) and a considerable amount of carbon (C) impurity that compensates for the distortion caused by the impurity. The object of the present invention can also be achieved by a method of manufacturing a high-strength silicon wafer in which the wafer is cut from a silicon single crystal ingot.

[Function] The feature of the present invention is that group 3B elements (B, Aff, GaIn)
The method is to dope a doped silicon single crystal ingot with a combination of an impurity of silicon (Si) and an impurity of a group 4B element, maintain the doped silicon single crystal ingot at a high temperature, and cool it at a predetermined cooling rate.

In order to explain the present invention, after explaining the yield stress τy (yield 5tress), which scientifically represents the strength of a crystal, we will explain how to suppress the movement of dislocations using the case of doping with B and Ge as an example. I will explain from a crystal engineering standpoint.

As shown in FIG. 9, when stress is applied to a crystal and it continues to be pulled in a certain direction, the stress and strain are in a nearly proportional relationship for a while (a in the figure). However, after reaching a certain stress τ, suddenly τ.

The strain increases with smaller stress (yield (13
) phenomenon). This τa is called the yield stress, and is a value sensitive to the internal structure of the crystal. A strain of less than ε8 is called an elastic (deformation) region, and a strain of ε8 or more is called a plastic (deformation) region. The mechanism of the yield phenomenon is as follows.

Existing dislocations contained in the crystal do not move (and therefore do not multiply) in the elastic region a even if the stress increases.
. However, when the stress matches τ, they suddenly start moving and multiply at the same time. That is, τa is the threshold for dislocations to start moving. As the dislocation density increases, lattice plane slips (faults) occur at various locations in the crystal, leading to progressive plastic deformation. Unlike elastic deformation, which displaces the entire two-dimensional lattice plane from its equilibrium position, this deformation displaces defects with a one-dimensional structure called dislocations, so the strain progresses with a stress considerably smaller than τa.

As is clear from the above explanation, a crystal with a large τ is a high-strength crystal in which dislocations are difficult to move.

It has been known that group 3B impurities (dopants) suppress the movement of dislocations. As shown in Figure 10,
According to Milvidskii et al.
4) IQIIIcm-: τ of FZ silicon when doped with l
, begins to rise by 2, but saturates with an increase of about 25% between 1016 and 1019 cm-3, and when doped with 1019 cm-3 and above, conversely, A starts to decrease (M, G, Mil
viclskii et al,, 5ovietPh
ysics-3olid 5tate 6 (1965)
253N Reference 11)).

The mechanism of the influence of 84 degrees on τ is considered as follows. B in the silicon crystal receives electrons from the valence band on one side and becomes negatively charged, and at the same time leaves holes in the valence band. On the other hand, since dislocations have unpaired electrons, they capture these holes (see "Crystal Plasticity, Vol. 8, Fundamentals of Metal Physics," edited by the Japan Institute of Metals, Maruzen, (1977) p. 524 (Reference 12)). It is thought that the dislocations then have a positive charge and electrically interact with the negatively charged B, making it difficult for the dislocations to move. Therefore, as the B concentration increases, the movement of dislocations is more strongly suppressed, so τa also increases.
Second ascension. In FIG. 10, the B concentration is 10'', 1
It is precisely for this reason that τ increases as the value increases to 017.1018 cm −3. but,
The atomic radius of a B atom is 0.90 people, which is considerably smaller than the radius of a Si atom, which is 1.17 people (I5). Therefore, when heavily doped, tensile strain accumulates in the silicon crystal, and under stress load, the generation and proliferation of dislocations due to the increase in tensile strain outweighs the suppression effect of dislocation motion by B, and the τ This is considered to lead to a decrease in In addition, B doped in a large amount electrically interacts with point defects existing in the crystal and combines to form microprecipitates (A, J, R, de Kock et al.
al,, 8pp1. Phys.

Lett, 34 (1979) 611 (Reference 13)
,8,J,R,de Kocket al,, J,C.
ryst, Growth, 49 (1980) 718
(Reference 14) and C, W., Pearce et al.
,, MR3symposia proceedings
s, Vol. 36 (1985) 23B Reference 15)). Therefore, it is considered that under stress loading, micro dislocation loops are punched out from micro precipitates caused by B-point defect bonding and multiply, leading to a decrease in τ. Due to the above two reasons, when the B concentration becomes i1019r''3 or more, τ.

It is thought that this will start to decline.

The basic idea of the present invention is to bring out the effect of suppressing dislocation motion by B to a greater degree than before, and to make the silicon single crystal higher in strength than before.

(16) Therefore, the accumulation of strain that inhibits the effect of B and the
Both causes of precipitate formation must be reduced as much as possible.

Regarding the relaxation of strain caused by B, as described above, B atoms are smaller than Si atoms, and therefore they accumulate tensile strain in silicon crystals. Therefore, it is considered that the strain caused by B can be alleviated by doping Ge, which is an impurity having a larger atomic radius than Si atoms and is electrically inactive in the silicon crystal, at the same time as B. Figure 11a shows the strain accumulated when only B is doped, and Figure 11b shows the strain accumulated when only Ge is doped.
The results of investigation using line diffraction method are shown. Since the strain caused by Ge doping is compressive strain, its sign is opposite to that caused by B. It can be seen from FIGS. 11a and 11b that the strain increases with the doping concentration, but the strain suddenly decreases after the concentration exceeds a certain value. The reason for this is that the introduction of dislocations causes strain relaxation.

Furthermore, from FIGS. 11a and 11b, it is thought that if B and Ge are doped simultaneously, a silicon crystal with small strain can be obtained (17). 1st], Figure c, B], 2 X
It shows how the strain of silicon crystal containing 10"c+n-3 decreases with Ge doping. From Fig. 11c, 5
X10-'The strain was 8XIO"cm-')
- It was revealed that the reduction was reduced to less than 1/10 when In other words, it was confirmed that the strain caused by B can be greatly alleviated by doping with Ge.

Strain relaxation can be similarly achieved by doping the same group element Sn or Pb instead of Ge.

Further, the strain caused by B can be alleviated by doping (adding) Aff, Ga, and/or In, which are elements in the same group as B and are larger than Si atoms. Furthermore, group 3B impurity dopane) (Aff, G
Relaxing the compressive strain that occurs when doping with a, In) is achieved by doping with C, which is a group 4B impurity smaller than a Si atom.

Regarding suppression of the formation of precipitates due to B, B in silicon crystals is negatively charged and interacts with positively charged vacancies to form minute precipitates (see References (18) 13 and 14). This precipitation occurs immediately after the crystals begin to solidify from the melt, and increases as the B concentration increases.

Furthermore, if oxygen is included in the crystal, oxygen will precipitate using these minute precipitates as nuclei, so a crystal with a high oxygen concentration will have a high precipitate density. Figure 12 shows the precipitate density with respect to the B concentration. This case is the result of heat treating an As-grown CZ method silicon wafer at 700° C. for 30 hours and then at 1100° C. for 2 hours.The formation of precipitates that is clearly dependent on the B concentration can be seen.

In the present invention, in order to stop the formation of precipitates by 1 iFt, B
A method is adopted in which a crystal grown by doping with and Ge is heat treated at a high temperature of 1200° C., preferably 1300° C.1 or higher, and then rapidly cooled at a predetermined cooling rate. The nucleus of the precipitate caused by B is
It is formed by the so-called thermal history from immediately after the crystal solidifies from the melt until it cools to room temperature.

Therefore, it is considered that by heating the crystal to melt the precipitate nuclei to just below the melting point, it is possible to prevent the formation of precipitates during subsequent heat treatment. In FIG. 12, the density of precipitates caused by B is shown by the mark (19) when a crystal grown by doping with B is heat treated at 1350° C. for 12 hours and then rapidly cooled at 60° C. per minute. For comparison, when there is no high temperature meditation PP (○
mark) is also shown at the same time. As is clear from FIG. 12, the precipitate density is reduced by 175 to 1/20 by high-temperature heat treatment, and in particular, the higher the concentration of B, the greater the effect of suppressing precipitation. Group 3B impurity dopants other than Naori (Affi,
Precipitate formation also occurs when doping with Ga, In) due to interaction with vacancies, but the precipitate density is reduced to 1/10 by similar high-temperature heat treatment and rapid cooling.

To summarize the above, in order to bring out the effect of suppressing dislocation motion by B to a greater extent than before, it is essential to dope Ge at the same time as B to relax the strain and to suppress the formation of precipitates caused by B by high-temperature heat treatment. This is a feature of the invention.

Figure 1 shows the yield stress τ measured for silicon single crystals doped with B and Ge simultaneously at 1350°C for 2 hours and rapidly cooled at 60°C/min, together with a conventional example (B doped only). show. From FIG. 11c, the doping amount of Ge was set to be seven times that of B.

As is clear from FIG. 1, by the method of the present invention (20
) of silicon single crystal, the B concentration is I x 1017 cm
-' or more, there is an improvement effect, which is 20% or more better than the conventional one.

[Example] Hereinafter, the present invention will be explained in more detail with reference to embodiment examples of the present invention.

A case will be explained in which a silicon single crystal (ingot) is produced by the CZ method and cut out to produce a silicon wafer.

35kg of polycrystalline silicon in a 16 inch diameter quartz crucible
, 210 g of Ge ingots were charged and heated and melted to form a melt, and then B was added to 3.0 g using a dopant introducing device.
Add 4g to the melt. Thereafter, a single crystal is grown in exactly the same manner as the normal CZ method. The reason for adding B immediately before crystal pulling without charging it together with polycrystalline silicon or Ge is to reduce B evaporation as much as possible and improve reproducibility. ! This is to obtain the fourth degree.

(21) Above (Da long method Niyori B 8 degrees (1,0±0.2)
X 10"cm-', Ge concentration (8.0±2.0)
Xl019cm-3 and oxygen concentration (1,0±0.2
) A silicon single crystal of Xl018cm-3 was obtained. Note that the old ASTM was used as the conversion factor for oxygen concentration.

FIG. 2 shows the shape of the single crystal ingot 1 grown in Example 1. The diameter d is d = 155 mm and the cone part IA
, the length of the tail part IB is approximately 150 mm, the length of the straight body part (the part with a constant diameter of 155 mm) IC is approximately 600 mm,
The total weight was approximately 30 kg. The reason why the length of the cone and tail parts is made almost the same as the diameter is due to the heat treatment described below →
This is to prevent the introduction of dislocations due to thermal stress during rapid cooling.

In order to suppress (reduce) precipitates caused by B in the grown single crystal, before taking out the single crystal ingot from the CZ furnace circle, high temperature heat treatment at 1350 ° C 12 hours was performed in the same furnace,
Then, the single crystal ingot 1- was lifted upward from the heating section of the furnace and quenched by blowing Ar gas onto it. The cooling rate at this time was 60°C/min.

The produced single crystal ingot was processed in the same manner as the normal processing process (22
) A silicon wafer of the present invention was obtained by processing (cutting, polishing) as shown in FIG.

When performing rapid cooling after high-temperature heat treatment, thermal stress acts on the single crystal and dislocations may be introduced. Therefore, the cone section,
When quenching was performed with the length of the tail portion set to 0.1/2d2/3d and d, it was found that no dislocations were introduced if the length was 2/3d or more. This is because the conical cone and tail parts prevent the crystal from changing its shape rapidly.
This is thought to be due to the reduction in thermal stress. Therefore, in this embodiment, the lengths of the cone portion and tail portion are made equal to the diameter.

FIG. 3 shows the results of investigating the relationship between the conditions of high temperature heat treatment and rapid cooling and the density of precipitates caused by B. High temperature heat treatment temperature is]
]OO℃ (mark), 1200℃ (・mark) and 130
The temperature was 0°C (○ mark), the high temperature holding time was 30 minutes, and the cooling rate was 20°C, 40°C, 60°C, and 8°C7 minutes.As-gr without high temperature heat treatment and rapid cooling
Silicon wafers obtained from single crystals in the own state (single crystal ingots obtained by the normal CZ method) (23) and silicon wafers from single crystals subjected to the above-mentioned high-temperature heat treatment and quenching were heated at 700°C for 30 hours and at 1100°C. After heat treatment for 2 hours, the state of occurrence of precipitates was examined.

From Figure 3, it was found that the precipitate density was small when the high-temperature heat treatment temperature was 1200°C or higher and the cooling rate was 20 to 6 minutes. Although 30 minutes was sufficient for the high-temperature heat treatment time,
It is considered that there is no problem even if the time is longer than this. For other B and Ge concentrations, a high-temperature heat treatment of 1200°C or higher → cooling rate of 20 to 6 (]°C/min is considered sufficient. This is because the suppression of precipitates caused by B is l')3-
This is due to the decomposition of precipitates through the reaction ``decapores←microprecipitates,'' but this reaction is not determined by the concentration of B or Ge, but when the temperature of the crystal at that time is T, e-KT
This is because they occur with a frequency proportional to . Here, E is the activation energy of the reaction (unrelated to T, Lk is Portzmann's constant.

Ru.

Furthermore, when oxygen is dissolved in interstitial positions, it fixes dislocations and increases the strength of the crystal, but when it forms precipitates (24), it conversely generates dislocations and reduces the strength. Therefore, it is sufficient to contain oxygen at the maximum concentration that does not precipitate.

This maximum oxygen concentration is believed to be approximately IXIO"cm"'. The reason is as follows. FIG. 4 shows the relationship between oxygen concentration and amount of oxygen precipitated. However, the crystal in this case has a B concentration of ~1015 cm-3 and a Ge concentration of ~0 (so-called
It is a normal silicon crystal without Ge additives), and from Figure 4 it can be seen that almost no precipitation occurs if the oxygen concentration is I x 1018cm-''.Recent research has shown that the core of oxygen precipitation is vacancies. I have come to understand that (H
, 1larada, T., Abe and J.

Chikawa, Sem1conductor 5i
licon Electrochemical 5o
ciety, Pennington N, J, (19
86) p., 76-85 (see reference 16)). However, there have been no examples of detecting holes themselves, and the current situation is that they are only indirectly estimated. The reason for the variation in the amount of precipitation even at the same oxygen concentration is thought to be due to the difference in vacancy concentration.

Therefore, the fact that almost no precipitation occurs in crystals with oxygen concentrations below IXl018cm-3 means that
This shows that vacancies cannot become precipitation nuclei (25) for oxygen concentrations below 0.018 cm-3. Of course, this does not mean that oxygen will not precipitate at any concentration of vacancies below I
It seems that there is no other choice but to set it to cm''3.

B is I Xl019cm-3, Ge is 7 Xl0I9c
1050 in a nitrogen atmosphere for a 4-inch diameter wafer according to the present invention containing +n-' and cooled at 60°C/min after heat treatment at 1350°C for 2 hours and a conventional CZ method silicon wafer.
After heat treatment for 1100 hours at 1100°C, a heat cycle of room temperature → 1100°C holding for 30 minutes → room temperature was repeated 1 to 5 times. The speed of loading and unloading into the furnace is 6 seconds. FIG. 5 shows the warpage of both wafers with respect to the number of thermal cycles. As is clear from FIG. 5, the wafer according to the present invention has 11% more warp than the conventional wafer.
It was reduced to 5 or less. Therefore, the wafer of the present invention is useful, for example, as a support substrate for SO2 devices.

The silicon wafer according to the present invention can also be used as a low-resistance substrate for evixaxial layer growth in the following ways compared to conventional silicon wafers. That is, (26) The substrate wafer for bipolar 1-transistor is usually 10Ω.
A low resistance wafer of about cm is used. At this time, since the B concentration of the substrate wafer is 10111 to 1019 cm-3, a considerable amount of lattice strain is accumulated. A silicon layer of about 10ΩcTrl (B concentration of about 1017cm''3) is usually grown on the surface of this substrate by CVD, but the mismatch in lattice constant caused by the extreme concentration difference between the substrate and the epitaxial layer Due to this, misfit dislocations are introduced into the epitaxial layer.
1019cm-' and Ge7X'l019cm-
3) is used as a substrate, the resistance is about 10-2
Since it is ΩcITl and there is almost no lattice strain in the substrate, it can be a high quality P substrate without causing lattice mismatch with the epitaxial layer. Figure 6a shows slip caused by misfit dislocations generated in the epitaxial layer of a conventional P'' substrate (B doped only), and Figure 6b shows slip occurring in the epitaxial layer when the wafer of the present invention is used as the P1 substrate. 6a.
As is clear from Fig. 6b and Fig. 6b, (27
) The wafer greatly reduces misfit dislocations.

Furthermore, when the wafer of the present invention was used, the warpage of the wafer that occurred after epitaxial growth was reduced to 1/3 of that when a conventional wafer was used.

Example 1 described above was doped with impurities in combination with B and Ge, but the silicon according to the present invention with small warp and precipitates could be similarly doped with the following combination of doping impurities. A wafer is obtained in the same manner as in Example 1.

Example 2-B (more than 10"cm-3), Sn Example 3-B (more than 1017cm-3) and Pb Example 4-
...Al2 (1017 cm-' or more) and C Example 5-
Ga (more than 1017 cm-') and C Example 6 =-I
The doping (addition) amounts of Sn, Pb, and C necessary for alleviating the strain due to the impurity tope in each combination of n (1017c+n-3 or more) and C are determined as follows.

The concentration of impurity atoms (i) in the silicon crystal is N1cm.
−', the strain ε caused by i is (2
8) Given by the following formula (Y, T, 1.ee et al,,
J,EIer,trochemSoc, Vol, 1
22 No, 4. (1975), pp, 530-
534 (see reference 17)).

ε, −〇, N,
(1) Here, No: Silicon density = 5 XIO
22cm-'γ. : Atomic radius of silicon - 1,17 people (
As is clear from equation 2), since B is smaller than the Si atom and rH<ro, the strain ε6 due to B becomes ε8>0 (tensile strain). Therefore, Ge has a larger atomic radius than Si.
, Sn, doped with Pb to form εGe, S,,t Pb
< 0 strain (compressive strain) is ε8+εce+ 311.
By generating Pb-0, the distortion caused by B can be alleviated. Also, All! , C; a and 1n are thicker than the Si atoms r AL, ca, +-> r o, so the strains caused by these elements εAL, Ga, and In are ε〈0
(compressive strain). Therefore, doping with C, which has a smaller atomic radius than Si, results in ε. 〉0 strain (tensile strain), (
3) (29) A/2. Ga.

It is possible to alleviate the distortion caused by

2 B+5n) Atomic radius of B rB = 0.90 Bi4 degrees Nl1 = I
When the atomic radius rsn = 1.40, the above (1), (2) and (3) concentration NSn
To find, we get 5n (4) ,',N, from the formula by In. -0,764'xNB 0.8XNe Therefore, the concentration (about 0.8 times the doped B concentration)
The strain caused by B can be alleviated by Sn doping amount (doping amount), and it is desirable to set Sn to 4 degrees (30) as shown in the table below, corresponding to the B concentration.

The atomic radius of Pb is γ, -1.46, and the Pb concentration N2. When calculating, Npb=0.6xN,
l is obtained. Therefore, approximately 0 of the doped B concentration
．． Strain caused by B can be alleviated by using Pb at a concentration (doping amount) six times higher, and it is desirable to set pb to 1 degree as shown in the table below, corresponding to the B concentration.

Atomic radius of Aff rAl = 1.25 in AQ concentration NAL = I 4) When calculating the degree Nc using the formula (J!f), Nc=0.3NA1 is obtained.

On the other hand, the solid solubility limit of C is 3.5Xl017cm-3 (T, Nozaki et al, J, EIectr
ochem, Soc, + Vol, 117. No, 2
(1970), pp, l566-1568 (Reference 1
8)), the strain caused by ApV can be alleviated with a C concentration (doping amount) approximately 0.3 times the doped Aff concentration, and the strain caused by /l can be alleviated. It is desirable to use the C concentration shown in the table below.

Since the atomic radius of Ga, γGa = 1.25, is the same as the atomic radius of Ap, the C concentration -〇, 3N is the same as in Example 4.
c-, and the C concentration corresponding to Ga@ degree is the same as the table in Example 4.

(32) Similarly, when calculating the C concentration Nc, Nc=1.6XN.

is obtained. Therefore, the doped In concentration is 1.
Strain caused by In can be alleviated with six times the concentration (doping amount) of C. Considering the solid solubility limit of C, it is desirable to set the C concentration as shown in the table below in accordance with the degree of Infi.

In the above-mentioned Examples 1 to 6, impurities of group 3B elements of the periodic table and 4B
In combination with group element impurities, silicon single crystals become nitrous, but 3B elements (B, A/!, Ga.

Even with the combination of In), a silicon wafer according to the present invention with small warpage and small precipitates can be obtained in the same manner as in Example 1.

Example 7...B+AI! .

Example 8...B+Ga Example 9...B+In (33) Example 10-B+ (2 types of Al1, Ga, In) Example 11-B+ (A, e, Ga, In
) in each combination alone or in total lXl017c
The doping (addition) amount of the impurity element necessary to ensure an impurity concentration (doping amount) of m −3 or more and to alleviate strain caused by impurity doping is determined as follows.

17 B+A℃ Atomic radius of B rB = 0.90 people, +1 atomic radius γA, = 1. ．． 25 Atomic radius γ of Si. = 1.17 people, and from (2) 2 mentioned above: Let the B concentration be N, and the Affi concentration be NA,
The relationship between NIl and I'JAt is determined from the above equation (1) and ε6+εA1-0: (34) ,', NAL-2,483XNB = 2.5XNa Therefore, approximately 2.5 times as much as B1 degree If the Al concentration is , the strain caused by B and the strain caused by An can be compensated and the overall strain can be alleviated. The effect of the present invention can be obtained if either B or Aj2 is contained at IXl017cm-3 or more, and the solid solubility limit of /l is about 2Xl019cm-3 (BellSyst
em Technical Journal, 39 (
1960), p., 210 (Reference 18)), it is desirable to set the doping amounts of B concentration N6 and A/2 concentration NA as shown in the table below.

(35) The relationship between Ga concentration N64 and 13i degrees N is shown in Example 7.
is the same as However, the Ga solid solubility limit is approximately 4 x 10 crr.
V” (see Document 1B), so the B concentration NR and G
It is desirable that the amount of a concentration N6a added be as shown in the table below.

Atomic radius of In TI, = ], , Example 7 as 50 people
Similarly, when the relationship between the In concentration N l n and 136 degrees NB is determined, it becomes N + n = 0.5 x NB. Therefore, about 0.5 times the In
Concentration N1. , the overall strain caused by impurities can be alleviated. The solid solubility limit of In is approximately 4 Xl017
c++r'' (see Reference 18), it is desirable to set the concentration as shown in the table below.

(36) In this example, AI! , Ga, and In, two of the three types are used, so there are the following three examples.

10-A B+ (An+Ga) Case above (1)
Considering equation (3), the following equation: ε, +ε4. +εG,
=0 (5), the strain caused by impurities can be alleviated.

And considering Examples 7 and 8, 2,5 Ni
l=NAL+NG, (6) is obtained, and A42? ! The sum of 4 degrees and the Ga concentration should be 2.5 times the B concentration. Then, the concentration of group 3B elements is I
If the effect of the present invention is to be achieved at Xl017cm-3 or higher, any one of the following seven conditions should be satisfied.

N8≧l×1017 (37) N, 1≧lXl0” Nca21 X10” NB +Nat≧I
) formula and ' NAL + Nca > I X1017cm
-", it is desirable to use a concentration as shown in the table below.

Note that the combination of Al concentration and Ga concentration is, for example, NA
L=0. I XIO”cm-3 and Nca= 0.9
X1017cm-”orNAL=0.2XIO”
cm-' and NGa=0.8XIO''cm-3or, and in other words, the total concentration of Affi and Ga should be 2.5 times the concentration of B.

Further, the upper limit of the amount added is /l as mentioned above.

(38) The solid solubility of Ga is ~2 ×] and 019 cm-3.
〜4×1019cm−3, so N At ten N ca
The maximum solid solubility of is considered to be ~6xlQ19cm-3.

Therefore, the concentration of B is (6 from Examples 7 and 9 NA1=2
．． There is an impurity concentration relationship of 5NB and N1n=0.5XNB, and considering ε8+εA1+ε1°-〇 which relaxes the impurity-induced flux, the following formula; % formula % can be found. Then, any one of the following seven conditions is satisfied.

NB≧IX1.017 N, l≧lXl0” NlTl≧1xto17 N [1+ N At≧1 xio” N A L + N , , l≧I Xl017N1.
, 10NB≧lXl0” NB1NAL 10N+fi≧I Bye.

10-CB+ Ga 10-In case Ga's atomic radius TGn is the same as Af's atomic radius γA1, and 1.
Since there are 25 people, this case is the same as Example 10-B, except that Aff and Ga are replaced.

Therefore, 5 NB = 2 NGa+1ON+,
B, Ga, and In may be doped (added) so as to satisfy the condition of (Formula (8)) and any one of the following seven conditions.

NB≧1×1017 NG11≧1×1017 N1,≧I XIO” NR+NG8≧I XIO” N6. +N,,≧lXl0” N,,+NB≧I
"cm-3 or higher is required. For this purpose,
B, At2. What is necessary is to dope (add) Ga and In.

Nn>1
．． ≧1×10'7 N6a ten N1. , ≧1×10'7 NG, +NB > I Xl017 N+l, +Ne≧I Xl017 NB+Nnt+Nca21 xlO" NAL +NG-+ N111k I XIO"NGa
+1ON+NB 21 xlOI7(9) (41) N,,,+Nll +NH no I Xl017NB +
NB (10N6. + N +,, ≧1XIO17 The above-mentioned embodiment is an example of a silicon single crystal produced by the CZ method, but the present invention can also be applied to a silicon single crystal produced by the FZ method. Although the single-crystal ingot is subjected to quenching, high-temperature heat treatment and quenching may be performed after it is made into a silicon wafer.

[Effects of the Invention] As described above, according to the present invention, by selecting the added impurity, regulating its amount, and regulating the conditions for high-temperature heat treatment and rapid cooling, it is possible to create a film with less warpage, fewer dislocations (including slips in the epitaxial layer), and A high-strength silicon wafer (substrate) with few precipitates can be obtained. If a silicon wafer with such good characteristics is used, the characteristics of the devices to be formed will be improved, and the yield will be improved.

[Brief explanation of drawings]

FIG. 1 is a graph regarding the yield stress of a silicon single crystal according to the addition of B and the addition of B and Ge (42), and FIG. 2 is a perspective view of a silicon single crystal ingot obtained by the CZ method. FIG. 3 is a graph showing the relationship between high-temperature heat treatment temperature and cooling rate and precipitate density, FIG. 4 is a graph showing the relationship between oxygen concentration of silicon single crystal and amount of precipitates, and FIG. 6 is a graph showing the relationship between the number of heat treatments applied to a silicon wafer and warping, FIG. 6a is a schematic plan view of an epitaxial layer formed on a conventional silicon wafer, and FIG. 7 is a schematic cross-sectional view of an epitaxial layer formed on such a silicon wafer, and FIG. 7 is a graph showing the relationship between the oxygen concentration in the silicon crystal and the warpage of the silicon wafer,
FIG. 8 is a graph showing the relationship between oxygen precipitate density and warpage of a silicon wafer, and FIG. 9 is a graph showing the relationship between strain and stress when a load is applied to a crystal. (43 ) Figure 10 is a graph showing the relationship between B'a degree and yield stress of silicon single crystal, Figure 11a is a graph showing the relationship between B concentration and strain of silicon single crystal, and Figure 11b is a graph showing the relationship between B concentration and strain of silicon single crystal. Figure 11c is a graph showing the relationship between GeFI degree and strain in a silicon single crystal, Figure 11c is a graph showing the relationship between Ge concentration and strain in a silicon single crystal with a constant B concentration, and Figure 12 is a graph showing the relationship between Ge concentration and strain in a silicon single crystal with a constant B concentration. FIG. 13 is a graph showing the relationship between B concentration and precipitate density in a silicon single crystal, and FIG. 13 is a graph showing the relationship between B concentration and precipitate density in silicon single crystal with and without high-temperature heat treatment. . 1...Silicon single crystal ingot (44) 5eed side 1A 1st flash 1.0 1.25 1.50 Oxygen concentration (01018cm-3) 1.75 1st round ○ 1 2 3 4 Number of thermal cycles 1st 6Q Oxygen concentration [cm-51 6th Figure 6b Oxygen precipitate density [cm-2] 2nd flash 9th 11a country 11b 10th Figure 2 4 6 8 B concentration (×1020cm-3) 0 Ge Relaxation of strain caused by B due to dope 1 Country 11c

Claims (1)

[Claims] A boron impurity of 1.1×10^1^7 cm^-^3 or more and an impurity of a group 3B or 4B element of the periodic table larger than a considerable amount of silicon atoms to compensate for the distortion caused by the boron impurity. After holding a silicon single crystal ingot containing the above at a temperature of 1200°C or higher in an inert atmosphere for 30 minutes or more, and cooling it at a cooling rate of 20 to 60°C/min, from the silicon single crystal ingot A method for manufacturing high-strength silicon wafers, characterized by cutting out the wafers. A silicon single crystal containing an impurity of any one of aluminum, gallium, and indium with a size of 2.1×10^1^7 cm^-^3 or more and a considerable amount of carbon impurity to compensate for strain caused by the impurity. holding the ingot at a temperature of 1200°C or higher in an inert atmosphere for 30 minutes or more,
A method for producing a high-strength silicon wafer, which comprises cutting the wafer from the silicon single crystal ingot after cooling at a cooling rate of 20 to 60° C./min.