JPH01246197A - Production of single crystal of compound iii-v semiconductor - Google Patents

Abstract

PURPOSE:To prevent cell growth and keep a single crystal not so as to cause rearrangement till high solidification ratio is attained, by controlling both of temperature gradient and pulling-up speed as a function of solidification so as to satisfy a specific relation equation and pulling up the single crystal of compound III-V semiconductor, when the single crystal is pulled up by a liquid encapsulated method. CONSTITUTION:Impurities having <=1 coefficient of segregation and high concentration, e.g., a raw material melt of compound III-V semiconductor containing In (having 5X10<19>-1X10<21>/cm<3> concentration of crystal head), e.g., GaAs is covered with a liquid capsule and high pressure is applied to the liquid capsule with an inert gas and a single crystal is pulled up using a seed crystal. In this case, the single crystal is obtained by controlling temperature ingredient G(g) and pulling-up speed R(g) used as a function of solidification ratio from g=0 to g=gm(0<gm<1) so as to satisfy the inequality (G: temperature ingredient of the melt, R: pulling-up speed, m: temperature gradient of liquid phase line, C0: initial concentration of impurity, K: coefficient of segregation in a raw material liquid of impurity, g: solidification ratio (ratio of solidified part to initial raw material melt amount).

Description

Detailed Description of the Invention (7) Technical Field The present invention is directed to the liquid encapsulated method (Liquid Encapsulated/L/ method) for preparing ■-V group compound semiconductor single crystals.
Czochralski) relates to improvements in the pulling method.

m-v group compound semiconductors include GaAs and GaP.
.

InSb%InP, - is a semiconductor composed of group Ⅰ and group V elements such as -.

As a method for producing bulk single crystals, a liquid-sealed Czochralski method (or simply a liquid capsule method), which is an improved version of the Czochralski method, is used.

This involves covering the raw material melt in a crucible with B2O3 and applying high pressure to prevent the volatilization of group V elements.

Group V elements have a high dissociation pressure at the melting point of the compound, so B2
Pressure is applied with O3 to suppress the dissociation of group (Ⅰ).

Since it is a pulling method, it has the advantage that it is easy to obtain cylindrical crystals.

However, in the liquid capsule method (LEC method and Renki), there is a large temperature gradient in the vertical direction, so thermal distortion is likely to occur. Therefore, a large number of dislocations occur in the pulled single crystal.

Two techniques are used to reduce dislocations.

One is to add impurities and harden them.
-4y hardening).

isoelectronic impurity
Add an impurity that is For example, GaA
Impurities such as In and B%kl are added to s. It is an isoelectronic impurity because it must not change the electrical properties of the crystal. Adding impurities makes the crystal harder and suppresses the generation of dislocations.

The other is to lower the temperature gradient above the solid-liquid interface. In this way, the crystallized portion will not be cooled down rapidly, and thermal strain will also be reduced. In order to reduce the temperature gradient, two or more heaters are used or a heat shield cylinder surrounding the crystal is provided.

(b) Conventional technology By adding impurities, it is possible to pull a single crystal with low dislocations.

In this case, the effect is weak unless impurities are added at a high concentration of about 10''cm'' to 10''cm-3.

In addition, pulling is performed under a low temperature gradient.

This improved LEC method introduces new difficulties.

That is, abnormal growth called cell growth occurs during growth. After this, it becomes polycrystalline and becomes useless.

Cell growth is caused by compositional supercooling or supercooling (s
supercooling;

In the pulling method, the conditions for cell growth (super cooling) to occur are expressed by the following equation.

Here, G is the temperature gradient in the melt near the solid-liquid interface, and R is the pulling rate. ko is the segregation coefficient of impurities in the melt. C is the impurity concentration in the raw material melt. D is the diffusion constant of impurities in the melt.

m is the temperature gradient of the liquidus line. In other words, the freezing point (melting point) is lowered by adding impurities, but this is the rate of freezing point lowering. That is, if the freezing point depression is denoted by ΔT and the impurity concentration is denoted by C, then ΔT = mc
There is the relationship (2). This is a relationship that always holds true when there is a solute dissolved in a solution and when the concentration is low.

What the conditions in equation (1) mean will be briefly explained.

This problem arises because impurities diffuse slowly in the melt. Since it is a fluid, even if there is fluctuation in impurity concentration, this is not immediately averaged out.

The segregation coefficient is the ratio of impurity concentrations contained in the solid and liquid that are in contact at the solid-liquid interface.

A problem arises when the concentration of impurities contained in the solid is lower, that is, when the segregation coefficient is less than 1.

This will be explained with reference to FIG. Let C be the impurity concentration in the raw material melt. Let AB be the interface between the crystal and the melt. Suppose that at a certain time, what was AB rises to the CD line after dt. The volume of this part is 5Rdt.

Here, S is the cross-sectional area of the crystal, and R is the pulling rate.

A melt having a volume of 5Rdt contains dQz = csRdt (3) of impurities. However, in a crystal, only dQ = ckSRdt (4) of impurities are contained for the same volume. These differences d Qt d Q , =c (1k) S Rd
t (5) becomes surplus at the solid-liquid interface.

In other words, impurities seep out at the solid-liquid interface.

Since the impurity flow is proportional to the concentration gradient, the impurity flow J is a, r = -5D-(6) z with two downward axes in the melt. The reason why an impurity flow is generated is that impurities are oozed out at the interface AB as shown in equation (5). The outflow of impurities per unit time is calculated by converting equation (4) to d
You can get it by dividing by t. Since this and the impurity flow J should be balanced, the following formula holds true in a steady state: cSD-+C(1-k)SR=Of force z.

(7) is obtained from equation (7). By integrating this with respect to 2, the impurity concentration with respect to 2 can be determined. Impurities decrease linearly in the depth direction.

Now, C multiplied by m is the melting point (freezing point) depression T. The melting point depression ΔT also decreases linearly in the depth direction.

The slope of melting point depression ΔT is given by.

The problem is that the melting point decreases.

A further problem is that the melting point depression varies as a function of depth 2.

Let us assume a surface IF in the melt.

Compare interface AB and melt surface EF. Of course, EF has a higher temperature. However, EFQ has a higher melting point. Assume that the difference in melting point between plane AB and plane EF is equal to the difference in temperature between AB and EF. Then, AB and EF can coagulate at the same time. Simultaneous coagulation is the cause of cell growth.

In other words, if the temperature gradient G between AB and EF is smaller than the melting point depression gradient, cell growth will occur. This is mc(1k)RG ((tO) where the temperature gradient is G. (10) and (1) are the same equation. Thus, (
This means that 1) has been proven.

In the pulling method, nearly all of the raw material melt is solidified, so the impurity concentration varies. If k is an impurity smaller than 1, the impurity concentration will gradually increase.

Therefore, when G and R are kept constant, as C increases, formula 00) is established and cell growth occurs.

The formula for how C changes with the increase in
Guide using Figure 9. For simplicity, the impurity concentration is assumed to be per volume. It is also assumed that the density does not change between the crystal and the melt. Let S be the volume of the crystal, and L be the volume of the melt. The sum of these is constant.

S + L = Lo Ql) Lo
is the amount of raw material melt before pulling.

Let m be the amount of impurities in the melt.

Suppose that the dS portion is crystallized. This includes c, k d
Contains S impurity. Therefore, a large amount of impurities are removed from the melt.

dm=-ckdS a3.

Multiplying the impurity concentration C by the melt scale gives the amount of impurities, so m=cL α3). (I
Differentiate I), differentiate α■, and get as = −dL
(14) dm = cdL +Ld
c (15), and from these, Ldc =-c(1-k)dL (15) This can be easily integrated. GO is the initial impurity concentration.

An independent variable called solidification rate g is used instead of L/Lo.

This represents the proportion of the solidified portion S of the initial raw material melt Lo.

Then, using g, ση becomes C= Co (1-g)k-1 (I'm. This is the impurity concentration of the melt. The impurity concentration C8 in the solid in contact with this is k Multiply by Cs = 'ok (1g
)k−” C'0.

The solidification rate g is the portion that is pulled up and solidified. It can be seen from (11) that the impurity concentration in the melt fluctuates as it is pulled up.If k(1), it diverges at the limit of g→1.

In reality, g→1 is never raised. However, ideally it should be close to 1.

By substituting 1 flaw into 00, the conditions for cell growth to occur are g
obtained as a function of

The conditions under which cell growth does not occur are as follows by reversing the inequality sign. That's what it means.

(1) Japanese Unexamined Patent Publication No. 61-31382 In order to solve the problem of Co, the present applicant has developed a method in which the pulling speed R is gradually lowered. It is disclosed in Japanese Patent Application Laid-Open No. 61-31382 (published on February 13, 1983).

Since g increases with pulling up, the right side of Co increases. In order to satisfy the Co inequality, the pulling speed R is decreased.

The temperature gradient G in the melt is kept constant.

Since the pulling speed R is approximately equal to the pulling speed of the pulling shaft, it can be easily controlled.

(Ko JP-A-62-191495.62-260794
On the other hand, by increasing the temperature gradient G in the melt, inequality I
Inventions have also been made to maintain 2D.

JP 62-191495 (S 62.8.21 Published) JP 62-260794 (S 62.11.13
(Published) All of these are things that increase the temperature gradient G.

The former is based on the G
(1-g)".

In the latter case, for the value of g=o, G is increased from the beginning like Go(1-g).

(b) The problem that the invention seeks to solve: The lower limit of G/R rises with the progress of growth.
A method was proposed to lower G or to raise G.

These are simply called the R control method and the G control method.

The R control method has the advantage of being simple, but also has the disadvantage of being too time consuming.

Since the G control method changes the temperature gradient in the melt by changing the relationship between the crucible and the heater, it is necessary to conduct extensive experiments to investigate these relationships in advance. However, the pulling time can be relatively short.

However, the inventor discovered that there is a more fundamental drawback.

In both the R control method and the G control method, the grown crystal has large dislocations in the portion where the solidification rate g is large (crystal tail).

FIG. 1 shows the results of measuring dislocation density (EPD) as a function of solidification rate.

The horizontal axis is the solidification rate. The vertical axis shows the etch pit density. A white circle indicates a result by G control, and a triangle indicates a result by R control.

The plate under R control is referred to as Comparative Example 2, and the plate under G control is referred to as Comparative Example Dish. The detailed conditions will be described later.

In either case, while the solidification rate g is low, there are few dislocations, but as the solidification rate g increases, the number of dislocations increases.

If the EPD must be 500 pieces/cIPI or less, the solidification rate must be 0.5 or less. In other words, only the top half of the crystal can be used.

The solidification rate indicates the percentage of the solidified portion of the melt, and it approximately corresponds to the distance from above the pulled crystal. The upper part of the crystal has a smaller g, and the lower part has a larger g. In other words, the fact that EPD increases with g means that EPD increases as you go down the crystal.

This means that in terms of R control, dislocations are more likely to occur because the pulling speed is too slow.

Considering dislocations, this means that H has a lower limit below which it should not be lowered.

Similarly, regarding G control, if the temperature gradient G in the melt is too large, dislocations are likely to occur. Considering dislocation, this means that there is an upper limit on G that should not be increased beyond that.

It is not enough to have low dislocations up to g=0.5;
It is desirable to have low dislocations even at high g values.

The present invention provides an improvement in the LECEC method-single crystal pulling method such that no cell growth occurs and low dislocations up to high g values.

(a) Method of the present invention First, the liquid capsule method will be briefly explained with reference to FIG.

The crucible 1 is supported by a susceptor 2. For example, the crucible 1 is made of PBN, and the susceptor 2 is made of carbon.

The lower shaft 3 supports the susceptor 2. The lower shaft 3 can be rotated up and down.

The crucible 1 contains a raw material melt 4 that is to be a raw material for a compound semiconductor, and a liquid capsule 5 that covers it.

A high-pressure inert gas 6 is contained and applies pressure to the liquid capsule 5. This prevents volatilization of group V elements.

The upper shaft 7 is provided so that it can rotate and move up and down. A seed crystal 8 is attached to the lower end. Seed crystal 8 is immersed in the raw material melt for seeding. The upper shaft 7 is pulled up while rotating the upper shaft 7 and the lower shaft 3. A single crystal 9 grows following the seed crystal 8.

A pressure container 10 surrounds these devices and is capable of applying a pressure of several atm to several tens of atm to the inert gas.

The heater LL12 is a resistance heater. This heats the raw material melt. In particular, the lower heater 11 heats the melt, and the upper heater 12 heats the crystal. In this way, the temperature gradient in the space where the crystal is pulled is lowered by using multiple heaters or by providing a heat shield plate.

Even though the temperature gradient in the space was low, a fairly high concentration of impurities was added, creating cell growth problems.

In order to solve this problem, a method of increasing the temperature gradient G in the melt and a method of decreasing the pulling rate were proposed. All of them can be evaluated in terms of inhibiting cell growth.

However, there was a problem that the low dislocation portion was too small.

Therefore, in the present invention, both G and H are changed.

That is, as the crystal is pulled, the pulling speed is lowered and the temperature gradient in the melt is increased. This prevents cell growth from occurring.

Varying both G and H is more difficult than varying either one.

However, if G and R are varied, the amount of variation in G and the amount of variation in H can be reduced. Therefore, the occurrence of dislocations is reduced.

For example, R control and GR control will be compared. As the solidification rate g increases, G/R must be increased.

If G is fixed and R is decreased, the decrease in H will be quite large.

However, if G is increased and R is decreased, the amount of decrease in R is small.

In other words, in terms of suppressing the occurrence of dislocations, the GR control method can reduce the decrease in R by a smaller amount than the R control method.
This means that dislocation can be reduced.

The same can be said when comparing the G control method and the GR control method. When G is increased while R is fixed, the increase in G is quite large. However, if R is decreased and G is increased, the amount of increase in G can be reduced. Since the increase in G is small, the occurrence of dislocations in the crystal is small.

The method of the present invention ultimately achieves the following by varying G and R.
This means ensuring that the inequality Cυ always holds true. That is, as a function of the solidification rate g, G (g, ), R (b)
The method of the present invention is to program .

The R control method is The G control method is That's what it means.

The method of the present invention is, after all, fully expressed by the (company) formula.

This will be explained more intuitively using drawings.

FIG. 2 is an (R,G) coordinate system in which the horizontal axis represents the pulling rate R and the vertical axis represents the temperature gradient G in the melt.

Here, we draw a group of straight lines that pass through the origin O and are υ with g as a parameter. The straight line g=0 is (c)
This is a straight line expressed by the equation where g=O.

If k(l), (1-g) increases with g, so the larger g is, the greater the slope of the straight line will be.

For a given value of g, the coordinates (R, G) must be above and to the left of the straight line for that g, using the formula C
υ〜(24) means.

For example, consider coordinates Q (Ro, Go) where GlR at the beginning of pulling is Go, Ro.

A straight line passing through point Q and parallel to the G axis is taken upwards and is defined as QS. This is to increase Q with g while setting 1 (=l"lo). This is G control. However, the coordinate for the value of g (Ro%G (2)) is the g of C straight lines.
It is required that it be on the upper left side than that of the object.

Pass through point Q and draw a straight line parallel to the R axis to the left.
, T. This is to reduce R with g while maintaining G=Go. However, the coordinate (R(d Go ) for the value of g must be on the upper left of the straight line for g of □□□.

The method of the present invention provides Q UI X Q U2, QU3, Q
For example, it can be indicated by U4 or the like. These are curves or straight lines going from point Q to the upper left.

In other words, (R, G) represented by an arbitrary curve included in /TQS is controlled.

However, the coordinates (R(2), G(2)) for g are C
It must be above and to the left of the line for g.

From these figures, it can be seen that the GR control method requires less variation in G and H than the method that controls only G or only R.

Assume that there is an upper limit for G in terms of dislocation density in the awn. This is designated as Gmax and is indicated by a broken line. If the temperature gradient in the melt is large, the temperature gradient above the melt is also large, which tends to cause thermal distortion in the crystal. It is natural that dislocations increase.

It is assumed that there is a lower limit for H in terms of dislocation density. This is designated as Rmin and is indicated by a broken line.

In order to lower the dislocation density, the coordinates (RlG
) must be below and to the right of these dashed lines.

However, as the solidification rate g increases, the condition Cυ~(24
) both push (R, G) upward to the left, making it difficult to satisfy the condition of low dislocation density.

When increasing the solidification rate g, G exceeds Gmax fastest under control along the control line QS.

In other words, it is G control. In this example, go, 5 to 0.6
G exceeds Gmax between. This corresponds to the curver shown in FIG.

When increasing the solidification rate g, R exceeds Rmin fastest under control along the control line QT. This is strictly R control.

In this example, g is between 0.5 and 0.6 and R exceeds Rmin. This corresponds to the curve In in FIG.

None of the control lines QU1 to QU4 of the present invention reach Gmax or Rmin at gm06. In particular, intermediate control lines QU2, QU3, etc. can have low dislocation densities even for fairly high g's.

The present invention has been described in detail with reference to the general formula (2) and FIG.

Next give some examples.

(G) Examples of G and R functions (Example 1) G(g:l= a + bg, R(g
)-c-dga, b, c, and d are positive constants. However, g must be υ between O'' gm. Since the right side of (to) diverges to infinity in the limit of g → 1, the equality sign holds in gm (< 1). When gm is 1 The closer to , the more cell growth can be delayed.

However, when gm approaches 1, G becomes large and R becomes small, which may increase the time required for pulling or increase dislocations.

(Example 2) G(g)=a −4−bg, R(g)=
(C+dg) -1a, bXcld is a positive constant. However, when g is between 0 and gm0, -1 (a+bg X c+dg)≧H(1-g)
It is assumed that (28) holds true.

Here, gm means a value between 0 and 1.

It is different from the previous GM. The same applies below.

(Example 3) G=a+bg, R=(a+bg)c(1
-g) 1-ha, b, c are positive constants, h is o<h<
is a constant of i. However, it is assumed that (1-g)''≧H(1-g)k-1C becomes glossy in the range of g=o-gm.

(Example 4) ""a+bg, R=(a-)-bg)c
(1-g)” This is Example 3, with h=.

The conditions imposed are cH≦1 (to).

(Example 5) G = (a-bg) C(1-g)h
=, R=a-bgalblc is a positive constant, and h is a constant of O~1. The conditions are C(1-g)h-''≧H(1-g)k-1c31) (Example 6) G=(a-bg)c(1-g)k- in the range of g-0 to g- 1, R=
a-b g condition is C≧HC33 (Example 7) G = (a+bg) (C-dg)
, R=c-dga, b, c, d are positive constants o
g"" within the range of 0-gm a-)-bg≧H(1-g)k-1C33) (Example 8
) G=a-)-bg: , R= (a-)-b
g) (d-cg) a, b, c, d are positive constants o
g=O” in grn (Example 9) G==a(1-g)
Below, R=b(1-g) a1 b is a positive constant. The conditions are: 1≧H (35) In addition to these, various combinations are possible.

G(g) increases with g, but does not have to be a monotonically increasing function. - May decrease over time.

R(2) decreases with g, but it does not have to be a monotonically decreasing function. - It may increase over time.

(ri) How to give temperature gradient G The pulling rate R is an independent variable that can be directly controlled.

However, the temperature gradient G is not. The temperature gradient G in the melt cannot be measured in practice.

The temperature of the melt cannot be measured unless the thermocouple is immersed in the melt. However, it is of course impossible to perform good crystal pulling with the thermocouple soaked.

If the crucible is empty, the temperature distribution inside the crucible can be determined as a function of the relative position to the heater.

The temperature distribution in the melt can be measured with the crucible filled with the raw material melt (without crystal pulling). Temperature distribution can also be determined by changing the relative position to the heater.

From these results, it was found that the temperature gradient G in the melt can be changed by changing the relative position of the heater and the crucible.

Then, a graph as shown in FIG. 3 is obtained.

The horizontal axis is the relative height of the crucible to the heater.

The power of the heater is constant. As the crucible is raised higher, only the area near the bottom of the crucible is heated.
The temperature gradient G increases.

In this example, L = Omn and G = 30'C/c
m, L = 70 mm, and G = 96°C/Cl11.

The relationship between L and G varies depending on the crucible, heater size, and heater power. Therefore, if the device changes, the LG curve will change. However, an LG curve can be obtained.

Since we know the relationship between L and G, we can control G by varying L.

(To) How to determine H The difficulty in carrying out the present invention is the method of controlling the temperature gradient G as described in the previous section. In order to accurately determine the GL curve, numerous tests must be performed.

Another difficulty appears in the inequality Cυ~(2a.

This means that the value of is unknown. m, k, and D should have fixed values for impurities, and co is determined from the design value. However, in reality, it can be said that there is no reliable data for m, Dl, and especially D.

D varies with temperature and may vary with equipment and crucible dimensions.

In such a case, the question is how to find the value of H. If we don't know H, we can't create a control program for G and H.

This, however, can be determined experimentally.

Single crystals of the same composition are pulled using the same apparatus while changing the impurity concentration, pulling rate R1, and temperature gradient G. Then, the solidification rate gmax when cell growth occurs is examined.

Repeating this experiment and using equation (25), we obtain (36)
H can be found. From this, mlD can be calculated.

Since mlD is a constant value for the device and compound, that value can be used from now on.

For example, when In is added as an impurity to a GaAs melt, the diffusion coefficient is D '';;10'ct/sec
C'37). Accurate values should be determined by measurements on the equipment and -1).

For example, when N grams equivalent of In is contained in the GaAs melt, T to lower the melting point is ΔT = 36ON
There is an evaluation that says that it is given by (to). GaAs molecular weight M = 72.3, density p = 5.3 jl/cra
, from Avogadro's number Lo=6.02 x 1023,
There is a relationship between impurity concentration C and N.

M ΔT=360-(4G ρLO i.e. ΔT=8.16 X io-”c
(411. The melting point depression does not depend on the type of impurity, and these constants are the same for any impurity. 'This is a chemical law. If it is dilute,
This relationship is established. In other words, if the melt is GaAs, m = 8.16 x 1.0-”°Ccra
(It would be 4'2J.

However, C37), (43glue, m) is a rough guide, and in reality, it is better to measure and determine these values for each device and each type of compound semiconductor.

However, even if D and m are not known individually, the present invention can be implemented if the parameter H including these is known. H
To find the cell t1., change the concentration C and actually set the cell t1. What is necessary is to measure the solidification rate gm at which the length increases.

(b) Examples and comparative examples Examples and comparative examples will be explained using examples ① to ②.

FIG. 4 is a graph showing the change in G/R depending on the solidification rate g during pulling. The horizontal axis is the solidification rate g. The pulling is done from g=0 towards g→1. The vertical axis is G/R.

Curve hi gives the limit at which cell growth occurs. This is based on the assumption that the equal sign holds in equation (21). In reality, it draws a curve. k=0.1.

(2) to V are examples shown in FIG. 4 when shown as a G/Rg curve.

In the curve jmop, G/R changes in a portion slightly above the cell growth generation region. This is a last-minute example. Three types of this curve will be explained: GR-controlled (I), R-controlled (1), and G-controlled (2).

In addition, the GR control along the curve jlqrOP (1
) will also be explained.

Furthermore, the GR control M along the jmsp in the cell growth generation region will also be explained.

Now, the common parts of the example and the comparative example will be explained.

GaAs polycrystal is doped with In to grow a GaAs single crystal.

6 kg of GaAs polycrystals produced by the horizontal Bridgman method (HB method) and In as an impurity were placed in a 6-inch PBN crucible. Additionally, B2O is used as a liquid sealant.
I put 3 into the crucible.

In, the impurity concentration is I at the front (upper part) of the crystal.
It is designed to be X 102102o.

This is heated with a heater using a device as shown in Figure 6.
After making it into a melt, it was pulled up using a seed crystal. The rotation speed of the lower shaft is 2 rpm, and the rotation speed of the upper shaft is 5 rpm. The direction of rotation is opposite.

The growth rate R was initially set to 0.9 cm/h.

The temperature gradient G in the melt was approximately 40° C./cm.

R□ = 0.9cm/h 1GO = 40°C/cm
This is common to all Examples 1 to V, but subsequent changes are different.

Then, using an automatic shape control device, a 3-inch diameter Ga
An As single crystal was created.

If you keep G and R constant and pull the single crystal in the rows “) and (
Normal LEG method) To find out what happens, G-40℃
/cm, R = 0.9 cm/h, and the pulling rate was 'f'f. This is from about g = 0.21,
Cell growth factor 1'0, I'l, 5←0 This single crystal is thinned (
The sea urchin... was etched with KOH, and the EP was released.
D was measured.

When g=0.1, EPD is 200cm, 1g=0.
2, the EPD was about 100 cm-2 and the dislocation was low. However, it becomes polycrystalline from 0.21, making it impossible to obtain a good product.

In all of the five examples described below, either G, H, or both are changed.

(b) Example ■ G/R is varied along the curve of jlrnnOp.

In other words, G/R= 44 0≦g≦0.2
(44) G/R= 36.3(1-g)-” 0.2
(g≦0.68 (49G/R= 100
0.68 < g < 1. (46). Furthermore, both G and R are changed here. The change is illustrated as a function of g in FIG. The solid line corresponds to ■.

For G, G = 4Q Ori, p, ・:<o2C1
ηG-4og-+-32o2<g≦0.68 (48
)a=60 o6s<g<1 (4
9) For H, 1==9.9 0≦g≦0.2-
(I) R=0.0275 (40g+32X1-g)
O-90,2 (g≦0.68 (51) R=0.6
A control such as 0.6 s (g(IN) is performed. According to this, cell growth did not occur until g = 0.68. Cell growth occurred above 0.68, resulting in polycrystalline formation. However, , so far it has been a single crystal.

The ingot was sliced into wafers and the etch pit density EPD was measured. The solidification rate g and the wafer can be related.

Curve I indicated by a black circle in FIG. 1 shows the average EPD within the plane of the wafer as a function of the solidification rate.

EPD is 100 to 200 pieces cm-2. It can be seen that the dislocation is extremely low. With this, the entire range of g from 0 to 068 can be used.

(-)) Comparative Example II [(R control) This is an example for comparing GR control and R control.

G/R is varied along the curve jmnop in FIG.

However, G is kept constant and only R is varied.

Changes in G and R are shown by the dashed dotted line in FIG.

G is 40°C/cm. R is R=0.9 0≦g≦0.2 (Foundation) R
= 1. I X (1-g)0°9 0.2≦g<o
, 6B (Foundation) R= 0.4 0.68
≦g<1).

As can be seen from the lower diagram in Figure 5, in the case of R control, R is set to 0.
．． The speed must be reduced to 4cm/h. With GR control, R only needs to be lowered to 0.6 cm/h.

Dislocation density is a problem. In FIG. 1, the triangles indicate the measurement results of dislocation density. Cell growth does not occur until g=0.68, but when g=0.5 or more, the dislocation density suddenly increases.

(S) Comparative inversion (G control) This is an example for comparing GR control and G control.

G/R is varied along the curve jmnop in FIG. R is kept constant and only G is varied. The broken lines in FIG. 5 indicate the values of G and H.

R is 0.9 cm/h.

G=40 0≦g (to 0,2)-〇,
9 G = 32.7X (1-g) 0.2≦g<0
68(5f)G = 90 0.68≦
g<t) The value of G increases at 90°C/cmi. 60°C/cn+ with GR control! That's fine.

The dislocation density becomes like the curve shown by the white circle in FIG. Also, when g exceeds 0.5, the number of dislocations increases significantly. Although cell growth does not occur until g = 0.68, such a large number of dislocations prevents the wafer from being used as a substrate for electronic devices.

(→ Example ■ G/R is changed along the broken line jlqrop (Fig. 4). This means that regarding G/R, G/R= 44.4 0≦g(0,1(E)
G/R=-g-1--0,1≦g (0,6■G/R=
100 0.6 (g (61). Both G and R were changed.

The state of the change is shown by the solid line in FIG.

For G, G=40 0≦g(0,1(60
G = 60 0.6≦g<1 (64
) For R, R=0.9 0≦g<0.1 (6
5) R= 0.96-0.6g 0.1≦g
<o, 6(66)R= 0.6 0.6
≦g<1 (6 times). This means that cell growth did not occur until g=0.68.

The measurement results of dislocation density are indicated by X marks in FIG.

According to this, it can be seen that there are fewer dislocations when g=0 to 068.

(S) Comparative Example ■ In FIG. 4, G/R is changed along the curve jlsp. This is within the region where cell growth occurs.

G/R= 44.4 0≦g(o, 1 ■
G/R=42. I X j; 7o, t≦g<0.82
(60G/R= 100 0.82≦g
< 1 (70) Furthermore, both G and H were varied. This is shown by the dashed line in FIG.

For G, G = 40 0≦g (0,1(71
)G=-g+-0,1≦g (0,82(7埠G=
60 0.82≦g<1 (c) For R, R=0.9 0≦g<0.1 (
74) R= 0.6 0.82≦g<1
(It is 7. It is GR control. However, g = 0.32
cell growth occurred.

Effect: In a method for producing a low-dislocation single crystal by pulling a compound semiconductor single crystal doped with impurities into a space with a low temperature gradient, the pulling rate R and the temperature in the melt are adjusted to avoid cell growth. Both temperature gradients G are changed.

When impurities with a segregation coefficient smaller than 1 are doped at a high concentration, cell growth tends to occur as the material is pulled up.

However, methods in which only or only G is varied in order to avoid cell growth tend to cause R to drop too much or G to rise too much, so that dislocations tend to increase instead.

In the present invention, since both G and H are varied in a complementary manner, R
, G changes gradually. Therefore, dislocations can be suppressed to a low level while preventing cell growth.

(Force Segregation Coefficient We have explained the example of doping GaAs with In (1 (==Q, l). However, for ■-■ group compound semiconductors, In, B, Sb, Si, Ge, Ga,
As1P and the like have an impurity hardening effect.

For example, in the case of InP single crystal, Ga1As, Sb
Doping methods such as these are effective. In the case of As-doped InP, the same problem occurs since k=0.4.

The segregation coefficient varies depending on the conditions, but approximate values are shown below. In reality, it should be measured depending on the equipment conditions.

(1) 5i(0,016) Ge(0,02)In(Q,
l) P(3) 5i(0,14)(2
1Ge(0,2) 5i(1) Te(0
,4)(3)Ge(0,07)5s(0,93)S in InAs
i (0,4)(4) Go(0,012) P(0,16) 5s in InSb
(0,17) (5) AJSb medium B (0,01) Pb (0,01) 5e
(0,4) sico, x) (6) As(0,4) Ga(5) in InP

[Brief explanation of the drawing]

FIG. 1 is a graph showing the results of measuring the dislocation density as a function of the solidification rate g for four methods 1 to 1 when pulling a GaAs single crystal doped with In by the liquid capsule method. FIG. 2 is a diagram for explaining how the temperature gradient G1 pulling rate R should be changed in order to avoid cell growth. FIG. 3 is a graph showing an example of actual measurement of the relative position of the heater and the crucible and the temperature gradient in the melt. FIG. 4 is a graph showing the relationship between solidification rate and G/R when GaAs is doped with In at a high concentration. Five examples of I to ■ are shown. Figure 5 shows the G/R change along jmop in Figure 4.
When performing GR control, R control, and G control, G and R are shown as functions of solidification rate g. FIG. 6 is a schematic cross-sectional view of a single crystal pulling apparatus using the liquid capsule method. FIG. 7 shows the control of G/H along jlqrp in FIG.
Graph showing changes in %G%H for control of G/R along lsp. FIG. 8 is an explanatory diagram of the vicinity of the solid-liquid interface for deriving a formula showing conditions for cell growth occurrence. FIG. 9 is a diagram for explaining fluctuations in impurity concentration. 1... Crucible 2... Susceptor 3... Lower shaft 4... Raw material melt 5... Liquid sealant 6... ... Inert gas atmosphere 7 ... Upper shaft 8 ... Seed crystal 9 ... Single crystal 10 ... Pressure-resistant container 11.12 ... ...Heater inventor Takashi Matsuura Patent applicant Sumitomo Electric Industries, Ltd. Assimilation rate g Fig. 2 (''(:7cm) Fig. 3r-1 Fig. 5 G Fig. 4 G/R

Claims (2)

[Claims] (1) Contains a high concentration of impurities with a segregation coefficient k of 1 or less
A liquid capsule is pulled by covering the raw material melt of III-V compound semiconductor with a liquid capsule, applying high pressure to the liquid capsule with an inert gas, and pulling a single crystal from the raw material melt using a seed crystal. In the above method, the temperature gradient in the melt is G,
The pulling rate is R, the temperature gradient of the liquidus line is m, the initial concentration of impurities is C_0, the segregation coefficient of impurities in the raw material melt is k,
When the solidification rate, which is the ratio of the amount of the solidified portion to the initial raw material melt amount, is g, both the temperature gradient G (g) and the pulling rate R (g) are expressed as a function of the solidification rate g by the inequality G (g )/R(g)≧{ mc_0(1-k)(1-g
) A method for manufacturing a III-V compound semiconductor single crystal, characterized in that control is performed from g=0 to g_m where 0<g_m<1 so that ^k^-^1}/D is satisfied. (2) The raw material melt is GaAs, the impurity is In, and the In concentration at the crystal head is 5×10^1^9 to 1×10
A method for manufacturing a III-V compound semiconductor single crystal according to claim (1), characterized in that the thickness is ^2^1 cm^-^3.