JPH0379319B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical field to which the invention pertains] The present invention relates to a method for controlling the diameter of a - group compound semiconductor single crystal using the Czyochoralski method.

[Prior art and its problems] Control of crystal diameter is an important issue in single crystal growth by the Czyochoralski method. Diameter detection methods include optical methods and gravimetric methods, and diameter control is performed by feeding back the diameter deviation detected by these methods to the heating power and pulling speed of the crucible. Generally, optical methods are widely applied to silicon single crystals, and gravimetric methods are widely applied to oxide single crystals. For - group compound semiconductor single crystals, many attempts have been made to control the diameter using the gravimetric method because a high-pressure vessel is required in the furnace body and the viewing window is clouded due to volatile components. However, in the case of - group compound semiconductor single crystals, the fixed density is lower than the density of the liquid, so when applying the gravimetric method, there is a transient inverse response between the increase/decrease in the crystal diameter and the output of the gravimetric detector. There are cases. To explain this specifically, the weight felt by the weight detector during pulling of a single crystal consists of the solidified crystal weight and the meniscus weight of the melt pulled by the crystal due to surface tension. When the temperature drops slightly in this state, the part of the melt that has spread due to surface tension solidifies, and the tail of the single crystal widens, resulting in a larger diameter. However, the weight detector will feel lighter by the amount of weight that is replaced by the less dense crystal from the denser melt. In this way, in a transient manner, as the crystal diameter increases, the output of the weight detector does not increase, but rather decreases, and an inverse response is observed. An example is W. Bardsley et al., J. Cryst. Growth 40 (1977).
It is described on pages 13 to 20 of .

At this time, if the loop gain is increased in an attempt to improve diameter control accuracy, control becomes unstable, hunting occurs, and the shape of the crystal becomes periodic uneven. If this phenomenon occurs, not only will it not be possible to obtain crystals with a uniform diameter, but the melt temperature will also fluctuate drastically, resulting in twins and polycrystals, and impurities in the crystals becoming non-uniform. There is also the problem that growth stripes occur. Furthermore, in extreme cases, there is a problem that the crystals may break off from the melt or the melt may solidify. Therefore, the conventional automatic diameter control technology using the gravimetric method has the drawback that it is not possible to stably grow high-quality - group single crystals with a uniform diameter with good reproducibility. To solve this problem, a method has been proposed in which a correction signal obtained by differentiating the output of the weight detector once or twice is added to the control loop for control. For example, it is described in Japanese Patent Application Laid-open No. 131683/1983. However, single crystals grown using this method have only been reported to have a small diameter (10 mm to 30 mm φ) and a small capacity (~300 g or less), and only those with a large diameter (50 mm φ or more) and large capacity (500 g or more). No reports have been found in which this method has been applied to crystals. Moreover, an automatic diameter control device based on this method is not currently commercially available and has not yet been put into practical use. Therefore, it cannot be said that this method is sufficient for growing large-diameter, large-capacity - group compound semiconductor single crystals. Therefore,
It has the disadvantage that it cannot be applied to - group single crystal growth such as large-diameter, large-capacity InSb and GaAs, which have recently attracted attention for use in infrared CCDs, ultrahigh-speed ICs, and FETs.

[Object of the invention] This invention was made in view of the above points,
The object of the present invention is to provide a method that eliminates the above-mentioned drawbacks when producing - group compound semiconductor single crystals and can stably produce high-quality single crystals with controlled diameters with good reproducibility.

[Summary of the Invention] The present inventor believes that the above-mentioned reverse response is an apparent one caused by measurement limitations, and therefore, it is necessary to compensate for the control operation based on the apparent reverse response information during the transition period. Ta. Figure 1 shows the inverse response process showing the relationship between melt temperature change and crystal weight for an InSb single crystal, measured as the time differential value of the crystal weight. In the figure, curve b shows the response when the pulling speed was 10 mm/h, and curve c shows the response when the pulling was temporarily stopped. As a result of further research, we found that when we calculated the difference between curves b and c, we obtained the circle mark a, which was the result of measuring the actual crystal diameter and the change in crystal diameter over time.
It was found that the response (curve a) when approximated by a next-lag transfer function and when the time constant was 5 minutes matched well. Furthermore, similar results were obtained by varying the crystal diameter and pulling speed from 20 to 50 mm and from 5 to 25 mm/h, respectively. Based on this, when we added the curve c shown in Figure 1 to the feedback loop as a compensation value added to the apparent measured value, no hunting phenomenon was observed in the control system, and stable diameter control was achieved. It was found to be sufficiently effective. Therefore, in the diameter control method of the present invention, the reverse response component is simulated from the relationship between the change in crucible heating power and the change in crystal weight.
It is characterized in that it has a compensation circuit, and by adding the output of the compensation circuit to the control loop, the reverse response component is canceled and the diameter is controlled. Furthermore, the present invention is characterized in that during single crystal pulling, an inverse response component is simulated based on the relationship between the change in crystal weight and the change in crucible heating power when pulling is temporarily stopped.

[Effects of the Invention] As explained above, according to the diameter control method of the present invention, (1) In - group compound semiconductor single crystal growth,
Extremely precise diameter control can be performed stably with good reproducibility.

(2) Compared to the conventional control method, the temperature fluctuation of the melt was reduced to about 1/10, and the occurrence of polycrystals and twins was hardly observed, and the crystal production yield was reduced to about 1/10.
Improve by 20%.

(3) Compared to crystals grown using conventional diameter control methods, high-quality crystals with less uneven impurity distribution and less uneven striation can be obtained.

(4) Productivity is improved by industrially applying the present invention.

There are other effects.

[Embodiment of the Invention] An embodiment of the present invention will be described in more detail based on the drawings. FIG. 2 is an example of a single crystal manufacturing apparatus equipped with the functions according to the present invention. In the figure, 1
is a container, 2 is a heater, 3 is a crucible, 4 is a melt, 5 is a crystal, 6 is a pulling shaft, 7 is a weight detector,
8 is a length detector, 9 is a differentiator, 10 is a controller,
11 is a tone adjuster, 12 is a heating device, 13 is a reverse response compensation filter, and 14 is a changeover switch. The weight of the crystal 5 being pulled from the melt 4 in the crucible 3 is measured by the weight detector 7, and after comparing it with the reference signal from the length detector 8, it is differentiated by the differentiator 9 to obtain a diameter deviation signal. . This deviation signal is input to the next diameter control controller 10 to obtain a crucible heating power control signal. This control signal is input to the tone adjuster 11 and the heating device 12 is used to control the heating power and perform diameter control. At this time, the proportion of controller 10,
It is necessary to set the integral and differential constants to optimal values.
This is usually done by determining the delay time of crystal growth and determining a constant using the Ziegler-Nichols method or the like. However, as described above, when diameter control is performed using the constant determined in this way, the reverse response component is also detected and input at the same time, making the control unstable and causing hunting. Therefore,
The constant must be set smaller than the optimal value to the extent that hunting does not occur. Therefore, by closing the switch 14 and operating the reverse response compensation filter 13 to cancel the reverse response component, it is possible to stably control the diameter using the optimum control constant. This reverse response compensation filter 13 simulates the reverse response process and predicts the reverse response component in advance from information on changes in heating power, and calculates the reverse response process G(S) from the actual measurement results shown in FIG. It can be approximated by the following transfer function.

G(S)=K ₁ /(1+T ₁ S) ² +K ₂ /1+T ₂ Se ^-LS ……(
1) Here, S is the Laplace operator, T ₁ , T ₂ , and L are constants determined from the thermal time constant of the melt and the delay time of crystal growth, and K ₁ and K ₂ are determined from the crystal diameter and pulling speed. It is a constant. This filter 1
3 can be configured by digital means as follows, for example. The transfer function of the filter 13 is D
(S), and the zero-order rolled circuit is GC(S), then D(S)=G _c G(S)=1-e ^-STS /SG(S)...(2). However, T _s is the sampling period.
Next, when D is Z-transformed and the compensation output Y is obtained, the following equation is obtained.

Y _o = _K 〓 ⁱ⁼¹ a _i U _oi − _K 〓 ⁱ⁼¹ b _i Y _oi ...(3) Here, Yn and Un are the compensation output and heating power at time n, respectively, and ai and bi are (2 ) is a coefficient determined by the Z transformation of the equation. FIG. 3 shows an example of sample output values for the step input of the filter 13 constructed in this manner.

Next, as a specific example, a case in which an InSb single crystal is manufactured using a single crystal manufacturing apparatus having the functions of the present invention will be described in detail. In Figure 2, 500 InSb polycrystalline raw materials are placed in crucible 3 with a diameter of 75 mm.
g and heated to ~650°C. Next, the <211>-axis seed crystal was brought into contact with the melt 4 while being rotated at 10 vpm, and then crystal pulling was started at a speed of 10 mm/h to grow a shoulder portion to a predetermined diameter of 35 mm. Here, the reverse response process is actually measured, T ₁ = 2.5 minutes,
T ₂ = 6 minutes, L = 5 minutes, K ₁ = 0.6 g/min, K ₂ = -
Since it was found to be 0.45 g/min, the coefficient of filter 13 is a ₁ = -7.39×10 ^-3 g/min, assuming T _s = 1 minute.
°C, a ₂ = −0.59×10 ^-3 , a ₃ = −4.79×10 ^-3 , a ₇ =
−13.82×10 ^-3 , a ₈ = 18.52×10 ^-3 , a ₉ = −6.21×
10 ^-3 , b ₁ = −4.37×10 ⁻³ , b ₂ = 3.17×10 ⁻³ , b ₃ = −
It was set in advance to 0.76×10 ^-3 (all other coefficients are 0). Next, when the predetermined diameter was reached, the controller 10 was operated to start diameter control. Here, when the comparison constant G was gradually increased from 0,
The control system started to become unstable at around 0.1, and at around 0.15
Hunting occurred at a cycle of ~20 minutes. Then, when the switch 14 was closed and the filter 13 was operated (A in FIG. 4), it was observed that the hunting was converging. Although the constant was further increased from 0.25 to 0.3, hunting did not occur, the diameter could be controlled stably, and 450 g of InSb single crystal could be pulled. FIG. 4 shows the effect of the diameter control method of the present invention. Using the above constants determined as above, InSb single crystal growth was performed 10 times in a row, and single crystals with a diameter of 35±1 mm and a weight of 450 g or more were obtained in all cases. All of these single crystals were extremely homogeneous single crystals without the occurrence of polycrystals or twins. On the other hand, when the filter 13 was not operated (FIG. 4B), due to hunting, the diameter changed periodically by as much as ±3 mm, and polycrystals were generated, so that good results could not be obtained.

It should be noted that the above equations (1) to (3) and each constant are not limited to the present embodiment, and other control equations or experimental equations may be used. The method of the present invention can also be applied to other - group compound semiconductor single crystals, e.g.
It can be similarly applied to GaAs, GaSb, GaP, etc., and the effects obtained are also great.

[Brief explanation of drawings]

FIG. 1 is a diagram for explaining the reverse response process showing the relationship between melting point temperature and coagulation weight, FIG. 2 is a diagram for explaining an embodiment of the present invention, and FIG. 3 is a diagram for explaining the reverse response compensation filter. FIG. 4 is a diagram for explaining the effect of the method of the present invention. 1... Container, 2... Heater, 3... Crucible, 4... Melting point, 5... Crystal, 6... Pulling shaft,
7... Weight detector, 8... Length detector, 9... Differentiator, 10... Controller, 11... Temperature controller, 12
... Heating device, 13 ... Reverse response compensation filter, 1
4... Changeover switch.

Claims (1)

[Claims] 1. When growing a single crystal using the Czyoklassky method, which controls the crystal diameter by detecting the weight of a substantially pulled - group compound semiconductor single crystal,
The amount of change in the weight of the single crystal with respect to the amount of change in crucible heating power that occurs when pulling of a single crystal having the same shape as the single crystal is temporarily stopped is determined in advance as an inverse response component,
The diameter is controlled by adding a signal of the difference obtained by subtracting the reverse response component from the amount of weight change of the single crystal in response to the amount of change in crucible heating power to a diameter control loop as the amount of weight change. A method for controlling the diameter of compound semiconductor single crystals.