JPH0469422B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for accurately predicting heat treatment conditions for obtaining a desired defect density when introducing crystal defects into a silicon single crystal wafer in a semiconductor device manufacturing process.

Currently, in the manufacturing process of semiconductor devices such as ICs and LSIs, crystal defects such as stacking faults and dislocations are generated inside the wafer as active sites that agglomerate contaminants that are inevitably generated during the process. In order to improve the electrical properties of the layer, crystal defects at a predetermined density are introduced into the wafer in advance. The introduction of crystal defects is performed by heat treatment.

This heat treatment is basically performed at a specific temperature and for a specific time in the temperature range of 600 to 1050 degrees Celsius, or by heating the wafer in the above temperature range by increasing the temperature from a low temperature at a specific temperature increase rate. In addition to introducing the desired density of defect nuclei within the
This is generally carried out by heat treatment at a temperature range of ~1200°C for a specific period of time to generate defects from the defect nuclei.

However, crystal defects generated by such heat treatment are variously affected by not only the above heat treatment conditions but also the thermal history that the wafer has undergone, initial solid solution oxygen concentration, carbon concentration, etc. Therefore, it is difficult to set heat treatment conditions to obtain a predetermined defect density, resulting in poor reproducibility and disadvantages in that efficient heat treatment conditions are not selected for the target wafer.

The present invention has been made in view of the above circumstances, and provides a method for controlling the introduction of crystal defects into silicon wafers that can accurately predict heat treatment conditions that can reliably generate a desired defect density in the target wafer. The purpose is to

The present invention will be explained in detail below.

The crystal defect introduction control method of the present invention is based on thermal history,
A first stage heat treatment is performed on a silicon single crystal wafer with a clear history such as initial solid solution oxygen concentration, etc. under heat treatment conditions of various temperatures and times to introduce defect nuclei, which is then further subjected to a second stage heat treatment. After growing the defective nuclei until they can be measured, the density of defective nuclei introduced by the first-stage heat treatment is determined, and this is determined for each history by the first-stage heat treatment. A standard graph is created by drawing it as an isopycnic curve on the temperature-time graph of (heat treatment conditions).

First, a silicon single crystal wafer with a clear history is prepared, but normally the initial solid solution oxygen concentration (hereinafter referred to as
It is abbreviated as O ₂ concentration. ) can be selected from several items for which the following information is known. This wafer is cut into test pieces of an appropriate size, and the test pieces are subjected to a first heat treatment. The temperature at this time is in the range of 650 to 1050°C, and if it is less than 650°C or exceeds 1050°C, it will take a long time to introduce defect nuclei or there will be a disadvantage that defect nuclei will not be generated. In addition, the time is 2 minutes to 300 hours, and if it is less than 2 minutes, it is difficult to control the wafer heat treatment;
If the time is exceeded, it becomes too long and becomes impractical. It is preferable to heat treat as many wafers as possible under as many different conditions as possible within this range of temperature and time conditions, as this will increase the accuracy of the standard graph obtained.

Next, a second stage heat treatment is performed to grow the defective nuclei thus introduced into the wafer specimen until the defective nuclei are actually formed. The heat treatment conditions are 1000 to 1200°C for 10 minutes to 10 hours.
At temperatures below 1000°C, it takes a long time for defects to form from defective nuclei, and at temperatures above 1200°C, defective nuclei re-dissolve into solid solution.
Disappearance of the generated defect occurs. Further, if the time is less than 10 minutes, it is not easy to detect or measure defects, and if the time exceeds 10 hours, the number of defects generated will become saturated or decrease, which is disadvantageous.

The density of the defects in the wafer thus generated can be determined by measurement methods such as counting defects from etching images obtained by selective etching, and counting defects from transmitted bright-field and dark-field images using a transmission electron microscope. It will be done. Note that it is also possible to determine the type of defect (stacking fault, dislocation, precipitate) from the shape of the image that appears on the surface using the selective etching method.

The defect density for each wafer specimen determined in this way is drawn as an isopycnic curve on the temperature-time graph of the first stage heat treatment, and a standard graph corresponding to each O ₂ concentration is obtained. It will be done.

In order to obtain this standard graph, as mentioned above, it is not possible to know the state of defect occurrence unless heat treatment is performed using a large number of wafer specimens under different conditions over a wide temperature range and a wide time range. In other words, although O ₂ concentration plays a dominant role in defect generation, there are many other factors that are thought to be influenced by other crystal pulling conditions and carbon concentration, and it is not possible to determine whether O ₂ concentration alone This is because it is difficult to accurately predict the occurrence of defects.

Next, a general procedure for predicting heat treatment conditions for forming a desired defect density in a wafer to be processed will be described using the standard graph obtained in this manner.

(b) First, select the graph corresponding to the O ₂ concentration of the wafer to be processed. Qualitatively, the higher the O ₂ concentration or the higher the processing temperature, the shorter the time for defects to occur. Although it is difficult to quantitatively predict the occurrence of defects in any wafer to be processed, recent improvements in crystal growth technology have led to stabilization of the defect occurrence situation. Therefore, if we know the defect occurrence conditions for silicon single crystals pulled under specific growth conditions, we can accurately estimate the defect occurrence for single crystals pulled under similar conditions by extrapolation.
By simply performing a simple confirmation heat treatment, it is possible to accurately and efficiently detect defects in the wafer being processed.

(b) Next, one of the wafers to be processed is heat-treated to check whether defects having the density shown in the selected standard graph actually occur. The heat treatment conditions at this time can be appropriately selected by looking at the standard graph.

(c) Examine the difference between the defect occurrence density of the wafer subjected to the above confirmation heat treatment and the defect density predicted from the selected standard graph. If this difference is small, the isodensity curve of defects in the selected standard graph may be corrected by shifting it to the longer time side or the shorter time side so that it matches the value obtained in the confirmation heat treatment. Also,
If this difference is large enough that such correction cannot be performed, perform heat treatment to determine whether the wafer is close to other standard graphs prepared in advance, and take measures to prevent defects in the wafer being processed. Select the standard graph to use.

(d) Compare the changes in the occurrence of defects in the wafer to be processed due to heat treatment conditions with the isodensity curve of the standard graph above to determine the most efficient and reliable heat treatment to obtain the desired defect types and defect density. Predict and set conditions. The actual processing of the wafer to be processed may be carried out by performing constant temperature heat treatment to rapidly cool it, or may be carried out by adjusting the cooling conditions from a high temperature.

When the wafer to be processed is heat treated using the heat treatment conditions obtained (predicted) in this way, the desired defect density can be introduced with high precision, and the variation in the generated defect density can be significantly reduced compared to conventional methods. . In addition, it is possible to know efficient (short-time) processing conditions to obtain the desired defect density.
Cost reduction is also possible.

Hereinafter, a specific explanation will be given by showing examples.

Example 1 To create a standard graph, the O ₂ concentration raised by the CZ method was 1.95×10 ¹⁸ atoms/cm ³ and 1.29×
^Three types of silicon single crystal wafers (diameter ^{104cm , thickness} 500μ
m) A total of 120 sample pieces with the same O ₂ concentration were prepared by dividing each of the 15 pieces into 8 pieces. Total number of specimen pieces is 360
It becomes one piece. These sample pieces were placed in a diffusion heat treatment furnace.
At various temperatures ranging from 650 to 1050℃ under nitrogen atmosphere,
Defect nuclei were introduced by changing the heat treatment time. (First stage heat treatment) In order to measure the density of defect nuclei, three types of heat treated samples were placed in a diffusion heat treatment furnace and heated under a steam atmosphere.
After heat treatment at 1100℃ for 1 hour, quenching
Defect density was actually measured by etching (step heat treatment) and transmission electron microscopy. Note that for the samples ○A and ○B, the density of not only stacking faults but also dislocations and precipitates was measured. The obtained defect density values were plotted on the temperature-time graph for the previous heat treatment, and the isopycnal points were connected to draw isopycnal curves to obtain FIGS. 1 to 5. Figures 1 and 2 are standard graphs for a wafer with an O ₂ concentration of 1.95 x 10 ¹⁸ atoms/cm ³ , where Figure 1 shows the stacking fault density and Figure 2 shows the dislocation and precipitate densities. The density curve is shown. Moreover, the white circles in FIG. 1 indicate the heat treatment conditions, and the broken line curve indicates the time required to reach the saturated density. Figures 3 and 4 are standard graphs for a wafer with an O ₂ concentration of 1.29 x 10 ¹⁸ atoms/cm ³ , and Figure 3 shows the stacking fault density.
FIG. 4 shows isopycnal curves of dislocation and precipitate densities. Moreover, the white circles in FIG. 3 indicate the heat treatment conditions, and the broken line curve indicates the time required to reach the saturation density. Figure 5 shows isopycnal curves of stacking fault density for a wafer with an O ₂ concentration of 1.65 x 10 ¹⁸ atoms/cm ³ , white circles indicate the heat treatment conditions performed, and the curve represented by a broken line reaches the saturation density. Show time. Figure 5 also shows cooling curves A, B, and C when heat treatment is performed by adjusting the cooling conditions from high temperature. , 1280 for C
This is a cooling curve from ℃.

Next, a method for introducing a desired defect density into a wafer to be processed will be explained using the standard graphs shown in FIGS. 1 to 5.

Consider a case where stacking faults of 10 ⁹ atoms/cm ³ are introduced into a wafer with an O ₂ concentration of 1.75×10 ¹⁸ atoms/cm ³ in the shortest possible time. First, a standard graph for determining the heat treatment temperature was selected. The thermal history of the wafer to be processed during crystal pulling is well controlled.
The detection sensitivity of the infrared spectrometer for carbon concentration (approximately 1.0×
10 ¹⁶ atoms/cm ³ ) or less, the tendency of stacking fault occurrence in the wafer can be roughly estimated by interpolation from the graphs in FIGS. 1 and 3. Specifically, for example, 1.95×10 ¹⁸ atoms/cm ³
Estimate the position of the isopycnic curve from the O ₂ concentration graph by moving parallel to the time axis. Then, a confirmatory heat treatment is performed to confirm whether this estimation is correct. FIG. 6 shows isodensity curves of the wafer to be processed obtained through this confirmation heat treatment. The standard graph in Figure 6 was obtained using 10 samples, but it is possible to reduce the number to about 4 samples. This is because the tendency of the isopycnic curve is predicted as described above. Compared to the approximately 60 samples needed to create the standard graphs shown in Figures 1 to 4, reliable predictions can be made with a much smaller number of samples. From the graph in FIG. 6, heat treatment conditions of 800° C. and 9 hours were predicted to introduce 10 ⁹ stacking faults/cm ³ into the wafer to be processed in the shortest possible time. When 1000 wafers were heat-treated under the above conditions, the results were 0.5×10 ⁹ wafers/cm ³ to 5×10 ⁹
Stacking faults/cm ³ occurred, and the defect density was normally distributed among wafers of the same type.

Example 2 Next, an example will be shown in which a cooling process from a high temperature can be used to introduce defects. Even if the O ₂ concentration is known, if the thermal history during crystal pulling is unclear or if the carbon concentration is higher than the aforementioned 1×10 ¹⁶ atoms/cm ² , defects will occur quickly and cannot be quantitatively determined. Since it is difficult to predict, it is necessary to investigate the defect generation behavior over a wide range of temperatures from 650 to 1050°C.

FIG. 5 is an isopycnal curve obtained for a wafer in which such defects occur quickly, and is for a wafer of Example 1 in which the O ₂ concentration of O 2 is 1.65×10 ¹⁸ atoms/cm ³ . For wafers where such defects occur quickly, it is possible to control the cooling rate after heating and make the cooling curve contact the isopycnal curve, without having to go to the trouble of performing heat treatment under the heat treatment conditions predicted from the isopycnal curve. If it is cooled so that the curves intersect, it is possible to introduce the defect density expected from the isopycnal curves that intersect. (This can also be said to be another application example of the standard graph.) The cooling curves A, B, and C show the wafers placed in a furnace at 1100°C for 5 minutes, 15 minutes, and 1280°C, respectively.
This shows the cooling rate when the product is taken out of the furnace after being placed in the furnace for 15 minutes. Cooling curve A
A wafer cooled along its cooling curve A
is in contact with the 10 ⁷ isopycnal curve, so 1×10 ⁷
It is predicted that stacking faults/cm ² will be generated.
Similarly, a wafer cooled along cooling curve B has stacking faults of 5×10 ⁷ /cm ³ because cooling curve B passes near the middle of the 10 ¹⁷ and 10 ⁸ isopycnic curves. Also, since the wafer cooled along the cooling curve C passes near the 10 ⁸ isopycnal curve, it is predicted that 7 to 8×10 ⁷ stacking faults/cm ³ will be generated. The average value of actual defect density is 1.5×10 ⁶ defects/cm ³ for A and 9.4×10 ⁶ defects/cm 3 for B.
For cm ³ and C, it was 8.5×10 ⁷ pieces/cm ³ , and for A and B, it was about 1/6 of the predicted value. However, the standard graph (Figure 5) created from the defect density generated by constant temperature heat treatment can also accurately predict the defect density generated by continuous cooling, which further increases the usefulness of the standard graph.

As explained above, the method of controlling the introduction of crystal defects into silicon wafers according to the present invention is based on thermal history, O ₂
A silicon single crystal wafer with a clear history of concentration, etc. is subjected to a first heat treatment under heat treatment conditions of various temperatures and times to introduce defect nuclei, and then a second heat treatment is performed to introduce defect nuclei. After growing to a point where it can be measured, this is measured to determine the density of defect nuclei introduced by the first heat treatment, and this is plotted as a temperature-time graph during the first heat treatment for each history. A standard graph is created by drawing an isopycnic curve on the top of the wafer, and this standard graph is used to predict the heat treatment conditions necessary to generate a desired defect density for a wafer into which new defects are to be introduced. It is something to do.

Therefore, according to this method, it is possible to easily and accurately predict the heat treatment conditions that can efficiently and reliably generate the desired defect density on the wafer to be processed, and in turn, the desired defect density can be generated on the wafer in a short time. It can be generated accurately. Therefore, defects can be introduced at a lower cost than in conventional methods, and wafers with less variation in defect density can be manufactured.

[Brief explanation of the drawing]

1 through 6 are standard graphs of the method of the present invention.

Claims (1)

[Claims] 1. A silicon single crystal wafer with a clear history of thermal history, initial solid solution oxygen concentration, etc. is subjected to the first heat treatment in an inert gas atmosphere in a temperature range of 650 to 1050°C for 2 minutes to 300 hours. Then, defect nuclei were introduced, and then a second stage heat treatment was performed at a temperature of 1000 to 1200°C for 10 minutes to 10 hours to grow the defect nuclei to an observable level, and the density of the defect nuclei was measured. , the temperature vs. time curves in the first stage heat treatment are illustrated as isopycnal curves of defect nucleus density for each history, and the isopycnal curves obtained in this way are used as standard graphs to determine the desired defect density in the silicon single crystal wafer. A method for controlling the introduction of crystal defects into silicon wafers, which is characterized by predicting heat treatment conditions that will allow the growth of crystal defects.