JPH06105702B2 - Method for manufacturing semiconductor substrate - Google Patents

Description

The present invention relates to a method for manufacturing a semiconductor substrate used in a power semiconductor device or the like, and more particularly to a low resistance diffusion region on one surface side. The present invention relates to a method for manufacturing a formed semiconductor substrate.

(Prior Art) In a semiconductor substrate (hereinafter referred to as a wafer) used in a power semiconductor device or the like, a low resistance diffusion region is formed on one surface side, and diffusion is performed on the other surface side to form an element. After the formation, the metal layer is formed on the diffusion region on the one surface side to take out the electrode. A wafer used for such an application (hereinafter referred to as an OSL (One Side Lapped) wafer) is conventionally manufactured from a starting wafer having a thickness of about 600 μm which is lapped. This O
Since the SL wafer is manufactured through a process such as slicing and lapping from a crystal rod (ingot), a material having a thickness of 1200 μm, which is almost twice that of the SL wafer, is required to manufacture a wafer having a thickness of 600 μm.

6 (a) to 6 (d) are cross-sectional views sequentially showing each step according to the conventional method. First, a wafer 30 with a thickness of 600 μm
Are prepared (FIG. 6 (a)). The wafer 30 is sliced from an ingot and then lapped to form a desired thickness. At this time, the slice allowance is about 350 μm and the lapping allowance is about 200 μm. Therefore, the material loss of the starting wafer is 550 μm. Next, diffusion regions 31 are formed on both sides of the wafer 30 (6th
(B)), then one side is ground by a grinder, and further lapping is carried out to remove one diffusion region (FIG. 6 (c)). Then, the removed surface is mirror-polished to make a mirror surface wafer. Is completed (Fig. 6 (d)). The total material loss due to grinding, lapping and mirror polishing is
It is about 300 μm.

In the case of the conventional method, the starting wafer has a thickness of 600 μm, whereas the OSL wafer after manufacturing has a thickness of 300 μm, and half of the material is discarded. This is inevitable with conventional methods. The cost allocation in the OSL wafer is
In the market, the profit rate is 10% and the indirect cost is 15%, and the remaining 75% is the direct cost. In addition, the raw wafer cost in direct costs will be 56.4% of the total. Therefore, OSL
The cost impact of raw wafers in wafers is very large, and the OS
This is a major barrier to cost reduction for L wafers.

However, in order to obtain a starting wafer of 600 μm, it is not completed unless it is performed through two steps of slicing and lapping, and a material loss of about 550 μm occurs in these steps. This has a diameter of 125 mm, resulting in a material loss of about 16 g or more.

(Problems to be Solved by the Invention) As described above, in the related art, since the final wafer is finished by removing many portions from the raw wafer, there is a drawback that the material is wasted and the cost is high.

The present invention has been made in consideration of the above circumstances, and an object thereof is to provide a method of manufacturing a semiconductor substrate, which can manufacture the semiconductor substrate at low cost.

[Structure of the Invention] (Means for Solving the Problems) A method of manufacturing a semiconductor substrate according to the present invention comprises: The semiconductor substrate of the diffusion regions are bonded to each other so as to face each other to form an ingot, and one end face of the ingot is bonded with a correction substrate for performing position correction between the flat surface of the ingot and the surface of the diamond blade,
In slicing each of the plurality of semiconductor substrates into two pieces in the thickness direction using the diamond blade, the correction substrate is sliced and a part thereof is removed to be parallel to the plane of the remaining portion of the correction substrate. Adjust the position of the sensor to remove the remaining part of the correction substrate,
After adjusting the position of the ingot so that the end surface of the ingot from which the correction substrate has been removed is parallel to the position-adjusted sensor, each semiconductor substrate is pitch-fed and sequentially sliced. I am trying.

(Operation) In the method of the present invention, the substrate having diffusion regions formed on both sides is sliced in the thickness direction and divided, and two substrates are taken, whereby the material loss per substrate is reduced as compared with the prior art.

Embodiment An embodiment method of the present invention will be described below with reference to the drawings. FIGS. 1 (a) to 1 (e) are sectional views sequentially showing each step according to the method of the present invention.

First, a starting wafer 10 having a thickness of 1200 μm, which is twice that of a conventional one, is prepared (FIG. 1 (a)). This wafer 10 is sliced from an ingot state as in the conventional case, and is further lapped to have a desired thickness.
This wafer 10 has, for example, a diameter of 125 mm ± 0.5 mm and a resistivity of
40 to 50 Ωcm, N-type P (phosphorus) as an impurity
Has been introduced, the surface is JIS # 1000 lap finish, and the parallelism is 5 μm or less. Also, the slice allowance is about 350 μm as before, and the lap allowance is 200 μm.
It is about m. Therefore, the material loss of the starting wafer is 550 μm as in the conventional case.

Next, diffusion regions 11 are formed on both surfaces of the starting wafer 10 (first
Figure (b)). This diffusion is performed in a 5-lot organization, for example, with 50 starting wafers as one lot. First, the starting wafer is pretreated with an alkaline cleaning solution to improve the surface cleanliness, and then placed on a quartz boat in units of one lot.
Then, after inserting into a diffusion furnace in an oxygen atmosphere at a temperature of 1200 ° C., vapor obtained by evaporating POcl ₃ with N ₂ gas is mixed with oxygen gas and supplied into the furnace. After maintaining the state for a predetermined time,
Stop the supply of POcl ₃ vapor and take the boat out of the furnace.
Next, the phosphorus glass formed on the surface is removed with an HF solution, thereby completing the so-called deposition, in which regions containing impurities at a high concentration are deposited on both surfaces of the starting wafer 10. After deposition, the surface impurity concentration of some wafers is evaluated to confirm that it is 10 ²⁰ / cm ³ or more. After deposition, the wafers are stacked on a quartz boat in a state of being closely attached to each lot. In this case, the efficiency is increased by stacking 5 lots at the same time. The stacked boats are inserted into a furnace in an oxygen atmosphere at a temperature of 1250 ° C, and slumping is performed for a predetermined time. After slumping, the boat is taken out of the furnace and the phosphorus glass formed on the surface is removed again with the HF solution. After that, the depth of the diffusion region 11 of a part of the wafer is measured, and it is confirmed that it is within 150 μm ± 5 μm, and the representative value for each lot is recorded.

Next, the wafers 10 that have undergone the diffusion process are bonded together as shown in the side view of FIG. 2 to form an ingot 12. At this time, paraffin having a melting point of 70 ° C. is used as an adhesive. At this time, the temperature of each wafer is raised in advance on the heater, paraffin is applied to the surface, and another wafer is stacked thereon to repeat all the wafers to form the ingot 12. . At this time, the orienteering flat portions provided on the respective wafers are superposed on each other with reference to alignment. In addition, dummy wafers 13 for reinforcement are bonded to both end surfaces of the ingot 12. This dummy wafer 13
Has a diameter of 125 mm ± 0.5 mm, a thickness of 600 μm, and the surface is JIS # 1000 lap finish. After these wafers are bonded, a pressure of 100 g / cm ² or more is applied to cool them while removing excess paraffin. Wipe off excess paraffin on the surface with trichlene. The length of the bonded ingot is measured, and the thickness of paraffin as an adhesive is converted per wafer and recorded.

Next, slice the ingot 12 as above,
As shown in FIG. 1C, the wafer 10 is divided into two wafers 18. This slice is performed as follows. First, the ingot is bonded onto the slicing carbon base with an adhesive resin (eg, epoxy resin). The ingot adhered on the base is mounted on a TS-27 type diamond blade type slicing machine manufactured by Mayer & Burge. The outer shape of the diamond blade used at this time is
27-inch blade, blade thickness 340μm, base metal thickness 150μ
m. Further, the cutting allowance when cutting with this blade is 400 μm.

Next, this slicing process will be described in detail with reference to the side views of FIGS. 3 (a) to 3 (d). First, a dummy wafer 15 for position correction is bonded in front of the ingot 12 bonded on the slicing carbon base 14. The dummy wafer 15 has a thickness of 3000 μm, for example. Next, the position detecting sensor 16 is made to face the dummy wafer 15 for position correction installed on the entire surface of the ingot 12 bonded on the slicing carbon base 14 (FIG. 3 (a)). Reference numeral 17 is a diamond blade of the slicing device. next,
Ingot bonded on carbon base 14 for slicing
The dummy wafer 15 located in front of 12 is sliced by the diamond blade 16 (FIG. 3 (b)). Next, the sliced dummy wafer 15 is removed, and the position of the sensor 16 is adjusted so as to be parallel to the plane of the remaining dummy wafer 15 (FIG. 3 (c)). After removing the remaining dummy wafer 15, the position of the slicing carbon base 14 is adjusted so that the plane of the reinforcing dummy wafer 13 on the end face of the ingot 12 is parallel to the sensor 16. This completes the alignment between the flat surface of the wafer and the diamond blade 16. After this, before turning on the automatic feeding mechanism in the slicing device, manually (cutting allowance / 2)
= 400 μm / 2 = 200 μm in advance, the automatic feed mechanism is turned on, and slices are sequentially performed at the feed pitch P. P =
T + ω, where T is the average thickness of each wafer 10, ω
Is the thickness of one paraffin wafer. Where T
Is 1200 μm, and when ω is 2 μm, the feed pitch P in this case is 1202 μm. At this time, the slicing speed was 50 mm / min, the blade rotation speed was 2000 rpm, and city water was used as the cooling water. A state in which such a slicing process is completed is shown in the side view of FIG. 2 for each wafer 10
It is divided into a plurality of wafers 18.

Next, the wafer 18 divided into two pieces is removed from the slicing carbon base, the adhesive resin and paraffin are removed by a chemical solution, and the wafer is peeled off to clean the surface. Then, the sliced surface of each wafer 18 is polished by an automatic grinder (FIG. 1 (d)). The polishing allowance at this time is determined according to the finished thickness of the so-called intrinsic layer in which the impurities left in the final step are not introduced. For example, this polishing allowance is set to 180 μm. As the polishing apparatus at this time, for example, SVG-502 manufactured by Shibayama Machinery can be used, and as the grindstone, # 400 is used for rough grinding, # 1200 is used for finish grinding, and the rough grinding speed is 150 μm / Min, and the finishing grinding speed was 10 μm / min.

Then, the polished surface of each wafer 18 is mirror-polished to be a mirror surface. This step is performed after each wafer 18 is adhered to the polishing plate by using wax. Abrasive is SiO ₂
The material for the mirror polish is 20 μm. The wafer after polishing is peeled from the plate and washed with a chemical to remove wax and the like, whereby the OSL wafer 19 having the diffusion region 11 on one surface is completed (FIG. 1 (e)).

FIG. 5 is a view showing the material loss generated in each step in comparison between the case of the conventional method and the case of the method of the present invention. As described above, the ratio of the raw wafer to the cost of the OSL wafer manufactured by the conventional method is high, and the cost reduction of the raw wafer directly leads to the cost reduction of the OSL wafer. In the raw wafer manufacturing process, 350 μm when slicing to the starting wafer and 200 μm when lapping, totaling 550 μm
Material loss of m occurs, which is independent of wafer thickness. Therefore, this loss has the same value as in the conventional method and the method of the present invention. Further, according to the method of the present invention, since one wafer is sliced into two after diffusion, a slice loss of 400 μm occurs, but the conventional method does not cause this loss. Furthermore, the conventional method requires 300 μm as the loss after grinding and mirror polishing.
With the method of the present invention, this loss can be reduced by the amount that it is not necessary to polish the diffusion region, and the value is 200 μm. In the conventional method, a material loss of 850 μm occurs as a total of the above losses for each wafer. However, in the method of the present invention, the loss shown in FIG. 5 occurs for every two wafers, so the material loss per wafer is 85% of that of the conventional method.
It can be reduced to 575 μm by 275 μm with respect to 0 μm.

As a result, waste of material can be reduced, and the OSL wafer can be manufactured at low cost.

[Effects of the Invention] As described above, according to the present invention, it is possible to provide a method for manufacturing a semiconductor substrate, which can manufacture the semiconductor substrate at low cost.

[Brief description of drawings]

FIG. 1 is a sectional view sequentially showing each step by the method of the present invention, FIG. 2 is a side view of an intermediate step of the method, and FIG. 3 is a sectional view showing the intermediate step of the method in detail.
FIG. 5 is a side view in the final step of the above method, FIG. 5 is a view showing material loss generated in each step, and FIG. 6 is a sectional view sequentially showing each step in the conventional method. 10 ... Starting wafer, 11 ... Diffusion region, 12 ... Ingot, 13 ...
Reinforcing dummy wafer, 14 ... Slice carbon base, 15 ... Position correcting dummy wafer, 16 ... Position detecting sensor, 17 ... Diamond blade, 18, 19 ... Wafer.

Claims (1)

[Claims] 1. A diffusing region is formed by diffusing impurities on both sides of a semiconductor substrate, and a plurality of semiconductor substrates having the diffusing region are bonded to each other so that the diffusing regions face each other to form an ingot. On one end face of the ingot, a correction substrate for correcting the position of the flat surface of the ingot and the surface of the diamond blade is adhered, and each of the plurality of semiconductor substrates in the thickness direction is formed by using the diamond blade. When slicing into, the correction substrate is sliced to remove a part, the position of the sensor is adjusted so as to be parallel to the plane of the remaining portion of the correction substrate, and the remaining portion of the correction substrate is removed. Then, after adjusting the position of the ingot so that the end surface of the ingot from which the correction substrate has been removed is parallel to the sensor whose position has been adjusted, The method of manufacturing a semiconductor substrate, characterized in that so as to sequentially sliced to feed pitch.