POLYCRYSTALLINE SILICON AND METHOD FOR SELECTING POLYCRYSTALLINE SILICON

Abstract

An object of the present invention is to provide a method for comparatively simply selecting polycrystalline silicon suitably used for stably producing single crystal silicon in high yield. According to the present invention, polycrystalline silicon having a maximum surface roughness (Peak-to-Valley) value Rpv of 5000 nm or less, an arithmetic average roughness value Ra of 600 nm or less and a root mean square roughness value Rq of 600 nm or less, the surface roughness values being measured by observing with an atomic force microscope (AFM) the surface of a collected plate-shaped sample, is selected as a raw material for producing single crystal silicon.

Claims

1. Polycrystalline silicon having a maximum surface roughness value Rpv (Peak-to-Valley) of 5000 nm or less, an arithmetic average roughness value Ra of 600 nm or less and a root mean square roughness value Rq of 600 nm or less, the surface roughness values being measured by observing with an atomic force microscope (AFM) a surface of a collected plate-shaped sample.

2. The polycrystalline silicon according to claim 1, wherein the value Rpv is 2500 nm or less, the value Ra is 300 nm or less and the value Rq is 300 nm or less.

3. The polycrystalline silicon according to claim 2, wherein the value Rpv is 2000 nm or less, the value Ra is 100 nm or less and the value Rq is 150 nm or less.

4. A method for selecting polycrystalline silicon, wherein a plate-shaped sample is cut out from a polycrystalline silicon mass; a surface of the plate-shaped sample is subjected to a lapping treatment with an abrasive; the surface of the plate-shaped sample resulting from the lapping treatment is subjected to an etching treatment with a mixture of hydrofluoric acid and nitric acid; surface roughness of the surface of the plate-shaped sample resulting from the etching treatment is evaluated through observation with an atomic force microscope (AFM); and when a maximum surface roughness value Rpv is 500 nm or less, an arithmetic average roughness value Ra is 600 nm or less and a root mean square roughness value Rq is 600 nm or less, the polycrystalline silicon mass is evaluated as good.

5. The method for selecting polycrystalline silicon according to claim 4, wherein the value Rpv is 2500 nm or less, the value Ra is 300 nm or less and the value Rq is 300 nm or less.

6. The method for selecting polycrystalline silicon according to claim 5, wherein the value Rpv is 2000 nm or less, the value Ra is 100 nm or less and the value Rq is 150 nm or less.

Description

BRIEF DESCRIPTION OF THE DRAWING

[0018] FIG. 1 is a graph illustrating correlation between a surface roughness value, which is obtained by observing with an atomic force microscope (AFM) a surface of a plate-shaped sample cut out from a polycrystalline silicon mass, and a crystal grain size value evaluated by an EBSD method.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Through analysis and study of polycrystalline silicon used as a raw material for stably producing single crystal silicon, the present inventors have found that the size of a crystal grain contained in polycrystalline silicon is varied in accordance with various conditions employed in separation of the polycrystalline silicon.

[0020] Differently from single crystal silicon, polycrystalline silicon contains crystal grains having random crystal orientations, and in general, the size of each crystal grain varies from roughly about several micrometers to several tens micrometers, and may be as large as several hundred micrometers in some cases.

[0021] As a method for measuring the size of a crystal grain of each crystal orientation in a polycrystal, an EBSD (electron backscatter diffraction image) method is known. In order to measure a crystal grain size by this method, however, it is necessary to introduce an expensive apparatus, which disadvantageously increases the production cost.

[0022] Alternatively, a crystal grain size can be measured with an optical microscope or an electron microscope, but it is necessary to obtain a size distribution by digital processing of an observed surface image in this case, and hence, the thus obtained value does not reflect a true crystal grain size in many cases. This is for the following reason: In binarization performed in image processing, if an image is observed with an optical microscope (a polarizing microscope using a metallograph), influence of lighting and light reflected on a sample surface cannot be ignored. Alternatively, if an electron microscope is used for the observation, it is necessary to change image processing conditions to be employed every time a portion of an image where crystal grains are continued is to be processed.

[0023] While conducting various studies on a simple method for measuring a crystal grain size, the present inventors have found that a surface roughness value obtained by observing with an atomic force microscope (AFM) a surface of a plate-shaped sample cut out from a polycrystalline silicon mass is well correlated with a crystal grain size evaluated by the EBSD method, and thus, the present invention was accomplished.

[0024] It was confirmed that high yield can be obtained in the CZ method when polycrystalline silicon having a maximum surface roughness (Peak-to-Valley) value Rpv of 5000 nm or less, an arithmetic average roughness value Ra of 600 nm or less and a root mean square roughness value Rq of 600 nm or less, the surface roughness values being measured by observing with an atomic force microscope (AFM) a surface of a collected plate-shaped sample, is used as a raw material for producing single crystal silicon.

[0025] Besides, it was confirmed that high yield can be obtained in both the CZ method and the FZ method when polycrystalline silicon having the value Rpv of 2500 nm or less, the value Ra of 300 nm or less and the value Rq of 300 nm or less is used as the raw material for producing signal crystal silicon.

[0026] Furthermore, it was confirmed that the yield obtained in the FZ method can be increased up to substantially 100% when polycrystalline silicon having the value Rpv of 2000 nm or less, the value Ra of 100 nm or less and the value Rq of 150 nm or less is used as the raw material for producing single crystal silicon.

[0027] Accordingly, in a method for selecting polycrystalline silicon according to the present invention, a plate-shaped sample is cut out from a polycrystalline silicon mass; a surface of the plate-shaped sample is subjected to a lapping treatment with an abrasive; the surface of the plate-shaped sample resulting from the lapping treatment is subjected to an etching treatment with a mixture of hydrofluoric acid and nitric acid; surface roughness of the surface of the plate-shaped sample resulting from the etching treatment is evaluated through observation with an atomic force microscope (AFM); and when a maximum surface roughness value Rpv is 5000 nm or less, an arithmetic average roughness value Ra is 600 nm or less and a root mean square roughness value Rq is 600 nm or less, the polycrystalline silicon mass is evaluated as good.

[0028] As described above, if polycrystalline silicon having the value Rpv of 2500 nm or less, the value Ra of 300 nm or less and the value Rq of 300 nm or less is selected as a raw material, high yield can be obtained in both the CZ method and the FZ method.

[0029] Furthermore, if polycrystalline silicon having the value Rpv of 2000 nm or less, the value Ra of 100 nm or less and the value Rq of 150 nm or less is selected as a raw material, the yield obtained in the FZ method can be increased up to substantially 100%.

EXAMPLES

[0030] Now, examples of the application of the present invention to a polycrystalline silicon rod synthesized by the Siemens process will be described. A core sample with a diameter of 19 mm (having a length of 130 mm) was collected from each polycrystalline silicon rod, produced by the Siemens process, in a direction vertical to the lengthwise direction (vertical direction). Besides, three core samples each with the same diameter (having a length of 130 mm) were similarly collected, in a direction parallel to the lengthwise direction, respectively from a region close to the core, a region corresponding to a half of the radius (R/2) of the polycrystalline silicon rod, and a region close to the outer surface thereof. Incidentally, four polycrystalline silicon rods A, B, C and D were prepared for the examples, and it is noted that these rods were obtained by separating polycrystalline silicon under different conditions.

[0031] From each of these core samples, plate-shaped samples each having a thickness of about 2 mm were cut out at equal intervals. These plate-shaped samples are regarded to respectively represent a distribution in the crystal growth direction and a distribution in the lengthwise direction.

[0032] One surface of each of these plate-shaped samples was polished with a #600 abrasive to remove a thickness of about 50 to 60 μm, followed by etching with fluonitric acid. A thickness removed by this etching was 20 to 30 μm. Thereafter, the surface roughness was measured with an AFM to evaluate a crystal grain of the sample. An apparatus used for the measurement with the AFM was Park NX200 manufactured by Park Systems Japan. As a cantilever, OMCL-AC160TS-R3 manufactured by Olympus Corporation, including a probe with a tip radius of 7 nm and made of silicon single crystal (n-doped and having resistivity of 0.1 to 0.4 Ω-cm), was used. Besides, the surface roughness was measured in the whole region of 90 μm×90 μm on the surface of the sample.

[0033] The measurement results are listed in Table 1. It is noted that values Rpv, Ra and Rq shown in the table as degrees of the surface roughness respectively correspond to the maximum surface roughness value Rpv (Peak-to-Valley), the arithmetic average roughness value Ra and the root mean square roughness value Rq.

TABLE-US-00001 TABLE 1 Surface parallel Surface vertical to lengthwise to lengthwise For direction direction solar For semi- Number Sample Rpv Ra Rq Rpv Ra Rq cells conductors of rods evaluated (nm) (nm) (nm) (nm) (nm) (nm) CZ CZ FZ evaluated Comparative A 9,583 1,235 1,458 9,254 1,120 1,298 ◯ X X 68 Example Example 1 B 4,784 498 554 4,687 408 591 ◯ ◯ X 41 Example 2 C 2,258 264 298 2,221 219 279 ◯ ◯ ◯ 87 Example 3 D 1,254 78 112 1,157 99 128 ◯ ◯ ⊚ 39

[0034] These polysilicon rods A to D were used as raw materials to grow single crystal silicon by the CZ method and the FZ method to examine the yields thus obtained.

[0035] In using the polycrystalline silicon rod A, there arose no problem in the pulling up performed in a CZ method for obtaining single crystal silicon for solar cells (hereinafter simply referred to as the CZ method for solar cells), but a crystal line disappeared in the middle of a CZ method for obtaining single crystal silicon for semiconductors (hereinafter simply referred to as the CZ method for semiconductors). This is probably because the polycrystalline silicon rod A had a large crystal grain.

[0036] In using the polycrystalline silicon rod B, there arose no problem in the pulling up performed in the CZ method for semiconductors, but a crystal line disappeared in the middle of the formation of single crystal in an FZ method for obtaining single crystal silicon for semiconductors (hereinafter simply referred to as the FZ method for semiconductors). This is probably because the polycrystalline silicon rod B had a crystal grain too large to be employed in the FZ method.

[0037] In using the polycrystalline silicon rod C, there arose no problem in the formation of single crystal by the CZ method and the FZ method for semiconductors, but a crystal line disappeared in the middle of the formation of single crystal by the FZ method, and the disappeared length was not 100% but 70% of the entire length.

[0038] In using the polycrystalline silicon rod D, a crystal line did not disappear in the middle of the formation of single crystal by the FZ method for semiconductors.

[0039] It is understood from these results that the size of a crystal grain of polycrystalline silicon used as a raw material is significant to stably produce single crystal silicon in high yield, that the crystal grain needs to have an optimal size in accordance with either of the production methods to be employed, and that the determination (selection) can be made by a comparatively simple method of surface roughness evaluation by a method using an AFM.

[0040] As a result of these various examinations, the present inventors have reached the following conclusions:

[0041] High yield can be obtained in the CZ method when polycrystalline silicon having a maximum surface roughness (Peak-to-Valley) value Rpv of 5000 nm or less, an arithmetic average roughness value Ra of 600 nm or less and a root mean square roughness value Rq of 600 nm or less, the surface roughness values being measured by observing with an atomic force microscope (AFM) a surface of a collected plate-shaped sample, is used as a raw material for producing single crystal silicon.

[0042] Besides, high yield can be obtained in both the CZ method and the FZ method when polycrystalline silicon having the value Rpv of 2500 nm or less, the value Ra of 300 nm or less and the value Rq of 300 nm or less is used as the raw material for producing signal crystal silicon.

[0043] Furthermore, the yield attained in the FZ method is increased up to substantially 100% when polycrystalline silicon having the value Rpv of 2000 nm or less, the value Ra of 100 nm or less and the value Rq of 150 nm or less is used as the raw material for producing single crystal silicon.

[0044] Accordingly, the following method is effective: A plate-shaped sample is cut out from a polycrystalline silicon mass; a surface of the plate-shaped sample is subjected to a lapping treatment with an abrasive; the surface of the plate-shaped sample resulting from the lapping treatment is subjected to an etching treatment with a mixture of hydrofluoric acid and nitric acid; surface roughness of the surface of the plate-shaped sample resulting from the etching treatment is evaluated through observation with an atomic force microscope (AFM); and when a maximum surface roughness value Rpv is 500 nm or less, an arithmetic average roughness value Ra is 600 nm or less and a root mean square roughness value Rq is 600 nm or less, the polycrystalline silicon mass is evaluated as good.

[0045] For example, if polycrystalline silicon having the value Rpv of 2500 nm or less, the value Ra of 300 nm or less and the value Rp of 300 nm or less is selected as a raw material, high yield can be attained in both the CZ method and the FZ method.

[0046] Besides, if polycrystalline silicon having the value Rpv of 2000 nm or less, the value Ra of 100 nm or less and the value Rp of 150 nm or less is selected as a raw material, the yield obtained in the FZ method can be increased up to substantially 100%.

[0047] In this manner, the present invention provides a method for comparatively simply selecting polycrystalline silicon suitably used for stably producing single crystal silicon in high yield.