EP3947280A1

EP3947280A1 - Method for producing and classifying polycrystalline silicon

Info

Publication number: EP3947280A1
Application number: EP19829497.7A
Authority: EP
Inventors: Thomas SCHRÖCK; Markus Wenzeis
Original assignee: Wacker Chemie AG
Current assignee: Wacker Chemie AG
Priority date: 2019-12-17
Filing date: 2019-12-17
Publication date: 2022-02-09
Also published as: TW202126578A; US20230011307A1; CN113727944A; WO2021121558A1; JP7342147B2; KR102657489B1; KR20220002624A; JP2022537014A; TWI758989B

Abstract

The invention relates to a method for producing and classifying polycrystalline silicon, comprising the following steps: - making a polycrystalline silicon rod by introducing a reaction gas, which contains, in addition to hydrogen, silane and/or at least one halosilane, into a reaction chamber of a vapour deposition reactor, wherein the reaction chamber contains at least one heated filament rod, on which silicon is deposited forming the polycrystalline silicon rod, - removing the silicon rod from the reactor, - where necessary comminuting the silicon rod to obtain silicon fragments, - creating at least one two-dimensional and/or three-dimensional image of at least one sub-region of the silicon rod or of at least one silicon fragment and selecting at least one analysis region for each created image, - generating at least two surface structure characteristics for each analysis region by means of image processing methods, wherein each surface structure characteristic is generated using another image processing method, - combining the surface structure characteristics to form a morphology characteristic. The silicon rods or the silicon fragments are classified in accordance with the morphology characteristic and provided for various further processing steps.

Description

Process for the production and classification of polycrystalline silicon

The invention relates to a method for the production and classification of polycrystalline silicon, the polycrystalline silicon being classified as a function of a morphology index determined on the basis of two- and / or three-dimensional images and being fed to various processing steps.

Polycrystalline silicon (polysilicon) is used as the starting material in the production of monocrystalline (monocrystalline) silicon, for example by means of crucible pulling (Czochralski or CZ process) or by means of zone melting (float zone process). Monocrystalline silicon is used in the semiconductor industry to manufacture electronic components (chips).

Furthermore, polysilicon is required for the production of multicrystalline silicon, for example by means of the ingot casting process. The multicrystalline silicon obtained in the form of a block can be used to manufacture solar cells.

Polysilicon can be obtained using the Siemens process - a chemical vapor deposition process. In this case, support bodies are heated in a bell-shaped reactor (Siemens reactor) by direct passage of current and a reaction gas containing a silicon-containing component and hydrogen is introduced. The silicon-containing component is usually monosilane (SiH4) or a halosilane with the general composition SiH _n X4- _n (n = 0, 1, 2, 3; X = CI,

Br, I). Usually it is a chlorosilane or chlorosilane mixture, usually trichlorosilane (SiHCl3, TCS). Mainly S1H ₄ or TCS are used in a mixture with hydrogen. The structure of a typical Siemens reactor is described, for example, in EP 2077 252 A2 or EP 2 444 373 A1. The bottom of the reactor (bottom plate) is generally provided with electrodes that hold the carrier bodies.

The carrier bodies are usually filament rods (thin rods) made of silicon. Usually, two filament rods are connected with a bridge (made of silicon) to form a pair of rods that form a circuit across the electrodes. The surface temperature of the filament rods is usually more than 1000 ° C during the deposition. At these temperatures, the silicon-containing component of the reaction gas decomposes and elemental silicon separates out of the gas phase as polysilicon. This increases the diameter of the filament rods and the bridge.

After the rods have reached a predetermined diameter, the deposition is stopped and the polysilicon rods obtained are removed. After removing the bridge, approximately cylindrical silicon rods are obtained.

The morphology of the polysilicon or the polysilicon rods and the fragments produced from them generally have a strong influence on the performance during further processing. In principle, the morphology of a polysilicon rod is determined by the parameters of the deposition process (eg rod temperature, silane and / or chlorosilane concentration, specific flow rate). Depending on the parameters, pronounced interfaces up to holes and trenches can form. As a rule, these are not distributed homogeneously inside the rod. Rather, by varying the parameters, polysilicon rods with different (mostly concentric) morphology areas can be formed, as is described, for example, in EP 2662 335 A1. The dependence the morphology of the rod temperature is expressed, for example, in US 2012/0322175 A1.

The morphology of polysilicon can vary from compact and smooth to very porous and fissured. Compact polysilicon is essentially free of cracks, pores, joints and crevices. The bulk density of such polysilicon can be equated with the true density of silicon or at least corresponds to this to a good approximation. The true density of silicon is 2.329 g / cm ³ .

A porous and fissured morphology, i.e. a very pronounced morphology, has a particularly negative impact on the crystallization behavior of polysilicon. This is particularly evident in the CZ process for the production of single-crystal silicon. Here the use of fissured and porous polysilicon leads to economically unacceptable yields. In general, particularly compact polysilicon leads to significantly higher yields in the CZ process. However, compact polysilicon is usually more expensive to produce because a slower deposition process is required. Furthermore, not all applications require the use of particularly compact polysilicon. For example, the requirements placed on the morphology in the production of multicrystalline silicon using the ingot casting process are significantly lower.

Accordingly, in addition to purity and chunk size, polysilicon is also differentiated and classified according to its morphology. Since various parameters such as porosity (sum of closed and open porosity), specific surface, roughness, gloss and color can be summarized under the term morphology, one reproducible determination of the morphology is a great challenge. An optical assessment of the polysilicon rods or fragments after the deposition by a person, ie the formation of a personal quality impression, as proposed in WO 2014/173596 A1, has, in addition to the disadvantages, one lack of reproducibility and accuracy also has the disadvantage of low throughput. Usually, only entire polysilicon rods or at least large sections of rods can be classified on the basis of personal quality impressions. Furthermore, only a random check can be carried out during normal operation.

The invention was based on the object of providing a method for determining the morphology of polysilicon after deposition, in particular to make the subsequent processing of the polysilicon more efficient.

This object is achieved by a method for the production and classification of polysilicon, which comprises the following steps:

- Generating at least one polycrystalline silicon rod by introducing a reaction gas that contains, in addition to hydrogen, silane and / or at least one halosilane, into a reaction space of a gas phase deposition reactor, the reaction space containing at least one heated filament rod on which silicon is deposited to form the polycrystalline silicon rod,

- removing the silicon rod from the reactor,

- if necessary, crushing the silicon rod to obtain silicon fragments,

- Generating at least one two-dimensional (2D) and / or three-dimensional (3D) image of at least a partial area of the silicon rod or of at least one silicon fragment and selecting at least one analysis area per generated image,

- Generating at least two surface structure indicators per analysis area by means of image processing methods, with each surface structure indicator being generated with a different image processing method,

- Combining the surface structure indicators to form a morphology indicator.

The silicon rods or the silicon fragments are then classified as a function of the morphology index and fed to various processing steps.

As described above, depending on the deposition parameters, polysilicon with different morphology can form, with areas of different morphology which are separated from one another by interfaces can also occur within the same poly silicon rod, in particular in the radial direction of its cross-sectional area.

Morphology is to be understood here in particular as the fissured nature of the polysilicon, which results from the frequency and arrangement of holes, pores and trenches. The morphology can also be understood as the porosity of the polysilicon.

During the deposition, the formation of pores and trenches is evident on the basis of a surface structure that is reminiscent of popcorn. Viewed in profile, a so-called popcorn surface is a collection of elevations (mountains) and ditches (valleys).

In general, the number of filament rods or silicon rods arranged in the gas phase deposition reactor for the Execution of the method according to the invention is insignificant. The gas phase deposition reactor is preferably a Siemens reactor as described in the introduction and for example in EP 2662 335 A1. The filament rod is preferably one of two thin rods made of silicon, which are connected to a pair of rods via a bridge made of silicon, the two free ends of the rod pair being connected to electrodes on the reactor bottom. In general, more than two filament täbe _S (a pair of rods) are arranged in the reaction space far. Typical examples of the number of filament täben _S (silicon rods) in a reactor are 24 (12 pairs), 36 (18 pairs), 48 (24 pairs), 54

(27 pairs), 72 (36 pairs) or 96 (48 pairs). The silicon rods can be described to a good approximation as cylindrical at any point in time of the deposition. This is particularly independent of whether the filament rods are cylindrical or, for example, square.

After the deposition has ended, the at least one silicon rod is removed from the reactor, usually after a cooling period. In the case of silicon rod pairs, the bridge is usually removed after removal. The region of the silicon rod with which it was connected to the electrode is also preferably removed. A device as described in EP 2 984 033 B1, for example, can be used for removal.

If comminution is planned, this can be done like a dog, for example with a hammer, or with a compressed air chisel. The comminution can be followed by sieving, classifying, pneumatic sorting and / or free-fall sorting. To generate the 2D and / or 3D images, the silicon fragments are preferably separated so that they are arranged next to one another. The separation is particularly preferably carried out in such a way that the silicon fragments do not touch one another and ideally have a distance from one another which corresponds, for example, to the average size of the silicon fragments.

The 2D and / or 3D images are preferably generated on a silicon rod as a whole, with the bridge and the area that was connected to the electrode generally being removed. However, the silicon rod can be broken into several cylindrical sections. The partial area from which the images are generated can also be a fracture surface (approximately cross-sectional area) in addition to the outer surface of the silicon rod. In particular, images are generated of both the surface and the fracture surface.

It is particularly preferred to generate 2D and / or 3D images of all silicon rods contained in the reaction space.

One or more cameras with appropriate lighting can be used to record 2D images. The camera can be, for example, a monochrome or color camera. It is preferably a digital camera.

Both area cameras (sensor as an array of pixels) and line sensors (with a corresponding advance of the object or camera to be recorded) can be used.

In general, the camera's sensors can cover different spectral ranges of light. Usually cameras are used for the visual light area. Cameras for the ultra- violet (UV) and / or infrared (IR) regions can also be used. X-rays of the silicon rods or fragments can also be produced. For cameras in the visible light range, it is possible to record the pure gray values, but also the color information (RGB cameras). In addition, special lighting with filtering can be used. For example, it can be illuminated with a blue light and the filtering can be set to precisely this light color in the transmission range. In this way, external light influences can be avoided.

In principle, one or more cameras can be used. If several images are to be related to one another, it must generally be ensured that the object to be recorded is at rest, at least in the event that the images are generated one after the other. If several cameras are used, the images are preferably recorded at the same time. If this is not possible, a movement of the objects between the recordings can usually be corrected using software.

In principle, different sources and different arrangements of these sources can be used for the lighting. Examples of different arrangements are reflected light, dark field, bright field or transmitted light, but also combinations thereof. These methods are described, for example, in Handbuch der Bildverarbeitung 2018, p. 49, ISBN: 978-3-9820109-0-8.

In general, sources of different spectral ranges can be used, for example white light, red or blue light, UV light, IR excitation. The sources preferably have as little change in brightness (drift) over time as possible. Ideally, LED lighting can be used. The Sources of the different spectral ranges can be flashed in order to increase the short-term intensity. For example, a flash control device can be used to readjust the intensity.

The 2D images are preferably generated when the dome is illuminated. Dome lighting is a diffuse light that falls on the object from all directions in the same way (Handbuch der Bildverarbeitung 2018, p. 51, ISBN: 978-3-9820109-0-8.). This enables homogeneous illumination. It can be preferred to control only individual segments of the dome in order to illuminate the object from different directions or viewing angles.

Preferably at least two, particularly preferably at least three, in particular at least four 2D images are generated from a different perspective. The individual images are preferably generated simultaneously, that is to say with two, three or four cameras.

According to a further embodiment of the method, at least two, preferably at least three, particularly preferably at least four, 2D images are generated in each case with a different illumination. This can be ensured, for example, by activating a different segment of a dome lighting for each image. In this way, a separation of surface structure and texture can be realized (Shape from Shading, see Handbuch der Bildverarbeitung 2018, p. 60, ISBN: 978-3-9820109-0-8.)

A 3-D image is generally understood to mean, on the one hand, images that map the height (z-direction) as a value for each pixel on a fixed grid (x and y direction). To the others are also generally to be understood as 3D point clouds, i.e. a collection of points with x, y and z values without a fixed grid in one of the directions.

The generation of three-dimensional images is preferably carried out with a laser as the light source.

In order to generate three-dimensional images, the scattering of a laser point and / or a laser line on a surface of the fragment or fragments is preferably evaluated.

The 3D images are preferably generated by means of laser triangulation (laser light section method), strip projection, plenoptic cameras (light field cameras) and / or TOF (time of flight) cameras. These methods are described under Handbuch der Bildverarbeitung 2018, pp. 263-68, ISBN: 978-3-9820109-0-8.

With laser triangulation, a laser line is usually projected onto the object and the image is recorded with an area camera that is at a defined angle to the object. This means that closer areas of the object are shown further above in the image. An algorithm then determines a height profile from the image. The 3D surface of the entire object can be recorded by moving the object or the sensors (laser and camera). In general, the laser and camera can be freely arranged in relation to one another and calibrated using software in combination with defined measuring objects. Integrated sensors that are already pre-calibrated are generally also available.

By projecting patterns (e.g. stripe patterns and a change in the phase position) onto the object and its Recording by one or more cameras, 3D information can be reconstructed.

Furthermore, 3D recordings of the silicon rods and / or silicon fragments can be generated by means of (computer) stereo vision. As a rule, several cameras are used that record the object from different angles. Using software (e.g. HALCON, MVTec), the images can then be related to one another and a 3D image can be constructed.

The silicon rod or the silicon fragments are preferably fed to the generation of 2D and / or 3D images via a conveyor belt. In particular, the conveyor belt has a constant advance. The image recording is particularly preferably carried out continuously while the belt is running, in particular using two or more cameras which are arranged in different positions. For example, images of a silicon rod can be generated continuously or at different positions along its longitudinal axis in this way. However, the conveyor belt can also be stopped for image generation if necessary.

According to a preferred embodiment, a dome lighting is arranged above the conveyor belt.

Furthermore, 2D and / or 3D images can also be generated in the free fall of the silicon fragments. For example, an opening can be provided in the dome lighting through which the fragments fall and are recorded by surrounding cameras. Line scan cameras can preferably be used in this variant. Furthermore, a pneumatic sorting system can be arranged after the conveyor belt, which sorts the fragments as a function of the morphology index determined with the aid of the dome lighting.

After the 2D and / or 3D images have been generated, they are usually subjected to image processing. The image processing can be done in particular by software, which is preferably integrated into the system of a processed _S tands. As a rule, the software is used to select the at least one analysis area for each image generated.

On the basis of the analysis area or areas, the surface structure indicators are generated with the aid of various image processing methods. Two, in particular three, different surface structure indicators are preferably generated for each analysis area.

Image processing, in particular to define the analysis area, can include the following steps:

- Processing of the image or analysis area with image filters, e.g. blurring or creation of directional derivations.

- Combination of different images to extract special information (e.g. ShapeFrom Shading, i.e. separation of structure and texture).

- Segmentation of partial areas of the image or analysis area, e.g. separation of a silicon fragment from the background, binarization with a fixed or dynamic threshold value or a method for finding the convex envelope.

- Calculation of key figures (e.g. gray value matrix (GLCM (gray level co-occurrence matrix) values or histogram values) for the analysis area. A first surface structure characteristic number is preferably generated by determining a gray value matrix (GLCM) as an image processing method. The gray value matrix describes the neighborhood relationships between individual gray value pixels in a certain direction. By combining individual probabilities for the neighborhood relationships (content of the gray value matrix) key figures can be calculated, eg energy, contrast, homogeneity, entropy. On the basis of this first surface structure characteristic number, statements can in particular be made on the surface texture (roughness).

A second surface structure characteristic number is preferably generated with a ranking filter, in particular a median filter, as an image processing method. A ranking filter is used here to search for local dark spots, for example. The median filter creates a basic gray value of the surroundings and the dark areas are evaluated relative to this. It is not the absolute gray value, but the relative gray value to the environment that decides whether a hole or crack is detected in the surface of the polysilicon.

A third surface structure characteristic number is preferably generated by the image processing method of recognizing depressions relative to a convex envelope. Here, an area around a depression in the polysilicon is first evaluated, e.g. by evaluating a gray value gradient (edge drop, steepness of the depression).

An averaging then takes place over all depressions in the analysis area and thus a determination of the mean steepness of the holes and trenches. The dimensions of a depression (eg hole or trench) can also be used become, for example, width, length, depth, volume, inner surface to volume.

A fourth surface structure characteristic number can also be generated by the image processing method of determining the width of a laser line (due to scattering). Structured lighting is carried out by means of a laser line and a recording is made with an area camera. Usually the width of the laser line is determined at each point on the silicon surface of the analysis area and a value is generated which correlates with the roughness of the silicon surface. For the calculation of the surface structure index, in particular, a mean value is formed over the scatter in the analysis area. The laser line appears fine and narrow on smooth surfaces, while it appears wider on rough popcorn surfaces. In addition, there is a reflection from different sides in the recesses and thus also a broadening of the laser line. Ideally, this process can be combined with a conventional laser light section process. In addition to the actual height (3D information), e.g. the intensity and the spread of the line (scatter) can be determined at the respective point.

The surface structure parameters obtained for the analysis area are now combined (offset) to form a (total) morphology number for the silicon fragment or the silicon rod. A morphology map (heat map) can also be created for the analysis area.

In general, various methods can be used to combine the surface structure indicators. The surface structure characteristic numbers obtained are preferably combined by a linear combination to form the morphology characteristic number.

Other methods that can be used are the formation of decision trees, support vector machines (SVM),

Regressions or (deep) neural networks are used.

The morphology index is, in particular, a dimensionless index, the value of which is greater the more rugged / porous, that is, the more pronounced the morphology of the polysilicon.

Using the morphology index for classification offers significant potential for quality assurance and maximizing productivity. In particular, different types of polysilicon (e.g. polysilicon for electronic semiconductor applications or for solar applications) can be identified on the basis of the morphology index and corresponding further processing steps can be carried out in a targeted manner.

For example, very compact polysilicon rods can be classified as suitable for the CZ method and assigned to a corresponding comminution device.

By permanently monitoring the morphology after the deposition, the process control can also be adapted in order to make the deposition more efficient overall.

The further processing steps can be selected from the group with shredding, packaging, sorting (e.g. pneumatic sorting or free-fall sorting), sampling for quality assurance and combinations thereof.

Fig. 1 shows an arrangement for determining the morphology after the deposition

Fig. 2 shows the segmentation of a polysilicon fragment. 3, the determination schematically shows a surface structure ^¬ key figure based on GLCM

Fig. 4 graphically shows the distribution of the GLCM-based surface structure characteristic numbers for different types of polysilicon

Fig. 5 shows schematically the determination of a surface structure characteristic number on the basis of the recognition of depressions

Fig. 6 graphically shows the distribution of the GLCM-based surface structure characteristic numbers for different types of polysilicon

Fig. 7 shows the distribution of the morphology index for different types of polysilicon

FIG. 1 shows an arrangement 10 comprising a conveyor belt 12, the direction of advance of which is indicated by two arrows. There are isolated polysilicon fragments 20 on the conveyor belt 12, which are to be classified on the basis of their morphology. A dome lighting 14, which comprises several cameras 18 and light sources 16, is arranged above the conveyor belt 12. The cameras 18 and light sources 16 are coupled to software and can each be controlled individually. For example, homogeneous light conditions can be generated with the light sources 16. However, incidence of light can also be generated from a certain direction. To determine the morphology, one or more of the fragments 20 is now moved under the dome lighting 14 and 2D images of the fragment or fragments 20 are generated in accordance with the selected image recording setup. The images are preferably generated continuously, that is, without stopping the conveyor belt 12. The software is used to determine the surface structure parameters from the images generated, which are then combined to form a morphology index, which is then used for classification. For example, a sorting system can be arranged at the end of the conveyor belt 12. In principle, a silicon rod can also be moved on the conveyor belt 12 along its longitudinal axis under the dome lighting 14.

example

Polysilicon rods of three different quality types were produced in a vapor deposition reactor.

Type 1 is a very compact polysilicon that is particularly intended for the production of semiconductors. As a rule, there are hardly any differences in terms of morphology between the surface and the interior of the rod.

Type 2 has a medium compactness and is used in particular for cost-optimized robust semiconductor applications and demanding solar applications with monocrystalline silicon.

Type 3 has a high proportion of popcorn. It has a relatively rugged surface and high porosity. It is used in particular for the production of multicrystalline silicon for solar applications. One rod of each type was comminuted and the morphology index of the fragments was determined using dome lighting as shown in FIG. First of all, after being crushed, the fragments were separated on a conveyor belt and moved at a constant speed (feed) under a dome lighting. The cathedral lighting was equipped with six area cameras in different positions. The 2D images were generated from several angles at the same time. A total of six images were taken per fragment. In the evaluation described below, for reasons of clarity, only one image per fragment (viewing angle perpendicular to the surface of the conveyor belt from above) was subjected to an evaluation, i.e. the morphology index was determined. A total of 4103 fraction were ^¬ pieces of type 1, 9871 silicon investigated 3 Poly type 2 and 6918-type.

An analysis area was defined for each image by segmentation. FIG. 2 shows, by way of example, a segmentation based on a type 3 polysilicon fragment for generating an analysis region. The segmented area, that is to say the analysis area, is shown on the right in FIG.

The fragment was segmented by the following steps: (1) Applying a filter (soft focus) to the entire image area in order to smooth out hard edges.

(2) Use of an additional filter (Sobel filter, independent of direction) to calculate the differences in brightness.

(3) Segmentation of the fragment from the outside to the inside by recognizing the area with a difference in brightness greater than a defined threshold value. Areas with too little difference in brightness of starting on the outside, it is omitted until only the relevant area (cf. FIG. 2, right) remains as an analysis area.

From this analysis area, a first surface structure index was generated by determining the gray value matrix (GLCM values) and a second surface structure index was generated by recognizing and evaluating depressions.

The scheme for calculating the GLCM values is shown in FIG.

The GLCM (gray value matrix) is determined by counting combinations of gray values. An entry is created in the GLCM for each pixel in the analysis area, where i is the gray value of its own pixel and j is the gray value of the pixel in the vicinity. Since a pixel in a typical 2D image has 8 neighboring pixels, it is common practice to determine the GLCM for all directions and take the mean value. It is also possible not to use direct neighboring values, but rather the neighboring value at a distance of n pixels. In the example, the direct neighbors were used. Usually it is then divided by the total sum of the matrix entries.

The values then correspond to a probability p for the respective gray value combination.

Consideration of the contrast (equation (I)): For this purpose, high contrasts (i.e. large differences in the gray values) are highly weighted. The term | ij | ² from equation (I) is large if the values are as far away as possible from the main diagonal. These are the values for which i and j are as different as possible, i.e. the gray values are as different as possible. Consideration of the homogeneity (equation (II)). Here the term 1 + \ ij | divided. So values close to the main diagonal are weighted more heavily. As a result, areas with very similar gray value areas are rated highly in this key figure. Basically, therefore, two surface structure parameters are obtained according to equations (I) and (II).

The graphical evaluation of the GLCM indicators for the three different types of polysilicon shown in FIG. 4 shows that the values obtained for the homogeneity and the contrast are in opposite directions. The distribution of the key figures for the individual polysilicon types is shown in the histograms. The values on the X-axis correspond to the values for the respective key figure. The density is the relative frequency for the occurrence of the respective value.

The generation of the second surface structure index on the basis of the recognition and evaluation of depressions is shown schematically in FIG. 5, whereby on the one hand the number of holes per area and on the other hand the hole thickness was determined as a mean gray value gradient at the edge of the hole. The depressions are displayed relative to their surroundings via a median filter. As a result, the areas with a value smaller than a defined threshold value and a defined minimum size in pixels can then be found and marked (see the rectangles of different sizes).

The evaluation for the second surface structure characteristic numbers is shown in FIG. Here the hole areas in the analysis area are counted and output relative to the pixel area. For type 1 (very compact) there are only very few holes, that is, the index has a value close to zero. There are slightly more holes for type 2. Type 3 (fissured) has a recognizable distribution of holes (see Fig. 6, below). To evaluate the holes, the hole thickness is viewed as the mean gradient at the edge of the hole (gray value drop), with the values being scaled. For type 1 this is less because the existing holes are less deep and pronounced and therefore do not appear as dark. For type 2 and type 3, the hole areas are more pronounced (steeper and therefore darker), which increases the value for the key figure.

In a final step, the determined surface structure characteristic numbers are combined (offset) with one another in order to obtain a morphology characteristic number, on the basis of which the relevant polysilicon fragment can then be subjected to sorting (that is to say classification), for example. This combination is made by means of a linear combination using the following equation in which

X _j , i - i-th index of the j-th fragment a ± = slope for the i-th index bi = base value for the i-th index y _j = morphology value of the j-th fragment.

The result of the linear combination is shown in FIG. 7 using the histogram. The resulting distributions are significantly different, and accordingly the three different types of polysilicon can be distinguished from one another. The combination of several key figures makes the process more robust and less dependent on individual outliers.

Claims

A method of making and classifying polycrystalline silicon, comprising

- Generating a polycrystalline silicon rod by introducing a reaction gas that contains, in addition to hydrogen, silane and / or at least one halosilane, into a reaction space of a gas phase deposition reactor, the reaction space containing at least one heated filament rod on which silicon is deposited to form the polycrystalline silicon rod,

- removing the silicon rod from the reactor,

- if necessary, crushing the silicon rod to obtain silicon fragments,

- Generating at least one two-dimensional and / or three-dimensional image of at least a partial area of the silicon rod or of at least one silicon fragment and selecting at least one analysis area per generated image,

- Generating at least two surface structure key figures per analysis area by means of image processing methods, each surface structure key figure being generated with a different image processing method,

Combining the surface structure indexes to form a morphology index, the silicon rods or the silicon fragments being classified as a function of the morphology index and fed to various further processing steps.

2. The method according to claim 1, characterized in that the generation of two-dimensional images takes place with a dome lighting.

3. The method according to claim 1 or 2, characterized in that at least two, preferably at least three, particularly preferably at least four two-dimensional images are generated from a different perspective.

4. The method according to any one of the preceding claims, characterized in that at least two, preferably at least three, particularly preferably at least four two-dimensional images are generated in each case with a different illumination.

5. The method according to any one of the preceding claims, characterized in that the generation of three-dimensional images takes place with a laser as the light source.

6. The method according to any one of the preceding claims, characterized in that the scattering of a laser point and / or a laser line on a surface of the fragments is evaluated in order to generate three-dimensional images.

7. The method according to any one of the preceding claims, characterized in that the generation of three-dimensional images takes place by means of laser triangulation and / or strip light projection.

8. The method according to any one of the preceding claims, characterized in that the silicon rod or the silicon fragments are fed to the generation of the two- or three-dimensional images via a conveyor belt.

9. The method according to any one of the preceding claims, characterized in that a first surface structure Key figure is generated by determining a gray value matrix as an image processing method.

10. The method according to any one of the preceding claims, characterized in that a second surface structure index with a ranking filter, in particular median filter, is generated as an image processing method.

11. The method according to any one of the preceding claims, characterized in that a third surface structure characteristic number is generated by the image processing method of recognizing depressions relative to a convex envelope.

12. The method according to any one of the preceding claims, characterized in that the surface structure indexes are combined by a linear combination to form the morphology index.

13. The method according to any one of the preceding claims, characterized in that the further processing steps are selected from the group with comminution, packaging, sorting, sampling for quality assurance and combinations thereof.