WO2013150424A1

WO2013150424A1 - Optical method of mapping the crystal orientation of a sample

Info

Publication number: WO2013150424A1
Application number: PCT/IB2013/052509
Authority: WO
Inventors: Jean-Daniel PENOT; Cyril Cayron; Benoît Marie
Original assignee: Commissariat A L'energie Atomique Et Aux Energies Alternatives
Priority date: 2012-04-02
Filing date: 2013-03-28
Publication date: 2013-10-10
Also published as: FR2988841A1

Abstract

Method of mapping the crystal orientation of a sample of polycrystalline material having undergone a step of mechanical shaping, the method comprising the steps consisting in: a) acquiring at least three images of said sample under directional lighting, each image corresponding to a different lighting and/or acquisition geometry; b) identifying crystal grains on said images, and associating with each of them a vector or matrix representative of luminous intensity values measured on various images in correspondence with said grain; and c) determining the crystal orientation of each said grain as a function of the corresponding vector or matrix; said steps being implemented after said step of mechanical shaping and before any operation of chemical etching or polishing of the surface of the sample.

Description

OPTICAL METHOD FOR CARTOGRAPHY OF THE CRYSTALLINE ORIENTATION OF A SAMPLE

The invention relates to a method for mapping the crystalline orientation of a sample of polycrystalline material. This process is based on an optical technique and applies to samples having undergone a mechanical shaping step, such as silicon wafers which are obtained by sawing an ingot. The invention applies in particular, but not exclusively. , the characterization of substrates in a photovoltaic cell manufacturing process,

Photovoltaic cells are mainly made from substrates of mono- or polycrystalline silicon. While monocrystalline substrates are, by definition, characterized by a constant crystallographic orientation, the polycrystalline wafers consist of a number of crystallites - or "grains" - of distinct crystalline orientations. Since silicon has a crystalline structure of the "diamond" type, it has an anisotropy: its physical characteristics depend on the crystalline orientation considered. In the case of polycrystalline slices the crystalline orientations of these grains are not uniformly distributed, which leads to a heterogeneity of the slice.

The anisotropy of silicon has multiple impacts on the characteristics of substrates and solar cells, in particular on their resistance to damage and fracture or on their electrical properties; see in this connection the article by B. A. Nesterenko et al "Electrical properties of clean silicon surfaces with different crystalline orientations", Surf. Sci., 18 (1969), pp.239-244.

Another practical consequence of this anisotropy concerns the so-called texturing step, performed during the manufacture of the photovoltaic cells to reduce the reflectivity of the surface of the cell. This step consists of developing a surface micrometric relief, usually pyramidal, by a chemical attack that reveals the dense planes - mainly the plans (111). In fact, the local crystalline orientation conditions the geometry of the etched surface patterns, which themselves directly influence on the absorption rate of incident solar radiation, and therefore on the energy conversion efficiency of the solar cell.

Knowledge of the crystalline orientation of the silicon wafers, and particularly the grains making up the polycrystalline silicon wafers (pofy-Si), is therefore essential.

Currently, crystal orientation determination is generally performed by electron or X-ray diffraction techniques, and sometimes by neutron diffraction or Raman spectroscopy. In particular, for poly-Si slices, X-ray diffraction by the Laue method or backscattered electron diffraction (EBSD) is commonly used. In addition to their high costs, these diffraction techniques have the major disadvantage of being slow. In the case of X-ray diffraction, specifically dedicated and optimized equipment for determining the crystal orientation of the grains of a poly-Si wafer requires 1 to 2 hours to measure the orientations of about 500 grains. Also very slow, the electronic diffraction also has a second significant drawback which is the small surface that can be analyzed. In a conventional scanning electron microscope (SEM), EBSD mapping can only be carried out on a surface of the order of 2 cm by 2 cm at most, in several hours or even tens of hours, with moreover the vacuum constraint imposed by the use of SEM.

These limitations inherent in diffraction techniques make them ineffective for the study of large-scale pofy-Si slices, especially on industrial photovoltaic cell production lines, whose rates are now from one to a few seconds per substrate. , or in microelectronics.

DE 197 25 535 discloses an optical method for identifying the (single) crystalline orientation of a monocrystalline sample by detecting light scattered from its surface.

No. 5,032,734 discloses an optical method for detecting surface defects of a sample, which may for example be a slice of semiconductor material. In the article by L. Sopori et al., "A reflectance spectroscopy-based tool for high-speed characterization of silicon wafers and solar cells in commercial production," 35th iEEE Photovoitaic Specialists Conference (PVSC), 2010, it is proposed to use a reflective optical imaging technique to obtain maps of crystalline orientations in a fast manner, that is to say in only a few seconds; The device for carrying out this process is described in US 6,275,295. The main drawback of this technique is that it presupposes texturing of the surface by anisotropic etching. Admittedly, the realization of such texturing is generally provided by the photovoltaic cell manufacturing processes, but it is not necessarily found in other processes - for example in microelectronics - also requiring the determination of a mapping of crystalline orientations . Even with respect to photovoltaic cells, in addition, it would be much better to know the crystal orientation (s) of the substrates more upstream in the production line. One could then, for example, select the most suitable texturing treatment according to the distribution of the crystalline orientations of the different grains, in particular the direction out of the majority plane. In general, the knowledge of the orientation of the crystallites constituting the poly-Si slices just after their creation would make it possible to characterize the crystallization and to sort at the earliest time the substrates in the production line of the photovoltaic cells. No known technique of the prior art allows this under conditions compatible with the constraints of an industrial manufacturing.

The invention aims to solve the aforementioned drawbacks of the prior art and to provide a method for mapping the crystalline orientation of a polycrystalline material sample which is simple, fast and inexpensive to implement, and which does not require not a step of preparing the surface to be mapped, in particular by chemical etching attack.

To achieve this goal, the invention exploits the fact that the manufacture of silicon wafers almost always involves one or more mechanical shaping processes - usually a cut, sometimes followed by a surface grinding. For the realization of cells In photovoltaic systems, Si-Si slices are generally sawn into the mass of an ingot by wire cutting methods. There are two main types of cuts depending on the type of abrasion involved. On the one hand, the free abrasive cut in which the abrasive grains are mixed with the lubricant ("slurry"), thus being independent of the two bodies in relative movement. On the other hand, fixed abrasive sawing for which the hard particles that constitute the abrasive are attached to the surface of the wire and "scratch" the surface of the body. In all these mechanical forming processes, the wear of the material is carried out by abrasion, that is to say caused by the sliding on its surface of another body. The topography of the sawed surface is often characteristic of the mechanism by which it wears progressively. In other words, the surface state of the part obtained is a function of the abrasion. However, silicon has anisotropic mechanical properties - that is to say which depend on the crystalline orientation - such as its hardness, the existence of preferential cleavage plane, its tenacity, its Young's modulus, etc. In fact, for a given shaping process, the surface condition obtained is a function of the orientation of the abraded grain. In the case of a purely mechanical process, which is not followed by fine chemical-mechanical polishing (CMP), the surface irregularities reach dimensions of the order of magnitude of the wavelength of visible light. or higher, i.e., about 0.5 μm or more.

The present inventors have realized that this makes it possible to reveal the zones of different crystalline orientations by observation under controlled illumination, without the need for the prior chemical etching texturing step provided in the aforementioned article by BL Sopori et al. The existence of a direct correlation between the reflectivity images of the mechanically shaped surfaces, without prior surface treatment, and the crystalline orientation of the grains is not obvious. Indeed, the alterations of the surface state induced by the mechanical shaping are due to complex and poorly understood physical mechanisms (scratch stripping of the anisotropic media, multiple indentations of the crystals, etc.). An object of the invention is therefore a method for mapping the crystalline orientation of a sample of polycrystalline material having undergone mechanical cutting, the method comprising the steps of:

a) acquiring at least three images of said sample under directional lighting, each image corresponding to a different scanning and / or acquisition geometry;

b) identifying crystal grains on said images, and associating with each of them a vector or matrix representative of luminous intensity values measured on different images in correspondence o said grain; and

c) determining the crystalline orientation of each said grain according to the corresponding vector or matrix;

said steps being carried out after said mechanical cutting and before any operation of etching or polishing the surface of the sample.

This method applies mainly to single-phase samples because the presence of several phases would introduce additional heterogeneity making its implementation more delicate.

According to various embodiments of the invention:

o - Said step a) may comprise the acquisition of a plurality of images of said sample with different lighting directions and for the same direction of observation of the sample.

Alternatively or additionally, said step a) may comprise acquiring a plurality of images of said sample with the same scanning direction and for different sample viewing directions.

Said step b) can comprise:

b1) identifying, from the images acquired during step a), crystal grains present on the surface of the sample; o b2) calculating a mean light intensity for each said image and for each said crystal grain; and b3) the association with each said crystal grain of a vector, each element of which is representative of the average light intensity of said grain, calculated on a respective image.

As a variant, said step b) can comprise:

b'1) identifying, from the images acquired during step a), crystal grains present on the surface of the sample;

b'2) calculating a mean light intensity for each said image, for each said crystal grain and for a plurality of colors and / or polarizations; and

b'3) the association with each said crystal grain of a matrix of which each element is representative of the average light intensity of said grain, calculated on a respective image and for a polarization and / or a respective color.

The method may also include a prior calibration step comprising:

performing said steps a) and b) for at least one reference sample having crystal grains of known orientation; and

for each of said crystal grains, relating said known orientation to the vector or matrix determined in said step b).

Said prior calibration step may also comprise the determination of the orientations of the crystal grains of said or of each reference sample by a technique chosen from:

a diffraction technique, especially of radii

X, electrons or neutrons; and

Raman spectroscopy.

Said step c) can be implemented by means of a statistical data processing method, in particular regression or classification, according to the data obtained during said calibration step.

The method may also include a prior step of mechanical cutting of the sample by an abrasion process. Said sample may be of semiconductor material, and in particular silicon or germanium.

Said sample may advantageously have crystalline grains of a size greater than or equal to 500 μm, preferably greater than or equal to 1 mm and more preferably greater than or equal to 5 mm. This means that at least 50%, and preferably at least 90%, of the analyzed surface is covered by crystalline grains whose exposed faces have a size greater than or equal to 500 μm (or 1 mm, or 5 mm , respectively), said dimension being defined as the minimum length of a segment traversing the exposed face of the grain through its centroid; it is generally accepted that the shape of the grains does not vary in a direction perpendicular to the surface of the sample (thickness). Indeed, the method of the invention makes it more difficult to identify nano- or micro-cristailites.

Another object of the invention is the application of such a method to the characterization of substrates in a process for manufacturing photovoltaic cells.

Other characteristics, details and advantages of the invention will emerge on reading the description made with reference to the accompanying drawings given by way of example, in which:

Figures 1A and 1B show two lighting configuration and observation of a sample that may be suitable for the implementation of the invention;

Figures 2A-2D show four images of the surface of a polycrystalline sample, corresponding to different illumination directions;

Figure 3 illustrates the identification of crystal grains from the images of Figures 2A-2D;

Figure 4 illustrates the numerical representation of a crystal grain; and

Figure 5 shows a mapping of crystalline orientation obtained by a method according to the invention. FIG. 1A illustrates, schematically, a lighting configuration and observation of a sample E according to a first embodiment of the invention. In this configuration, three light sources S1, S2 and S3 are arranged to illuminate the surface of said sample from three different directions. An image sensor (in this case, a camera) C1 observes the sample from above, in a direction perpendicular to the surface. The three light sources are lit in turn, and each time the camera acquires an image of the surface. The three images are different from each other because they correspond to different lighting directions.

FIG. 1B illustrates a lighting configuration according to a second embodiment of the invention. In this configuration, a single light source S is arranged to illuminate the surface of said sample from above, in a direction perpendicular to the surface. Three image sensors (cameras) Cil, CI2 and CI3 observe the sample from three different directions. The three image sensors are activated in turn or simultaneously to acquire respective images of the surface illuminated by the source S. The three images are different from each other because they correspond to directions of acquisition, or observation, different. This second embodiment is more expensive to implement than the first, because it requires a plurality of image sensors, but is also faster, because all the images necessary for the identification of crystalline directions can be acquired. at the same time.

Other configurations are suitable for the implementation of the invention. For example, in the case of Figure 1A, the image sensor may not be oriented perpendicular to the surface; likewise, in the case of FIG. 1B, the illumination may not be in normal incidence. The number of different lighting / observation configurations need not be equal to three, but may be higher, for example between 3 and 20, preferably between 3 and 8, more preferably between 3 and 6. It is also possible to combine the principles of the two embodiments described above and to use both a plurality of light sources and image sensors. Checking and observation configurations can differ in their direction of incidence relative to the sample frame of reference and / or in their angle of incidence relative to the plane of the surface. The incident light may be collimated by an optical device to produce a beam of light by a beam of parallel light rays. The incident light can be polarized or not and be mono- or polychromatic; it is preferably spatially incoherent. The light sources may be, for example, light-emitting diodes or lamps. Image sensors can be in black and white or in color; they can use in particular CCD or CMOS technology.

Typically, the lateral dimensions of the sample E can be between 5 mm x 5 mm and 450 mm x 450 mm. For samples with "large" grains (with an average characteristic dimension greater than or equal to 1 cm), the dimensions of the sample are preferably between 100 mm × 100 mm and 160 × 160 mm. For small-grain samples (average characteristic dimension less than 1 cm) the analyzed area is preferably between 5 x 5 mm and 30 mm x 30 mm. For such dimensions, the image sensors may comprise optical objectives, in particular by performing the reflectivity measurements on an optical microscope or through a magnifying glass.

Prior to the acquisition of the images, the sample E must have undergone a mechanical shaping, for example a wired cut with free abrasive grains, a wire cut with abrasive grains integral with the wire, a circular saw cut, a rectification.

The sample may also have been cleaned or rinsed after mechanical shaping, but it must not have been finely polished to erase surface irregularities caused by shaping.

FIGS. 2A-2D show four images of a Poly-Si sample, corresponding to four different lighting and / or acquisition geometries, in accordance with the principle of the invention. A graphic scale is drawn next to Figure 2A. These images are in black and white, digitized on 32 bits (more generally, 8 bits or more). It can be noted that the different crystalline grains have gray levels that are different from one image to another: the grains that appear light on one of the images are much darker on another, and vice versa. As explained above, this is due to their surface condition which depends on their crystalline orientation, revealed by abrasions produced by mechanical shaping.

The grain contours are detected by image processing, for example using the ImageJ software (http://rsb.info.nih.gov/ij/). The detection of the contours is advantageously preceded by a pretreatment including a noise suppression operation ("Despecklïng") to erase small surface defects such as dust and a smoothing of the hue within a grain, without damaging its properties. outlines ("Threshoided Blur").

It should be noted that a single image, corresponding to a specific lighting / observation geometry, may not be sufficient to identify the contours of all the grains: in fact, in some cases two adjacent grains may have the same level of gray. It is therefore necessary to carry out the step of identifying the contours on several images (advantageously, all), then to superimpose the contour matrices to form only one.

In FIG. 3, five grains referenced A1, A2, A3, A4, B1 have been highlighted. In a real application, all the grains visible in the image would be identified, but this is only to illustrate the operating principle of the invention.

The grains identified are numerically in the form of a set of pixels whose gray level is defined by a value for example between 0 and 255, as illustrated in FIG. 4, where the interior of the grain is distinguished. G and the outline C.

To smooth the measurement noise, the gray level of this grain is defined by the average value of the pixels it contains. In the case of the example of FIG. 4, this value is 240 (average of the values of the pixels of the zone G). Thus, a numerical value is assigned to each grain and for each image. If we start with N> 3 images, we associate with each grain a vector with N components (a number, corresponding to a mean gray level, that is to say to a mean luminous intensity, for each image). Starting from a color image, each gray level should be replaced by three values, corresponding to the red, green and blue components. So, we would associate with each grain a matrix Nx3 (or, equivalent, a vector of dimensions 3N) N being the number of images (N> 3) and 3 the number of chromatic components. It is also possible to work with polarized light.

The table below contains the average gray levels of A1 - A4 and B1 grains. The grain A1 is associated with the vector {96, 183, 203, 207} and so on.

To show that there is a correlation between these vectors and crystal orientation directions of the grains, a crystalline orientation mapping has been carried out by electron diffraction (EBSD). The result is presented by the left-hand part of FIG. 5. The crystalline orientations of grains A1-A4 have been plotted on the orientation scale on the right of the figure. These orientations can be used to predict the - supposedly unknown - grain B1, through the corresponding gray level vectors.

In this case, given the small number of grains used, one can proceed qualitatively, noting that the vector associated with grain B1 - {79, 96, 95, 69} - corresponds substantially to that associated with grain A2 - {81, 99, 99, 65} - with an error of less than ± 5 on each component, which can be attributed to the measurement noise. It can be deduced that the crystalline orientation of the grain B1 is identical to that of the grain A2, which is confirmed by the direct measurement by EBSD.

In a real application a calibration will be carried out using a large number of grains of known orientation (500 - 10000, for example 2000 for an expected angular resolution of the order of the degree) forming one or more reference samples so to build a model regression or classification statistics, using known data analysis techniques.

The method of the invention has the advantage of being very fast (of a few seconds per sample), of being able to be used on the production lines of semiconductor substrates or of solar cells and to require no steps surface preparation.

Claims

A method of mapping the crystalline orientation of a sample (E) of polycrystalline material having undergone a mechanical shaping step, the method comprising the steps of:

a) acquiring at least three images of said sample under directional lighting, each image corresponding to a different lighting and / or acquisition geometry;

b) identifying crystal grains (B1) on said images, and associating with each of them a vector or matrix representative of light intensity values measured on different images in correspondence of said grain; and

said steps being carried out after said mechanical shaping step and before any operation of chemical etching or polishing of the surface of the sample,

The method of claim 1 wherein said step a) comprises acquiring a plurality of images of said sample with different illumination directions and for the same viewing direction of the sample.

3. Method according to one of the preceding claims, in said step a) comprises the acquisition of a plurality of images of said sample with the same lighting direction and for different viewing directions of the sample.

4. Method according to one of the different claims, wherein said step b) comprises:

b1) identifying, from the images acquired during step a), crystal grains present on the surface of the sample;

b2) calculating a mean light intensity for each said image and for each said crystal grain; and b3) the association with each said crystal grain of a vector, each element of which is representative of the average light intensity of said grain, calculated on a respective image.

5. Method according to one of claims 1 to 3 wherein said step b) comprises:

6. Method according to one of the preceding claims, also comprising a preliminary calibration step comprising:

performing said steps a) and b) for at least one reference sample having crystal grains of known orientation; and for each of said crystal grains, relating said known orientation to the determined vector or matrix in said step b).

7. The method of claim 6 wherein said prior calibration step also comprises determining the crystal grain orientations of said or each reference sample by a technique chosen from:

a diffraction technique, especially X-rays, electrons or neutrons; and

Raman spectroscopy.

8. Method according to one of claims 6 or 7 wherein said step c) is implemented by means of a statistical data processing method, including regression or classification, according to the data obtained in said step calibration.

9. Method according to one of the preceding claims, wherein said mechanical shaping step comprises at least one operation selected from an abrasive cut and a grinding.

10. Method according to one of the preceding claims, wherein said sample is of semiconductor material, and in particular silicon or germanium.

11. Method according to one of the preceding claims, wherein said sample has crystal grains with a size greater than or equal to 500 μm, preferably greater than or equal to 1 mm and more preferably greater than or equal to 5 mm.

12. Application of a method according to one of the preceding claims to the characterization of substrates in a photovoltaic cell manufacturing process.