TW201330136A - Qualification of silicon wafers for photo-voltaic cells by optical imaging - Google Patents

Abstract

During a solar cell manufacturing process a wafer is characterized by capturing an optical image of light scattered from a surface of the wafer after an etching step that leave a surface of the wafer exposed, for example without emitter diffusion or covering layers on that surface. An image processing operation is applied to the captured image to obtain processed image, the image processing operation being of a type that emphasizes edge features. The processed image to estimate a density of dislocations on the wafer and the density is used as input of a performance prediction algorithm. The application of one or more of the further steps of processing to the wafer is controlled dependent on a prediction of a property obtained from the performance prediction algorithm. Processing may be stopped, or the wafer may be rotated for example.

Description

矽 wafer for photovoltaic cell inspection by optical image quality

The present invention relates to the verification of wafers by optical image quality. Furthermore, the present invention relates to a method of measuring the properties of a semiconductor substrate used in the manufacture of photovoltaic cells and a method of fabricating a photovoltaic cell.

Photovoltaic cells, such as solar cells, can be fabricated from crystalline wafers of germanium-like semiconductor materials. The crystalline wafer contains, for example, misalignment due to thermal stress of the ingot during the crystal growth process. The misalignment density varies with the ingot, and the misalignment density and spatial distribution vary from wafer to wafer. After the wafer is processed into a solar cell, the above misalignment shortens the life of a few charge carriers. The decrease in lifetime is due to imperfect crystal structure or high concentration of metal ions that are not absorbed or passivated during the manufacture of solar cells. The shortened life of a few charge carriers results in a reduction in the performance of solar (photovoltaic) cells.

Therefore, it is possible to identify wafers or wafer regions that may contain high dislocation density, and can be used to reject wafers or based on wafer properties for classifying manufactured cells by quality. In terms of cost effectiveness, wafer quality verification should be performed as early as possible in the solar cell manufacturing process, and Time (ie, except that the data is obtained without delaying the manufacturing process) and is performed in a non-intrusive manner.

Error bit regions such as photoluminescence (PL), optical image (OI), and electroluminescence (EL) can be identified by various methods. Photoluminescence and optical imaging are methods that can be applied to bare wafers. Electroluminescence requires the presence of an electrode and is therefore typically only applied to solar cell products. In the case of photoluminescence, sample preparation is not absolutely necessary, but a better image is usually obtained after some steps (e.g., etching) of the solar cell process or application of the emitter layer.

Optical images can be easily combined with well-known digital image processing methods. The quality characterization of the optical image properties can then be coupled to a model that predicts the performance based on the presence of the misalignment.

The B.Sopori team has been working on NREL to characterize wafers by optical imaging. "A reflectance spectroscopy-based tool for high-speed characterization of silicon wafers and solar cells in commercial production" (Photovoltaic Specialists Conference (PVSC)" is described in B. Sopori et al. , 2010 35th IEEE, 002238-002241 (2010)). The point is that extensive information is available in this way. Sopori has explained the misplacement effect through a two-dimensional network model. In the document entitled "Use of optical scattering to characterize dislocations in semiconductors" (Appl. Opt., 27, 4676-4683 (1988)), Sopori has shown that the degree of scattering can be misalignment density, misalignment size, and misalignment mapping. The basis of the measurement method.

Recently, Korte et al. describe the use of a platform scanner to measure wafers in the "Measurements of effective optical reflectivity using a conventional flatbed scanner-Fast assessment of optical layer properties" (Solar Energy Materials and Solar Cells, 92, 844-850 (2008)). Optical reflectivity. One of the features is that the platform scanner or photocopier captures scattered light, ie does not include specular reflection. Photocopiers and platform scanners are similar A device that quickly scans the entire surface. Alternatively, the camera can be used. The camera has the opposite property: the light falls on the misalignment from all directions, and the scattering properties cause the misalignment to have some reflection in the direction of the camera.

In terms of optical imaging, an etching process is necessary. Specially designed etchings (containing mixtures such as HF, nitric acid, acetic acid) have been developed to optimize optical image quality. The etching process results in a rougher texture of the misaligned portions, meaning that the light reflected at these locations is highly scattered. Without such an etching process, prior art techniques do not provide reliable misalignment measurements.

Electroluminescent methods and optical images that require special etching are not suitable for use in manufacturing processes. The photoluminescence method is too costly and does not seem to distinguish between permanent defects and certain defects that will disappear in subsequent processes when applied to the stage before the emitter diffusion.

US2011025839 relates to the use of luminescent images to detect defects in solar cells, that is, luminescent images produced by exciting solar cell materials. The literature preferably uses photoexcitation, ie photoluminescence, but also mentions electroluminescence. Light scattering images are not discussed. The literature discloses photoluminescence images of wafers after surface damage etching, after emitter diffusion, after SiN deposition, and after completion of the battery process. The image after emitter diffusion and SiN deposition shows a noteworthy structure. The literature reveals that although misalignment can also be detected by photoluminescence of the original dicing wafer, measurement after the diffusion step can be advantageous because the photoluminescence intensity usually allows for shorter time data acquisition after this step. Lower quality devices or cause higher spatial resolution images or any combination thereof.

It is an object of the present disclosure to provide a method of measuring the misalignment density on a wafer that is compatible with the use of wafers in the fabrication process and produces reliable misalignment density measurements.

A solar cell manufacturing method according to item 1 of the patent application scope is provided. Here, optical inspection is used during the process. It is found that when a scattered light image is obtained (instead of specularly reflected light) and the image reinforced edge is processed (for example, the detection of the image position of the detected edge), the defect after the etching step used as part of the solar cell manufacturing process can be evaluated. density. For example, the number of defects can be evaluated as the number of image points at which the edge is detected. In this manner, online testing can be used to control the application of subsequent processing steps during solar cell fabrication.

In one embodiment, the application of the metallization pattern during the process is varied based on the performance prediction. In this manner, the orientation of the metallization pattern and/or metallization pattern that produces the best predictive performance can be selected, and the remaining process steps can be varied to apply the selected pattern and/or orientation.

10‧‧‧ Wafer Support Table

12‧‧‧Light source

14‧‧‧Detector

16‧‧‧Scanning agency

17‧‧‧ computer system

18‧‧‧ Wafer

21-27‧‧‧Steps

The above and other objects and advantages will be apparent from the following description of the drawings.

Figure 1 shows an optical inspection system.

Figure 2 shows a flow chart of the manufacturing process.

1 shows an optical inspection system including a wafer support table 10, a light source 12, a detector 14, a scanning mechanism 16, and a computer system 17. For example, wafer 18 is displayed on wafer support table 10. Light source 12 can be, for example, a linear light source that produces light along a line on wafer 18. Similarly, detector 14 can be a line detector for measuring the intensity of light of wafer 18 along a series of locations along the line. The light source 12 and the detector 14 are positioned relative to each other such that the light direction from the light source 12 to the wafer And the direction of the light to the detector 14 relative to the normal of the surface of the wafer 18 is at a different angle from each other (that is, the angle at which the detector 14 does not capture the specularly reflected light, the specularly reflected light occurs in the detective The detector 14 and the source 12 are in the plane of the opposite side of the normal plane of the surface of the wafer 18, at the same angle with respect to the normal plane). The detector 14 has an output coupled to the computer system 17. The computer system 17 is coupled to the scanning mechanism 16 to control scanning or at least to receive information indicative of the location during the scanning.

The scanning mechanism 16 is used to scan the support table 10 and the combination of the light source 12 and the detector 14 with respect to each other. The support table 10 can be moved in a lateral direction of the line on the wafer 18.

Figure 2 shows a flow chart of the manufacturing process. In a first step 21, the wafer is diced. In a second step 22, wet etching is performed using a wet etch process using an acidic etchant containing a mixed solution of, for example, an acid. This etching process creates pockets on the surface of the misaligned locations in the germanium wafer. This can be achieved using an etching process known in the art of solar cell fabrication. For example, the etching step can be an etching step that etches away the cleavage damage and/or produces a two-dimensional texture on the surface of the wafer. For example, a wet chemical process using a mixture of acids such as hydrofluoric acid, nitric acid or acetic acid can be used. Such a process forms a surface texture (a highly mutated pattern) that is sensitive to the misaligned areas. The etchant used preferably forms an etched pocket on the surface of the wafer containing the misaligned regions.

In a third step 23, computer system 17 captures the output signals from detector 14, which are for points along the line on wafer 18 and for successive lines during the scan. The captured signal defines an image of wafer 18.

In a fourth step 24, the computer system 17 applies an edge detection operator to the image. As such, the edge is a sudden, abrupt change in light intensity in the image. A well-known method for detecting edges is the Canny edge detection algorithm, which is described in R.C. Gonzalez and R.E. Woods in 2008 by Pearson Education International, "Digital Image Processing". 719-725 pages. Canny edge detection involves filtering images using a complex gradient filter (eg, four filters), selectively combining flattened images, such as by a Gaussian filter or an FIR filter that approximates a Gaussian filter response. The gradient filter is designed to produce an output signal for the image position that is proportional to the light intensity gradient in a predetermined direction. For example, a FIR filter with 5x5 pixels can be used. In the Canny edge detector, the complex gradient filter contains filters for gradients in different directions (for example, horizontal, vertical, and two diagonal directions (/ and \)). The edge signal is obtained by combining the magnitudes of the results obtained by applying the gradient filter to the image at each pixel position. This result represents a function of the edge magnitude as a pixel location. The resulting value in the suppressed pixel position is not the result of the local maximum. Double (low and high) thresholds can be applied to these results to reduce the number of false edge points. In fact, the parameters of these operations can be optimized for wafer characteristics. These operations convert the image into a processed image called a Canny edge image. The pixel has a first value (the value between the threshold values) at the position where the edge is detected and a different position at the position where the edge is not detected. Two values.

In the fifth step 25, the Canny edge image is used to calculate the Canny edge fraction (CEF=edge pixel number in the image/the total image pixel number) or the (binary) map used to create the misalignment distribution. . In an embodiment, mapping the misalignment distribution may include dividing the image into sub-blocks and calculating the number of edges of each sub-block of the image; and identifying the block whose edge density exceeds another threshold as the misalignment area.

Edge signals (eg, from the output of the gradient filter) can also be substituted for binary edge detection, and edge signal values can be used as weights, thereby weighing the effect of corresponding edge points on edge counts. In the alternative embodiment, in a fourth step 24, an edge detection algorithm other than the Canny algorithm can be used. The parameters of this edge detection algorithm can also be adjusted.

In an embodiment, the fifth step 25 may include a weighting operation, wherein the image is The position multiplies the number of edges by the weight. In another embodiment, the selected image area (corresponding to a weight that is not one or zero) may be masked.

Before calculating the Canny edge fraction, a line filter can be applied to the Canny edge image to remove linear edges that are unrelated to misalignment during detection by, for example, testing the Canny edge image values of the pixel location group along the line. And if the edge points have been detected for all of the pixel positions of the predetermined length segment along the line or not all of the pixel positions of the predetermined number of pixel positions that do not form the continuous sub-segment, the edge detection is suppressed.

In the alternative embodiment, the fourth step 24 includes replacing the Canny edge detection using a gradient method. Gradient methods are known, for example, from pages 706-714 of the book by R. C. Gonzalez and R. E. Woods. The fourth step 24 can then include applying a high pass filter (eg, a Laplacian mask) to the captured image, and the fifth step 25 can include: calculating a histogram of the filtered image (eg, frequency versus intensity values) or for different The sub-block calculates a complex such histogram; for each histogram, the integral of the intensity values of the histogram between the low light intensity value and the high light intensity value of the histogram is calculated.

In the sixth step 26, the information obtained in the fifth step 25 is used to obtain a performance prediction. Correlation can be used for this purpose. Different predicted performance values may be pre-stored with different pre-stored different density values, and the results of the fifth step may be used to retrieve relevant predicted performance values. In one embodiment, the average dislocation density obtained from the fifth step 25 (the number of edge pixels in the image divided by the total number of pixels in the image) is used to capture the predicted performance value. In another embodiment, the average density calculated by dividing the number of error bits of the sub-block by the total number of sub-blocks or all sub-block spatial mappings is used to obtain the predicted performance value. This import can adjust two additional parameters, namely the block size and the gate that selects the misaligned block Depreciation.

In a seventh step 27, an evaluation is performed based on the predicted performance to control, for example, further use of the wafer. Predictable thresholds for performance and performance can be compared.

If the wafer meets the threshold, the wafer can be processed using other steps of a predetermined process for fabricating the solar cell. If not, the wafer is processed differently. Furthermore, if the wafer can be removed before the other steps are applied according to the test results, the cost can be saved.

Other steps may include, for example, an emitter diffusion step in which the substrate is doped to create a p-n junction. The etching step can be the last etching step before the emitter diffusion. The etching step of the second step 22 of the image capturing step 23 is followed by the earliest etching step, such as the step of removing the cutting damage.

The emitter diffusion can have the effect of reducing the misalignment of the surface, but the misalignment may be a through-wafer, so problems may occur even if the misalignment is reduced. Similarly, the step of subsequently adding an additional layer to the surface of the crucible has the potential to reduce misalignment. Therefore, it is preferred to capture the image when the captured surface of the crucible does not contain the emitter diffusion, or more preferably without other diffusion layers (for example, the back side or the front side surface diffusion) or without the subsequent addition of the overlay layer (eg, Electrical layer or conductor layer).

In some processes, doping of the p-n junction is performed by doping on both sides of the substrate, and then the doped layer on either side is removed by etching to obtain a material having a single junction. The method can also perform an image capture step 23 to capture an image of the etched surface obtained by such a one-sided etching step. Although this is after the application of the emitter, the one-sided etching ensures that the surface without the emitter diffusion can be captured.

Performance prediction algorithm

Battery performance prediction predicts the value of performance properties, such as once the sun is manufactured The open circuit potential Voc of the battery, once the manufactured battery efficiency η or both. Efficiency is the most relevant characteristic for evaluating performance, but Voc is usually calculated earlier and may capture effects that are associated with at most a low charge sub-life time.

In one embodiment, a "heuristic model" can be used in which linear regression of the performance of the upper misplacement fraction is used. The misplacement fraction can be directly derived from the Canny edge calculation or from the intermediate mapping of the Canny image-based sub-block, or the correlation performance property value calculated by the Canny edge calculation based on the linear regression.

In one embodiment, an analytical model can be used in which the physical properties of the photovoltaic cell are represented. This can be done based on the equivalent circuit of the photovoltaic cell.

The first example applies an equation for a one-sided equivalent circuit for a solar cell. The area without misalignment (which specifies the first (low) diode dark saturation current density) and the region with the misalignment (which specifies the second (higher) diode dark saturation current) are distinguished. The surface average of the dark saturation current of the diode is then used as a parameter of the equivalent circuit, thereby calculating the value of the performance property.

Furthermore, the improved performance value η can be performed by the same procedure but now includes the series resistance of the emitter layer and the metallization pattern. In addition, the locally generated photon current density may be determined by the presence of misalignment. This can be done in a manner similar to the following dark saturation currents: assigning a value to an area without misalignment and assigning a lower value to an area with misalignment. Models can be extended by additional components of the equivalent circuit, such as splitters, additional diodes, series resistors, contact resistors, and the like.

In a second example, a model that describes a solar cell as a two-dimensional network of parallel solar cell single diode circuits can be used. The network has a first type and a second type (bad and good) diodes, and the diodes at different positions in the network are based on whether the sub-blocks corresponding to the positions in the image are edge blocks or do not correspond to the above. The map is selected to be the first type or the second type.

In the model, this diode network can be combined with a metallized pattern model that can be applied during manufacturing. In this model, the circuit is connected by series resistance caused by the combined effect of metallization and emitter resistance. This is defined in a partial differential equation in which the battery voltage fixed at the current collection point is used as a boundary condition. The differential equation can be numerically solved using an appropriate method such as the Finite Element Method.

Dark saturation current can be used as a property of characterization misalignment, but in this case, the position dependent photons can be used to generate current and additional components can be added to the equivalent circuit model to extend the model (see 2c above). This particular model type facilitates the use of misaligned spatial resolution information.

This predicts the possible different effects of the metallization pattern depending on the misalignment. "Performance limitations of mc-Si solar cells caused by defect clusters" by Sopori et al. (ECS Trans., 18, 1049-1058 (2009)) and "Influence of distributed defects on the photoelectric characteristics of a large-area device" (J. Cryst. Growth, 210, 375-378 (2000)) has described the effect of such a network model on the effect of misalignment on solar cell performance.

Evaluation step

The predicted performance is obtained based on the images captured after the early etching step of the process and before the remaining process steps of fabricating the solar cell. Compare predicted wafer performance to preset thresholds. It can be tested whether the predicted efficiency is above the threshold, whether the predicted Voc is above this threshold or both. If the predicted wafer performance meets the threshold, the remaining process steps are performed in a predetermined manner.

If the predicted wafer performance does not meet the preset performance threshold, one of many approaches can be taken:

1) The wafer can be removed, that is, the remaining process steps required to manufacture the solar cell are not performed.

2) For inferior wafers, wafers can be transferred to different production lines.

3) The wafer can be used for production procedures based on predicted performance changes.

In an embodiment of the latter case, the process steps of applying metallization to the wafer can be varied. For example, when using the "H" pattern or the metallization of the "E" pattern, the pattern can be applied to the wafer with zero or 90 degree rotation relative to the wafer (ie, can be applied These steps are preceded by metallization by rotating the wafer, or by rotating the device for applying the pattern (eg, screen printing), or a different printed pattern can be used). The rotation angle can be selected according to the performance prediction obtained by the metallized pattern model with different rotation angles, wherein the rotation angle with the best prediction performance is selected. This gives better performance. Research on 2D model networks has shown that misalignment is significant for metallization.

In a variant embodiment, the spacing of the fingers of the metallized pattern can be varied. The spacing can be selected based on a metallized pattern model of fingers having different pitches, wherein the spacing with the best predictive power is selected. This can be combined with the choice of angle of rotation or can be applied to the use of a predetermined angle of rotation.

In one embodiment, metallization application techniques, such as inkjet printing, can be used to impart more variation in the metallization pattern not only in rotation and/or spacing. The pattern can be selected based on performance predictions obtained using the model of the metallization pattern, wherein the pattern with the best predictive power is selected.

Monitoring wafer quality with simultaneous performance predictions can identify instability or problems in production, and can distinguish such problems from wafer quality variations. In an embodiment, in completion Performance (eg, Voc or efficiency) measurements are taken after the manufacturing process, or during the manufacturing phase following the capture image phase. Comparing the measured wafer performance to the predicted wafer performance, and if the deviation exceeds the threshold, a warning is generated indicating that an error may occur in the production process. If you do not use predictions, you may not be able to detect such errors because they cannot be distinguished from unknown wafer defect effects.

Quality issues can be fed back to the wafer supplier to provide recommendations for improving the crystal growth process.

10‧‧‧ Wafer Support Table

12‧‧‧Light source

14‧‧‧Detector

16‧‧‧Scanning agency

17‧‧‧ computer system

18‧‧‧ Wafer

Claims (12)

A solar cell manufacturing process, wherein the manufacturing process includes an etching step and characterization of a wafer, the manufacturing process comprising: etching a wafer as part of solar cell fabrication using one of an acid etchant wet chemistry Etching; capturing an optical image of light scattered from a surface of the wafer after the etching step and prior to other processing steps of the wafer; applying an image processing operation to the captured image to obtain a processed image Image, the image processing operation is an enhanced edge feature mode; using the processed image to evaluate a misalignment density on the wafer; using the misalignment density as an input to a performance prediction algorithm; obtaining from the performance prediction algorithm One of the properties predicts the application of one or more of the other processing steps of the wafer. The manufacturing process of claim 1, wherein the optical image is captured at a manufacturing stage in which the surface of the wafer has no emitter diffusion after the etching. The manufacturing process of claim 1, wherein the image processing operation is a Canni edge detection operation. The manufacturing process of claim 1, wherein the image processing operation comprises applying a high pass filter to capture the image, the manufacturing process comprising calculating a histogram of the filtered image, and calculating the histogram One of the intensity values of the histogram between the low light intensity value and a high light intensity value is integrated. The manufacturing process of any of the preceding claims, wherein one or more of the other are controlled This step of applying the steps includes selecting not to have the wafer perform the other steps or to have the wafer perform the other steps. The manufacturing process of any of the preceding claims, further comprising changing the application of a metal pattern in accordance with the prediction. The manufacturing process of claim 1, wherein the misalignment density on the wafer is evaluated as a function of position on the wafer. The manufacturing process as described in claim 7 includes: using the dislocation density as an input to a performance prediction algorithm, using a solar cell performance model having a different metallization pattern and/or a metallization pattern orientation, And combining the evaluated spatial distribution of the dislocation density; selecting one of the orientations of the metallization pattern and/or the metallization pattern corresponds to an optimal prediction performance; according to the selected metallization pattern and/or The orientation of the metallization pattern processes the wafer. The manufacturing process of any one of claims 1 to 8 includes the step of not allowing the wafer to perform the application of the other step if the prediction of the property does not meet a predetermined threshold. The manufacturing procedure of any of the preceding claims, comprising measuring the property after applying the other step, comparing the property of the prediction and measurement of the property, and if the property is predicted and measured for the property When the difference exceeds a predetermined amount, a warning signal is generated. A system for performing solar cell characterization during a process, the system comprising: a support table for a wafer; a light source directed to a position on the support table; a light detector directed to the position Positioned relative to the light source, but only detected from the wafer Scattered light; an image processing system configured to apply an edge detection processing operation to an image captured by the detector to obtain a processed image, the image processing system configured to be based on the processed image Evaluating a misalignment density as a function of position on the wafer to use the dislocation density as an input to a performance prediction algorithm, and predictively controlling the wafer based on one of the properties obtained from the performance prediction algorithm Application. The system of claim 11, wherein the edge detection operation is a Canni edge detection operation.