WO2011017772A1

WO2011017772A1 - Detection of discontinuities in semiconductor materials

Info

Publication number: WO2011017772A1
Application number: PCT/AU2010/001041
Authority: WO
Inventors: Thorsten Trupke; Juergen Weber; Ian Andrew Maxwell; Robert Andrew Bardos; Graham Roy Atkins
Original assignee: Bt Imaging Pty Ltd
Priority date: 2009-08-14
Filing date: 2010-08-13
Publication date: 2011-02-17
Also published as: TW201125057A; CN102575993A; CN102575993B

Abstract

Methods and systems are disclosed whereby light scattered laterally within a semiconductor sample is imaged to detect a discontinuity such as a crack. The light can be introduced into the sample using an external light source, or generated in situ as long wavelength photoluminescence. The methods are described with respect to crack detection in silicon wafers and photovoltaic cells, but are applicable in principle to any semiconductor wafer or thin film material.

Description

Detection of discontinuities in semiconductor materials

Field of the Invention

The present invention relates to methods for detecting discontinuities, and in particular cracks, in semiconductor materials. The invention has been developed primarily for detecting cracks in semiconductor wafers or photovoltaic cells or modules either during or after manufacture, however it will be appreciated that the invention is not limited to this particular field of use. Background of the Invention

Any discussion of the prior art throughout this specification should in no way be considered as an admission that such prior art is widely known or forms part of the common general knowledge in the field.

Micro-cracks, or cracks in general, are a source of reduced yield in the production of semiconductor devices. In the manufacture of silicon-based photovoltaic cells and modules in particular, cracks are a significant risk because of the fragility of silicon wafers combined with the high throughput (of the order of one wafer per second) and the mechanical handling required at several stages, such as the screen printing of electrical contacts and the tabbing and stringing of individual cells during module manufacture. Severe cracks can cause cell breakage, with breakage rates in

photovoltaic cell production being on the order of several percent of total

throughput. Less severe cracks can still degrade the performance of photovoltaic cells, for example by disrupting the flow of electrons and holes in the cell material or breaking electrical traces, and have the potential to grow either during manufacture or in use (e.g. from thermal cycling). In silicon a distinction can be drawn between 'natural' cracks that can occur in multicrystalline wafers, for example as a result of thermal issues either in casting or in overly -rapid load or unload of wafers from diffusion furnaces in photovoltaic cell lines, and 'mechanical' cracks induced in monocrystalline or multicrystalline wafers by mechanical stress. Both types of crack are however problematic. Various techniques are used currently to detect the position and/or presence of cracks in semiconductor materials, including ultrasonic resonance, visible (VIS) and

infrared (IR) light transmission, and electroluminescence. These techniques have limitations with regard to at least one of measurement speed, accuracy, reliability of crack identification and applicability to a wide range of samples; for example VIS/IR transmission is inapplicable to finished screen printed silicon photovoltaic cells because of their fully metallised rear surface, electroluminescence requires

mechanical contact (which can cause further damage) and can only be applied to metallised cells, and ultrasonic vibration, which also requires mechanical contact, provides little or no information on the location of detected cracks. This latter point is significant because cracks close to an edge of a wafer or cell are more likely to grow and cause cell breakage. A reliable crack detection method is thus required with sufficient speed for current photovoltaic cell manufacturing (from about one to a few seconds per sample), that detects the position, shape and/or distribution of cracks and is applicable to wafers, finished cells and photovoltaic modules. It is an object of the present invention to overcome or ameliorate at least one of the disadvantages of the prior art, or to provide a useful alternative. It is an object of the present invention in its preferred form to provide improved methods for detecting cracks in photovoltaic cells. Summary of the Invention

According to a first aspect the present invention provides a method for detecting a discontinuity in a semiconductor material, said method comprising the steps of:

a. generating and guiding light through said material;

b. acquiring an image of said material, wherein said image comprises light scattered or transmitted from said material; and c. identifying a light intensity differential in said image thereby detecting said discontinuity.

Preferably the discontinuity, or a feature of the material, is a crack or an inclusion. Preferably more than one discontinuity is identified. The discontinuity may be naturally occurring defects, e.g. inclusions, or those which are induced, e.g. cracks.

Preferably the light which is generated and guided through the material, which is preferably substantially planar, is externally generated light which is shone into and guided through the material, or is generated within the material (i.e. PL or EL) and then guided through the material.

The discontinuity is detected by determining a differential in light intensity in the image, or a contrast difference or change in the image. For example, the light intensity differential may be a bright linear or curvilinear feature in a relatively darker background of the image, or a dark linear or curvilinear feature in a relatively brighter background of the image. Alternatively, the light intensity differential is an abrupt increase or decrease in the intensity of light scattered or transmitted out of the material.

In one embodiment step (b) comprises acquiring two or more images, wherein the light is generated and guided so as to propagate laterally through the substantially planar semiconductor material in two or more directions. In this embodiment step (c) preferably comprises monitoring variations in said light intensity differential between the two or more images.

In one embodiment step (a) comprises coupling the light into the material at one or more locations. Preferably the light is coupled into the material through a textured surface of the material.

In another embodiment step (a) comprises illuminating a surface of the material at one or more locations with above band gap radiation, to generate the light as

photoluminescence.

Preferably the one or more locations are substantially linear in shape and step (b) comprises acquiring, with one or more line cameras, images of said material when said material is in relative motion with respect to said one or more line cameras. Preferably the one or more line cameras are adapted to capture light from areas that are substantially parallel to said locations. In another embodiment step (b) comprises acquiring, with two - A - or more line cameras adapted to capture light from areas on either side of each of said locations, images of said material.

In yet another embodiment step (a) comprises applying a forward bias to said material, to generate said light as electroluminescence.

Preferably the method of the invention further comprises the steps of:

(d) acquiring an optical image of the material; and

(e) comparing said optical image with the image acquired in step (b).

Preferably the method of the invention further comprises the steps of:

(f) acquiring a photoluminescence image of said material; and

(g) comparing said photoluminescence image with the image acquired in step (b).

Preferably the photoluminescence in the photoluminescence image is generated with an illumination intensity of order 100 Suns or higher.

In one embodiment, the photoluminescence image is acquired when the material is at elevated temperature following a high temperature processing step. For example the high temperature processing step is selected from the group consisting of emitter diffusion and metal contact firing.

In another embodiment the method of the invention further comprises the step of: (h) calculating, for a discontinuity detected in step (c), one or more parameters selected from the group consisting of length, width, position and shape.

Preferably the method of the invention is applied to a thin film, wafer, or photovoltaic cell comprised of the semiconductor material, or alternatively is applied to a wafer or photovoltaic cell comprising multicrystalline or monocrystalline silicon.

According to a second aspect the present invention provides a system for detecting a discontinuity in a semiconductor material, said system comprising:

i. an optical source adapted to couple light into the material at one or more locations such that the light is guided within the material; and

ii. an imaging device sensitive to the light, for acquiring one or more images of the material, whereby the discontinuity is detected on the basis of a light intensity differential in the one or more images.

Preferably the optical source is adapted to couple light into the material through a textured surface of the material. According to a third aspect the present invention provides a system for detecting a discontinuity in a semiconductor material, said system comprising:

i. a source of above band gap radiation adapted to illuminate a surface of the material to generate photoluminescence, at least a portion of which is trapped as light guided within the material; and

ii. an imaging device sensitive to the light, for acquiring one or more images of said material, whereby the discontinuity is detected on the basis of a light intensity differential in said one or more images.

Preferably the imaging device comprises a line camera for acquiring images of the material as the material is moved relative to the line camera.

More preferably the imaging device comprises two line cameras for acquiring images of said material, adapted to collect light from areas of said material on either side of a location where light is launched into or generated within said material.

Alternatively the imaging device comprises a line camera adapted to collect light from an area of said material between two locations where light is launched into or generated within said material.

According to a fourth aspect the present invention provides a system for detecting a discontinuity in a semiconductor material, said system comprising:

i. a power supply adapted to apply a forward bias to said material to generate electroluminescence, at least a portion of which is trapped as light guided within said material; and

ii. an imaging device sensitive to said light, for acquiring one or more images of said material, whereby said discontinuity is detected on the basis of a light intensity differential in said one or more images.

Preferably the system of the invention further comprises:

iii. a source of visible light for illuminating said material; and

iv. a first camera for acquiring an optical image of said material.

Preferably the system of the invention further comprises:

v. a source of above band gap radiation for illuminating said material to generate photoluminescence from said material; and

vi. a second camera for acquiring a photoluminescence image of the material. Preferably the source of above band gap radiation is adapted to illuminate the material with an intensity of the order of 100 Suns or higher. For example, the source of above band gap radiation is a flash lamp.

Preferably the system of the invention further comprises a processor adapted to analyse the one or more images for a light intensity differential and to detect, based on said light intensity differential, the presence of the discontinuity in the material.

Preferably the processor is further adapted to analyse said one or more images for variations in said light intensity differential when said light is generated so as to propagate in different directions and to detect, based on said variations, the presence of said discontinuity in said material.

In one embodiment the discontinuity is a crack, and said processor is further adapted to calculate, for said detected crack, one or more parameters selected from the group consisting of length, width, position and shape of said crack.

Preferably the semiconductor material is a thin film, wafer, or photovoltaic cell. More preferably, the semiconductor material is a wafer or photovoltaic cell comprising multicrystalline or monocrystalline silicon.

According to a fifth aspect the present invention provides an article of manufacture comprising a computer usable medium having a computer readable program code configured to conduct the method according to the first aspect or operate the system according to the second, third or fourth aspects.

Brief Description of the Drawings

Benefits and advantages of the present invention will become apparent to those skilled in the art to which this invention relates from the subsequent description of exemplary embodiments and the appended claims, taken in conjunction with the accompanying drawings, in which:

Fig 1 illustrates conventional transmission methods for identifying cracks in a

semiconductor wafer;

Fig 2 shows a plot of absorption length versus wavelength for silicon;

Figs 3A and 3B illustrate the interruption by a crack of long wavelength light propagating inside a wafer with smooth surfaces, and how this interruption can be detected; Figs 4A to 4C illustrate the interruption by a crack of long wavelength light propagating inside a wafer with textured surfaces, and how this interruption can be detected;

Fig 5 illustrates the coupling of light from an external source into a wafer having a textured surface;

Figs 6A and 6B illustrate in plan view and side view a crack detection apparatus according to an embodiment of the invention;

Fig 7 illustrates in side view an illumination unit suitable for use in the embodiment shown in Figs 6A and 6B;

Fig 8 illustrates in side view a crack detection apparatus according to another embodiment of the invention;

Figs 9A, 9B and 9C show respectively a photoluminescence image, a visible light image and a long wavelength scattered light image of a fragment of a multicrystalline silicon photovoltaic cell containing a crack;

Fig 10 illustrates in plan view a crack detection apparatus according to yet another embodiment of the invention;

Figs 1 IA and 1 IB illustrate how the intensity differential caused by a crack can vary with the direction of light propagation with the sample;

Fig 12 illustrates image contrast effects of variations in grain texture in a multicrystalline sample; and

Figs 13 A and 13B show PL images of a monocrystalline silicon wafer containing several cracks, acquired with ~1 Sun and -100 Suns illumination respectively.

Detailed Description

Preferred embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings.

The present invention relates primarily to methods for detecting cracks in semiconductor materials, and in particular to methods for detecting cracks in photovoltaic cells and modules formed from semiconductor wafers. The inventive methods will be described with respect to crack detection in silicon wafer-based photovoltaic cells, but are not limited to this particular field of use. For example the methods may also be applicable to thin film semiconductor materials, particularly if the substrate has a lower refractive index than the semiconductor material to encourage lateral guidance of long wavelength light in the thin film. The methods may also be applicable to the detection of other forms of discontinuities that disrupt the guidance of long wavelength light, such as large inclusions of silicon carbide in silicon wafers. The methods are of course not limited to silicon materials, but can also be used to detect features of interest (e.g. cracks, inclusions, discontinuities, etc) in a wide variety of other compound and single element

semiconductor materials including for example germanium, gallium arsenide, indium gallium arsenide, gallium nitride, indium phosphide, cadmium telluride, zinc selenide and copper indium gallium selenide.

Conventional visible or IR transmission techniques for crack detection rely on

measurements with illumination from one side of the sample and detection from the other side. As illustrated in Fig 1 , illumination 2 at wavelengths where the sample 4 is strongly absorbing (e.g. < 900 nm for 100-200 μm thick silicon wafers) will be transmitted by an open crack 6 but absorbed elsewhere in the wafer, so that open cracks will appear as bright features to a detector 8 such as a conventional silicon-based camera. Partially open or closed cracks 9 however cannot be detected reliably. A partially open wedge-shaped crack 10 on the other hand will scatter light differently from non-cracked regions, and can be detected by a contrast change in the transmission of moderately absorbed light 12 that has a penetration depth approximately equal to the wafer thickness. Light that is strongly absorbed will not penetrate to the camera 14 at all, while light that is weakly if at all absorbed will provide insufficient contrast at the camera. The methods of the present invention use light that is sufficiently weakly absorbed within a semiconductor material to allow it to travel distances of several millimetres (or even centimetres) within the sample before its intensity would be below the detection limit of a suitable detector. In silicon a suitable wavelength range is 1150-1700 nm, which is very weakly absorbed at room temperature and within the sensitivity range of near- infrared cameras such as InGaAs cameras or silicon cameras in combination with an InGaAs photocathode or some other IR-enhancing component. Fig 2 shows the absorption length of photons in silicon at room temperature showing that for wavelengths > 1180 nm the penetration depth is greater than 10 cm. For the purposes of this specification, we will refer to suitably weakly absorbed light as 'long wavelength light' irrespective of its actual wavelength range, which will of course depend on the sample material. The inventive methods rely on this long wavelength light being guided laterally within the sample over distances of millimetres to centimetres before escaping and being detected, and on the disruption of this lateral guidance by a crack or some other discontinuity, such as a large inclusion, which creates an intensity differential in an image of light escaping from the sample. The refractive index of silicon is ~3.5 in the near infrared region, creating a large index difference with respect to the air gap within a crack, which will be present even in 'closed' cracks that are difficult to detect with conventional optical reflection or transmission techniques. Even if cracks are filled with a dielectric material such as silicon nitride (n ~1.8-2.2 depending on composition), commonly used as a passivation and antireflection layer in silicon-based photovoltaic cells, the refractive index difference will still be sufficiently large to influence the light travelling laterally through the wafer.

In one example illustrated in Fig 3A, long wavelength light 16 is guided laterally in a silicon wafer 4 with smooth surfaces 18 until it encounters a crack 6 that reflects a portion of the light out of the wafer. Assuming the crack is oriented such that the reflected light 20 is in the field of view of a suitable camera 22, the crack will appear bright against a dark background as shown in the intensity versus position line scan 23 shown

schematically in Fig 3B. This picture is, however, not representative for most silicon wafers used in photovoltaic applications. As shown in Fig 4A, silicon photovoltaic cell wafers 24 typically have textured surfaces 26 to trap incident solar radiation within the wafer, increasing the likelihood of it being absorbed rather than reflected or transmitted. In this situation, portions 28 of the guided long wavelength light 16 may be scattered out of the wafer at each encounter with a textured surface in a more or less random fashion. Lateral propagation of long wavelength light will therefore be a sequence of scattering events with portions 28 coupled out of the wafer at each bounce, causing a gradual loss of intensity. If the long wavelength light propagates from the left as shown in Fig 4A, an imaging camera 22 may see a gradually decaying lateral intensity profile 29 across the region 30 to the left of a crack 6, as in the intensity line scans 23 shown schematically in Figs 4B and 4C. Propagation of the long wavelength light is interrupted by the crack, so that the region 32 to the right of the crack will appear dark to the imaging camera, resulting in a sharp intensity differential 33. Similar to the situation shown in Fig 3A, light 20 reflected by an appropriately angled crack may reach the camera, in which case the crack will appear as a bright linear or curvilinear feature 31 in a darker background as shown in Fig 4B. In other circumstances, e.g. in a sample with a textured surface that efficiently scatters light out of the sample, and/or with light propagating in a variety of directions, a crack may appear as a dark linear or curvilinear feature in a brighter background. Intensity differentials may be enhanced if a large fraction 34 of the long wavelength light is reflected by the crack back into the wafer, resulting in higher photon density and higher intensity of light scattered out of the wafer in the region to the left of the crack. We note that even if a crack does not extend completely through a wafer, it will still disrupt the propagation of long wavelength light.

The samples depicted in Figs 3A and 4A are planar silicon wafers with parallel surfaces, which are well-suited for lateral guidance of light. However it will be appreciated that lateral guidance can also occur in non-planar samples, e.g. bowed wafers or samples with monotonically decreasing thickness.

The question now arises as to how the long wavelength light can be made to propagate laterally inside the sample. In preferred embodiments it is launched or coupled into the sample from an external light source, while in alternative embodiments it is generated as luminescence within the sample. The first approach allows the wavelength range to be chosen according to available light sources, whereas the second approach is limited to the wavelength range of the luminescence that can be generated from the sample material. Preferably this luminescence is photoluminescence, i.e. luminescence generated by external illumination with a source of above band gap light such as, for silicon, an 805 nm laser or LED array, or a flash lamp. In alternative embodiments, applicable to finished or partially completed cells with electrical contacts, the luminescence is electroluminescence, i.e. luminescence generated by applying a forward bias to the sample. At room temperature, spontaneous emission via band-to-band recombination in crystalline silicon has its peak at about 1140 nm, and contains significant contributions up to about 1250 nm. Luminescence measurements with a suitable long pass filter (e.g. an 1180 nm LP filter) in front of the detector are thus suitable to exploit the long wavelength scattering effects.

Considering firstly a wafer with smooth parallel surfaces 18 as shown in Fig 3A, it is difficult to launch a significant amount of laterally propagating light quickly and efficiently into such a wafer. Elementary waveguide theory shows that light impinging on a smooth wafer surface can never be within the acceptance angle of the slab waveguide formed by parallel surfaces, and will tend to be reflected from or pass straight through the wafer. Significant amounts of light can be 'end- fired' through an edge of the wafer or coupled in through the surface via a prism, but these techniques require careful alignment or direct contact with the sample and are not preferred for rapid measurements such as for in-line characterisation of wafers in a photovoltaic cell line. With textured wafers however, advantage can be made of the textured surface to couple long wavelength light into the wafer. As shown in Fig 5, a significant amount of incident long wavelength light 35 will be scattered or refracted by a textured surface 26 and captured by the wafer 24, within which it can propagate laterally as represented by the arrows 16. A similar effect can occur at the rough surface of an as-cut silicon wafer.

Since measurements rely on lateral light propagation, the illumination is preferably non- homogeneous across the sample, i.e. localised, with the camera detecting light scattered out of the sample from surface texture, cracks or inclusions in non-illuminated regions. Many combinations of illumination intensity profiles and image acquisition configurations can be used, such as a periodic checkerboard pattern or a line pattern in combination with line scan cameras or area scan cameras. The common aspect is that one or more images of a sample surface will be generated with one or more non-homogeneous illumination patterns, with cracks identified by a contrast in the amount of light escaping from the sample at the site of a crack or in the vicinity of a crack. In certain embodiments the images are analysed manually, e.g. by an operator looking for intensity differentials that may indicate the presence of cracks in the sample. In other embodiments the image analysis is automated, using image processing algorithms for pattern recognition performed on single images or on combinations of more than one image, e.g. image comparisons, ratios or differences. Image processing could also occur by fitting a curve/model to a lateral light decay profile and thereby detecting a crack on the basis of the camera image showing large deviations relative to the fitted curve/model. Automated image analysis is preferred for high speed applications such as the in-line inspection of wafers or photovoltaic cells on a photovoltaic cell or module process line.

In the described embodiments the illumination and detection are from the same surface of the sample, which is advantageous since it allows the technique to be applied to finished or partially processed photovoltaic cells with a fully metallised rear surface. However the invention is not limited to this specific embodiment, as illumination and detection from opposite surfaces is also possible and may be applied to partially processed cells. This geometry has the advantage that light reflecting off the incident surface cannot contribute to the signal. Special measures to be described below with reference to Fig 7 may be required to mitigate the influence of reflected light when detection and illumination are from the same surface.

A schematic of a crack detection system according to one embodiment of the invention is shown in Figs 6A (plan view) and 6B (side view). A textured silicon wafer 24 carried on transport belts 36 passes under an illumination unit 48 that illuminates a substantially linear portion 40 of the sample with light in the 1150-1700 nm spectral range. The illumination unit comprises a linear light source 38, which may for example be an array of LEDs, a laser, an array of lasers, a laser diode, an array of laser diodes, arc lamps or flash lamps, in combination with filters to select the required wavelength range if necessary, and a housing (baffle) 42 that prevents light escaping from the top and sides.

Alternatively the light source(s) could be located remotely, with the light guided to the housing by an optical fibre bundle. One or more line cameras 44 (not shown in Fig 6A) generate images of the sample by repeatedly capturing line images of regions 46 as the wafer is moved along on the transport belts. The imaged regions are preferably parallel to the illumination line 40 at a distance R. The precise magnitude of R is not particularly important, although it should be large enough to reduce reflected light issues (discussed further below) without exceeding the absorption length of the long wavelength light. Optionally, a reflective sheet (not shown) may be located adjacent the rear surface of the wafer to reflect escaped light back into the wafer; this would not be required if the rear surface were metallised.

In one preferred embodiment the system includes two line cameras 44 and one

illumination unit 48 as illustrated in Figs 6A and 6B, with the line cameras imaging regions 46 in front of and behind the illumination line 40 with respect to the direction of motion 54 so that the entire surface of the wafer 24 can be inspected for cracks. Another preferred embodiment uses the reverse configuration, i.e. two illumination units illuminating lines either side of a region imaged by a single line camera. It will be appreciated that a system including only one camera and one illumination unit will be unable to detect cracks in a portion of the wafer extending a distance R from the leading or trailing edge.

It is preferable that the light detected by the line scan camera(s) 44 is largely light that has entered the wafer and been guided laterally within the wafer before escaping at the region(s) 46 imaged by the camera(s). One possible structure of an illumination unit 48 is shown in side view in Fig 7, with a baffle 42 having light trapping sections 50 with highly absorbing surfaces positioned on either side of a linear light source 38 to trap 'first bounce' light, i.e. light reflected off the incident surface (line 52). Since the wafer is in motion as indicated by the arrow 54, there needs to be a gap between the illumination unit and the wafer. This gap should be as small as possible to minimise the chance of light (line 56) reflecting off the textured surface 26 at a large angle to reach the camera, although realistically the proximity is limited by fluctuations in wafer thickness, the extent of wafer bowing, and the flatness or stability of the belt system or other wafer moving mechanism. Even if the wafer is stationary, there will preferably be a gap between the illumination unit and the wafer to prevent mechanical damage. The amount of reflected light escaping can be further reduced if necessary by attaching a smooth and flexible absorbing material 58 (such as black silk) to the underside edges of the illumination unit (shown schematically only for the left hand side). This material would glide smoothly over the surface as the wafer moves on the belt, without scratching or otherwise damaging the wafer. Alternatively, a freely rotating roll with a soft surface could be pressed lightly against the wafer. In another method for reducing the impact of reflected light 56, the light from the source can be polarised either intrinsically or with a polariser. The reflected light will remain largely polarised and can be blocked with a cross polariser placed in front of the camera, whereas guided and scattered light 62 will have its polarisation scrambled. As illustrated in Fig 7, the illumination unit may also include a diffuser 60 to homogenise the intensity along the illuminated line. Note that the surfaces of the wafer 24 are textured not smooth, as indicated by the irregular reflection angles of each illustrated path including the desired light path 62. A schematic of a crack detection system according to another embodiment is shown in side view in Fig 8. This system, suitable for stationary samples, comprises one or more long wavelength light sources 38, preferably housed within a baffle 42, positioned proximate to the textured surface 26 of a silicon wafer 24, and an area camera 22 imaging a substantial portion of the wafer surface. Many variations are possible with the shape and location of the light source; for example a point source could be positioned close to a corner or the centre of the wafer, or a linear source could be positioned at an edge of the wafer. More than one image could be acquired, with the one or more light sources positioned proximate different locations on the wafer surface, to maximise the likelihood of detecting a crack.

Fig 9A shows a photoluminescence (PL) image of a fragment 76 of a multicrystalline silicon photovoltaic cell, where the luminescence is generated by illuminating the fragment with an intensity of ~ 1 Sun from a near IR laser and detected with a silicon CCD camera as described in detail in published PCT patent application No WO

07/041758 Al , the contents of which are incorporated herein by reference. One Sun is defined herein as 100 mW/cm². The PL image shows a number of low carrier lifetime (and therefore low PL intensity) features, including dislocation-rich regions 78 and a curvilinear stripe 80 surrounding a crack. Note that the low intensity stripe is

considerably wider than the crack, which is not itself discernible, because carrier recombination at the crack surfaces acts as a sink drawing charge carriers in from the surrounding material. The metallisation pattern also appears dark in the PL image, because the metal lines block both the illumination and the luminescence.

Fig 9B shows an image of visible light reflected from the cell fragment 76, acquired with a silicon CCD camera, in which the metallisation pattern appears bright. The reflection image also reveals some grain structure in the silicon, but does not reveal the crack. The shadow in the middle of the cell fragment is cast by a 1300 nm LED 82 sitting on the sample surface, but not activated. Fig 9C shows an image of the cell fragment 76 acquired with an InGaAs camera when the 1300 nm LED is activated. No attempt was made to prevent reflected light from reaching the camera, as evidenced by the bright arc 84 surrounding the LED 82. It can be seen however from the bright edge 86 that 1300 nm light has been launched into and guided within the fragment, and the presence of the crack seen in the PL image (Fig 9A) is clearly revealed by an intensity differential in the form of bright signal 88 and the absence of any signal from the area 90 beyond the crack.

Comparison between the Fig 9C image and either of the images shown in Figs 9A and 9B reinforces the conclusion that the bright signal 88 is caused by a crack, not by an edge of the sample.

As described above with reference to Fig 4A, the basis for the methods of the present invention is that cracks intercept the path of light scattered laterally within the sample. It is possible then that linear cracks that happen to extend substantially parallel to the direction of lateral light propagation may cause little or no contrast in light scattered out of the wafer. This potential problem may in fact be relatively insignificant because many cracks, particularly 'mechanical' cracks, are non-linear (for example mechanically- induced cracks in monocrystalline silicon are generally cross-shaped because of the crystallographic axes), and because the in-coupled long wavelength light will generally propagate in a variety of directions. Nevertheless the sensitivity of the inventive method to cracks with various orientations can be enhanced by applying the previously described procedures two or more times with the wafer held at different orientations. For example the wafer may be rotated by 90 degrees with respect to the light source and camera orientation and the procedure repeated with the same measurement system, or the wafer may be rotated en route between two measurement systems. In an alternative embodiment that avoids having to rotate the wafer (involving handling and therefore increasing the risk of damage), two measurement systems (each with one or more line sources and line cameras) at different orientations image the wafer sequentially. For example Fig 10 shows a configuration with two measurement systems 64, each comprising an illumination unit 48 and two line cameras (not shown) imaging the regions 46, oriented at +45 degrees and -45 degrees relative to the movement of a wafer 24 on transport belts 36. This configuration avoids having to rotate the wafer, and increases the likelihood that cracks with different orientation will be revealed in images acquired from the regions 46 by one or other of the measurement systems. A minor disadvantage compared to the

configuration shown in Fig 6A is that for a given number of pixels in each camera, the spatial resolution of the images will be reduced by \2 because of the diagonal orientation of the cameras. The configuration shown in Fig 10 is merely exemplary, with many other configurations possible. For example three or more sequential measurement systems with different orientations can further increase the likelihood of identifying cracks of all orientations.

For systems with area cameras instead of line cameras, as shown in Fig 8 for example, the problem of identifying cracks with all orientations can be addressed by comparing two or more images taken with different illumination patterns that will cause long wavelength light to propagate through the sample in different directions.

As described previously with reference to Fig 4A, a crack 6 will intercept the path of long wavelength light 16 propagating inside an illuminated wafer 24, resulting in a distinct intensity differential in the image at the position of the crack, with higher intensity of out- coupled light 28 in the area 30 on the illumination side of the crack. While the presence of a crack can determined from a single sharp contrast feature as shown in Fig 9C, the reliability of crack detection can be increased by acquiring one or more images with light propagating laterally in different directions. For example two images could be acquired with cameras 44 on either side of an illumination source 38 as shown in Figs 6A and 10, or a single image could be acquired with long wavelength light injected from two illumination sources at opposite ends of a wafer. In particular, the crack-induced contrast between bright and dark areas is expected to be reversed for light propagating in opposite directions. As illustrated by the two schematic images of out-scattered long wavelength light shown in Figs 1 IA and 1 IB, the position of bright and dark areas 30 and 32 on either side of a crack 6 depends on the direction of long wavelength light propagation 16. This contrast reversal can be analysed by an operator or with specific image processing algorithms, and in short features that appear with opposite contrast in the two images can be identified as cracks with greater confidence.

Furthermore, depending on the resolution of the imaging camera(s), the acquisition of images with long wavelength light propagating in different directions offers a means for determining the width of a crack. For example with light propagating from the left as shown in Fig 1 IA the bright area 30 will end at the left hand edge of the crack 6, whereas with light propagating from the right as shown in Fig 1 IB the bright area 30 will end at the right hand edge of the crack. Alternatively, if a crack deflects long wavelength light into an imaging camera to produce a bright feature 31 as shown in Fig 4B, the leading edge of the bright feature may correspond to an edge of the crack.

In applying the inventive methods to surface textured or as-cut wafers, a distinction may need to be made between wafers with different types of surface texture or roughness. For samples with a substantially homogeneous surface roughness or texture, including as-cut monocrystalline or multicrystalline silicon wafers, monocrystalline silicon wafers after acid or alkaline etching, or multicrystalline silicon wafers after acid etching, long wavelength light will be scattered out of crack-free areas in a more or less statistical manner, resulting in smooth variation of the escaping light intensity across the sample. For alkaline etched multicrystalline silicon wafers however, grain- specific features are expected in images of the scattered long wavelength light. This is because the texturing in an alkaline etch depends on the crystal orientation of the grains, resulting in a patchy surface where highly textured grains have a rough appearance and weakly textured grains appear shiny. This is illustrated schematically in Fig 12, which shows an alkaline-textured multicrystalline wafer 66 with a weakly textured (and therefore smooth and shiny) grain 68 surrounded by a strongly textured (and therefore rough) grain 70. The effect is also seen in the visible light image shown in Fig 9B, where different grains have differing reflectivities. Comparing Figs 3A and 4A, a smoothly surfaced grain will tend to confine the long wavelength light laterally with minimal scattering, so will appear dark in an image of the out-scattered light, and conversely a strongly textured grain will appear bright. Consequently a long wavelength image will show an intensity differential at the grain boundary 72 that may be mistaken for a crack.

Several different approaches can be used to distinguish between intensity differentials caused by texture variations and those caused by cracks. Firstly, because cracks are usually linear, curvilinear or branched features with two or more ends, contrast features or intensity differentials that appear as closed loops (such as the rectangle 72 in Fig 12) are likely to be due to grains rather than cracks. Secondly, one can use two or more cameras to capture images with illumination from different sides of a contrast feature. An intensity line scan across the smooth grain 68 shown in Fig 12 will show a bright-dark-bright pattern irrespective of the propagation direction of the long wavelength light, whereas an intensity line scan across the crack 6 shown in Figs HA and HB will appear as dark- bright or bright-dark depending on the propagation direction. The method as described above where two line cameras are used for each orientation of the sample may therefore be particularly useful for alkaline-textured multicrystalline cells and wafers. Finally, as demonstrated by Fig 9B, other inspection methods such as optical imaging (reflection or transmission) can detect the grain structure in a multicrystalline silicon sample, allowing correspondingly located contrast features in the long wavelength images to be identified as grain boundaries rather than cracks. In alkaline-textured multicrystalline samples the situation is further complicated by the fact that the texture variations affect not only the lateral transport of long wavelength light through the wafer and the fraction coupled out of the wafer and into the imaging camera, but also the fraction of light coupled into the wafer from the illumination source. Since illumination and imaging of a specific area occur at different wafer/belt positions, a given grain will influence an image twice, firstly at the position of the light source, and secondly in the area imaging by the camera. For example a smoothly textured grain will reduce the amount of light coupled both in and out of the wafer, causing the grain structure to influence the image twice in a convoluted manner. Since the same effect does not occur for cracks, suitable deconvolution techniques may assist in distinguishing cracks from boundaries between differently textured grains.

As an alternative to in-coupling from an external light source, long wavelength light can be generated directly inside a sample by either photoluminescence or electroluminescence as mentioned briefly above. This generation technique is essentially independent of the surface quality of the sample, and can be applied to samples with either smooth or textured surfaces. Photoluminescence has the advantage of being contact- less and can therefore be applied to partially processed wafers, whereas electroluminescence requires electrical contacts. Photoluminescence has the additional benefit that the lateral variation of the luminescence generation can be freely controlled by the lateral illumination profile, whereas in electroluminescence the lateral generation profile is largely determined by the positions of the metal contacts of the device/sample under test. Compared to the above described in-coupling method, photoluminescence also has the advantage that the excitation light is at shorter wavelengths than the luminescence used for detection, and can therefore be filtered out at or before the camera/detector. This relaxes any requirement for an essentially light-tight housing/baffle for the illumination source as shown in Figs 6A, 6B and 7, but a higher sensitivity detector is required to measure the relatively weak luminescence signal, particularly for indirect band gap materials such as silicon.

In the photoluminescence method the short wavelength excitation light is absorbed inside the sample within a length typically less than the sample thickness, and the laterally guided long wavelength luminescence detected with a suitably sensitive camera. Area or line scan cameras can be used to detect cracks via light scattering in much the same way as described above for the in-coupling method. For the specific case of a silicon wafer, the excitation source could be in the wavelength range 500-1000 nm, more preferably 700-900 nm, and the detection would be limited to the spectral range >1100 nm to provide a relatively long propagation length within the wafer, to enhance the sensitivity to cracks.

With automated image analysis, the above described crack detection methods can be performed on a timescale compatible with the throughput of current photovoltaic cell manufacturing lines, of order one sample every second or every few seconds. In certain embodiments, as described above with reference to Figs 6A and 10 for example, they are performed in-line with the sample moving on a belt or other continuously moving support, and images captured with one or more line cameras. In alternative

embodiments, images can be captured with one or more area scan cameras either with the sample temporarily stationary, or with the sample in motion and the illumination and/or detection optics moving synchronously with it, or with the illumination provided from a pulsed source such as a flash lamp to prevent blurring.

The crack detection methods of the present invention can be performed off-line as an R&D application, or in-line on a large number of selected samples or even on every sample, by several industries including wafer manufacturers, cell manufacturers and module manufacturers. Wafer manufacturers could use the methods to identify wafers with cracks, with the results used for wafer rejection or quality binning, and to provide feedback to the wafer production process. Photovoltaic cell manufacturers could use the methods for incoming wafer inspection and subsequent binning or rejection, or for crack detection in partially processed wafers for wafer quality binning or rejection, for process monitoring and control, or for crack detection in finished cells resulting in quality binning or rejection. Module manufacturers could use the methods for incoming cell inspection and subsequent rejection or quality binning, or to monitor cell handling and to identify cracks in cells that have been tabbed or integrated into strings or modules. In all three cases the information about cracks obtained from the inventive methods can be used for process monitoring and process control, e.g. repair or modification of unsuitable processing or handling equipment that introduces or expands cracks. Each industry is likely to be interested in different sets of information on cracks, but in general wafer manufacturers, cell manufacturers and module manufacturers will all be interested in parameters or factors such as crack length and width, the orientation, shape and position of a crack, and whether a given crack is 'natural' (i.e. thermally induced) or mechanically induced, with the latter factor giving guidance as to the origin of a crack. Shape is perhaps the single best indicator of whether a given crack is natural or mechanically induced, since natural cracks tend to be linear while mechanical cracks tend to be three-armed in multicrystalline silicon and cross-shaped in monocrystalline silicon. Crack position is an important factor, recalling that cracks close to a wafer or cell edge are more likely to grow and cause breakage, while crack orientation can also be useful information since a crack extending predominantly parallel to a bus bar is more likely to isolate part of a cell electrically than a crack extending predominantly normal to a bus bar. Information on crack size, i.e. length and/or width, is important if a cell or module manufacturer wishes to monitor the growth or origin of cracks in wafers or cells as they pass through the various stages of a cell or module line, similar to the photoluminescence imaging-based methods described in published PCT patent application No WO

09/026661 Al. Information on crack size may also be of value simply as an indicator of the seriousness of a crack; for example crack theory predicts that cracks under a certain size act simply as stress relief sites, whereas cracks over a certain size are likely to grow if the sample is subject to differential stress.

These and other parameters or factors can be calculated and reported to a human operator or a process line decision module, who/which will decide what to do with the information. There may be a set of user-definable criteria, including for example crack length and position, that will determine whether a detected crack is serious enough for the sample to be discarded or diverted to a lower efficiency cell line.

The long wavelength lateral scattering methods of the present invention are suitable for obtaining information on the length, shape, position and orientation of cracks, and perhaps also width information as described above with reference to Figs HA andllB, depending on the resolution of the imaging device. Photoluminescence imaging is already known to be of use for crack detection, as evidenced by the PL image in Fig 9A, and may provide information on several crack parameters including crack width. It is important to note however that the stripe 80 in the Fig 9A image represents the carrier depletion region along each side of the crack and not the crack itself, and its width is therefore more indicative of carrier transport properties than crack width. PL images acquired with much higher illumination intensities than the ~1 Sun used for the Fig 9 A image provide a better indicator of crack width, since lateral carrier transport is suppressed, particularly in emitter layers, at high injection level conditions (i.e. high carrier densities) produced by intense illumination. To demonstrate, Figs 13A and 13B show PL images of a monocrystalline wafer with several intentionally produced and characteristically cross-shaped cracks 6, where the PL was generated with ~1 Sun and -100 Suns illumination intensities from a near IR laser and a flash lamp respectively and captured with a silicon CCD camera. The cracks are revealed much more sharply in the high illumination PL image, and the associated carrier depletion features are more likely to provide a measure indicative of the actual widths of the cracks themselves.

Illumination intensities greater than 100 Suns are expected to reveal cracks even more sharply. Another method for enhancing the contrast of cracks in PL imaging is to hold the sample at elevated temperature. For example in H. Nagel et al 'Luminescence imaging - a key metrology for crystalline silicon PV, 20th Workshop on Crystalline Silicon Solar Cells and Modules: Materials and Processes, August 1-4, 2010, Breckenridge, USA it was shown that carrier lifetime-related contrast features such as dislocations and grain boundaries were suppressed in a PL image of a multicrystalline wafer held at 220⁰C, such that cracks became much more clearly visible. This effect, which is also expected to occur with electroluminescence imaging, is believed to be due to the temperature dependence of the Shockley-Read-Hall recombination lifetime, which causes the bulk lifetime of defected areas to increase with increasing temperature. The above described temperature effect could find practical implementation in a PV cell manufacturing line by acquiring PL images of wafers at positions in the line where they are already at elevated temperature, e.g. after emitter diffusion or after firing of metal contacts. Of these two possibilities it is probably simpler to acquire PL images after firing since the wafers go through the process tool linearly whereas diffusion is either a batch process or processes several wafers in parallel. Since the samples would be cooling, i.e. not held at constant temperature, it would be advantageous to acquire PL images rapidly, e.g. with flash lamp excitation, at one or more positions chosen for optimal or maximum temperature effect. The high illumination intensity of a flash lamp would, of course, further increase the crack contrast.

In general, while the lateral scattering methods of the present invention can detect cracks in semiconductor wafers and provide a significant amount of information about those cracks, the methods are likely to be more powerful when combined with other imaging techniques, such as optical imaging and PL imaging, and in particular PL imaging with high intensity illumination and/or elevated sample temperature. In one example, PL imaging is more likely than the lateral scattering technique to provide complete information on crack shape, especially for multi-armed cracks where the long wavelength light may not reach all of the arms. In another example, comparison between the optical reflection and long wavelength light images of Figs 9B and 9C is informative because of the disappearance of the area 90 in Fig 9C.

Although the present invention has been described with particular reference to certain preferred embodiments thereof, variations and modifications of the present invention can be effected within the spirit and scope of the following claims.

Claims

Claims:

1. A method for detecting a discontinuity in a semiconductor material, said method comprising the steps of:

a. generating and guiding light through said material;

b. acquiring an image of said material, wherein said image comprises light scattered or transmitted from said material; and

c. identifying a light intensity differential in said image thereby detecting said discontinuity.

2. A method according to claim 1 wherein said semiconductor material is

substantially planar.

3. A method according to claim 1 or claim 2, wherein said discontinuity is a crack or an inclusion.

4. A method according to any one of the preceding claims, wherein said light

intensity differential is a bright linear or curvilinear feature in a relatively darker background of said image, or a dark linear or curvilinear feature in a relatively brighter background of said image.

5. A method according to any one of claims 1 to 3, wherein said light intensity

differential is an abrupt increase or decrease in the intensity of light scattered or transmitted out of said material.

6. A method according to any one of claims 2 to 4, wherein step (b) comprises

acquiring two or more images, wherein said light is generated and guided so as to propagate laterally through said substantially planar semiconductor material in two or more directions.

7. A method according to claim 6, wherein step (c) comprises monitoring variations in said light intensity differential between said two or more images.

8. A method according to any one of the preceding claims, wherein step (a)

comprises coupling said light into said material at one or more locations.

9. A method according to claim 8, wherein said light is coupled into said material through a textured surface of said material.

10. A method according to any one of claims 1 to 7, wherein step (a) comprises illuminating a surface of said material at one or more locations with above band gap radiation, to generate said light as photoluminescence.

11. A method according to any one of claims 8 to 10, wherein said one or more

locations are substantially linear in shape and step (b) comprises acquiring, with one or more line cameras, images of said material when said material is in relative motion with respect to said one or more line cameras.

12. A method according to claim 11 , wherein said one or more line cameras are

adapted to capture light from areas that are substantially parallel to said locations.

13. A method according to claim 11 or claim 12, wherein step (b) comprises acquiring, with two or more line cameras adapted to capture light from areas on either side of each of said locations, images of said material.

14. A method according to any one of claims 1 to 7, wherein step (a) comprises

applying a forward bias to said material, to generate said light as

electroluminescence.

15. A method according to any one of the preceding claims, further comprising the steps of:

(d) acquiring an optical image of said material; and

(e) comparing said optical image with the image acquired in step (b).

16. A method according to any one of the previous claims, further comprising the steps of:

(f) acquiring a photoluminescence image of said material; and

(g) comparing said photoluminescence image with the image acquired in step (b).

17. A method according to claim 16, wherein the photoluminescence in said

photoluminescence image is generated with an illumination intensity of order 100 Suns or higher.

18. A method according to claim 16 or claim 17, wherein said photoluminescence image is acquired when said material is at elevated temperature following a high temperature processing step.

19. A method according to claim 18, wherein said high temperature processing step is selected from the group comprising emitter diffusion and metal contact firing.

20. A method according to any one of the previous claims, further comprising the step of:

(h) calculating, for discontinuities detected in step (c), one or more parameters selected from the group consisting of length, width, position and shape.

21. A method according to any one of the previous claims, when applied to a thin film, wafer, or photovoltaic cell composed of said semiconductor material.

22. A method according to claim 21 , when applied to a wafer or photovoltaic cell composed of multicrystalline or monocrystalline silicon.

23. A system for detecting a discontinuity in a semiconductor material, said system comprising:

i. an optical source adapted to couple light into said material at one or more locations such that said light is guided within said material; and ii. an imaging device sensitive to said light, for acquiring one or more images of said material, whereby said discontinuity is detected on the basis of a light intensity differential in said one or more images.

24. A system according to claim 23, wherein said optical source is adapted to couple light into said material through a textured surface of said material.

25. A system for detecting a discontinuity in a semiconductor material, said system comprising:

i. a source of above band gap radiation adapted to illuminate a surface of said material to generate photoluminescence, at least a portion of which is trapped as light guided within said material; and

26. A system according to any one of claims 23 to 25, wherein said imaging device comprises a line camera for acquiring images of said material as said material is moved relative to said line camera.

27. A system according to any one of claims 23 to 26, wherein said imaging device comprises two line cameras for acquiring images of said material, adapted to collect light from areas of said material on either side of a location where light is launched into or generated within said material.

28. A system according to claim 26, wherein said line camera is adapted to collect light from an area of said material between two locations where light is launched into or generated within said material.

29. A system for detecting a discontinuity in a semiconductor material, said system comprising:

30. A system according to any one of claims 23 to 29, further comprising:

iii. a source of visible light for illuminating said material; and

iv. a first camera for acquiring an optical image of said material.

31. A system according to any one of claims 23 to 30, further comprising:

v. a source of above band gap radiation for illuminating said material to

generate photoluminescence from said material; and

vi. a second camera for acquiring a photoluminescence image of said material.

32. A system according to claim 31 wherein said source of above band gap radiation is adapted to illuminate said material with an intensity of the order of 100 Suns or higher.

33. A system according to claim 32, wherein said source of above band gap radiation is a flash lamp.

34. A system according to any one of claims 23 to 29, further comprising a processor adapted to analyse said one or more images for a light intensity differential and to detect, based on said light intensity differential, the presence of said discontinuity in said material.

35. A system according to claim 34, wherein said processor is further adapted to analyse said one or more images for variations in said light intensity differential when said light is generated so as to propagate in different directions and to detect, based on said variations, the presence of said discontinuity in said material.

36. A system according to claim 34 or claim 35, wherein said discontinuity is a crack, and said processor is further adapted to calculate, for said detected crack, one or more parameters selected from the group consisting of length, width, position and shape of said crack.

37. A system according to any one of claims 23 to 36, wherein said semiconductor material is a thin film, wafer, or photovoltaic cell.

38. A system according to claim 37, wherein said semiconductor material is a wafer or photovoltaic cell comprising multicrystalline or monocrystalline silicon.

39. An article of manufacture comprising a computer usable medium having a

computer readable program code configured to conduct the method of any one of claims 1 to 22 or operate the system of any one of claims 23 to 38.