Field of the Invention

The present invention relates to illumination systems, and methods using these systems, for the characterisation of semiconductor materials using photoluminescence imaging. The illumination systems have particular application to the characterisation of silicon-based photovoltaic cells and cell precursors.

Related Applications

The present application claims priority from Australian provisional patent application Nos 2010900018, 2010903050 and 2010903975, the contents of which are incorporated herein by reference.

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.

The semiconductor industry has for decades used photoluminescence (PL), the generation of luminescence with above band gap excitation, as a non-destructive method for investigating direct band gap semiconductor materials, especially for the presence of defects. Generally, the PL has been generated using a laser focused onto a small area of the sample, and to investigate a large area the laser beam or sample is raster scanned to generate a map of the PL emission. Focused beam excitation has traditionally been used because of a desire for high spatial resolution, e.g. when mapping defect distributions, and because the associated high intensity generates a stronger PL response. AT&T Bell Labs have considered the possibility of PL imaging of direct band gap semiconductors using broad area illumination (G. Livescu et al, Journal of Electronic Materials 19(9) 937-942 (1990)), but despite the advantage of rapid measurement apparently concluded it was inferior to scanning systems because of problems with non-uniform illumination and low sensitivity (G.E. Carver, Semiconductor Science and Technology 7, A53-A58 (1992)). In any event PL has usually been used to investigate high value semiconductor samples during or as part of the manufacture of computer chips and the like, where rapid measurement is not essential.

To a lesser extent, PL mapping with focused beam excitation has also been used to characterise indirect band gap semiconductor samples, as disclosed for example in published PCT patent application No WO 98/11425 Al. As with direct band gap semiconductors, the choice of focused beam illumination has been motivated for reasons of higher spatial resolution and greater PL response, with the latter being particularly significant for indirect band gap materials because of their much lower radiative quantum efficiency. In the photovoltaic (PV) cell industry, dominated by silicon-based cells, the throughput is currently of order 1-2 seconds per wafer, so that measurement speed is critical for in-line inspection. Even in off-line sampling the sheer number of wafers handled by PV cell production facilities necessitates fast measurement, making broad area PL imaging far more attractive than PL scanning/mapping. In 2005 the founders of the present applicant demonstrated that it was in fact possible to inspect silicon-based PV cells and wafers with broad area PL imaging despite the low radiative quantum efficiency, and disclosed suitable methods and systems in published PCT patent application No WO 07/041758 Al entitled 'Method and system for inspecting indirect bandgap semiconductor structure', the contents of which are incorporated herein by reference. Most if not all commercially available PL imaging systems for silicon- based photovoltaics utilise laser illumination because they provide the required illumination intensity over a narrow wavelength band, enabling rejection of stray illumination at the camera with commercially available filters, and the laser beam can be readily expanded and homogenised for the purposes of illumination uniformity across an entire cell, typically 15.6 x 15.6 cm .

There are however a number of drawbacks with the current laser-based PL imaging systems. Firstly there are significant light safety issues, especially if the excitation light is in the 750 to 1000 nm region (as is typical for silicon samples) which can be focused onto the retina with no protective blink response. Consequently the PL measurement chamber must be optically isolated, requiring shutters, doors or equivalent mechanisms, adding complexity and cost to the sample transfer mechanisms into and out of the chamber.

Secondly, current PL imaging systems require the sample to be stationary to prevent blurring of the PL image. This is because with broad area excitation the PL emitted from many silicon samples, and raw or unpassivated silicon samples in particular, can be of such low intensity that even the most sensitive silicon-based CCD cameras require an exposure time of order at least 1 second to acquire a sufficient PL signal. Although this 'stationary sample' requirement is acceptable for off-line applications, it is a complication for in-line applications and it would be advantageous, especially with fragile wafers or on large or fast production lines, to avoid any stop/starting of individual wafers.

Current systems are also limited in terms of their ability to differentiate between various material properties or defects, e.g. shunts, dislocations and cracks, in a semiconductor material or a PV cell produced from semiconductor material, or to identify the presence of certain defects amongst other features.

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 a preferred form of the present invention to provide methods and systems for acquiring photoluminescence images of semiconductor devices during their production process without interrupting the motion of the devices through the production line. It is another object of a preferred form of the present invention to provide methods and systems for acquiring photoluminescence images of semiconductor devices with reduced light safety requirements. It is another object of a preferred form of the present invention to provide a method for classifying features and defects in a semiconductor device by comparing photoluminescence images acquired with widely differing illumination intensities. It is another object of a preferred form of the present invention to provide a method for classifying features and defects in a semiconductor device with photoluminescence images acquired with high intensity illumination. Summary of the Invention

In accordance with the first aspect, the present invention provides a method of analysing a semiconductor material comprising applying illumination from a high intensity illumination source to the semiconductor material for a sufficient time and at sufficient intensity to induce photoluminescence, obtaining an image of said photoluminescence and analysing said image to determine the location and/or nature of features and defects in the semiconductor material.

In a preferred embodiment, the illumination source is an eye-safe light source, more preferably a non-laser light source. In another embodiment, the illumination is provided by a pulsed light source, LED or laser system. In a particularly preferred embodiment, the light source is a flash lamp. In another embodiment, the light source comprises one or more light emitting diodes (LED). In another embodiment a laser- based illumination system is developed, the overall output of which is eye-safe. As mentioned above, the laser-based PL imaging systems of the prior art have significant light safety issues. The potential hazard of laser light sources, and of illumination systems incorporating them, arises from the fact that they may be much brighter than other light sources, where the brightness (in units of power per unit area per unit solid angle) may be defined for example as the optical power passing through an aperture (e.g. a laser output aperture) divided by the aperture area divided by the solid angle subtended by the optical beam in the far field. When an extremely bright light source is viewed with the eye, either directly or via intermediate optics such as a collimating lens, the image formed on the retina can be extremely intense, resulting in virtually instantaneous and permanent damage. However although there is less likelihood of this occurring with near IR light from non-laser sources, e.g. from high power LEDs, and the regulatory requirements are less onerous, it needs to be understood that because brightness is a key parameter, light safety issues cannot simply be ignored just because a system uses non-laser light sources. Rather, the key factors are the intensity of light entering the pupil and the image size on the retina. With this in mind, it is preferred for a PL imaging system to use relatively low brightness, eye-safe illumination systems. Importantly, we will show that it is possible to provide the high intensity illumination required for silicon and other indirect bandgap materials with eye-safe illumination systems. Furthermore even if an illumination system itself is not eye-safe, it is possible for the PL imaging system as a whole to be eye-safe without resorting to stringent laser safety measures such as safety shutters and interlocks.

The use of high intensity illumination for generating photoluminescence provides significant advantages in terms of time and cost. For example, flash lamp equipment is generally cheaper than laser light sources and if applied correctly, can provide sufficient energy to produce a photoluminescence effect within a very short space of time, up to 100 milliseconds preferably up to 10 and most preferably a few milliseconds as compared with conventional laser based PL imaging systems.

The term 'high intensity' refers to an intensity of illumination which is high relative to that which would be used to obtain a conventional photoluminescence image with conventional sensor and laser technology. To explain, using a conventional silicon CCD camera, a photoluminescence (PL) image can be obtained with laser illumination at 1 or 2 Suns, where 1 Sun is considered to be equivalent to 100 mW/cm . The applicant has found that by applying illumination with flash lamps, pulsed lasers, LEDs etc at levels (e.g. 50 Suns or greater) that are one or two orders of magnitude higher, relative to the conventional intensities, the PL response of the wafer is quite different and the resulting PL image is quite different from a conventional PL image. This allows operators to determine other attributes of the semiconductor material which would not be determined by conventional techniques. With other systems e.g. using a different camera, 'high intensity' may mean illumination at the level of 6-8 Suns or greater. For instance, as will be disclosed below, for a PL imaging system with a hybrid flash illumination-Time Delay Integration (TDI) arrangement utilising silicon CCD cameras, relatively high intensity illumination of around 8-16 Suns is used compared with only 2 Suns using a silicon CCD camera in a standard 'stop sample and expose' image mode. Again, using alternative sensors such as a MOSIR (photocathode CCD based) camera the absolute intensity applied to the wafer may be lower than 50 Suns but it would still be regarded as a 'high intensity' when viewed in combination with the sensor. Accordingly the term 'high intensity' should not be regarded as an absolute value but rather regarded as 'high' relative to a conventional system where samples are stopped for measurement of the PL response. It will be clear to the person skilled in the art that this is affected by the measurement technique and sensor used.

In some respect the 'high intensity' can be considered as a ratio of the required illumination intensity relative to camera and sample sensitivity. For example, in a previous system 1 Sun illumination was used with a 1 second capture time for raw wafers with CCD cameras, whereas in the new system only a 1 millisecond capture time is required for about 10 Suns or more illumination for pre-diffusion wafers and with an InGaAs camera. Similarly, a 1 millisecond capture time can be used for a 100 Sun illumination for pre-diffusion wafers and with a Si CCD camera. When the relative sensitivity to the silicon PL emission is improved, for example by using an InGaAs camera, the required illumination intensity drops accordingly by a factor of 10 or more. This has an additional benefit of allowing a drop in the required time for the same intensity level. Preferably the associated optics are modified to remove stray light, incident light etc, i.e. optimised for each camera and light source combination. In other words, the illumination intensity is a function of the relative camera sensitivity to PL emission, preferably with optimised optical filtering.

By 'flash' illumination we refer to short duration intense illumination preferably provided by a flash lamp, pulsed LED etc. Typically this is between 50 and 1,000 Suns (5-100 Watts/cm ) preferably 75 to 200 Suns and most preferably around 100 Suns of illumination applied to the semiconductor material. This illumination is typically provided within fractions of a second, e.g. 100 milliseconds, preferably up to 10 milliseconds and most preferably within a few milliseconds. In some embodiments such illumination can be provided in 1 millisecond or less.

Illumination can also be provided by pulsed lasers which still provide high intensity illumination. In the interests of eye safety, such pulsed lasers can illuminate the sample with very short and therefore low energy pulses, e.g. less than 1 microsecond. Alternatively, longer pulses or even cw laser emission can be provided in a less hazardous form by mechanically agitating one of the optical components, e.g. mirrors in the optical system, to smear out any image formed on the retina, reducing the effective intensity at the retina, and therefore the thermal effects. In a second aspect, the present invention provides a method for producing a photoluminescence image of a semiconductor material, said method comprising the steps of: applying to said semiconductor material at least one pulse of illumination with an intensity of between 50 and 1,000 Suns-within 100 milliseconds; and capturing the photoluminescence response as an image.

In a preferred embodiment, the high intensity illumination produced by a flash lamp is applied to a moving semiconductor material. The rapid energy application and image capture system provides the opportunity for obtaining a PL image of the semiconductor material while still moving. Due to the length of time required to produce photoluminescence and capture the image, in a conventional system using a silicon based camera it is almost always necessary to stop the semiconductor material to avoid blurring of the image. Using the aforementioned flash system allows inducement of photoluminescence and capture of the photoluminescence effect as an image within fractions of a second, e.g. 10 or preferably 5 and most preferably up to a few milliseconds. We note that it may take considerably longer to read out the image, depending on the camera technology, but this does not affect the ability to acquire images with minimal blurring.

In a third aspect, the present invention provides a method for producing an image of a photoluminescence response in a semiconductor material, said method comprising the steps of: applying sufficient illumination to the semiconductor material to produce a photoluminescence response; and capturing an image of the photoluminescence response within 100 milliseconds with a silicon camera.

In a fourth aspect, the present invention provides a method for producing an image of a photoluminescence response in a silicon sample, said method comprising the steps of: applying sufficient illumination to the silicon sample to produce a photoluminescence response; and capturing an image of the photoluminescence response within 10 milliseconds with a camera that captures substantially all of the photoluminescence response of silicon. Preferably the photoluminescence response is captured with a camera that captures most or all of the PL emission spectrum which, subject to the type of sample, could be for example a silicon sensor based camera, a compound semiconductor sensor based camera or a compound semiconductor photocathode based camera. Preferably the semiconductor material is an indirect band gap semiconductor material such as silicon. The inventive method can be applied to such a semiconductor material in the form of an ingot, a block, a wafer or a complete or partially completed photovoltaic device.

As mentioned above, the high intensity illumination source can be selected from laser sources or incoherent sources such as flash lamps, LEDs etc. Irrespective of the nature of the source, the overall brightness of the illumination system is preferably sufficiently low or the exposure sufficiently short, e.g. up to a few milliseconds, to mitigate the need for light safety shielding.

To some extent the upper limit of the time for the above mentioned process depends upon belt speed. Current state of the art belt speeds are between 100mm per second and 200mm per second on a cell line and approximately twice that on a wafer line. To reduce or avoid 1-pixel blurring it is necessary either to stop or slow down the sample or, preferably, to expose the sample and capture the PL response as an image in less than a few milliseconds, preferably less than 1 millisecond, for an on-sample pixel size of 160μιη and a belt speed of 150mm per second. The preferred exposure time depends on line speed, pixel size and the acceptable level of blurring.

In a fifth aspect, the present invention provides a method of identifying defects or features in a semiconductor material, said method comprising the steps of: obtaining a first image of a photoluminescence response from said semiconductor material generated with a first, higher illumination intensity; obtaining a second image of a photoluminescence response from said semiconductor material generated with a second, lower illumination intensity; and comparing said first and second images.

In a sixth aspect, the present invention provides a method of differentiating defects or features in a semiconductor material, said method comprising the steps of: applying to the semiconductor material a predetermined level of high intensity illumination - Si - adapted to obtain a photoluminescence response characteristic of a predetermined defect or feature; capturing the resultant photoluminescence response; and analysing said response to determine the presence and/or location of such defects or features.

In a seventh aspect, the present invention provides a method of identifying defects or features in a semiconductor material, said method comprising the steps of: obtaining an image of a photoluminescence response from said semiconductor material generated with illumination having an intensity of at least 6 Suns; and processing said image to obtain useful information.

The applicant has determined a significant and surprising advantage of using high intensity illumination over conventional low intensity PL imaging systems. It has found that the PL response of certain defects in a semiconductor material or PV cell differs with the illumination intensity. That is, certain defects appear differently in PL images generated with high or low intensity light. Such a differential effect may improve the ability of operators to classify defects in a semiconductor material than would otherwise be the case using conventional low intensity PL imaging alone.

The applicant has also determined that an image of PL generated by high intensity illumination can, in itself, be more accurate in indicating the precise location of certain defects. For example cracks show up more clearly and sharply in an image of PL generated with high intensity illumination. This is significant and quite surprising and, as will be appreciated, provides significant advantages over the prior art.

The aforementioned high intensity photoluminescence imaging systems can be used throughout the production line of a photovoltaic device. They can be used by themselves or, preferably, in combination with other imaging and testing equipment.

In this regard, in an eighth aspect, the present invention provides a production line for the production of a photovoltaic device comprising a plurality of process steps to convert a semiconductor material to said photovoltaic device, said production line including at least one analysis device comprising a high intensity illumination source for applying illumination with intensity of at least 6 Suns to a semiconductor material, and an image capture device for obtaining an image of photoluminescence emanating from said illuminated semiconductor material. The aforementioned photoluminescence imaging systems can be used throughout the production line of a photovoltaic device. They can be used by themselves or, preferably, in combination with other imaging and testing equipment.

In a ninth aspect, the present invention provides a method of analysing a semiconductor material, said method comprising the steps of: inducing photoluminescence from said material by applying a high intensity illumination for a sufficient time and at an intensity of at least 6 Suns; and obtaining an image of said photoluminescence, wherein said illumination is provided by an eye-safe illumination system. As mentioned above, flash, LED or pulsed lasers provide a high intensity light source suitable for PL imaging. PL imaging systems containing high intensity laser-based illumination systems which are not eye-safe, as discussed above, generally require expensive and complex safety devices and optics. The use of eye-safe yet high intensity illuminations systems including sources such as flash lamps, LEDs, low pulse energy lasers etc not only produce good PL images but substantially reduce the costs associated with ensuring containment of illumination.

The aforementioned high intensity photoluminescence imaging systems can be used by themselves or, preferably, in combination with a low intensity photoluminescence imaging system, and optionally with other measurement techniques. In this regard, in a tenth aspect, the present invention provides an apparatus for determining the quality of a finished photovoltaic cell comprising (a) a high intensity photoluminescence system, and at least one of (b) a low intensity photoluminescence system, and (c) components for determining the Series Resistance of the cell.

The aforementioned photoluminescence imaging systems can be used as separate measurement tools, or they can be integrated into a process tool. In this regard, in yet a further aspect, the present invention provides a production tool comprising at least one PL measurement system and one process production system.

The present invention also provides an article of manufacture comprising a computer readable medium having a computer readable program code configured to carry out the aforementioned method and/or operate the production line or apparatus. The present invention also provides in yet a further aspect, an image of a photoluminescence effect from a semiconductor material illuminated flash illuminated by high intensity illumination as hereinbefore described.

The present invention also provides substantial advantages when applying additional stimuli to a semiconductor material. One such stimulus can be the application of thermal energy to the material, since it is known that the electronic properties of different features have different temperature dependence. For instance, in some applications a photoluminescence image may be obtained of a multicrystalline silicon cell at an elevated temperature, e.g. 200°C, where lifetime related variations in the image are strongly suppressed while cracks, for example, remain clearly visible. Above about 200°C however, thermal emission from the cell (and heating element) becomes the dominant signal and saturates the sensor, making conventional or low intensity PL measurement impossible since for an image capture time of around 1 second the thermal emission simply saturates a CCD camera. The present application, on the other hand, provides for high intensity illumination of the cell which does not require long image exposure or capture times. For instance, where a flash lamp is used to illuminate the semiconductor material, a PL response can be captured in around 1 millisecond. Such a short response or capture time means that the thermal emission signal from a heated wafer is 1000 times lower than for conventional systems, i.e. 1 millisecond compared to 1 second.

This is a particularly useful process when using a silicon CCD sensor. Using an InGaAs camera also allows a shorter measurement time, but it measures longer wavelengths than a Si CCD camera and is therefore relatively more sensitive to thermal emissions from the wafer or cell. This embodiment of the present inventive process is particularly important for wafers at high temperatures at two critical stages of production, namely diffusion and firing. For instance, a PL imaging system can be integrated after diffusion or after firing. Providing such an imaging system after the firing step would be preferable since generally wafers pass through the firing tool linearly, i.e. one by one, whereas diffusion is either a batch process or several wafers are processed in parallel. The resultant high intensity images, preferably obtained by flash lamp illumination, could be extremely useful. Also the samples need not necessarily be held at a constant temperature during the process. Indeed, a high intensity (e.g. flash) image could be taken during the cooling steps if it was desired. In such a way it could be determined whether the firing/diffusion step and/or cooling produce any effects on the wafer, detrimental or otherwise.

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 in side view a system suitable for in-line PL imaging of semiconductor samples;

Fig. 2 illustrates in side view a PL imaging system according to an embodiment of the present invention;

Figs 3(a) and 3(b) show in plan view and side view an arrangement of a flash lamp and camera according to a preferred embodiment of the invention;

Figs 4(a) and 4(b) show PL images of a silicon-based PV cell precursor acquired with illumination intensities of ~1 Sun and -100 Suns respectively; Figs 5(a), 5(b) and 5(c) illustrate how a finger shunt in a PV cell appears differently in images of PL generated with different illumination intensities;

Figs 6(a), 6(b) and 6(c) illustrate the different responses of shunts and dislocations to varying illumination intensity;

Fig. 7 illustrates schematically the incorporation of flash-based PL imaging systems into a semiconductor device production line;

Figs 8 and 9 illustrate in side view PL imaging systems with illumination from an LED bar array; and Figs 10(a), 10(b) and 10(c) are diagrammatic views of a PL imaging system using time delay integration with high intensity illumination.

Detailed Description

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

Photoluminescence (PL) imaging is known to be a rapid and convenient technique for characterising silicon ingots, blocks, wafers, as well as silicon-based photovoltaic (PV) cells both during and after manufacture, using systems and methods described in the abovementioned published PCT patent application No WO 07/041758 Al. The PL emission from silicon arises primarily from band-to-band recombination in the wavelength range 900 to 1300 nm, and can provide information on many material, mechanical and electrical parameters of relevance to PV cell performance including minority carrier diffusion length and minority carrier lifetime, and the impact of certain materials impurities and defects on the these properties. Figure 1 shows a PL imaging system 2 suitable for acquiring PL images of semiconductor devices such as silicon PV cells during their production process without removing them from the production line. This system includes two outer transport belts 4 to interface with a continuous on-belt production line and an inner transport belt 6 to bring a sample 8 to a stop in a measurement chamber 10. Inside the measurement chamber, photoluminescence 12 generated from the sample with broad area photo- excitation from a coherent laser source 14 of above-band gap light 16 is directed onto an imaging detector 18 such as a silicon CCD camera via collection optics 20, with the system preferably including homogenisation optics 22 to improve the uniformity of the broad area excitation and a long-pass filter 24 in front of the detector or the collection optics to block excitation light. The system may also include one or more filters 26 to select the wavelength range of the photo-excitation, and light-tight shutters 28 that open to allow samples in and out of the measurement chamber 10 to satisfy light safety requirements if necessary, i.e. if the system would not otherwise be eye-safe. In standalone systems with manual sample handling, e.g. for off-line inspection of silicon wafers or PV cells, the transport belts are not required, but any light safety issues remain would remain. As mentioned above however, it would be advantageous to avoid having to bring the samples to a stop for measurement. It would be especially advantageous to have a PL imaging system that just included a camera, a light source and optics as the major hardware components, that could be placed anywhere in a PV cell production line, for example above a transport belt bearing samples along the line, without requiring special modifications e.g. to satisfy stringent light safety requirements. The invention will be described with reference to various systems and methods for acquiring PL images of silicon-based PV cells and cell precursors, but the systems and methods are also applicable to PV cells and cell precursors or other devices based on other indirect or direct band gap semiconductors.

In a first embodiment, referred to hereinafter as a 'high intensity-based PL' system and illustrated in Figure 2, a substantial area (preferably at least 1 cm x 1 cm, more preferably the entire area) of a silicon-based PV cell or cell precursor 8 moving on a transport belt 36 is illuminated with excitation light 16 from a high intensity light source 30 such as a xenon flash lamp, pulsed LED, or other non-laser optical illumination source, or a laser light source with a short pulse period (say less than 1 ms), and the resulting PL emission 12 acquired with a silicon CCD camera 18 (or any other camera that can capture part or all of the photoluminescence band between 900nm and 1300nm). Other components that will generally be present include an excitation filter 26, PL collection optics 20 and a long pass filter 24 as in the Figure 1 system, and a reflector 32 may be present for directing a greater amount of the excitation light onto the sample. If the excitation light is from a broad band source such as a flash lamp, the excitation filter 26 is a more critical component than for laser excitation because of the necessity to prevent excitation light in the PL emission band from reaching the camera. The entire high intensity PL system 27 is preferably enclosed in a box 29, to be discussed later.

The image acquisition time in the Figure 2 system will be determined by the overlap of the illumination pulse and the camera shutter time, and it is generally advantageous for both to be short. The illumination time should be short to reduce power consumption and avoid excessive heating of the excitation filter and the sample, bearing in mind that high illumination intensity is generally required to generate sufficient PL signal within a short acquisition time. For preference, the camera shutter is substantially synchronised with the pulsed excitation source, subject to the limitations of shutter speed; for example the activation times for commercially available mechanical shutters are typically in the millisecond range. Leaving the camera shutter open too long may cause image blurring if the radiative lifetime is sufficiently long for the sample to move a significant distance (e.g. by a distance corresponding to several camera pixels) before the PL emission has decayed, although this is only likely to be a problem for very high carrier lifetime samples such as passivated monocrystalline silicon where the lifetime can exceed several milliseconds. This effect is expected to be negligible for typical multicrystalline silicon wafers where the carrier lifetime is of order hundreds of microseconds at most. In preferred embodiments the image acquisition time is sufficiently short that the sample moves by a distance of no more than that corresponding approximately to one or two rows of pixels in the imaging camera or less. This guideline depends on the speed of movement of the samples and on the number of pixel rows in the camera, but by way of example only, for a PV cell line throughput of 1 wafer per second (i.e. line speed of order 15 cm/s) and a 1 megapixel camera (1024 x 1024 pixels), this guideline would suggest an image acquisition time of duration 1 ms or less, compared to the -I s acquisition time permitted for in-line inspection of samples brought to a stop on a 1 wafer per second line.

This great reduction in permitted acquisition time obviously presents challenges for measuring a sufficient PL signal, but can to a large extent be compensated by the high intensity of commercially available flash lamps for example. We have demonstrated illumination intensities of up to 1000 Suns (100 W/cm ) across a standard silicon PV cell (15.6 cm x 15.6 cm) with a flash lamp, and although the PL signal does not necessarily scale linearly with illumination intensity, it is generally true that more intense illumination generates a greater PL signal. To explain further, the dependence is essentially linear at low injection levels (i.e. low illumination conditions) because the PL signal is proportional to the minority carrier concentration, whereas at high injection levels the dependence becomes quadratic, convoluted with loss mechanisms such as Auger recombination. In any event, it will be seen that we have been able to acquire PL images of silicon PV cell precursors with illumination exposures on the ms timescale, which is encouraging for in-line inspection. The time required to process an image is of course longer than the illumination/exposure time, since it includes factors such as read out time which may for example be around 700 to 1000 milliseconds for Si CCD cameras. However with most cameras it is certainly possible to keep up with a line speed of order one wafer per second.

As mentioned above, pulsed excitation is also possible. PL intensity is proportional to the product of the electron and hole concentrations, i.e. IPL °C n*p. With increasing illumination intensity the PL response to illumination thus changes from being linear in the excess carrier density An (at low injection conditions) to being proportional to the square of An (at high injection conditions) subject to various loss mechanisms. In many cases it is desirable to achieve a specific PL intensity with minimal exposure time. Using high illumination intensity can be beneficial in cases where the intensity is sufficiently high to reach high injection conditions, due to the quadratic PL response to excess carriers.

Flash illumination is one potential approach to get high illumination intensity. Another is to use a pulsed light source which, compared to a continuous wave (cw) light source with the same average optical power, reaches much higher instantaneous intensities. Higher PL intensities can therefore be achieved with the same average optical power. Since quantitative analysis of PL is more difficult at high injection, because the PL response no longer depends linearly on excess carrier density, the above approaches are therefore of particular interest where one only needs to analyse spatial features (patterns) rather then absolute PL intensities, for example for characterisation of as-cut multicrystalline wafers.

One specific preferred embodiment with flash lamp illumination, illustrated in Figure 3(a), includes on the illumination side a Broncola ring flash C 30 producing a 1 millisecond pulse that, after passing through an excitation filter 26 comprising a 6 mm thick KG1 Schott glass short pass filter, illuminates a silicon sample 8 with an intensity of 10-100 W/cm (100 to 1000 Suns). On the imaging side this system comprises collection optics 20, a long pass filter 24 and a 1 Megapixel silicon CCD camera 18 for acquiring an image. The system may also include a cylindrical reflector 37 if greater illumination intensity on the sample is required (e.g. of order 1000 Suns), and a shroud 39 to prevent excitation light entering the camera. As shown schematically in plan view in Figure 3(b) the ring-shaped flash lamp 30 allows the camera 18 to be centrally mounted, enabling both to be pointed orthogonally to the surface of a sample for greater illumination and imaging uniformity compared to configurations such as those shown in Figures 1 and 2 where one or both of the illumination source 14 and camera 18 is angled with respect to the surface of the sample 8. This arrangement also has the benefit of allowing an overall more compact system and, more importantly, the camera and flash lamp can both be closer to the sample without obstructing the field of view or casting a shadow. Having the flash lamp and camera closer to the sample will generally improve the efficiency of both the illumination and imaging systems. Flash lamps are preferable over laser illuminators, especially near IR laser illuminators, because of reduced light safety considerations. Although flash lamps are high intensity sources, their extended size means that under typical viewing conditions a much larger image is produced on the retina than with prior art laser sources, with consequently reduced thermal hazard to the retina. The thermal hazard is further reduced by the short pulse lengths of typical flash lamps. Flash lamp illuminators may also offer advantages of reduced cost and footprint.

As mentioned above in relation to Figure 2, the excitation filter 26 is an important component with broad band flash lamp illuminators or other high intensity light sources because of the necessity to remove excitation light in the PL emission band, and there are a number of factors to be considered. Although dielectric filters have sharper transitions from high to low transmission than absorption filters, which is especially important for indirect band gap materials where the PL emission is orders of magnitude weaker than the illumination, their transmission has a strong angular dependence causing the cut-on/cut-off wavelength to vary with incidence angle. The coherent, directional emission from lasers is readily collimated for efficient filtering with dielectric filters, but this is much more difficult to achieve with the incoherent and essentially isotropic emission from flash lamps or LEDs, favouring absorption filters or a combination of absorption and dielectric filters. A KG1 glass filter from Schott is an example of a short pass absorption filter suitable for PL imaging of silicon samples. We note that lamps that emit over a narrow wavelength range, such as low pressure sodium lamps that emit an extremely narrow doublet around 590 nm, may be advantageous in that the illumination can be easily separated from the silicon PL emission.

Apart from having less abrupt transitions from high to low transmission, absorption filters may also suffer from a heating problem, especially for the in-line inspection of PV cells/precursors where the flash lamp may need to be activated at a frequency of order 1 Hz or higher. There are several possible ways for dealing with such a heating problem, including efficient air or liquid cooling of a solid absorption filter, and using liquid filters where an absorbing liquid is re-circulated through a flow cell, composed of glass for example, and if necessary through a heat exchanger. Solutions of organic dyes, for example a combination of the IRA 955 and IRA 1034 infrared absorbers from Exciton, Inc, may be suitable for removing excitation light in the PL emission band. UV stability of organics may be an issue when filtering flash lamp emission, but most UV light can be blocked with a judicious choice of glass flow cell material, or by addition of UV absorbing material in the filter or in the cooling liquid if used, and in any event the optimal solution for a given system of flash lamp, sample material and camera technology may well involve a combination of filters and cooling techniques.

Figure 4(a) shows a PL image of a commercial passivated monocrystalline silicon wafer 8 with an emitter layer, acquired with a conventional PL imaging system where the wafer was illuminated at ~1 Sun (100 mW/cm ) with a near IR diode laser and the image acquired with a silicon CCD camera with an exposure time of two seconds. The image reveals four deliberately introduced cracks 40A to 40D and several dark patches 42 indicative of low carrier lifetime material. Figure 4(b) shows a PL image of the same wafer acquired with a flash-based high intensity PL imaging system, where the wafer was illuminated through a 650 nm short pass absorption filter with a 1 ms pulse having an estimated intensity of 100 Suns from a xenon flash lamp. Comparison of the two images immediately shows that all features are sharper and the low lifetime regions less dark in the image acquired with flash lamp illumination. The large, slightly brighter feature 43 appearing in the Figure 4(b) image is a support post visible through the wafer, which has substantial transparency at the imaging wavelengths. The blurring in the Figure 4(a) image arises from lateral currents flowing from high lifetime regions into adjacent defect-rich (i.e. low lifetime) regions to equilibrate the charge carrier distribution. This current flow occurs primarily in the emitter layer which for efficient PV operation is designed with a sheet resistance that enables low loss carrier transport under ~1 Sun illumination. Higher illumination intensities generate more charge carriers, resulting in larger lateral current flows, in which case the emitter sheet resistance causes greater transport losses, effectively isolating defect- rich regions from surrounding high lifetime areas. As shown in Figure 4(b), at 100 Suns illumination the lateral carrier transport is reduced to such an extent that the cracks 40A to 40D stand out much more clearly, lessening the chance of a crack detection algorithm missing cracks or reporting false positives. For example the crack 40A at the top right in Figure 4(b) is more likely be missed in the Figure 4(a) image because of the proximity of other low lifetime features.

Shunting is another cause of decreased efficiency in PV cells, and can cause more severe problems later on with hot spots and module failure. The most common cause of shunting is wrap-around of the emitter layer, but shunts are also caused by material inclusions or metal fire-through. Severely shunted cells are currently detected at the end-of-line IV tester, but earlier detection, additionally with spatial precision, would be beneficial. For example the ability to locate shunts, rather than just detecting their presence from their global effect on cell efficiency as done in IV testing, is important for remedial actions such as laser isolation currently performed by some PV cell manufacturers.

It is known that the location of a shunt can be determined with thermal (mid IR) cameras, but we believe that PL imaging offers a cheaper and quicker alternative. The applicant has observed that shunts in PV cells appear quite differently in PL images acquired with different illumination intensities, again because of limited lateral carrier transport at high illumination. At lower illumination shunts appear as blurred dark patches because they draw charge carriers in from surrounding regions, whereas they become more localised at higher illumination. This differential visibility effect is particularly marked for shunts because the current flow into them depends logarithmically on illumination intensity, compared to a linear dependence for most other recombination active defects including dislocations, cracks and impurities, meaning that comparison of high/low illumination intensity PL images may be particularly valuable for discriminating shunts from other defects. To demonstrate, Figures 5(a), 5(b) and 5(c) show PL images of a silicon PV cell with a shunted finger, acquired with on-sample illumination intensities of order 1 Sun, 100 Suns and 1000 Suns respectively. In the low intensity image (Figure 5(a)) the finger shunt is clearly identified by the large blurred dark area 60, but its precise location is difficult to determine and the blurred dark area obscures other features such as dislocation clusters 62 that may be of interest. As the illumination intensity is increased (Figures 5(b) and 5(c)) the finger shunt 64 appears progressively more localised, in clear distinction from the dislocation clusters which appear more or less the same in each image.

Similarly, Figures 6(a), 6(b) and 6(c) show PL images of a silicon PV cell with a group of shunts 66 within a large dislocation cluster 62, acquired with illumination intensities of order 1 Sun, 100 Suns and 1000 Suns respectively. As in the previous example, the shunts appear more localised as the illumination intensity increases while the dislocation clusters appear essentially unchanged. In Fig 6(c) (highest illumination intensity) the locations of two individual shunts 66 within the large dislocation cluster 62 are indicated. Clearly the combination of images enables an operator or image analysis routines to distinguish between these two defect types.

The present inventive technique is also useful for finding defects and other features in selective emitter structures. This aspect utilises the reduced lateral blurring that occurs at high illumination intensity in samples with an emitter. Selective emitter structures use various processing methods to create highly doped regions (usually lines) within an otherwise homogeneous lightly doped surface region, which are subsequently metalised using either screen printing or plating for example. Shunts and other defects can be introduced locally into the sample during either the doping or the metalisation processes, or both. PL imaging with high illumination intensity such as with a flash, or a comparison between high and low illumination intensity PL images, allows the exact position and shape of such defects and shunts to be identified much more accurately than, for instance, conventional low-intensity PL imaging, for example a one-Sun image.

The high intensity illumination of the above described process may also be used to monitor the diffusion and post diffusion processes in cell production as will now be explained. Recombination in the emitter is one of the loss mechanisms of a solar cell. Emitter recombination is a process that reduces the effective excess carrier lifetime and thus the PL signal, particularly at high excess carrier densities. Apart from the excess carrier density, the magnitude of the emitter recombination is dependent on the background doping of the base, the emitter doping profile and the surface passivation. Since emitter recombination increases relative to bulk recombination (and may become dominant) at higher excess carrier density, a PL image taken at high illumination intensity is affected more strongly by the emitter recombination then a PL image taken at lower illumination intensity. Qualitative or quantitative information about the emitter quality can therefore be gained either from a single PL image taken with high illumination intensity or from a comparison of two PL images, one taken with high illumination intensity, the other one with low illumination intensity. When this is carried out after the diffusion step, it provides a monitor of the diffusion process. When carried out after a post-diffusion step, it provides a monitor of the cumulative effect of the processes between that step and the diffusion step inclusive. When carried out before and after a post-diffusion step, the comparison of the emitter quality determined before and after the step provides a monitor of that process step.

Turning now to industrial implementations of high intensity PL imaging, we believe that high intensity-based systems are potentially well- suited to in-line inspection of devices on a PV cell production line. As shown schematically in Figure 7, high intensity-based PL imaging systems 27 can be mounted in simple boxes 29 over a transport belt 36 carrying semiconductor devices 8 along a production line, and located before or after all or selected process stations 38. In a production line for silicon-based PV cells, the processes in individual stations typically include saw damage etch, emitter diffusion, silicon nitride deposition, screen printing of metal contacts, thermal treatment, edge isolation and IV testing. All or a substantial fraction of devices entering or leaving selected process stations can be inspected in a contact-less fashion and without interrupting their motion along the production line, offering powerful means for quality control of the devices and process control of the various stations. The high intensity-based PL imaging systems are preferably mounted in some form of enclosure as shown in Figure 7, to provide a significant distance between the light source and an operator, and at the least to prevent direct viewing of the light source (i.e. the light must bounce off the sample or another object). However this simple enclosure is to be distinguished from the complex systems of light-tight shutters that could be required under laser safety regulations; these will be unnecessary if the PL imaging system as a whole is eye-safe, simplifying the automation and integration into a production line. In certain embodiments the PL imaging system as a whole is eye- safe because the illumination system is eye-safe. In other embodiments the PL imaging system as a whole is eye-safe even if the illumination system is not, for example via measures such as the prevention of direct viewing of the illumination system output, or the presence of production line guarding that provides some minimum distance between an operator and the PL imaging system. We note that more sensitive cameras (e.g. InGaAs cameras as noted below) may allow use of the configuration as shown in Figure 7 with lower intensity light sources or for even shorter illumination periods.

Before describing further embodiments of PL imaging systems, and in particular those preferred embodiments with reduced or no light safety requirements, it will be useful to include some discussion of current laser safety standards and some strategies for producing PL imaging systems with reduced light safety requirements. As mentioned previously the brightness of a source, which can be defined as the optical power passing through an aperture divided by the aperture area divided by the solid angle subtended by the optical beam in the far field, is a key parameter, and light safety issues cannot simply be ignored just because a system uses non-laser (incoherent) light sources.

In Australia and New Zealand, the standards for laser classification and safety requirements are provided by AS/NZS 2211.1:2004 and its associated guidelines (AS/NZS 2211.10:2004), based on the international standard IEC 60825-1:2001. An important concept in laser safety is the 'Maximum Permissible Exposure' (MPE) level, which is defined in the standard as 'that level of laser radiation to which, under normal circumstances, persons may be exposed without suffering adverse effects'. The definition further states that 'MPE levels represent the maximum level to which the eye or skin can be exposed without consequential injury immediately, or after a long time, and are related to the wavelength of the radiation, the pulse duration or exposure time, the tissue at risk and, for visible and near infra-red radiation in the range of 400 nm to 1 400 nm, the size of the retinal image'.

Since the wavelengths of light suitable for generating PL from silicon are within this 400 to 1400 nm range, it follows that retinal image size is a key factor for light safety in PL imaging systems. Within certain limits, the MPE level increases with increased image size on the retina, although there is no decrease in the MPE below a certain minimum image size and no increase above a certain maximum image size. For quantitative purposes the standard uses an angular measure of the retinal image size, the angle subtended by the source at the eye, a. This is generally referred to as the 'angular subtense' and is given approximately by the source size divided by the distance between the source and the eye. The angular subtense representing the image size below which there is no further decrease in the MPE is referred to as 'oimin' (1.5 mrad), and exposure conditions below this are referred to as 'point source viewing'. 'Extended source viewing' conditions apply at angular subtenses above a _m j _n , and as the angular subtense increases from a _m i _n the MPE level increases until it reaches a maximum at a = a _max ( 100 mrad), beyond which the MPE is constant. It is important to note that if the source radiation is modified by illumination optics, as shown in Figs 1-3 for example, the 'apparent source' for MPE purposes is the image, real or virtual, that produces the smallest retinal image. For the purposes of this specification, the term 'illuminator' will be used to refer to the portion of a PL imaging system that provides optical excitation to a sample. An illuminator will include one or more optical sources, possibly in combination with a number of other components including filters and focusing optics.

In the standards, laser products are classified in a system ranging from Class 1 , 'safe under reasonably foreseeable conditions of operation', to Class 4, 'generally powerful enough to burn skin and cause fires', using limits known as 'accessible emission limits' (AELs). AELs are derived from MPEs using limiting apertures and may be expressed as a power limit, an energy limit, an irradiance limit, a radiant exposure limit, or a combination thereof. The limiting aperture is usually taken to be 7 mm, representing a dilated pupil as a 'worst case scenario'. Although meeting Class 1 AELs is necessary but not sufficient for making a laser product Class 1, there being other constraints, for the purposes of this specification a PL imaging system as a whole will be considered to be 'eye-safe' if it meets Class 1 AELs. Similarly, the illuminator portion of an imaging system will be considered to be 'eye-safe' if it meets Class 1 AELs.

Relatively high brightness sources, typically required for acquiring PL images of silicon PV samples on a timescale suitable for in-line applications, are potentially hazardous because they can result in a relatively high intensity at the eye, even at a distance, or a relatively small retinal image (and correspondingly low MPE level). However to determine the actual hazard, it is necessary to consider brightness in combination with the viewing conditions, in particular the angular subtense. The importance of viewing conditions is demonstrated by the following specific example. According to the calculation methodology prescribed in IEC 60825-1:2001, an 808 nm cw laser product can only be classified as Class 1 (i.e. does not exceed the Class 1 AEL) if its emission under point source viewing conditions (i.e. angular subtense a < oCmin) does not exceed 0.64 mW through a 7 mm diameter limiting aperture. In contrast, for extended source viewing conditions where a > a _max (100 mrad), the Class 1 AEL is 42 mW (i.e. 65x higher) through a 7 mm diameter limiting aperture. The brightest light sources in common use are laser sources, which have high temporal coherence (or equivalently, coherence length) compared to non-laser (i.e. thermal) sources. Since coherence is an inherent aspect of the lasing process, higher coherence than thermal sources may be considered a necessary condition for achieving the highest brightness practical light sources. However coherence does not imply brightness, as it is not a sufficient condition. In general, coherence length varies widely (over orders of magnitude) between different laser types, but this does not necessarily correlate with brightness. For example the coherence length of a laser source can be increased by using a high quality factor (Q) resonator at the expense of output power, meaning that while the beam collimation (the 'per unit solid angle' part of the brightness definition) may be increased, the reduced output power reduces the 'power per unit area' part of the brightness definition, counteracting the potential increase in brightness. Optics can be added to a light source to reduce the brightness without altering the coherence, a trivial example being an absorbing filter which may be used to reduce the brightness arbitrarily without altering the coherence. Of significant practical relevance for PL imaging systems of the present invention are illuminator designs which have reduced brightness without significantly reducing the intensity on the sample, typically all or part of a wafer or PV cell. In certain embodiments this is achieved in a second illuminator ('system 2') compared to an unimproved, prior art illuminator ('system ) by one or both of: (i) Increasing the solid angle filled by the light output from system 2 relative to that of system 1. This may be expressed as decreasing the 'f number' or increasing the Numerical Aperture of the illuminator, and essentially the excitation light is made to diverge more rapidly so that its intensity at a distance is reduced. (ii) Increasing the size of the source (real or apparent, as discussed above in the context of illumination optics) in system 2, for example by dividing a single beam in system 1 into one or more beams, or an array of beamlets, in system 2, or by mechanically agitating a component of the illumination system (e.g. a mirror). If system 1 already uses a number of beamlets, their number may be significantly increased in system 2.

Approach (i) decreases the intensity of light at the eye, while approach (ii) increases the angular subtense a which, subject to the limits described above, may increase the MPE level as follows:

(a) If a for system 1 was greater than (Xmin and less than a _max , then the MPE level for system 2 is greater than for system 1.

(b) If a for system 1 was less than a _m i _n and a for system 2 is greater than a _m i _n , then the MPE level for system 2 is also greater than for system 1.

(c) If a for system 1 was less than < _m i _n and a for system 2 is also less than a _m i _n , then the MPE level for system 2 is the same as for system 1.

By means of one or both of these measures, it is possible for an illuminator to meet Class 1 AELs (i.e. be eye-safe) even when the source itself is rated as high as Class 4. If the illuminator does not meet Class 1 AELs, with or without these measures, it is still possible for a PL imaging system as a whole, or such system integrated into a production line or other wafer/cell handling system, to meet Class 1 AELs without resorting to stringent laser safety measures such as safety shutters and interlocks. This represents a significant simplification for the system integration; for example the configuration shown in Fig. 3 would be simplified considerably if the light-tight shutters 22 were not required and the measurement chamber 24 did not have to enclose the imaging system on all sides. Instead, the PL system itself or the production line guarding may provide some minimum human access distance from the illuminator, and the PL system can prevent direct viewing of the illuminator output, i.e. viewing will be limited to reflections from a wafer or solar cell or some object in the PL system or production line. Reflections off sample edges are of particular concern, since broken wafers may present mirror-like edge surfaces at unpredictable angles. Reducing the illuminator brightness by increasing the divergence angle of the excitation light (approach (i) described above) is particularly useful in combination with measures that provide a minimum human access distance. All these details need to be considered in determining if a PL imaging system meets Class 1 AELs.

To summarise, it is preferred for a PL imaging system as a whole, or such system integrated into a production line or other wafer/cell handling system, to meet Class 1 AELs without resorting to stringent laser safety measures such as safety shutters and interlocks. More preferably, the illuminator meets Class 1 AELs. With these light safety considerations in mind, we now turn to the description of certain preferred embodiments of PL imaging systems for in-line inspection of silicon solar cell samples. For both area illumination schemes and line illumination schemes, the above described approaches can be applied to reduce light safety requirements.

Some PV cell process steps, including IV testing, laser edge isolation and the laser- based selective emitter processing used for some high performance cell designs, require the PV cell to be brought to a stop, in which case a high intensity-based PL imaging system could be incorporated within the particular station to inspect the cell before, during or after that process step. Because high intensity-based PL images can be acquired in a matter of milliseconds, this will have essentially no effect on the process flow.

We note that high intensity illumination is applicable at all stages of PV block, wafer and cell production, but requires adjustments to the wavelength range of the illumination. In one particular example, the surface region of as-cut silicon wafers is damaged to such an extent as to be essentially useless for PL imaging purposes since there is extremely high non-radiative recombination in this region, so that the shorter wavelength (visible) portion of a flash lamp spectrum, being absorbed close to the surface, generates little PL response. Excitation light closer to the 1000 nm band edge of silicon penetrates deeper and generates a measurable PL response, but careful filtering of the illumination is required to pass light in this region and yet reject wavelengths overlapping with the detectable PL emission band, which is approximately 900 nm to 1160 nm for silicon CCD cameras. The options for achieving this include absorption filters and dielectric filters. For silicon-based PV cells, reliable crack detection is a highly desirable capability. While some types of defects in PV cells can be remedied, e.g. laser isolation of shunts, or at least do not become worse over time, cracks have the potential to grow from tiny and difficult to detect micro-cracks, ultimately causing catastrophic failure of a PV cell. Furthermore one defunct cell can compromise an entire module. Figure 4(b) clearly shows the value of flash-based PL imaging for crack detection, and the differences in contrast between high and low lifetime regions (Figures 4(a) and 4(b)) and between shunts and dislocations (Figures 5(a) to 6(c)) suggests that comparison (e.g. image subtraction) between PL images acquired with high and low illumination intensities could also be of value. In particular, image subtraction or the like could be used to highlight defects such as shunts that appear differently under different illumination intensities.

While a qualitative comparison of PL images such as those shown in Figures 4(a) and 4(b) is relatively straightforward, quantitative image comparison (i.e. subtraction or the like) is more difficult because the results are sensitive to any grey level change. There are many imperfections in imaging systems that can interfere with defect detection algorithms, and quantitative image comparison requires that these imperfections match. It is also desirable to align the two images and for the gain and offset calibrations and flat field corrections to match. Therefore in one embodiment it is desirable to capture the two images without moving the sample to simplify the image comparison, and this is more easily achieved if the total image capture and read out time is less than 1 second or so, so as to keep up with line speeds. Clearly there are benefits in acquiring PL images under different illumination conditions, and we now turn to consideration of suitable PL imaging configurations for this purpose. In one embodiment, one or more reflectors such as the cylindrical reflector 37 shown in Figure 3(a) can be used to enhance the illumination intensity, while in another embodiment neutral density filters can be used to attenuate a flash lamp emission as required for lower intensity excitation. Alternatively the drive power of the flash lamp can be adjusted as required; if the signal-to-noise ratio is too low at reduced flash intensities, the PL response of several low-intensity flashes can be accumulated into one image by reading out and averaging the individual images, or by firing a rapid sequence of low intensity flashes during one image exposure cycle. In other embodiments, a PL imaging system contains a separate low intensity illumination source, such as one or more LEDs or even a laser system. A wide range of LEDs are available with different powers and emission wavelengths, and LEDs are in general inexpensive light sources and, like flash lamps, are extended sources so have lesser light safety concerns than do lasers. Notwithstanding light safety issues, we also envisage situations where laser sources are also of value, either in combination with flash lamps or LEDs, or indeed by themselves. For example lasers, with their narrow emission bands, may be better suited for samples such as as-cut silicon wafers where the excitation wavelength needs to be close to the PL emission band to avoid being absorbed in the surface damage layer. Additionally, for certain choices of camera, e.g. an InGaAs camera, a larger portion of the PL spectrum is sampled, so the illumination pulse can be many times shorter for the same response. Preferably the pulse is sufficiently short for the energy per pulse to be low enough for eye safety to be of much lower concern, so that laser safety mechanical systems may not be required.

LEDs may be used to generate extremely bright ms pulses like flash lamps, and in all the descriptions above the use of the term 'flash' should be used interchangeably with pulsed LED. Even lower intensity LED light can still be adapted for in-line inspection of semiconductor devices without interrupting their motion. For example in the configuration shown in side view in Figure 8, excitation light 16 from an LED bar array 44 passes through an excitation filter 26 and is focused with a cylindrical lens 46 onto a semiconductor device 8 moving on a transport belt 36, and the PL emission 12 from the illuminated region 47 imaged by a system of collection optics 20, that may for example include an optical fibre bundle or one or more cylindrical lenses or double Gaussian lenses, through a long pass filter 24 onto a line camera 48. LEDs can also be pulsed for short periods, at approximately lOx their rated power. In an alternative configuration shown in Figure 9, the LED bar array 44 and line camera 48 are on opposite sides of the sample 8, with a gap 50 in the transport belt 36 allowing the sample to be inspected across its entire width. The 'back side illumination' configuration shown in Figure 9 has more complicated engineering, but the sample acts as a long pass filter (being largely transparent in the PL emission band) so a separate long pass filter 24 may not be required. As the price of LEDs continues to fall, we also envisage cost-effective PL imaging systems where arrays of high power LEDs generate enough power to produce intensities of order 100 Suns across a broad area, as required for the high power source 30 in the configuration shown in Figure 2. LED arrays are also eminently suitable as less optically hazardous replacements for lasers in broad area PL imaging systems where the sample is brought to a stop for measurement. Again, the LED array and the camera can be on the same side of the sample or on opposite sides. LEDs are commonly packaged with an in-built lens, so that the emission is confined to a relatively narrow range of angles. Consequently, it should be possible to use a sharp cut-off dielectric filter as the excitation filter 26. While it would be advantageous to have PL imaging systems located at several points along a PV cell production line for early detection of process problems or defective wafers, end of line testing is particularly important. At present finished cells are primarily subjected to an IV test where their electrical characteristics are measured under simulated solar illumination, but there is clearly much additional useful information that could be gained besides a global measure of performance. In one example, high intensity-based PL imaging can be integrated into an IV test unit, with the resulting spatially resolved defect information used to determine why particular cells show poor performance in the IV test instead of just sorting them into quality bins. The existing ~1 Sun source of an IV test unit could be used to generate low illumination intensity PL images, for comparison with high illumination intensity PL images as discussed previously, e.g. for improved crack and shunt detection. End of line PL imaging, in combination with PL imaging at earlier stages, could also be used as part of a difference or manufacturing execution system (MES) style image system. Another advantage of combining a PL imaging system with an IV test unit is that the camera can be used to acquire electroluminescence (EL) images, where luminescence is generated by electrical excitation, while the subject cell is electrically contacted for the IV test. It is advantageous to gather as much information as possible at this stage, because there is a risk of breakage of fragile silicon PV cells every time they are contacted.

Series resistance (Rs) imaging is a topic of great interest in the PV cell industry at present, because of its ability to locate electrical problems such as broken fingers and excessive contact resistance, as well as cracks that disrupt current flow. Luminescence systems can be used to determine quantitative series resistance with EL, PL and various hybrid means, with most of these methods requiring the capture of at least two images under different conditions. Incorporation of a flash-based PL imaging system with its associated camera into an IV test station would enable PL and/or EL-based Rs imaging techniques, such as those described in published PCT patent applications WO 07/128060 Al and WO 09/129575 Al, to be performed. This would be a significant improvement over the current situation where IV testing can only measure an average Rs value across a cell.

However there is a limit as to how many additional tests can be combined with IV testing without slowing down the entire production line, recalling that in a sequential process the flow rate is limited by the slowest step. In the alternative, Rs imaging can be performed in a separate stage before or after IV testing, which may require cells to be contacted again.

However the preferred end of line test device combines the following attributes: (a) a high intensity PL source (>50 Suns); (b) a low intensity PL source (<10 Suns); and (c) one or more spatially inhomogeneous optical filters to illuminate selected portions of a sample as described in published PCT patent application No WO 2010/130013 Al. Together with an appropriate camera this enables the non-contact module to measure: (a) shunts and other features from the difference of the high and low Sun images; (b) quantitative series resistance using the inhomogeneous filters; (c) defects (e.g. impurities) that impact cell performance and that are more discernible in low Sun measurements; and (d) features relating to the emitter discernible in high Sun PL images. This non-contact module would be placed before or after the IV tester, and preferably used in conjunction with the IV tester to identify specific faults, causes of specific faults, poor materials and process errors, as well as provide data that can be used to improve the manufacturing process via MES systems and the like.

PL imaging can also be used to inspect individual PV cells and groups of cells during moduling, where a number of cells are connected into a single module for installation in the field. For example a module manufacturer may use PL imaging for quality control of incoming PV cells, looking for problems such as cracks, shunts and series resistance issues, or to monitor the cells during the moduling process, looking for cracks or potential hot spots for example. As before, the advantages of reduced light safety requirements with using a non-coherent or short pulse illuminator apply. Flash lamp illumination may also be useful at the moduling stage because of its high power; for example instead of illuminating a single cell with an intensity of -100 Suns, a flash lamp could be used to illuminate a large number of cells with an intensity of ~ 1 Sun.

The above description is generally written with silicon CCD or CMOS cameras in mind. However we note that other cameras, e.g. InGaAs cameras, can be used. As described in PCT patent application No AU2010/001045 entitled 'Photoluminescence imaging systems for silicon photovoltaic cell manufacturing', incorporated herein by reference, InGaAs cameras and the like have advantages and disadvantages compared to silicon-based cameras. The primary benefit is that such cameras are sensitive throughout the silicon PL emission spectrum, so that significantly more of the PL signal can be captured, potentially up to 30x more. By capturing more of the PL signal, the time requirements on the light source are dramatically reduced for the same measurement time result. For example a laser pulse can be many times shorter, and even down to 1 ms if the compound semiconductor camera pixel sensor size is much larger than the equivalent silicon camera. At 1 ms a pulsed laser may not require laser safety shutters because the total energy in the pulse may be sufficiently small not to constitute a hazard. In principle, the safety hazard for any pulsed laser can be reduced by reducing the pulse length and therefore the total energy per pulse. The main drawback of such cameras is the image smearing caused by the fact that the longer wavelength PL light that these cameras detect may move an excessive distance within the sample prior to escaping from the surface, due to silicon transparency at these wavelengths.

One of the challenges for PL imaging systems is as-cut wafers or silicon blocks. With surface damage and low quality silicon at least in the surface region, these types of samples, especially as-cut wafers, have poor PL response. In applications such as incoming wafer sorting, there is a need to acquire PL images rapidly without stopping the samples, preferably with an illumination system that requires no light safety shutters, and at reasonable cost. The present invention therefore provides several options, including short pulse broad area illumination with a suitably filtered flash lamp and an InGaAs or Si CCD camera, and laser/LED (1-2 Suns) illumination with an InGaAs camera, e.g. with 640x512 pixels, using a number of full frame 1ms images in a frame averaging type approach.

If any of these systems requires too long an exposure for PL response, again the present invention provides additional embodiments to allow non-stop measurement, including (a) tracking of the sample with a mirror during measurement, or (b) frame averaging a number (say up to 10) 1 millisecond images to reduce the signal to noise to an acceptable level.

Another key benefit of fast measurement in a non-contact system is potential integration into a process tool. One example is in laser processing of partially processed solar cells, where rapid PL measurements taken before and after processing would reveal all positive and negative changes on the sample due to processing by simple subtraction of the before and after images, thus allowing quality control and process control. Other process steps where this technique could be used include screen printing, IV testing, diffusion and passivation. Integration into the process line is aided by the simplicity of the PL measurement system - a light source with optics and a camera that can be arranged in many configurations. Images can be taken before or after processing, or both, and processing can be continuous or stop-start.

The flexibility, accuracy and speed provided by the present invention finds utility in a number of environments. Yet another approach to rapid analysis of a semiconductor such as a silicon wafer is to combine the use of high intensity illumination with a Time Delay Integration (TDI) based PL imaging system to produce a combined mechanism/process that is suitable for non-stop in-line use. In this case, the 'high intensity' illumination is high relative to the intensity used in the prior art for stationary PL imaging of silicon wafers. TDI typically uses a full frame CCD with up to several hundred lines (rows) of pixels. For the following discussion we will consider the use of a full frame TDI-CCD with 1024 columns and 128 rows, the performance of which is otherwise similar to CCDs used in the prior art for stationary PL imaging of silicon wafers, for PL imaging of p- type raw wafers of approximately 1 Ohm.cm resistivity. In this case, a continuous illumination source with intensity approximately in the range of 8 to 16 Suns is required over an area of approximately 13% of the wafer to measure 3600 wafers per hour with 1 Megapixel resolution.

Figures 10(a)- 10(c) show a top view of wafers 52A and 52B and their movement relative to a TDI sensor 54. Essentially the shaded areas 56A and 56B within the sensor show the pixels that are capturing wafer images, with the relative CCD height (number of rows) exaggerated compared to the sample being measured.

In Figure 10(a) the image of the leading edge of a wafer 52A has integrated on the TDI sensor 54 for half the total time available, assuming constant wafer velocity, as indicated by the shaded area 56A extending halfway across the TDI sensor. In Figure 10(b) the leading edge of the wafer 52A has finished integration and its PL signal read out. In Figure 10(c) imaging of the first wafer 52A is almost complete and imaging of the next wafer 52B has started, as indicated by the shaded areas 56A and 56B.

In a further embodiment, a standard CCD camera (i.e. not optimised for TDI) as used in the prior art for stationary PL imaging of silicon wafers may be used for the combined high intensity-TDI PL imaging of silicon wafers. For a throughput of about 1 wafer per second for instance, this embodiment requires a readout speed of about 1 MHz to acquire a 1 Megapixel image. This is well within the capability of standard CCD cameras used in the prior art for stationary PL imaging of silicon wafers. In this embodiment, no signal to noise compromise is required to apply the TDI technique to PL imaging of silicon wafers.

The limit of TDI gain (number of rows over which an image can be synchronously built up without significant blurring) is affected by a number of parameters including:

- Lens distortion;

Belt position synchronisation accuracy or velocity measurement accuracy;

Alignment of the CCD columns with the belt direction.

The abovementioned TDI method may be less desirable for diffused wafers since the emitter will bias up parts the wafer that are not being imaged, potentially reducing the photoluminescence response from the strip being imaged at any given moment. This can be overcome by over-illuminating the camera field of view, i.e. providing substantial additional illumination to the target area to ensure a sufficient PL response. Of course there may be some of additional equipment cost and increased power consumption.

The aforementioned TDI method is also advantageous for raw wafers, since the illumination of the camera field of view does not have to be uniform in the 'vertical' direction (i.e. the CCD column direction), which corresponds to the belt motion direction, since images are integrated over this direction. This increases optical design freedom to address laser safety issues or the design of illumination systems with horizontal uniformity (i.e. in the CCD row direction). In a further embodiment, applied particularly to the case of raw wafers, there may be no need at all for uniformity of illumination in the horizontal direction. Rather, the illumination may, for example, consist of staggered rows of discrete light spots, for example staggered arrays of illuminated squares, enabling simple projection of LED array light sources onto the wafer.

As mentioned above, in one embodiment of the present invention flash lamp illumination is used to produce a PL response. This flash method of in-line PL imaging relies on the high intensity light pulse being short enough to eliminate blurring caused by sample movement. On the other hand for quantitative analysis of PL data, and/or for measurement of certain features or parameters of importance to PV devices, such as shunts, it may be preferable to use lower intensity illumination. However these illumination levels require a sample to be stationary to prevent blurring. One approach that largely overcomes the limitations of both methods separately is a hybrid of flash and TDI based PL imaging, referred to hereinafter as 'FTDI imaging'. A standard square CCD array may be used in this approach. Since in a normal TDI operation a gain of about 10 would be used, the following embodiment will be discussed in similar terms. In this case the height of the image on the CCD plane must be less than the height of the CCD active area by at least 10 rows; for example for a 1024 x 1024 CCD, when the top of the image is aligned with the first row, the bottom of the image must not extend below row 1014.

In practice, a margin of about 55 rows is allowed for sample alignment - this may be reduced or the true number of rows in the wafer image may be reduced, or both. In this embodiment, the wafer image moves from top to bottom of the CCD and the alignment margin is taken to be zero. The flash is triggered when the leading edge of the wafer reaches a position on the belt where the top of the wafer image reaches or passes Row 1. The CCD vertical clock (row-clock) is synchronised to the motion of the wafer until the image has moved down 10 rows. After that time, the flash lamp source is off and the CCD may be read out in normal frame-readout mode (typically using a faster vertical clock than during the FTDI mode for a 15cm per second belt speed). An FTDI-PL system could also be used in conventional flash mode. For example this may be used where high light intensity is advantageous and a short (non-blurring) pulse is adequate (i.e. for emitter quality measurements on diffused wafers). The FTDI-PL mode could be used at lower peak intensity (10 or 20 times lower for example) when bulk lifetime is being measured, or it could be used at the full intensity for raw wafers where the full pulse energy (i.e. long pulse) capability of the flash system is desired to achieve adequate signal to noise. Depending on CCD details, an FTDI-PL system may be usable in standard TDI mode, with the number of rows being the maximum possible TDI gain; in typical practice about 100 rows is the practical maximum. The large pixel size (on the sample) used in PL imaging in solar cell inspection offers some scope to increase the typical maximum TDI gain since the row synchronisation is not as demanding as for smaller sample pixels on faster moving belts in other industries. Although the description above refers to flash TDI, we note that the embodiment is not restricted to flash but can be any high intensity short light pulse as described above, e.g. pulsed laser, LED etc.

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.