Raman spectroscopy apparatus and method

Abstract

There is provided a Raman spectroscopy apparatus (20) comprising: a deep ultra-violet electromagnetic radiation source (30) for generating a deep ultra-violet electromagnetic radiation beam; an optical arrangement configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source (30) along an irradiation path towards an object (32), the optical arrangement comprising at least one parabolic optical element (22a, 22b); and an optical detection device (34, 36) configured to, in use, detect at least one optical signal resulting from the irradiation of the object (32).

Description

[0001] RAMAN SPECTROSCOPY APPARATUS AND METHOD

[0002] The present disclosure relates to a Raman spectroscopy apparatus, a waste inspection apparatus, a method of using a Raman spectroscopy apparatus and a method of operating a waste inspection apparatus.

[0003] It is known to identify and sort plastic waste according to waste disposal and recycling needs.

[0004] According to a first aspect of the present invention, there is provided a Raman spectroscopy apparatus which comprises a deep ultra-violet electromagnetic radiation source (e.g. laser) for generating a deep ultra-violet electromagnetic radiation beam; an optical arrangement configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object, the optical arrangement comprising at least one parabolic optical element; and an optical detection device configured to, in use, detect at least one optical signal resulting from the irradiation of the object.

[0005] The or each parabolic optical element may be, but is not limited to, a parabolic optical reflector and / or an off-axis parabolic optical element. The optical arrangement may include, but is not limited to, a plurality of offset parabolic optical elements. A diameter of the at least one parabolic optical element may be at least 25 mm.

[0006] The at least one parabolic optical element may be arranged at an end of the irradiation path so that, in use, the at least one parabolic optical element directly guides the deep ultra-violet electromagnetic radiation beam to the object.

[0007] The working distance of the optical arrangement may be at least 12 cm, preferably in the range of 12 cm to 50 cm.

[0008] According to another aspect of the invention, there is provided a Raman spectroscopy apparatus comprising a deep ultra-violet electromagnetic radiation source for generating a deep ultra-violet electromagnetic radiation beam; an optical arrangement configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object, wherein a working distance of the optical arrangement is at least 12 cm, preferably in the range of 12 cm and 50 cm; and an optical detection device configured to, in use, detect at least one optical signal resulting from the irradiation of the object. The Raman spectroscopy apparatus may further include an object identification module configured to, in use, process the at least one detected optical signal, or at least one component of the at least one detected optical signal, so as to identify the object. The identification of the object by the Raman spectroscopy apparatus may include identification of the object as a polymer or plastic object or as not a polymer or plastic object. The identification of the object by the Raman spectroscopy apparatus may include distinguishing the identified object from another object.

[0009] The object identification module of the Raman spectroscopy apparatus may be configured to, in use, compare the at least one detected optical signal, or at least one component of the at least one detected optical signal, to a corresponding optical signature so as to identify the object.

[0010] The comparison of the at least one detected optical signal, or at least one component of the at least one detected optical signal, to a corresponding optical signature may include the use of a look-up table including a plurality of stored optical signatures or the use of a classification algorithm.

[0011] The object identification module may include a processor and memory including computer program code, the memory and computer program code configured to, with the processor, enable the object identification module to perform its function(s). The object identification module may be, may include, may communicate with or may form part of one or more of an electronic device, a portable electronic device, a portable telecommunications device, a microprocessor, a mobile phone, a personal digital assistant, a tablet, a phablet, a desktop computer, a laptop computer, a server, a cloud computing network, a smartphone, a smartwatch, smart eyewear, and a module for one or more of the same.

[0012] The optical arrangement of the Raman spectroscopy apparatus may include at least one optical lens configured to be movable to adjust a working distance of the optical arrangement. The optical lens may be a quartz lens, a composite lens, a CasF? lens, a MgF? lens, or a lens made of another material.

[0013] When included, the at least one optical lens of the Raman spectroscopy apparatus may be arranged between the deep ultra-violet electromagnetic radiation source and the at least one parabolic optical element in the irradiation path. The Raman spectroscopy apparatus may include a distance sensor configured to, in use, measure a distance between the optical arrangement and the object. The optical arrangement may be reconfigurable to adjust a working distance of the optical arrangement responsive to the measured distance between the optical arrangement and the object. The optical lens may be configured to be movable to adjust a working distance of the optical arrangement, preferably responsive to the measured distance between the optical arrangement and the object.

[0014] The optical detection device may include, but is not limited to, a spectrograph and / or an imaging device.

[0015] The imaging device may be configured to, in use, adjust an exposure time responsive to a measured dimension of the object.

[0016] According to a second aspect of the invention, there is provided a waste inspection apparatus comprising a Raman spectroscopy apparatus according to any one of the first aspect of the invention and its embodiments, wherein the optical arrangement is configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object in a collection of waste (such as a waste stream or pile), wherein the optical detection device is configured to, in use, detect at least one optical signal resulting from the irradiation of the object in the collection of waste.

[0017] The waste inspection apparatus may further include a conveyance device (such as a conveyor belt) configured to, in use, convey the collection of waste into an irradiation and detection region of the Raman spectroscopy apparatus.

[0018] The waste inspection apparatus may include a waste sorting device configured to, in use, sort the collection of waste in accordance with the detected at least one optical signal. The waste sorting device may be automated or may be manually operated by a human operator.

[0019] The waste inspection apparatus may further include a waste composition analysis module configured to, in use, process the at least one detected optical signal, or at least one component of the at least one detected optical signal, so as to analyse a composition of the collection of waste. The result(s) of the analysis may be used by an algorithm or computer program for further analysis or processing. The result(s) of the analysis may be used to aid the sorting of the collection of waste or as an indicator of quality control. In other embodiments, the analysis may be viewed by a human operator. In further embodiments it may be both viewed by a human operator and used by an algorithm or computer program for further analysis or processing.

[0020] The waste composition analysis module may include a processor and memory including computer program code, the memory and computer program code configured to, with the processor, enable the waste composition analysis module to perform its function(s). The waste composition analysis module may be, may include, may communicate with or may form part of one or more of an electronic device, a portable electronic device, a portable telecommunications device, a microprocessor, a mobile phone, a personal digital assistant, a tablet, a phablet, a desktop computer, a laptop computer, a server, a cloud computing network, a smartphone, a smartwatch, smart eyewear, and a module for one or more of the same.

[0021] According to a third aspect of the invention, there is provided a method of using a Raman spectroscopy apparatus according to any one of the first aspect of the invention and its embodiments, the method comprising: by the deep ultra-violet electromagnetic radiation source, generating a deep ultra-violet electromagnetic radiation beam; by the optical arrangement, guiding the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object; and by the optical detection device, detecting at least one optical signal resulting from the irradiation of the object.

[0022] According to a fourth aspect of the invention, there is provided a method of operating a waste inspection apparatus according to any one of the first aspect of the invention and its embodiments, the method including performing a method according to any one of the third aspect of the invention and its embodiments, the method further comprising: by the optical arrangement, guiding the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object in a collection of waste; and by the optical detection device, detecting at least one optical signal resulting from the irradiation of the object in the collection of waste. The method of operating a waste inspection apparatus may further include, by the waste sorting device, sorting the collection of waste in accordance with the detected at least one optical signal.

[0023] In a recycling facility, polymer waste objects must first be sorted into streams of polyethylene, polypropylene and so on in order to obtain chemically pure recyclates. Conventionally optical instruments / spectrometers in recycling facilities have difficulty identifying black / dark plastic material due to either light absorption or accompanying fluorescence.

[0024] In the invention, the deep ultra-violet (UV) Raman spectrometer is used to aid identification of plastic material, which are not readily identifiable with currently used techniques. Raman spectroscopy is an identification / analysis technique, which can be used to identify polymer materials based on their molecular vibrational spectral fingerprint. Operating at deep-UV wavelengths (<250 nm) not only makes the detection insensitive to highly fluorescent compounds such as plastic additives, pigments / dyes, food remains and dirt but also prevents interference from environmental visible lighting, such as factory lights and sunlight.

[0025] The invention is preferably configured as an in-line stand-off spectrometer for object identification (optionally in combination with near-infrared (NIR) cameras) to achieve high recycling and sorting rates. The invention can be additionally used for post-sorting validation of a recycled waste stream or a sample thereof.

[0026] The invention is applicable to black, non-black, dark, light and coloured object identification, particularly black, non-black, dark, light and coloured plastics identification. Expanding the range of object identification not only enables a large percentage of waste recycling from a collection of waste but also increases the purity of the sorted recyclates, thus adding reusability and market value to the recyclates.

[0027] Powerful deep-UV electromagnetic radiation sources may be used to improve the accuracy and speed of object identification and / or enable parallel identification of several objects.

[0028] Within the scope of this application it is expressly intended that the various aspects, embodiments, examples and alternatives set out in the preceding paragraphs, and the claims and / or the following description and drawings, and in particular the individual features thereof, may be taken independently or in any combination. That is, all embodiments and all features of any embodiment can be combined in any way and / or combination, unless such features are incompatible. The applicant reserves the right to change any originally filed claim or file any new claim accordingly, including the right to amend any originally filed claim to depend from and / or incorporate any feature of any other claim although not originally claimed in that manner.

[0029] Preferred embodiments of the invention will now be described by way of non-limiting examples with reference to the following drawings, in which :

[0030] Figure 1 shows a Raman spectroscopy apparatus according to an embodiment of the invention;

[0031] Figure 2 shows a comparison between the Raman spectra of polymers measured by deep ultraviolet Raman spectroscopy and Raman spectroscopy with a conventional 532 nm green laser;

[0032] Figure 3 shows a comparison of Raman spectra of polymers of different colours measured by deep ultraviolet Raman spectroscopy and Raman spectroscopy with a conventional 532 nm green laser;

[0033] Figure 4 shows exemplary Raman spectra of different samples collected according to an embodiment of the invention;

[0034] Figure 5 shows the effect on signal strength when altering the working distance of an embodiment of the invention;

[0035] Figure 6 shows the effect on signal strength when altering the speed of a conveyance device according to an embodiment of the invention;

[0036] Figure 7 shows a comparison of two classification algorithms for determining plastic composition from the Raman spectra.

[0037] The figures are not necessarily to scale, and certain features and certain views of the figures may be shown exaggerated in scale or in schematic form in the interests of clarity and conciseness.

[0038] The following embodiments are described with reference to a waste inspection apparatus for plastic recycling or sorting, in particular black and dark plastics identification for sorting and recycling with deep-UV Raman spectroscopy. It will be appreciated that the following embodiments are applicable to non-black, light and coloured object identification and to other types of apparatuses and applications that do not involve waste inspection and plastic recycling or sorting. Plastic recycling entails many challenges, including the processing of large volumes of inhomogeneous plastic waste of different colours, sizes and polymer type. Polyethylene (PE), polypropylene (PP) and polyethylene terephthalate (PET) are of particular interest. Conventional hyperspectral imaging in the near-infrared (NIR) range is one of the leading techniques for plastics identification for sorting purposes, but still falls short when measuring dark or black plastics. Conventional Raman technology at visible wavelengths is accompanied by a very strong fluorescence background, which makes identification of these plastics challenging.

[0039] A waste inspection apparatus according to an embodiment of the invention may include a Raman spectroscopy apparatus for outputting an irradiation beam and receiving an optical signal. The waste inspection apparatus may further include an object identification module or a conveyance device or both. An object identification module may be included in the Raman spectroscopy apparatus. An embodiment of a Raman spectroscopy apparatus according to the invention is shown in Figure 1 and is generally designated by the reference numeral 20.

[0040] The Raman spectroscopy apparatus 20 includes an optical arrangement comprising at least one parabolic optical element. In the embodiment of Figure 1, the Raman spectroscopy apparatus 20 comprises two off-axis parabolic mirrors 22a, 22b. In another embodiment, there may be only one parabolic optical element, or there may be additional parabolic optical elements. In further embodiments, the at least one parabolic optical element may be a parabolic optical reflector. In yet further embodiments, the at least one parabolic optical element may have a diameter of 25 mm. In other embodiments, the diameter of the at least one parabolic optical element may be greater than 25 mm, or may be smaller than 25 mm.

[0041] The Raman spectroscopy apparatus 20 comprises a deep ultra-violet electromagnetic radiation source 30 for generating a deep ultra-violet electromagnetic radiation beam. Figure 1 illustrates an irradiation path ('excitation path') of the Raman spectroscopy apparatus 20 from the deep ultra-violet electromagnetic radiation source 30 towards an object 32, wherein the object is a polymer or plastic sample or waste product to be scanned. The deep ultra-violet electromagnetic radiation source 30 may be a laser, and the deep ultra-violet electromagnetic radiation beam may be a laser beam. The irradiation path of the pictured embodiment travels through the optical arrangement and onto the object 32. That is to say, the optical arrangement is configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source 30 along the irradiation path towards the object 32. In the pictured embodiment, the deep ultra-violet electromagnetic radiation beam travels along the irradiation path from the radiation source 30, is reflected by mirrors 24a, 24b, 24c and by the long pass filter 26 before passing through a lens 28. Each mirror is preferably a planar mirror. The lens may be, for example and not limitation, a quartz lens. The lens 28 may be an adjustable or movable lens. After passing through the lens 28, the radiation beam is then redirected by mirror 24d to the parabolic optical elements 22a, 22b at the end of the irradiation path. The parabolic optical elements 22a, 22b directly guide the deep ultra-violet electromagnetic radiation beam to be incident on the object.

[0042] Optical arrangements of further embodiments may include a different number of mirrors or parabolic optical elements. In yet further embodiments, the optical arrangement may not comprise a lens 28, or may not comprise any planar mirrors, according to the size or footprint requirements of the Raman spectroscopy apparatus 20. In some embodiments, the deep ultra-violet electromagnetic radiation may be guided directly to the object by a parabolic optical element, and the object may be illuminated or irradiated. In other embodiments, there may be other optical elements between a parabolic optical element and the object 32, for example but not limitation, a lens, a planar mirror or a filter.

[0043] The irradiation of the object results in the generation of an optical signal in the form of Raman backscattered light, which travels along a Raman signal return path ('collection path')- In the embodiment shown, the Raman signal return path is aligned (i.e. colinear) with the irradiation path until the Raman signal return path reaches the long pass filter 26, whereupon the returning Raman signal passes through the filter instead of being reflected. The returning Raman signal passes through the lens 28 and line filter 26 into an optical detection device of the Raman spectroscopy apparatus 20. In the pictured embodiments, the optical detection device comprises a spectrograph 34 and a camera 36 (such as a charge-coupled device (CCD)). Other embodiments may have a combined spectrograph and camera, a different optical detection device, or a different method of detecting the Raman signal and recording the resultant Raman spectra.

[0044] Figure 2 shows a comparison between the recorded Raman spectra of reference polymers achieved using a deep ultra-violet laser beam according to an aspect of the invention (indicated by "UV") and using a conventional 532 nm green laser beam (indicated by "Green"). The y-axis numbers are the counts of the DUV spectrometer, while the conventional Raman spectra are shown on a comparable scale in arbitrary units. The PE and PP spectra integration time was 10 sec, while the PET spectra acquisition was integrated over 200 sec. The asterisks indicate two air peaks: O2 at 1555 cm , and N2 at 2329 cm4. The full circles indicate two overtones / combinations of the PET peaks. The background signal is reduced in the returning Raman signal from the deep ultra-violet laser beam when compared with the 532 nm green laser beam. This effect is especially prominent in the spectrum of the PET reference polymer, where many peaks in the spectrum resulting from illumination with the 532 nm green laser are indistinct, though the background in the signal is reduced for every reference polymer measured.

[0045] Figure 3 shows the Raman spectra on white and black HDPE using a deep ultra-violet laser according to an embodiment of the invention, and a visible 532 nm green laser beam for comparison. The Raman spectra achieved using a visible 532 nm green laser on black coloured PE has a high background due to fluorescence, and a weaker Raman signal due to absorption such that no peaks are visible and the polymer composition of the object cannot be identified. The same measurement when performed on a white coloured PE sample results in a lower background, but the background is still notably higher than the background signal in the spectra taken using a deep ultra-violet laser beam. The Raman spectra taken according to the present invention has a far smaller background fluorescence, and even very weak signals are possible to capture and utilise for the determination of sample polymer composition. In DUV Raman, the Raman signal from the white HDPE is stronger due to lower absorption. The black HDPE spectrum shows more fluorescence at high wavenumbers.

[0046] Exemplary Raman spectra achieved using an embodiment of the Raman spectroscopy apparatus 20 are shown in Figure 4. The spectra were taken from real waste products with reflective or black surfaces which are difficult to identify by existing means. The "chocolate metallized package" was measured for 20 seconds to achieve the pictured spectra. The "chips metallized package" was measured 3 times for 1 second and an average was taken to achieve the pictured spectra.

[0047] A waste inspection apparatus according to an embodiment of the invention may further include a conveyance device (e.g. a conveyor belt) having a conveyance speed and a distance from the Raman spectroscopy apparatus 20. The conveyance device is configured to, in use, convey a waste stream into an irradiation and detection region of the Raman spectroscopy apparatus 20. In a particular embodiment, the Raman spectroscopy apparatus 20 is arranged above the conveyance device so that the optical arrangement is configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source 30 along an irradiation path towards an object in the waste stream, and so that the optical detection device is configured to, in use, detect the resulting Raman signal resulting from the irradiation of the object in the waste stream.

[0048] The optical arrangement of the Raman spectroscopy apparatus 20 has an optimal working distance from the object 32. The effect of working distance on the strength of the Raman signal is shown in Figure 5, which illustrates the determination of focal depth of the standoff design in Figure 1. Measurements of PE powder (reference material) were taken at different distances from the last parabolic mirror. Each measurement (1 sec) was repeated three times and averaged. Distance 0 is the optimal working distance of the last parabolic mirror (19 cm from its centre). In Figure 5 (right side), the maximum of each spectrum from Figure 5 (left side) is plotted against the distance. The error bars represent the standard deviation.

[0049] The optimal working distance results in the strongest signal and the highest peak counts in the Raman spectra. It is shown in the figure that it is still possible with an embodiment of the Raman spectroscopy apparatus 20 to receive a usable signal from objects 32 placed further away or closer to the optical arrangement than the optimal working distance. An object placed at the optimal working distance from the optical arrangement yields the strongest Raman signal.

[0050] The distance of the conveyance device from the Raman spectroscopy apparatus 20 is such that the objects 32 to be scanned are within the range of acceptable working distances away from the optical arrangement. To this end, the waste inspection apparatus may further include height sensors, monitors or detectors to detect the height of incoming objects on the conveyance device and thereby measure the distance between the optical arrangement and the object. An embodiment of the invention may further include a height adjustment system, such as a vertically movable platform, which adjusts the distance between the objects 32 and the Raman spectroscopy apparatus 20 to ensure that an incoming object 32 is within the range of acceptable working distances from the optical arrangement of the Raman spectroscopy apparatus 20. This may include adjusting the height of the Raman spectroscopy apparatus 20, or adjusting the height of the conveyance device, or a combination. The height adjustment may be carried out responsive to a measured distance between the optical arrangement and the object falling outside an acceptable range. In an embodiment, the acceptable working distance may be at least 12 cm, preferably between 12 cm and 50 cm. In another embodiment, this range of acceptable working distances may be different. In further embodiments, the ideal working distance and the range of the acceptable working distances of the Raman spectroscopy apparatus 20 may also be adjusted by moving a movable lens in the optical arrangement. This lens may be provided in addition to the optical arrangement shown in Figure 1, or the lens 28 may be configured to be movable so as to adjust the working distance.

[0051] The conveyance device's speed is selected to allow the Raman spectroscopy apparatus 20 to achieve a strong and accurate returned Raman signal from the object 32. The returned Raman signals for different conveyance device speeds are shown in Figure 6, which shows measurements of a black polyethylene object (in this case a bottle), 10 cm long, at different conveyor belt speeds; 0 represents stationary. Each spectrum is an average of 3 independent measurements with integration time of Is. Figure 6 (left side) shows the full spectra of the black bottle at an increasing speed. In Figure 6 (right side), the signal at 2874 cm1from each averaged spectrum is plotted against the speed. Error bars show the standard deviation.

[0052] The slower the conveyance device speed, the stronger and less noisy the signal. This is to be balanced with the speed required by a waste inspection, identification and sorting facility. The signal strength may also be adjusted by adjusting the number of samples, sampling rate and integration time. An object length monitor may be used to determine the length of the object 32 to be scanned. An object length monitor may be integrated with a height monitor or may be a separate device or camera. The conveyance device speed may be adjusted depending on the length of the object 32 to be scanned to allow to the Raman spectroscopy apparatus 20 to receive a clear enough signal. If the object is small, a slowing of the conveyance device may be advantageous. If the object is large, a faster conveyance may be advantageous.

[0053] The Raman spectroscopy apparatus 20 may further include an object identification module and a waste composition analysis module. This object identification module may process a detected optical signal, for instance a Raman signal, or a component of a detected optical signal, in order to identify the object. In some embodiments, the spectrograph 34 and camera 36 are integrated with the object identification module. In other embodiments they may be separate. For example, the object identification module may form part of a processor remote from the spectrograph 34 and camera 36 by way of wired or wireless communication. In further embodiments, the object identification module may be configured to distinguish the identified object 32 from another object. This may include identifying the object 32 as a plastic object or distinguishing the identified object 32 from a non-plastic object. The waste composition analysis module may be configured to, in use, determine the plastic composition of the object. The object identification module and the waste composition analysis module may be separated modules or integrated as a single module.

[0054] The identification of the object may be achieved by comparison of the detected optical signal with a reference signal (e.g. an optical signature) from a known object. In some embodiments the identification of the object may be performed visually by a human operator. In other embodiments, the identification of the object, either by comparison with a reference signal or otherwise, may be performed by the object identification module. The object identification module may include a classification algorithm or a look-up table including a plurality of stored optical signatures for this purpose, or may include both.

[0055] Object identification according to an aspect of the invention was able to identify the plastic composition of the objects for which the Raman spectra are shown in Figure 4. The "chocolate metallized package" was identified from the pictured spectra as being made from PET. The "chips metallized package" was identified as PE. The "yoghurt bottle with a black coated sleeve" was identified as being PET.

[0056] Figure 7 shows a comparison of two classification algorithms which may be included in the object identification module. The simple partial least square (SPLS) and the neural network (NN) were both given access to the spectra of known reference materials. The algorithms were tested on both stationary objects and objects on a moving conveyance device. The object identification module may include either one of these algorithms, another classification algorithm, or a plurality of classification algorithms which are used to determine plastic composition.

[0057] Errors in plastic identification and inconsistencies between the identification of plastics classification algorithms shown in Figure 7 may be improved by increasing signal collection. This may be achieved, for example and not limitation, by increasing the power of the deep ultra-violet electromagnetic radiation source or by increasing the efficiency of collection optics. Providing algorithms with a more extensive training set may also yield improvements. Further embodiments may include a calculated quality factor in object identification, such as a comparison of peak Raman count rate with background or air count rate, for quality control and providing a confidence estimate of the object identification. The waste inspection apparatus may further include a waste sorting device which is configured to sort the waste stream in accordance with the detected optical signal, or with the determination of the polymer content by the object identification module. The waste sorting device may be automatic or may be operated by a human operator. For sorting the waste stream, the waste sorting device may include, for example but not limitation, one or more handling equipment (e.g. gripper), mechanical levers, mechanical valves, air blowers, and / or conveyance-based sorting devices (e.g. belts). Waste sorting performed as a result of waste stream analysis may be performed automatically in a computer-controlled process or it may be performed by a human operator, or a combination.

[0058] In some embodiments the Raman spectroscopy apparatus 20 or the waste inspection apparatus may include an imaging device or object detector for measuring or assessing the dimensions of the object 32. The imaging device may adjust the measurement time, illumination time, exposure time or another measurement parameter in response to a measured dimension of the object 32.

[0059] The Raman spectroscopy apparatus 20 is able to receive a clear Raman signal from materials which are otherwise difficult to assess, including colourful or black plastics. The inclusion of a deep ultra-violet electromagnetic radiation source in the apparatus minimises potentially problematic fluorescence which can "drown out" the weaker Raman signals received when measuring colourful or black plastics. Conventional Raman spectroscopy using lasers in the visible spectrum, such as the 532 nm green laser used in the comparisons of Figures 2 and 3, induces fluorescence in the samples which decreases the signal quality of black and colourful plastics when compared to the signal quality of white plastics. For this reason, visible spectrum lasers are not ideal for use in recycling facilities, waste inspection apparatus and other waste categorisation and sorting systems, methods, apparatuses, or facilities. Deep ultraviolet Raman spectroscopy does not require isolation from ambient visible spectrum light or factory lights, unlike visible light Raman spectroscopy. The Raman spectroscopy apparatus 20 according to the present invention works in well-lit conditions, which allows for facility monitoring and quality checking by visible light spectrum sensors, monitors or cameras as well as visual monitoring by human operators and monitoring staff. The spectra shown in Figures 4, 5 and 6 were acquired under ambient factory conditions. A method of operating a waste inspection apparatus according to the present invention may further include a quality checking system or assessment which validates or checks the results. The quality checking system may be achieved by a second or further inspection by an embodiment of the Raman spectroscopy apparatus 20, or may be performed by another apparatus or system, or may be performed by a human operator. The quality assessment and checking may be any combination of these.

[0060] Experiments

[0061] Materials

[0062] Reference materials of PET, PP pellets and PE powder were purchased from Sigma Aldrich (429252, 428175, 332119, respectively). Other measured objects were selfcollected consumer items, bought in local stores, or received from the National Test Centre Circular Plastics (NTCP) recycling research facility. Only plastic materials identifiable with a recycling code were included. Conventional (visible) Raman spectra were collected with a Renishaw Invia spectrometer using a 532 nm excitation green laser and 1800 lines / mm grating.

[0063] Setup

[0064] The experimental setup is based on the use of a deep-UV (DUV) laser beam at 248.6 nm to irradiate the sample ('object') and collecting the Raman backscattered light, while rejecting the Rayleigh wavelength with an optical filter, as per the apparatus shown in Figure 1. The laser (NeCu 70-248, Photon Systems, Covina, USA), equipped with a laser line filter system to reject plasma lines, emits 10-80 ps long pulses at a maximum repetition rate of 80 Hz. The average power of the beam was 0.8 mW at the laser output, and 0.6 mW at the sample. The beam was sent to a long pass filter (LP02-248RS-25, Semrock), positioned at 10° to reflect the light so that the excitation and collection paths become colinear, thus serving as a dichroic mirror. The filter's 10° tilt does not change its cut-off wavelength significantly so that it still rejects the laser light effectively and enables collection of the Raman signal above 650 cm1.

[0065] The beam then passed through a quartz lens and reflected off a pair of large off-axis parabolic mirrors with a diameter of 50 mm and an effective focal length of 19 cm (#37-996, Edmund Optics). The light scattered back from the sample was coupled to the spectrograph (Semrock SR-303i - A spectrograph with a 2400 line / mm grating) and an Andor™ technology NewtonEMEMCCD camera (DU-970 P-UVB). The slit width was set to 80 pm and the electron multiplier DAC was set to 200 at full vertical binning. The parabolic mirrors and lens pair were positioned in a 4f setup so that the collected light was focused on the slit. The setup was built on a 50x100 cm breadboard, thus making it transportable between locations (e.g. lab and recycling research facility, NTCP).

[0066] For experiments simulating a factory environment, the setup was mounted above a conveyor belt at the NTCP recycling research facility (Heerenveen, the Netherlands). The setup was shielded for eye safety reasons and from ambient light, by enclosing it with black covers at the top and sides. The only way for light to enter the setup was from below where the parabolic mirrors were mounted, but this did not affect the background in the DUV range. The focal point of the setup (19 cm below the middle of the last parabolic mirror) was positioned 3 cm above the conveyor belt. The free space under the setup was 18 cm, meaning that only items smaller than that were allowed to pass under the setup.

[0067] Analysis

[0068] Before processing with any of the classification methods, the Raman spectra underwent pre-processing that includes spectrum truncating between 1000 and 3200 cm , Savitzky-Golay smoothing (5 pixels window), background subtraction (with 6thorder polynomial fitting) and normalization using standard normal variate (SNV).

[0069] The simple partial least squares (SPLS) method integrated over seven spectral regions relevant for PP, PE and PE vibrational modes (1268-1303 cm4, 1309-1375 cm4, 1590- 1635 cm1, 1695-1734 cm1, 2820-2855 cm1, 2860-2935 cm4and 2935-2995 cm4). Especially interesting is the C-H stretch range, since it shows more counts per exposure time than the fingerprint region, and a complex, albeit dense, spectral structure. Preliminary lab testing showed good identification when items were kept stationary at the optimal working distance.

[0070] A neural network (NN) was trained using a MATLAB 2021a classification trainer. Out of all methods (including others like classification tree), an NN with one narrow layer showed the best result for the trained set (layer size 10 nodes, Activation ReLU). The first 6 components of principal component analysis (PCA) were used as input for the NN. The training set included 36 waste items (mostly black items) and white reference materials. The items were measured under different conditions, changing angles, height and exposure time, amounting to a data set of 101 measurements. The lab training showed satisfactory results (100 % identification). Testing under lab conditions (stationary objects at optimal distance) agreed as well with this result.

[0071] Experimental results

[0072] DUV Raman spectra of reference plastics showed largely the same fundamental peaks as conventional Raman spectra with 532 nm green laser excitation. Figure 2 shows a comparison between the DUV and visible wavelength Raman spectra of PET, PE and PP. Although the spectral resolution is not the same in both modalities, the shape of the spectrum is still indicative for the material type. In order to obtain the reference spectra, DUV Raman measurements were 10 s long, except the PET DUV measurement which was 3 minutes long.

[0073] The Raman signal of PET in the DUV region is weaker than that of the other polymers, so the air peaks can be seen and have a comparable intensity. Moreover, unlike in conventional Raman, the DUV Raman spectra of PET show strong overtones at high Raman shifts, originating from the aromatic ring vibrations of PET. In spite of this resonance enhancement, the measurement time for PET was longer and the Raman signal lower. The inventors attribute this to strong absorption of DUV light by PET, both the laser beam and the Raman photons, effectively generating Raman signal from only a thin layer at the top of the PET sample.

[0074] Comparing the DUV spectra of black and white HDPE waste items in Figure 3, these materials show very similar spectral signatures. However, the white HDPE Raman signal is slightly stronger than the signal of the black one, presumably due to lower absorption Both samples show the onset of fluorescence above 3400 cm , which corresponds to 270 nm. This fluorescence background does not interfere with the DUV Raman spectrum.

[0075] Further experiments were performed at the NTCP plastic recycling research facility to test the performance under real conditions, such as noise, vibrations and factory lights. The latter did not show any influence on the noise level of the DUV measurements compared to the measurements in the dark or lab measurements. Since the plastic material sent to recycling is highly inhomogeneous in size, the DUV spectrometer standoff setup in Figure 1 was designed for flexibility in working distance. The optimal working distance was 19 cm from the centre of the last parabolic mirror. In order to examine the effective working distance, the inventors measured the reference PE powder sample at different heights above and below the optimal focal distance. The resulting spectra are shown in Figure 5 (left side), with the maximum intensities as a function of distance plotted in Figure 5 (right side). Analysing the DUV Raman response curve in Figure 5 results in a FWHM of 3.75 cm. This range corresponded with plastic waste objects of different heights, roughly from 1 to 5 cm, traveling on the conveyor belt.

[0076] To test the data collection on a moving conveyor belt, the inventors measured the same black polyethylene object, 10 cm long, at different conveyor belt speeds, where 0 means stationary (see Figure 6). All measurement times were set to Is. The measurement was started manually as the object edge reached the irradiation and detection region of the Raman spectroscopy apparatus 20. A stationary speed and a slow speed (0.1 m / s) allowed a full Is integration of the Raman signal, which decreased at higher conveyor belt speeds. Although at high speeds the effective integration time is only a fraction of Is, the sample was successfully identified as PE, even at the maximum speed of 0.9 m / s. The air Raman peaks were constant throughout these measurements. In Figure 6 (right side), the maximum of the PE C-H stretch peak at 2874 cm1is plotted against the conveyor belt speed, and the decrease in signal shows 1 / speed behaviour.

[0077] Additionally, the DUV setup was tested on other materials, which are also difficult to identify with NIR. For example, snack packages are usually metalized on the inside, and are a challenge for NIR-based identification. A chocolate wrapper was measured for 20 seconds and identified as PET. The potato chips bag, a multi-layered material, measured for Is and averaged for 3 times, had a Raman signal similar to PE. The yoghurt bottle with a black coated sleeve around it, was identified as PET. The DUV Raman spectra are shown in Figure 4.

[0078] In order to be able to use the DUV spectrometer above a conveyor belt, an automated identification algorithm was realised. The inventors tested two algorithms, representing two different approaches and shown in Figure 7. One was a simple partial least square (SPLS) for the detection of PET, PE and PP, as well as denoting 'Other' for un-identified objects. The SPLS algorithm was calibrated on the reference materials. The second algorithm was a neural network (NN) based algorithm trained by measurements in the lab. Both algorithms were tested over the conveyor belt in stationary mode and moving mode (at 0.1 m / s). In total, 45 items were measured in triplicate in stationary mode, and 37 items were measured in triplicate in moving mode. Acquisition time for all measurements was 1 second. Figure 7 illustrates a comparison of two algorithms used for the analysis of the measurements on the conveyor belt. Figure 7 (left side) show the results of SPLS analysis for stationary and moving measurements above the conveyor belt, respectively. In Figure 7 (right side), the neural network was used to analyse the same data. The bar height shows the percentage of identified objects, with the imprinted recycling code taken as the ground truth.

[0079] The results of SPLS analysis identification of non-moving (stationary) samples on the conveyor belt are shown in Figure 7 (left side). The results of SPLS analysis identification of moving samples on the conveyor belt are shown in Figure 7 (right side). The results show that above 80% of PE and about 50% of PP samples were identified as such, and less for PET. The moving samples showed lower identification rates, and some misidentifications (false positives). The PET identification rate remained about the same.

[0080] Using the neural network, the inventors were able to identify PE and PP in the stationary measurements with similar success rates as with SPLS. For both polymers, these rates decreased when the conveyor belt was moving.

[0081] Discussion

[0082] The DUV Raman spectroscopy apparatus presented here was designed to identify plastics, with emphasis on black plastics, in order to assess the potential for future use in plastic recycling. In Figure 2, it is seen that, for PE and PP, two of the most common plastic types in packaging and recycling, the DUV spectra are similar to the conventional Raman spectra in the visible wavelength range. The PET spectrum does show differences. Firstly, the C-H stretch peaks around 3000 cm1were not detected. Secondly, overtones of the 1615 cm1peak and the combination band of this peak with the weaker 1725 cm1peak were observed. Thirdly, the photon counts per unit time were lower than that of PP and PE, and therefore required more integration time. A longer integration time, in turn, made the air Raman peaks more apparent in the spectrum. The relatively weak Raman signals are attributed to more absorption by the aromatic groups in PET, resulting in a smaller focal volume generating signal. Air peaks could be a good measure for quality control, since they are mostly constant throughout the measurements. This is especially true for black samples, where there is only little backscattering from the sample. Lastly, the PET reference material showed some fluorescence background in visible Raman, and no background in the DUV spectrum. The DUV spectra of black and white of common objects (non-reference materials) are comparable, as shown in Figure 3. Differences arise from the weaker back scattering and increased absorption of the black plastic, as well as increased fluorescence at high wavenumbers above 3400 cm . Unlike the case of conventional (visible wavelengths) Raman shown in Figure 3, with DUV excitation the fluorescence does not overlap with the Raman spectrum.

[0083] The system is built with F / 4, which is a high F / # compared to common Raman microscopes. Due to the low Raman cross-section, efficient collection optics are required for Raman spectroscopy, which usually leads to the implementation of high NA objectives (low F / #). The inventors balanced this requirement with the need to implement a "stand-off" design. Using 50 mm parabolic mirrors with a long (e.g. 19 cm) effective focal length allowed the implementation of the Raman spectroscopy apparatus above a conveyor belt with objects of varying heights. The F / 4 additionally allowed some flexibility of the working distance. Figure 5 (right side) shows that the FWHM of the working distance is almost 4 cm.

[0084] Figure 7 shows the implementation of two different approaches to analyse factory collected data based on calibration / training on lab data. Both approaches lead to similar results, which might indicate that the low rate of identifying plastics is linked to the poor collection of the Raman signal for objects of varying height (see Figure 5), and even more so when these objects are moving and the effective integration time decreases (see Figure 6). Overcoming these issues, to allow real time identification, will require different solutions, non-limiting examples of which are outlined as follows.

[0085] Increasing the Raman signal reaching the spectrograph would improve the Raman spectrum. To that end, first and foremost, a stronger laser should be used to compensate the limitations imposed by the stand-off design and inhomogeneous nature (e.g. size differences) of the measured objects. The average output power of the NeCu laser used here was only 0.8 mW. For waste objects on a moving conveyor belt, the downside of higher laser powers (e.g. photodegradation) should not be an issue. Secondly, combination with other sensors (e.g. a machine vision or an RGB camera) could allow adaptation of the exposure time to the object's length (in the direction of the conveyor belt's movement). A longer exposure for a long object will allow more integration of the Raman signal, while for a shorter object less exposure time will limit the noise level. Thirdly, a combination with a height sensor could enable focal adaptation. Adapting the focal distance to the measured object could recover the signal over a longer range than shown in Figure 5, and achieve identification rates closer to the lab results. Lastly, even with increased signal, the classification methods discussed here should be addressed.

[0086] Two classification methods were tested, SPLS and NN. SPLS is a simplistic method which relies on the reference spectra acquired in the lab to classify the lesser quality spectra acquired under more challenging conditions. For the task in question, it cannot identify objects when the signal-to-noise (SNR) is low. Similarly, the NN-based classification also fails when the SNR is poor but is expected to improve with more training data. Additionally, the classification algorithms do not account for the quality of the spectra analysed. A quality score could indicate the strength of classification, for example, by comparing the air peak counts to the sample Raman peaks counts. The inventors consider the bottle neck to lie in the signal collection, rather than the classification method.

[0087] An advantage of performing Raman at DUV wavelengths is avoiding the fluorescence at longer wavelengths, as demonstrated in Figure 3. Furthermore, the Raman DUV detection occurs away from the visible spectrum, which allows this setup to work under ambient, well-lit conditions with minimal or no added background noise. This is in contrast with visible wavelength (ex. 532 nm) Raman spectroscopy, which requires either isolation from ambient light, or gating of the signal to minimize external background. Notably, time gating is a costly solution. Also, in spite of the rather low power and low NA used for this study, the wavelength dependency of the Raman signal (scaling with VL4) is expected to yield about 20 times more Raman signal than visible wavelength Raman, which somewhat compensates for those factors. Lastly, as shown in Figure 4, the inventors demonstrated the potential of DUV Raman to acquire spectra of more complex objects such as metalized plastics.

[0088] Summary

[0089] The inventors have discovered that Raman spectrometry in the DUV wavelength range is suitable for black plastics, as the fluorescence does not overlap with the Raman spectrum. The inventors have devised a dedicated DUV setup, based on a 248.6-nm HeCu laser and designed for stand-off black plastics detection above a conveyor belt, with the main focus being polymers such as PE, PP, and PET.

[0090] Dark and especially black plastics have a strong absorption that may extend from the visible (hence they appear black) to the NIR wavelengths. This is especially the case for black carbon, which is commonly used to colour many commercial plastics. The strong absorption reduces the back scattering of the light from the object, hampering NIR identification. The detection of such strongly absorbing materials is also challenging for other techniques, such as conventional Raman spectrometry. When using visible wavelengths for excitation, the Raman scattered light is accompanied by a very strong fluorescence background in the same wavelength range, such that its noise surpasses the Raman signal intensity. Hence, avoiding this fluorescence can be key for the detection of black plastics.

[0091] DUV excitation light also creates fluorescence. However, the Raman spectrum will always have the same Raman shift, which means that the wavelength will change with the excitation wavelength and Raman photons will also be detected at very short wavelengths. Even in very complex samples there is usually no fluorescence at wavelengths shorter than ca. 270 nm. The essence of using DUV excitation wavelengths (<250 nm) is that black plastics could be made detectable and identifiable in the wavelength range of <270 nm, and the Raman and fluorescence spectra would not overlap. With short-wavelength excitation, one also profits from the fact that the Raman intensity scales with the laser wavelength to the power (-4), and that for some polymers there is also resonance enhancement. Additionally, the detection at DUV wavelengths means that overhead factory light, sun light or any other visible light sources would not interfere with the measurement.

[0092] Plastic sorting for recycling is made under certain conditions that impose unique requirements. Plastic waste streams for mechanical recycling often pass on a conveyor belt. Detection should be fast, robust and automated, to allow large volumes to be sorted efficiently and effectively. The objects are inhomogeneous in size. To accommodate that, the measuring setup should be of a stand-off design, meaning that it should hang above the conveyor belt, leaving enough room for objects to pass under. The Raman photon collection should not depend critically on distance.

[0093] The listing or discussion of an apparently prior published document or apparently prior published information in this specification should not necessarily be taken as an acknowledgement that the document is part of the state of the art or is common general knowledge.

[0094] Preferences and options for a given aspect, feature or parameter of the invention should, unless the context indicates otherwise, be regarded as having been disclosed in combination with any and all preferences and options for all other aspects, features and parameters of the invention.

Claims

CLAIMS1. A Raman spectroscopy apparatus comprising: a deep ultra-violet electromagnetic radiation source for generating a deep ultraviolet electromagnetic radiation beam; an optical arrangement configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object, the optical arrangement comprising at least one parabolic optical element; and an optical detection device configured to, in use, detect at least one optical signal resulting from the irradiation of the object.

2. A Raman spectroscopy apparatus according to Claim 1 wherein the or each parabolic optical element is a parabolic optical reflector and / or an off-axis parabolic optical element.

3. A Raman spectroscopy apparatus according to any one of the preceding claims wherein the optical arrangement includes a plurality of offset parabolic optical elements4. A Raman spectroscopy apparatus according to any one of the preceding claims wherein a diameter of the at least one parabolic optical element is at least 25 mm.

5. A Raman spectroscopy apparatus according to any one of the preceding claims wherein the at least one parabolic optical element is arranged at an end of the irradiation path so that, in use, the at least one parabolic optical element directly guides the deep ultra-violet electromagnetic radiation beam to the object.

6. A Raman spectroscopy apparatus according to any one of the preceding claims wherein a working distance of the optical arrangement is at least 12 cm.

7. A Raman spectroscopy apparatus comprising: a deep ultra-violet electromagnetic radiation source for generating a deep ultraviolet electromagnetic radiation beam; an optical arrangement configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object, wherein a working distance of the optical arrangement is at least 12 cm; andan optical detection device configured to, in use, detect at least one optical signal resulting from the irradiation of the object.

8. A Raman spectroscopy apparatus according to any one of the preceding claims including an object identification module configured to, in use, process the at least one detected optical signal, or at least one component of the at least one detected optical signal, so as to identify the object.

9. A Raman spectroscopy apparatus according to Claim 8 wherein the identification of the object includes identification of the object as a polymer or plastic object.

10. A Raman spectroscopy apparatus according to Claim 9 wherein the identification of the object includes distinguishing the identified object from another object.

11. A Raman spectroscopy apparatus according to any one of Claims 8 to 10 wherein the object identification module is configured to, in use, compare the at least one detected optical signal, or at least one component of the at least one detected optical signal, to a corresponding optical signature so as to identify the object.

12. A Raman spectroscopy apparatus according to Claim 11 wherein the comparison of the at least one detected optical signal, or at least one component of the at least one detected optical signal, to a corresponding optical signature includes the use of a look-up table including a plurality of stored optical signatures or the use of a classification algorithm.

13. A Raman spectroscopy apparatus according to any one of the preceding claims wherein the optical arrangement includes at least one optical lens configured to be movable to adjust a working distance of the optical arrangement.

14. A Raman spectroscopy apparatus according to Claim 13 when dependent on any one of Claims 1 to 6, wherein the at least one optical lens is arranged between the deep ultra-violet electromagnetic radiation source and the at least one parabolic optical element in the irradiation path.

15. A Raman spectroscopy apparatus according to any one of the preceding claims including a distance sensor configured to, in use, measure a distance between the optical arrangement and the object, wherein the optical arrangement is reconfigurableto adjust a working distance of the optical arrangement responsive to the measured distance between the optical arrangement and the object.

16. A Raman spectroscopy apparatus according to Claim 15 when dependent from Claim 13 or Claim 14, wherein the optical lens is configured to be movable to adjust a working distance of the optical arrangement responsive to the measured distance between the optical arrangement and the object.

17. A Raman spectroscopy apparatus according to any one of the preceding claims wherein the optical detection device includes a spectrograph and / or an imaging device.

18. A Raman spectroscopy apparatus according to Claim 17 wherein the imaging device is configured to, in use, adjust an exposure time responsive to a measured dimension of the object.

19. A waste inspection apparatus comprising a Raman spectroscopy apparatus according to any one of the preceding claims, wherein the optical arrangement is configured to, in use, guide the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object in a collection of waste, wherein the optical detection device is configured to, in use, detect at least one optical signal resulting from the irradiation of the object in the collection of waste.

20. A waste inspection apparatus according to Claim 19 including a conveyance device configured to, in use, convey the collection of waste into an irradiation and detection region of the Raman spectroscopy apparatus.

21. A waste inspection apparatus according to Claim 19 or Claim 20 including a waste sorting device configured to, in use, sort the collection of waste in accordance with the detected at least one optical signal.

22. A waste inspection apparatus according to any one of Claims 19 to 21 including a waste composition analysis module configured to, in use, process the at least one detected optical signal, or at least one component of the at least one detected optical signal, so as to analyse a composition of the collection of waste.

23. A method of using a Raman spectroscopy apparatus according to any one of Claims 1 to 18, the method comprising:by the deep ultra-violet electromagnetic radiation source, generating a deep ultra-violet electromagnetic radiation beam; by the optical arrangement, guiding the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object; and by the optical detection device, detecting at least one optical signal resulting from the irradiation of the object.

24. A method of operating a waste inspection apparatus according to any one of Claims 19 to 22, the method including performing a method according to Claim 23, the method further comprising: by the optical arrangement, guiding the deep ultra-violet electromagnetic radiation beam from the deep ultra-violet electromagnetic radiation source along an irradiation path towards an object in a collection of waste; and by the optical detection device, detecting at least one optical signal resulting from the irradiation of the object in the collection of waste.

25. A method according to Claim 24 when dependent on Claim 21, the method including: by the waste sorting device, sorting the collection of waste in accordance with the detected at least one optical signal.

Images