A FREQUENCY^CONVERSION^DEVICE^FOR^USE^WITH^AN INFRARED^ IMAGING^SYSTEM^ Technical^Field The present disclosure relates to a frequency conversion device for use with a detector for imaging a sample. The disclosure also relates to an infrared imaging system using the frequency conversion device and a corresponding method for imaging a sample. Background Typical infrared (IR) detectors or thermal imagers form an image of an object on a photocathode, leading to a discharge of electrons. The electrons are multiplied and strike a phosphor screen. The phosphor screen luminesces, and the resultant visible image is viewed by the eye or by an imaging sensor. Such detectors are bulky, heavy and monochrome. Infrared imaging and spectroscopy based on nonlinear metasurfaces has been suggested as a potential substitute for commercial IR imaging detectors, for instance see WO 2018/204991 and R. Camacho-Morales, et al., Advanced Photonics, 2021, 3, 036002. These metasurfaces are formed of arrays of densely packed nanostructures designed to manipulate the properties of incident light, which may include its amplitude, directionality, phase, polarization, and frequency. The metasurfaces can also enable strong light concentration inside them to convert the frequency of an image. The use of this process for frequency conversion eliminates the need for conversion of the optical to the electronic domain. WO 2018/204991 describes a frequency conversion device based on an array of semiconductors islands composed of a III-V semiconductor compound. III-V semiconductor compound are non-centro-symmetric materials which are employed for second order non-linear interactions, such as Second Harmonic Generation (SHG), and Sum Frequency Generation (SFG). The latter process (SFG) allows employing an extra laser (known as pump) beam to generate a new visible wave. Yet existing approaches remain limited either due to a lack of tools to improve the conversion efficiency or by the inability to extend the coverage band to longer wavelength-IR (up to 50 µm). Some metasurfaces made of Silicon and Germanium have been demonstrated for third-order nonlinear interaction, such as Third-Harmonic-Generation (THG), or Four Wave- Mixing (FWM) see for instance S. Chen, et al., Acs Photonics, 2018, 5, 1671- 1675; W. Tong, et al., Optics Express, 2016, 24, 19661-19670; M.P. Fischer, et al., Light: Science & Applications, 2018, 7, 1-7; L. Xu, et al., New Journal of Physics, 2022, 24, 035002. A so called meta-lens has also been proposed (C. Schlickriede, S. S. Kruk, L. Wang, B. Sain, Y. Kivshar and T. Zentgraf, Nano Letters, 2020, 20, 4370– 4376). This setup is limited to specific wavelengths and working distance. Also, it is not fully functioning if any part of the meta-lens is blocked. However, no third order susceptibility metasurfaces have been designed and exploited for nonlinear imaging, to date. Currently, in order to detect visible, near-infrared (NIR), and mid-infrared (MIR) lights, one need three separate cameras that employ different technologies. While visible detectors/cameras, that detect up to 1000 nm, are small, low-power consumption, and inexpensive, the other two detectors (near-infrared, and mid-infrared) are much more expensive, and consume much more space and power. It is an object of the disclosure to address one or more of the above mentioned limitations. Summary According to a first aspect of the disclosure, there is provided a frequency conversion device for use with a detector for imaging a sample; the frequency conversion device comprising a structure made of a third-order susceptibility material adapted to convert a first beam of electromagnetic radiation having a first wavelength in the infrared region into an output beam having a wavelength of less than about 1000 nm via a nonlinear process. For example the frequency conversion device may be an upgrading component to expand the operational bandwidth of the detector. The output beam may have one or more wavelengths of less than about 1000 nm obtained via one or more nonlinear processes such as four wave mixing or higher order mixing nonlinear processes. Optionally, the third-order susceptibility material is adapted to generate the output beam by mixing the first beam with at least one of a first pump beam having a second wavelength and a second pump beam having a third wavelength. Optionally, wherein at least one of the second wavelength of the first pump beam and the third wavelength of the second pump beam is greater than about 1000 nm. For instance, both the second and the third wavelengths may be greater than 1000 nm. Optionally, the structure comprises an arrangement of individual elements, wherein each individual element has a nano or a micro scale. Optionally, wherein the arrangement of individual elements is a periodic arrangement. For instance the periodic arrangement may be an array or may form a geometric lattice such as a square grid, or a hexagonal lattice etc… Optionally, the arrangement of individual elements supports resonant light matter interaction. For example the arrangement may create bound states in continuum or quasi-bound states in continuum. For example the individual elements may be arranged to obtain a plurality of resonances. Optionally, the individual elements are holes formed within a layer of third order susceptibility material. For example the holes may be dimmer holes or dimmer-airy holes. Optionally, the individual elements are nano or micro elements formed of a third-order susceptibility material. For instance the nano or micro elements may include various protruding three dimensional shapes such as disks or cylinders. The individual elements may also be nano particles or aggregates of particles. Optionally, the individual elements are provided on a substrate. For example the substrate may be made of glass, silicon oxide, quartz, sapphire, or magnesium fluoride MgF. The substrate may be a thin layer or slab. Optionally, the structure comprises several layers of individual elements. Optionally, the nonlinear process is one of a four wave mixing process and a third harmonic emission process. Optionally, the third-order susceptibility material comprises at least one of a dielectric material, a semiconductor material, an oxides based material, a transition metal dichalcogenide material, and a nitrides based material. For example the third-order susceptibility material may comprise silicon or germanium. Optionally, the frequency conversion device has an active area equal or greater than about 20 µm x 20 µm. For example the active area may be greater than 30 µm x 30 µm, or greater than about 50 µm x 50 µm or greater than about 100 µm x 100 µm. The active area may be designed to cover the area of the detector. Optionally, the individual elements within the arrangement have a same shape and a same size. Optionally, the individual elements within the arrangement have different shape and sizes. Optionally, the structure comprises multiple regions of individual elements. Optionally, each individual element within a region has a same shape and a same size, wherein the shape and size of individual elements are different for each region. Optionally, the individual elements within a region have different shape and sizes; wherein the shape and size of individual elements are different for each region. According to a second aspect of the disclosure, there is provided an imaging system for imaging a sample, the system comprising a frequency conversion device according to the first aspect; an optical arrangement configured to collect the first beam of electromagnetic radiation having the first wavelength in the infrared region and to direct the first beam onto the frequency conversion device to generate the output beam of radiation at a wavelength of less than about 1000 nm; wherein the first beam of electromagnetic radiation passes through or arises from the sample; and a detector configured to detect the output beam of radiation to obtain an image of the sample. For example the sample may be an object or a subject. For example the frequency conversion device may be integrated with the detector. For instance the structure made of a third-order susceptibility material may be provided onto a surface of the detector. Optionally, the imaging system comprises one or more radiation sources adapted to generate at least one of a first pump beam of electromagnetic radiation having a second wavelength and a second pump beam of electromagnetic radiation having a third wavelength. Optionally, the frequency conversion device is configured to resonate at one or more of the first wavelength of the first beam, the second wavelength of the first pump beam, the third wavelength of the second pump beam, or the wavelength of the output beam. Each one of the first wavelength, the second wavelength and the third wavelength may form a wavelength range extending over several wavelengths. Optionally, the system forms a sensor adapted to sense visible, near infrared and/or mid infrared radiations. According to a third aspect of the disclosure, there is provided a method of imaging a sample, the method comprising providing a frequency conversion device according to the first aspect comprising a structure made of a third-order susceptibility material adapted to convert a first beam of electromagnetic radiation having a first wavelength in the infrared region into an output beam having a wavelength of less than about 1000 nm via a nonlinear process; collecting the first beam of electromagnetic radiation; wherein the first beam of electromagnetic radiation passes through or arises from the sample; directing the first beam onto the frequency conversion device to generate the output beam of radiation; and detecting the output beam of radiation to obtain an image of the sample. Optionally, the method comprises generating a first pump beam of electromagnetic radiation having a second wavelength; and mixing the first beam and the first pump beam using the frequency conversion device to obtain the output beam. Optionally, the method comprises generating a second pump beam of electromagnetic radiation having a third wavelength; and mixing the first beam, the first pump beam and the second pump beam using the frequency conversion device to obtain the output beam. Optionally, the method comprises varying at least one of the second wavelength and the third wavelength to adjust the wavelength and intensity of the output beam. Optionally, at least one of the second wavelength of the first pump beam and the third wavelength of the second pump beam is greater than about 1000 nm. Optionally, the third-order susceptibility material is configured to generate a nonlinear signal, wherein the nonlinear signal is one of a degenerate four wave mixing emission, a non-degenerate four wave mixing emission or a third harmonic emission. According to a fourth aspect of the disclosure there is provided use of a frequency conversion device for imaging a sample; wherein the frequency conversion device comprises a structure made of a third-order susceptibility material adapted to convert a first beam of electromagnetic radiation having a first wavelength in the infrared region into an output beam having a wavelength of less than about 1000 nm via a nonlinear process. The options described with respect to the first aspect of the disclosure are also common to the second, the third and the fourth aspects of the disclosure. Description of^the^drawings^ The disclosure is described in further detail below by way of example and with reference to the accompanying drawings, in which: figure 1 is a flow chart of a method for imaging a sample according to the disclosure; figure 2A is a diagram of a system for implementing the method of figure 1; figure 2B is an exemplary target sample for use in the system of figure 2A; figure 3A is another diagram of a system for implementing the method of figure 1; figure 3B is a diagram showing direct incidence of the first beam and the pump beam onto the frequency conversion device; figure 4A is a diagram illustrating a first example of the four wave mixing process; figure 4B is a diagram illustrating the different frequencies that can be generated with a degenerate FWM; figure 4C is a diagram illustrating the different frequencies that can be generated with a non-degenerate FWM; figure 5 is a diagram of an exemplary frequency conversion designed for third-order nonlinear frequency conversion; figure 6A is a plot of the transmission spectra obtained with a frequency conversion device based on an array of nanopillars; figure 6B is a cross sectional view of a nanopillar on a substrate; figures 7A-D is a set of FWM emissions obtained for different values of pump and signal frequencies; figure 8A is an image of a target measured under white light illumination; figure 8B is an image of the target measured using the set-up of figure 3 with a signal beam at 1600 nm and a pump beam at 1100 nm; figures 9A-9C is a set of images obtained for a same target and measured using the set-up of figure 3 with different combinations of signal and pump beams; figure 10A is a plot of the transmission spectra obtained with a frequency conversion device based on another array of nanopillars; figure 10B is a cross sectional view of a nanopillar on a substrate; figure 11A-D is a set of FWM emissions obtained for different values of pump and signal frequencies; figure 12A-12D show images of a same target measured using the set- up of figure 3 with different combination of signal and pump beams; figure 13A is a plot of the transmission spectra obtained with a frequency conversion device based on an array of nano disks; figure 13B is a cross sectional view of a nano disk on a substrate. figure 14 is a table providing the simulated output wavelength λ _FWM obtained for various signal and input beams of the disk metasurface of figure 13; figure 15 is a plot showing the simulated conversion efficiency obtained for the various scenarios presented in figure 14; figure 16 is a simulation showing the electric filed enhancement obtained at different resonance frequencies; figure 17 is a diagram of an exemplary frequency conversion device including an array of micro/nano elements on a substrate; figure 18A is a plot of the transmission spectra obtained with a frequency conversion device based on an array of nano disks formed on a slab; figure 18B is a cross sectional view of a nano disk and slab on a substrate; figure 19 is a table providing the simulated output wavelength λFWM obtained for various signal and input beams of the metasurface of figure 18; figure 20 is a plot showing the simulated conversion efficiency obtained for the various scenarios presented in figure 19; figure 21 is a simulation showing the electric filed enhancement obtained at different resonance frequencies; figure 22 is a diagram of another exemplary frequency conversion device based on an array of nano holes formed within a slab; figure 23A is a plot of the transmission spectra obtained with a frequency conversion device with slab-hole configuration; figure 23B is a cross sectional view of slab with nano hole provided on a substrate; figure 24 is a table providing the simulated output wavelength λFWM obtained for various signal and input beams of the hole metasurface of figure 23; figure 25 is a plot showing the simulated conversion efficiency obtained for the various scenarios presented in figure 24; figure 26 is a simulation showing the electric filed enhancement obtained at different resonance frequencies; figure 27A is a diagram of an exemplary polarisation-sensitive frequency conversion device for THG; figure 27B is a partial top view of a metasurface showing two dimer-airy holes; figure 28 is a plot showing THG spectra measured from metasurfaces designed with different offset x0; figure 29 is a plot showing the calculated band structure for the metasurface when x0 =150 nm; figure 30 is a simulation of the nearfield electric distributions for the mode TE(3,1,1) and TE(3,2,1), respectively; figure 31 is a plot of the measured linear transmission spectrum of the metasurface under a white light source illumination; figure 32A is a set of plots showing the measured transmission spectra of metasurfaces designed with different offsets; figure 32B is a set of plots showing the angular dependence of the transmission spectrum by changing the incident angle θ, of the y-polarised pump light along xz plane. Figure 33 illustrates THG imaging using a metasurface; Figure 34A is a plot of the transmission spectra obtained experimentally with a frequency conversion device with a metasurface as described in figure 17; Figure 34B is a cross sectional view of a nano disk and slab on a substrate; Figure 34C is an electron microcopy image of the metasurface formed of the nano disks of figure 34B; Figure 35 is a set of FWM emissions obtained for different values of signal frequencies using the metasurface of figure 34C; Figure 36 is a set of FWM emissions obtained for different values of pump frequencies using the metasurface of figure 34C; figure 37A is an image of a target measured under white light illumination; figure 37B is an image of the target of figure 37A measured using the set-up of figure 3 with a signal beam at 2250 nm and a pump beam at 1130 nm; Figures 38A-38H is a set of images obtained for the same target of figure 37A and measured using the set-up of figure 3 with different combinations of signal and pump beams; Figure 39A is an image of a ‘NTU’ target measured under white light illumination; figure 39B is an image of the ‘NTU’ target measured using the set-up of figure 3 with a signal beam at 2200 nm and a pump beam at 1130 nm; figure 40 is a schematic diagram of a structure having several regions or sub-metasurfaces. Description Figure 1 is a flow chart of a method for imaging a sample according to the disclosure. The sample may take different forms, it may be an object or a subject. At step 110 a frequency conversion device is provided. The frequency conversion device includes a structure made of a third-order susceptibility material adapted to convert a first beam of electromagnetic radiation having a first wavelength in the infrared region into an output beam having a wavelength of less than about 1000 nm via a nonlinear process. For instance the first beam may be a beam in the short, near or mid infrared region. For instance the first wavelength may be greater than about 1000 nm, for instance between about 1000nm to about 50000 nm. The third-order susceptibility material may be configured to generate a degenerate four wave mixing emission, a non-degenerate four wave mixing emission or a third harmonic emission. For FWM, two pump beams at different wavelengths can be used, i.e. in total, three input frequencies and one output frequency, for each FWM process. At step 120 the first beam of electromagnetic radiation is collected. The first beam of electromagnetic radiation passes through or arises from the sample. At step 130 the first beam is directed onto the frequency conversion device to generate the output beam of radiation at a wavelength of less than about 1000 nm. At step 140 the output beam is detected to obtain an image of the sample. The output beam may have a wavelength between about 400nm and 1000nm. Therefore the output beam may be detected using a conventional visible detector. Conventional visible detectors/cameras can detect radiation in the range of 400-1000 nm, which goes beyond the traditional visible spectrum (400-700nm). The first beam may be mixed on the frequency device with one or more pump beams having different frequencies/wavelengths to generate the output beam. Therefore the method may further include the steps of generating one or more pump beam(s) of electromagnetic radiation. For instance generating a first pump beam having a second wavelength and a second pump beam having a third wavelength. Then the method may include mixing the first beam and the second or more pump beams using the frequency conversion device to obtain the output beam of electromagnetic radiation. The wavelength of the pump beam(s) may have a wavelength greater than about 1000 nm, to avoid generating noise in the detection band (400-1000nm). For instance the one or more pump beams may be IR pump beams. The structure made of the third-order susceptibility material may be adapted to be functional independently of the polarisation of the first wavelength in the infrared region. It will be appreciated that if the subject is a living animal or individual, the power of the light beam(s) being used should be set to prevent harming the subject. Figure 2A is a diagram of a system for implementing the method of figure 1. Figure 2A illustrates the case in which the frequency conversion device is used to perform third harmonic generation (THG). the first beam of IR electromagnetic radiation is the input, and the output beam having a wavelength less than about 1000nm is generated through THG in the frequency conversion device. The system 200 can be also used for imaging though other odd-order harmonic generations, such as fifth harmonic generation, seventh harmonic generation, etc. The system 200 includes a frequency conversion device 210, an optical arrangement formed of optical elements 222, 224, 226, and a detector/camera 230. In this example, a first beam of electromagnetic radiation S1 in the infrared IR transmits, scatters or reflects through a target sample 205. Figure 2B is an exemplary target sample for use in the system of figure 2A. In this example the target sample is a wheel having multiple arms. However the target could take another form and have an arbitrary pattern. A first lens 222 is used to direct the first beam S1 onto the frequency conversion device 210 to generate a beam of visible radiation Svis. Depending on the application the lens 222 may be used to focus or defocus the first beam S1. A second lens 224, and a third lens 226 are used to image the visible beam onto the detector 230. It will be appreciated that the optical pathway between 210 and 230 may be implemented in different ways. In some systems, the lens 224 or the lens 226 alone can suffice. The detector 230 may also be provided with its own lens. The detector 230 may be couped to a processing device such a computer 240 to collect and analyse the detector data. For instance the detector 230 may be a charge coupled device CCD. Figure 3A is another diagram of a system for implementing the method of figure 1. Figure 3A shows an experimental setup for imaging a target via one or more additional pump beams. Such pump beams may be used to perform for example four wave mixing (FWM), self-phase modulation (SPM), or cross- phase modulation (CPM). The system 300 is similar to the system of figure 2. The same reference numerals have been used to represent corresponding components and their description will not be repeated for sake of brevity. In this case the target 205 has been replaced by a target 305, which may be located at point A or point B on the transmission axis of S1’. A first source 360 is used to generate the IR signal S1’. For instance the IR signal S1’ may have a wavelength between about 1000nm to about 50µm. Additional radiation source(s) can be used to generate one or more pump beams Spump(s) of electromagnetic radiation having additional wavelength(s). A pump source 370 is used to generate a pump beam having a wavelength greater than about 1000 nm. The radiation sources can be selected from various technologies and do not need to be lasers, as there is no requirement for coherence. The setup can be modified to include additional pump beams in order to perform higher order wave mixing process. In this example, a beam splitter 250 is used to combine the first beam S1’ and the pump beam Spump. The beams S1’ and Spump are then mixed using the frequency conversion device 210 to obtain the output beam of electromagnetic radiation Sout. Alternatively, the pump beam(s) Spump(s) can be directly directed to the surface of the frequency conversion device 210 through other incident angles. In such cases the beam splitter 250 is not needed. Figure 3B is a diagram showing direct incidence of the first beam and the pump beam onto the frequency conversion device 210. Both S1’ and Spump can be illuminated on the metasurface with different angles represented as Θ1 and Θ2, even from the backside of the metasurface sample. The frequency conversion device 210 provides an extension to the visible detector 230, enabling it to detect wavelength in the near-infrared, and mid- infrared regions. On some occasions, a filter can be used in front of the CCD/detector to filter the near-infrared signal and pass the output signal Sout. The frequency conversion device 210 has a small footprint and may be implemented in different fashions. It is also relatively inexpensive to manufacture and can therefore save space, power consumption and cost of today’s detectors such as the ones used in mobile phones, cars, and other applications. The structure made of the third-order susceptibility material, for instance a metasurface, is very thin, in the order of 0.05 µm to 1µm and can be fabricated on any surface. For instance the metasurface can be fabricated on a substrate or directly on the surface of the detectors or cameras being used. The substrate or surface on which the metasurface is provided does not need to be flat. In some applications the metasurface may also even be fabricated on thin membranes (~100 nm thickness) or it can be embedded in a thin flexible layer. The metasurface may be designed to cover an area that is sufficient for the desired application and what the optical system is designed to image. Metasurfaces having an area smaller than about 20 µm x 20µm may cause lattice perturbations at the array's edge breaking the coherence and leading to strong scattering of light into free space. To avoid such perturbations the active area of the frequency conversion device may be designed to be greater than about 20 µm x 20µm, for instance a few 100 µm ² or a few mm ² or a few cm ² or more. The structure made of the third-order susceptibility material, for instance metasurface, may be located in an appropriate area of the optical system, to avoid perturbations at the edge of the structure. Four-wave mixing (FWM) is a third-order nonlinear effect where two or three different incident photons are mixed together to generate a new photon at a different frequency. If the input consists of only two different frequencies, a degenerate FWM takes place and four different frequencies can be generated based on the equation ω _FWM = 2ω _1,2 ± ω _2,1 . For the three-photon case, i.e. using three different frequencies as the input (ω1, ω2 and ω3), a nondegenerate FWM process takes place, and four different frequencies can be generated at ω1 + ω2 + ω3, ω1 + ω2 – ω3, ω1 + ω3 – ω2, and ω2 + ω3 – ω1. Third-harmonic generation (THG) is a sub-category of FWM, where all the input photons possess the same frequency. Figure 4A is a diagram illustrating a first example of the four wave mixing process. The frequency of the output beam can be expressed as a function of the frequency of the signal beam ^ _^ =^ _^ and the frequency of the pump beam ^ _^ =^ _^ = 2 ^ _^ + ^ _^ . Therefore the frequency of the generated output beam can be adjusted by varying the frequency of the pump beam. In photonics applications, the infrared spectral region can be divided into Near-Infrared (NIR) for wavelengths between 0.78-3µm, Mid-Infrared (MIR) for wavelengths between 3-50µm, and Far-Infrared (FIR) for wavelengths between 50-1000µm (See ISO 20473 scheme). Using the proposed approach an output beam (for instance a visible beam) can be generated by changing the wavelength of the pump beam regardless of whether the signal beam is in the NIR or the MIR region. Figure 4B is a diagram illustrating the different frequencies that can be generated with various FWM types, so-called degenerate FWM. Figure 4C is a diagram illustrating the different frequencies that can be generated with various FWM types, so-called non-degenerate FWM. FWM can be generated by array of nanoparticles on substrate, array of nano- holes within a thin film on a substrate, array of nano-particles on a slab on a substrate, or any other nanoscale design, which resonate with the one or more frequencies involved in the FWM process. The dimensions of the elements and the spacing between them can be chosen to achieve nonlinear imaging. These parameters can also be adjusted to achieve multiple resonances. Various resonant modes, generated by metasurfaces, can facilitate trapping light and enhancing the nonlinear processes. For instance the resonances may be chosen at the pump and signal wavelengths. Figure 5 is a diagram of an exemplary frequency conversion device 500 including an array of micro/nano elements 510 on a substrate 520, configured for third-order nonlinear frequency conversion. The substrate 520 may be a transparent substrate. The substrate may be selected depending on the application so that the substrate is transparent to the wavelength or range of wavelengths of the output beam. Example of substrates include glass, silicon oxide, quartz, sapphire, or magnesium fluoride MgF among others. For instance the nano elements may be nanoparticles such as silicon nanoparticles or nano islands which may include various protruding three dimensional shapes such as nano disks or nano cylinders. Based on the material, dimension, and arrangement of micro/nano elements, such a configuration can exhibit multipolar resonances (including electric dipole, magnetic dipole, and other higher- order electric and magnetic multipoles). The frequency conversion device 500 can be used in the systems of figures 2 and 3. The nano elements 510 are arranged to form an array which can be referred to as a metasurface. Metasurfaces are planar arrays of densely packed nanoantennas designed to manipulate the properties of incident light, including its amplitude, directionality, phase, polarization, and frequency. Since the nano elements 510 are nanopillars, the metasurface S1 is referred to as pillar metasurface. The metasurface S1 may be fabricated by regular electron beam lithography technique. The frequency conversion device 500 may be used in various optical configurations, that is with the substrate 520 towards the IR beam or with the metasurface / nanoelements 510 towards the IR beam. Figure 6A is a plot of the transmission spectra obtained with a frequency conversion device based on an array of nanopillars. Figure 6B is a cross sectional view of a nanopillar on a substrate. In this example the frequency conversion device was fabricated out of commercial silicon on sapphire SoS substrate with silicon nanopillars - having a height of about 550 nm. The diameter and centre-to-centre separation of the nanopillars in this sample are 650 nm and 1630 nm, respectively. The frequency conversion device exhibits several Mie-type resonances, making it suitable for third-order nonlinear imaging. In this particular case, the device is designed to be resonant at 1600 nm and 1100 nm, which are the frequencies of frequency ^ _^ and frequency ^ _^ in Figure 4B (i) and (ii), respectively. Figure 7A-D shows FWM emissions obtained for different values of pump and signal frequencies. The FWM emission for third-order nonlinear imaging has a frequency ^ _^ or ^ _^ ^‘ that can be generated using a pump beam of frequency ^ _^ and signal beam of frequency ^ _^ via type (i) and type (ii) FWM processes (shown in Figure 4B). By changing the pump frequency ^ _^ , one can control the wavelength of the FWM signal (^ _^ and/or ^ _^ ^‘ ) emission. Also, multiple FWM emissions can be obtained. Some of the FWM processes are quadratically dependent on the pump intensity. One can boost the emission to a much higher level by controlling the pump power. Figure 8A shows an image of a target measured under white light illumination. The image is obtained using the setup of figure 3 but with a single white light source and no signal or pump beams. The detector 230 measures visible light passing through the target and through the frequency conversion device 210. Figure 8B shows an image of the same target measured using the set-up of figure 3 with a signal beam at 1600 nm and a pump beam at 1100 nm. The visible image of the target arises from visible light generated by the frequency conversion device 210 via FWM process as shown in figure 7A. To demonstrate the effect of the FWM process alone for imaging, the THG frequency has been filtered out. Figure 9A shows an image of the same target measured using the set-up of figure 3 with a signal beam at 1500 nm and a pump beam at 1100 nm. Figure 9B shows an image of the same target measured using the set-up of figure 3 with a signal beam at 1700 nm and a pump beam at 1100 nm. Figure 9C shows an image of the same target measured using the set-up of figure 3 with a signal beam at 1600 nm and a pump beam at 1150 nm. The images of figure 9 demonstrate the importance of the designed metasurface and chosen frequency involved in the quality of the output image. To demonstrate the effect of FWM process alone, for imaging the THG frequency was filtered out. Figure 10A is a plot of the transmission spectra obtained with a frequency conversion device based on another array of nanopillars (metasurface S2). Figure 10B is a cross sectional view of a nanopillar on a substrate. The metasurface S2 may be fabricated by regular electron beam lithography technique. The substrate and the height are similar to Metasurface 1. However, in this case the diameter and centre-to-centre separation of the nanopillars are 720 and 1700 nm, respectively. The Metasurface 2 exhibits several Mie- resonances (magnetic quadrupole, electric quadrupole, magnetic dipole, electric dipole), making it suitable for third-order nonlinear imaging. In this particular case, this device is designed to be resonant at 1150nm and 1650nm, which are the frequencies of frequency ^ _^ and frequency ^ _^ in Figure 4B (i) and (ii). Figure 11A-D show FWM emissions obtained for different values of pump and signal frequencies. Figure 12A-12D show images of a same target measured using the set-up of figure 3 with different combination of signal and pump beams. The images of figure 12 demonstrate the importance of the designed metasurface and chosen frequency involved in the quality of the output image. Figure 13A is a simulation of the transmission spectra obtained with a frequency conversion device based on an array of nano disks. Figure 13B is a cross sectional view of a nano disk on a substrate. In this case the height and diameter of the nanopillars are 400nm and 400nm, respectively. The separation (centre to centre) between nanopillars is 600 nm. This third Metasurface S3 also referred to as disk metasurface is configured to generate a combination of broad and sharp resonances. A high-Q (sharp) resonance is excited at ωR1 = 1128nm (Fano resonance) and two broad resonances (electric and magnetic dipole resonances) ωR2 and ωR3 are excited by the first electromagnetic frequency. Figure 14 is a table providing the simulated output wavelength λFWM obtained for various signal and input beams combinations of the disk metasurface S3 of figure 13. The FWM process corresponds to figure 4B(i). The output wavelength / frequency can be tuned, even though the first (incident) frequency is still at the resonance positions of the imaging device. Figure 15 is a plot showing the simulated conversion efficiency obtained for the various scenarios 1-6 presented in figure 14. The conversion efficiency is the highest when the pump frequency ω1 = ωR1 see cases 1 and 2. The nonlinear intensity is quadratically dependent on the pump intensity. When the pump is enhanced by the strong resonance, the conversion efficiency can be enhanced quadratically. The conversion efficiency is medium when the signal frequency ω ₂ = ω _R1 , see cases 3 and 4 (different FWM process compared with 1 and 2). The nonlinear intensity is linearly dependent on the pump intensity. When the pump is enhanced by the strong resonance, the conversion efficiency can be enhanced linearly. The conversion efficiency is the lowest when there is no pump resonance involved in, see cases 6 and 5. Figure 16 is a simulation showing the electric filed enhancement obtained at different resonance frequencies. The simulation illustrates where the field is enhanced by pumping the light at the resonances. This explains the various conversion efficiency obtained in figure 15. Figure 17 is a diagram of an exemplary frequency conversion 1700 device including an array of micro/nano elements 1710 on a substrate 1720, configured for third-order nonlinear frequency conversion. Compared with the example of figure 5, the micro/nano elements are provided on the slab 1730 of third order susceptibility material. This configuration may be referred to as slab-disk configuration. The slab 1730 with the array of nano elements 1710 forms a metasurface. The slab of third order susceptibility material 1730 is provided on the substrate 1720. The frequency conversion device 1700 may be used to generate resonances, beyond pure Mie resonances. Figure 18A is a simulation of the transmission spectra obtained with a frequency conversion device based on an array of nano disks formed on a slab. Figure 18B is cross sectional view of a nano disk on a slab provided on a substrate. In this case the slab and nano disks are made of Silicon. The slab has a height of 250 nm. The nano disks have a height of 160 nm and a diameter of 300nm. The separation centre to centre between nano disks is 500nm. The substrate is made of silicon dioxide. This Metasurface S4 is configured to generate very sharp resonances (so-called high-Q) including guided resonances, Mie resonances, and bound states in the continuum, which enhance the device efficiency. Figure 19 is a table providing the simulated output wavelength λ _FWM obtained for various signal and input beams of the metasurface S4 of figure 18. The FWM process corresponds to figure 4B(i). The first (incident) frequency can be converted to various output frequencies. Figure 20 is a plot showing the simulated conversion efficiency obtained for the various scenarios 1-6 presented in figure 19. With more resonances involved, the conversion efficiency increases. With less resonances involved; the conversion efficiency decreases. When the signal resonance is stronger than pump resonance, the process with signal resonance involved has a higher efficiency. Figure 21 is a simulation showing the electric filed enhancement obtained at different resonance frequencies. The near field is enhanced by pumping the light at the resonances with high-Q factors. Figure 22 is a diagram of another exemplary frequency conversion device 2200 configured for third-order nonlinear frequency conversion. The device 2200 includes an array of nano holes 2210 formed within a slab 2230 provided on a substrate 2220. This configuration may be referred to as slab-hole configuration. The slab 2230 with the array of nano holes 2210 forms a metasurface also referred to as hole metasurface. The slab 2230 is made of a third order susceptibility material, for instance Silicon. The frequency conversion device 2200 may be used to generate resonances, beyond pure Mie resonances. Figure 23A is a simulation of the transmission spectra obtained with a frequency conversion device based on an array of nano holes formed on a slab referred to as slab-hole configuration. Figure 23B is a cross sectional view of slab with nano hole provided on a substrate. In this case the slab and nano holes are made of Silicon. The slab has a height of 300 nm. The nano holes have a height of 300 nm and a diameter of 300nm. The separation between nanoholes is 600 nm. The substrate is made of silicon dioxide. This Metasurface S5 is configured to generate very sharp resonances including guided resonances, Mie resonances, and bound states in the continuum, which enhance the device efficiency. Figure 24 is a table providing the simulated output wavelength λFWM obtained for various signal and input beams of the hole metasurface S5 of figure 23. The FWM process corresponds to figure 4B(i). The first (incident) frequency can be converted to various output frequencies. Figure 25 is a plot showing the simulated conversion efficiency obtained for the various scenarios 1-6 presented in figure 24. When the pump resonance ω ₁ is stronger than signal resonance ω ₂ , the process has a higher efficiency. The efficiency of case 1 is greater than the efficiency of case 2 because the resonance at ωR1 is greater than the resonance at ωR2. Similarly the efficiency of case 4 is greater than the efficiency of case 6 because the resonance at ωR2 is greater than the resonance at ωRn. Figure 26 is a simulation showing the electric filed enhancement obtained at different resonance frequencies. The near field is enhanced by pumping the light at the resonances with high-Q factors. The implementations of the frequency conversion devices described above with respect to figures 5 onwards are some examples that can be reconfigured and redesigned for various applications. For example, these devices are polarisation-insensitive but could be designed for polarisation- sensitive applications. Figure 27A is a diagram of an exemplary polarisation-sensitive frequency conversion device for THG. The frequency conversion device 2700 is similar to frequency conversion device 2200 of figure 22. The device 2700 includes an array of nano holes 2710 formed within a slab 2730 provided on a substrate 2720. The slab 2730 is made of a third-order susceptibility material for generating third-order nonlinear interactions, e.g. THG and FWM. For instance the third-order susceptibility material may comprise Silicon or Germanium. The substrate may be a glass or Sapphire substrate. The slab comprises a periodic array of nanohole dimers (also referred to as dimer-airy holes) whose dimensions and periods are small compared with the wavelength of an incoming radiation. This structure may be referred to as a hole metasurface or membrane metasurface. Like in the previous examples the dimensions of the nanohole elements in this configuration and the spacing between them are chosen to achieve nonlinear imaging via FWM interaction. Figure 27B is a partial top view of the metasurface 2730 showing two dimer-airy holes. The radius r ₀ of each hole is shown with an arrow labelled and the offset between the two holes is given by the parameter x0. The offset parameter x0 is defined as the distance from the centre of one airy hole to the centre of the dimer hole. By changing the value of the offset (either increasing or decreasing) the dimensions of the periodic slice in the x and y direction (Dx and Dy) are shortened or lengthened. The value of the offset will affect the resonances that can be achieved within the material and hence the frequencies the conversion device will be able to convert in order to obtain a visible image. The hole metasurface 2730 is designed to support multiple symmetry- protected bound states in the continuum (BICs). The strong light confinement within the metasurfaces is an important factor in achieving strong light- matter interactions and generating nonlinear signals with higher intensity under relatively low-power input light. The bound states in the continuum (BICs) can be used to manipulate the light-matter interactions (C. W. Hsu, et al, Nature Reviews Materials, 2016, 1, 1–13.). Compared with dielectric metasurfaces composed of an array of nanoparticles, metasurfaces based on nonlinear material with an array of holes can obtain a larger mode volume for enhancing light matter interaction (L. Xu, et Al, New Journal of Physics, 2022, 24, 035002). By tuning the gap between two airy holes, the symmetry-protected BICs can be transformed into the quasi-BICs with high Q-factor, resulting in a strong electromagnetic field confinement within the silicon membrane via plane-wave illumination. Compared with disk metasurfaces, the hole metasurfaces provide more options for designing the light-matter interactions due to their unique optical response. In a specific example the amorphous silicon (a-Si) membrane metasurface is formed of a 235-nm-thick silicon slab with dimer airy holes arranged periodically on a glass substrate. The radius of the airy hole is fixed as r ₀ = 100 nm. The period along x and y directions D _x and D _y are set to be 600 nm and 300 nm, respectively. Thus when the offset x0 =150 nm, the dimer-hole membrane metasurface is the same as a hole membrane metasurface with the unit cell period Dx = Dy =300 nm. The system of the a-Si membrane metasurface is invariant under the 180 degree rotation around the z-axis. which is defined as the operation ^^ ^ . The characteristics of this membrane metasurface are described in figures 28 to 32. Figure 28 is a plot showing THG spectra measured from metasurfaces designed with different offset x ₀ . The TH emissions are dramatically enhanced at the resonance spectral position. With the offset increase from 120 nm to 150 nm, the measured TH emission first increase to a maximum value when the offset is 135 nm, and then gradually decrease. The width of resonance is experimentally estimated as 20 nm for x0 =135 nm. This matches well with the width of the pump laser around 1500 nm, suggesting the best coupling scheme for obtaining large nonlinear emission power. When the offset is 150 nm, no significant enhancement of the TH emission is observed. This observation agrees with the fact that the ideal BIC is formed and the coupling of the BIC with external radiation has been suppressed when x ₀ =150 nm. The THG conversion efficiency is estimated as the collected THG power divided by the incident pump power: ^ _^^^ = ^ _^^^ /^ _^^^^ . A maximum THG conversion efficiency is obtained around 3.6 × 10 ⁻⁶ experimentally when the offset x ₀ =135 nm. The mode properties of the membrane metasurface are characterised by calculating the band structure based on the Massachusetts Institute of Technology Photonic-Bands (MPB) open sources. Figure 29 shows the calculated band structure for the metasurface when x0 =150 nm. Figure 30 shows the nearfield electric distributions for the mode TE(3,1,1) and TE(3,2,1), respectively. For the considered unit cell with pitch sizes of D =600 nm and D =300 nm, two typical odd mod ^ x y es are observed under ^ _^ , TE(3,1,1) and TE(3,2,1) with their electric pattern M1 and M2 (see Figure 10B)). However, when x0 =150 nm, the pitch sizes (Dx × Dy) of unit cell changes from 600×300 nm to 300×300 nm. With the transformation of the unit cell, the two modes, TE(3,1,1) and TE(3,2,1), become even under ^^ ^ , decoupling to the odd leaky channel to the far-field. The features of TE(3,1,1) and TE(3,2,1) indicate that they are symmetry protected BICs with infinite Q-factor when x ₀ =150 nm, and can be converted into quasi-BICs with finite Q-factor when the offset x0 is tuned. By controlling the offset x0, one can transform such an ideal BIC into a quasi- BIC. This phenomenon has been explained from the view of the symmetry. This explanation is intuitive and concise to summarize the formation of the BIC, but hardly shows the process of the energy exchange from the bound modes to the external modes. Taking Mode 1 as an example, the modes are then characterised via spherical and Cartesian multipolar analysis. This enables an observation from the view of multipolar transformations, manifesting itself as a Fano feature in the optical response spectrum. Spherical and Cartesian multipolar analysis are performed for three different cases: x0 =120 nm, 135 nm and 150 nm, respectively. The ED, MD, EQ, MQ, EO, MO, py, and TD represent the electric dipole (ED) and magnetic dipole (MD), electric quadrupole (EQ) and magnetic quadrupole (MQ), electric octupole (EO) and magnetic octupole (MO), the electric dipole moment along the y-axis (py), and electric Toroidal moment (TD). The sample was fabricated using standard electron beam lithography. The membrane metasurface is fabricated via etching an array of nano holes in the amorphous silicon film on a glass substrate. Figure 31 is a plot of the measured linear transmission spectrum of the metasurface under a white light source illumination. Two pronounced asymmetric Fano line shapes are observed around λ =980 nm and λ =1510 nm, respectively. This indicates the excitation of the two quasi-BICs. The estimated widths of the two states are around 50 nm and 20 nm, respectively. Figure 32A shows the measured transmission spectra of metasurfaces designed with different offsets. The width of the resonance narrows as the offset increases from 120 nm to 150 nm. This corresponds to the decrease in geometric asymmetry. The Fano feature in the spectrum vanishes when x ₀ =150 nm, corresponding to the inaccessibility of the ideal BIC sup-ported by the metasurface. A slight redshift of the resonance is also observed along with another Fano resonance at around the wavelength of 1000 nm. Similarly, it vanishes when x ₀ =150 nm. By performing the multipolar analysis, one can conclude that this Fano resonance is mainly formed by the interference between the ED, EQ and MO resonances. Figure 32B shows the angular dependence of the transmission spectrum by changing the incident angle θ, of the y-polarised pump light along xz plane. With the increase of the incident angle θ, the redshift of the two resonances is observed. Meanwhile, the resonance MR2 is blue shifted. This matches well with the calculated band structure. Figure 33 illustrates THG imaging using a metasurface. This system is used to demonstrate experimentally the conversion of near-infrared image into visible based on the THG process in the membrane metasurfaces. Figure 33(a) is schematic diagram for THG imaging. A pump beam pass passes through a target (NBS 1963A resolution test target), which is then imaged by the focal lens onto the metasurface. A THG signal is then collected for measurement. Figures 33(bi-ci-di-ei) show the visible image when the metasurface is overlapped with the target at different positions under a white light source illumination. The metasurface sample (offset x0 =135 nm) is then overlapped with the image of the target near the focal plane and illuminated with a 1512- nm IR laser. Figures 33 (bii-cii-dii-eii) show the images captured by the CCD camera (CS165MU/M, Thorlabs). The pattern of the overlapping area between image of the target and the metasurfaces can be converted to visible images clearly. A clear multi-stripe pattern is obtained based on THG process. The use of THG therefore allows performing infrared imaging without the need for an additional pump light. The various structures / metasurfaces presented in the disclosure show the potential of various design to manipulate/enhance the efficiencies of THG or FWM. It will be appreciated that these embodiments may be modified to suit a particular application. Figures 34 to 39 present various experimental results obtained with a frequency conversion device including an array of elements on a substrate as described in Figure 17 with a slab-disk configuration. Figure 34A is a plot of the transmission spectra obtained with a frequency conversion device based on an array of nano disks formed on a slab (metasurface SA). Figure 34B is a cross sectional view of a nano disk on a slab provided on a substrate. The metasurface SA may be fabricated by regular electron beam lithography technique. Figure 34C shows a top image of the metasurface SA taken via electron microscopy. In this example the frequency conversion device was fabricated out of commercial silicon on sapphire SoS substrate with silicon nano disks on top of silicon film. The height of silicon nano disks and silicon film are 200 nm and 300 nm, respectively. The diameter and centre-to-centre separation of the nano disks in this sample are 300 nm and 500 nm, respectively. The frequency conversion device exhibits several Mie-type resonances and guided resonances, making it suitable for third-order nonlinear imaging. In this particular case, the device is designed to be resonant at 1130 nm. A pump beam at 1130 nm has a of pump frequency ^ _^ as shown in Figure 4B (i) and (ii). Figure 35 shows the FWM emissions obtained for different values of signal frequencies. The FWM emission for third-order nonlinear imaging has a frequency ^ _^ or ^ _^ ^‘ that can be generated using a pump beam of frequency ^ _^ and signal beam of frequency ^ _^ via type (i) and type (ii) FWM processes (shown in Figure 4B). By changing the signal frequency ^ _^ , one can control the frequency of the FWM signal (^ _^ and/or ^ _^ ^‘ ) emission. Also, multiple FWM emissions can be obtained. Some of the FWM processes are quadratically dependent on the pump intensity. One can boost the emission to a much higher level by controlling the pump power. Figure 36 shows the FWM emissions obtained for different values of pump frequencies. By changing the pump frequency , one can control the frequency of the FWM signal (^ _^ and/or ^ _^ ^‘ ) emission. Figure 37A shows an image of a target measured under white light illumination. The image is obtained using the setup of figure 3 but with a single white light source and no signal or pump beams. The detector 230 measures visible light passing through the target and through the frequency conversion device 210. Figure 37B shows an image of the same target measured using the set-up of figure 3 with a signal beam at 2250 nm and a pump beam at 1130 nm. The visible image of the target arises from visible light generated by the frequency conversion device 210 via FWM process as shown in figure 37A. To demonstrate the effect of the FWM process alone for imaging, the THG frequency has been filtered out. Figures 38A-38H show the images of a same target measured using the set- up of figure 3 with different combinations of signal and pump beams. To demonstrate the effect of FWM process alone for imaging, the THG frequency was filtered out. Figure 39A shows an image of a target with ‘NTU’ marking measured under white light illumination. The image is obtained using the setup of figure 3 but with a single white light source and no signal or pump beams. The detector 230 measures visible light passing through the target and through the frequency conversion device 210. Figure 39B shows an image of the same target measured using the set-up of figure 3 with a signal beam at 2200 nm and a pump beam at 1130 nm. The visible image of the target arises from visible light generated by the frequency conversion device 210 via FWM process as shown in figure 39A. To demonstrate the effect of the FWM process alone for imaging, the THG frequency has been filtered out. The metasurfaces described in the disclosure show a periodic arrangement of individual elements. However it will be appreciated that the separation between individual elements does not need to be mutual so that individual elements are randomly separated. Such metasurfaces are called disordered metasurfaces. Similarly, the metasurfaces described in the disclosure show a periodic arrangement of individual elements having a same shape and a same size. However, it will be appreciated that metasurfaces may be designed with a structure that includes individual elements of different shape and sizes. Depending on the application, this approach may be used to control the phase of the output beam. A structure could also include multiple sets or regions of individual elements. Figure 40 shows a structure 4000 having several sets or regions of individual elements. Each set or region represent a single sub-metasurface, having its own characteristics. In this example nine sub-metasurfaces are represented and labelled 4010-4090. Each sub-metasurface may have different types of individual elements. Each individual element within a sub-metasurface may have a same shape and a same size. The shape and size of individual elements may therefore be different for different sub-metasurfaces. Alternatively, the individual elements within a sub-metasurface may have different shape and sizes. Depending on the application one or more of the sub-metasurfaces may be the same; or they could all be different. In figure 40 the nine sub-metasurfaces 4010-4090 have a rectangular shape and are juxtaposed next to one another. It will be appreciated that the sub- metasurfaces may be designed with other geometrical shapes and arranged in a tessellated fashion. For instance, the sub-metasurfaces may have a polygonal shape. It will also be appreciated that depending on the application and design of the metasurface, the output beam may have one or more wavelengths of less than about 1000 nm obtained via one or more nonlinear processes such as four wave mixing or higher order mixing nonlinear processes. A skilled person will therefore appreciate that variations of the disclosed arrangements are possible without departing from the disclosure. Accordingly, the above description of the specific embodiments is made by way of example only and not for the purposes of limitation. It will be clear to the skilled person that minor modifications may be made without significant changes to the operation described.