GB2586268A - Photodetector array - Google Patents

Abstract

A photodetector array, and method of forming, comprising at least one semiconductor structure 1, such as a nanowire, having first and second locations 3, 4. The semiconductor structure 1 comprises at least two elements, such as cadmium (Cd) and sulphur (S). The composition and bandgap of the semiconductor 1 varies between the first location and the second location. A number of electrodes 5 are connected to and spaced apart along the semiconductor 1 between the first location and the second location so as to provide a series of photodetector units 6. In response to illumination of the semiconductor 1 with light, each photodetector unit generates a respective photocurrent response. Also described is a method of operating the photodetector array comprising: receiving calibrated spectral responses for each photodetector unit; receiving photocurrent responses for each photodetector unit; reconstructing an incident spectrum in dependence upon the calibrated spectral responses and the photocurrent responses. The method may use Tikhonov regularization.

Description

Photodetector array

Field

The present invention relates to a photodetector array, in particular to a broadband visible and near infrared light photodetector array.

Background

Optical spectroscopy is one of the most ubiquitous, versatile characterization techniques across industrial processes and fundamental scientific research. A variety of /0 miniaturized, portable spectrometers have been developed for applications where reduced footprint and weight takes precedence over high resolution. These microspectrometers have typically been inspired by conventional bench-top spectrometers, centering around interferometers or gratings, with miniaturized or integrated optics. When minimizing physical dimensions towards sub-millimeter scales /5 desired in, for example, lab-on-a-chip systems, such designs are inherently limited by adverse effects associated with scaling their optical components or path lengths. Microspectrometers that employ computational spectral reconstruction circumvent these constraints by addressing a full range of spectral components simultaneously at multiple detectors. However, they have thus far been based on complex millimeter-scale arrays of individually prepared filters arranged over CCD or CMOS detectors, which are challenging to miniaturize.

Microspectrometers with ever-smaller footprints are demanded in a range of spectroscopy applications where minimized size and weight is paramount, such as in-situ characterization. Spectrometers based on a single nanomaterial structure are reported here. Representing arguably the simplest, most compact microspectrometer design possible, this platform is independent of the complex optical components, cavities or CCDs that constrain further miniaturization of current systems. it is demonstrated that incident spectra can be computationally reconstructed from different spectral response functions and measured photocurrents along an individual compositionally-engineered nanowire. This versatile platform can straightforwardly be expanded across the infrared to ultraviolet range. Despite their simplicity, these devices are capable of accurate monochromatic and broadband light reconstruction, as well as spectral imaging from centimeter-scale image planes down to lensless, single-cell-scale in-situ mapping.

Summary

According to a first aspect of the present invention there is provided a photodetector array. The photodetector array comprises at least one semiconductor structure, having first and second locations. At least a section of the semiconductor structure comprises at least two elements. The composition of the semiconductor structure varies between the first location and the second location such that bandgap of the semiconductor structure varies between the first location and the second location. The photodetector array further comprises a number n of electrodes connected to and spaced apart along the semiconductor structure between the first location and the second location so as to jo provide a series of photodetector units along the semiconductor structure. In response to illumination of the semiconductor structure with light, each photodetector unit generates a respective photocurrent response.

The elements are elements from the periodic table of elements. The number n of electrodes is a positive integer.

The semiconductor structure may be, for example, a nanostructure. The semiconductor structure may be, for example, a thin film and/or a nanobelt.

The semiconductor structure may be a semiconductor nanowire.

The semiconductor structure may lie on a surface of a substrate. The photodetector array may further comprise a dielectric layer disposed on the substrate arranged to be coplanar with at least a section of the semiconductor structure.

The semiconductor structure may lie, in other words, prone, on a principal surface of a substrate and the photodetector array further comprises a dielectric layer disposed on the substrate arranged to planarize at least a section of the semiconductor structure.

The dielectric layer may be disposed directly on the substrate.

The photodetector array may further comprising first and second dielectric layer portions laterally coating or laterally abutting opposite sides of the semiconductor structure. The first and second dielectric layer portions may comprise co-planar portions of the same dielectric layer. -3 -

The dielectric layer may comprise an inorganic dielectric material. The inorganic dielectric material may be an oxide. The dielectric layer may comprise an organic dielectric material. The dielectric layer may be a 2D monolayer. The dielectric layer may have t thickness of at least 1 nm. The dielectric layer may have a thickness of at least 10 nm.

Each photodetector unit may be a single electrode. Each photodetector unit may comprise a pair of electrodes. Each photodetector unit may comprise two or more electrodes.

Spaces between adjacent electrodes may be between 10 nm and 5 pm. Spaces between adjacent electrodes may be between 10 nm and 40 nm. Spaces between adjacent electrodes may be between 40 nm and loo nm. Spaces between adjacent electrodes may be between loo nm and 500 nm. Spaces between adjacent electrodes may be between 500 nm and 5 pm.

Electrodes may have a width of between 10 nm and 5 pm. Electrodes may have a width of between 10 nm and 40 nm. Electrodes may have a width of between 40 nm and 100 nm. Electrodes may have a width of between loo nm and 500 nm. Electrodes may have 20 a width of between 500 nm and 5 pm.

The electrodes may comprise a semiconductor material. The electrodes may comprise a metallic alloy or metal.

A passivation layer may cover at least a portion of at least one surface of the photodetector array.

The number 71 of electrodes may be greater or equal to eight so as to allow reconstruction of an incident spectrum comprising at least two peaks.

The bandgap may increase monotonically between the first and the second locations of the at least one semiconductor structure.

The at least one semiconductor structure may comprises at least one heterostructure. -4 -

The bandgap at the first location of the at least one semiconductor structure may be between 0.1 and 2 eV.

The bandgap at the second location of the at least one semiconductor structure may be 5 between 2 and 5 eV.

The semiconductor structure may comprise CdS.Selsx, wherein o x and wherein, at the first location, x = xl, and at the second location, x = x2, wherein x1 x2.

/0 The semiconductor structure may comprise Zn,Cdi 2SySel " wherein o and wherein, at the first location, x = xt, and at the second location, x = x,, wherein xi_ x2 and wherein o y land wherein, at the first location, y = y1, and at the second location, y = y2, wherein y, y2.

The semiconductor structure may comprise In"Ga1_2N wherein o x and wherein, at the first location, x = x1, and at the second location, x = x2, wherein x1 x,.

The semiconductor structure may comprise Si,,Ge. wherein 0 X 1 and wherein, at the first location, x = x,, and at the second location, x = x,, wherein x, x2.

The semiconductor structure may include a mixture of a Group IV compound, a Group II-V1 compound, a Group II-V compound, a Group III-V1 compound, a Group III-V compound, a Group IV-VI compound, a Group II-IV-V1 compound, or a Group 11-IV-V compound, for example, Si, Ge, Sn, ZnO, ZnS, ZnSe, ZnTe, CdO, CdS, CdSe, CdTe, or MgO, MgS, MgSe, MgTe. Hg0, Hg.S, HgSe, HgTe, MN, AlP, AlAs, AlSb, GaN, GaP, GaAs, GaSb, InN, InP, InAs, InSb, TiN, TiAs,TiSb, PbS, PbSe, PbTe, Hg0, HgS, HgSe, HgTe, or mixtures thereof. The semiconductor structure may include one or more organic semiconductors.

There may be any number of electrodes on the photodetector array. The number of electrodes may be at least 20 electrodes. The number of electrodes is between 20 and 5o. The number of electrodes is between 5o and loth The number of electrodes is between loo and 200. The number of electrodes is between 200 and moo. The number of electrodes may be greater than moo. -5 -

According to a second aspect of the present invention, there is provided a method of operating a photodetector array comprising a series of photodetector units. The method comprises receiving calibrated spectral responses for each photodetector unit, receiving photocurrent responses for each photodetector unit and reconstructing an incident spectrum in dependence upon the calibrated spectral responses and the photocurrent responses.

Reconstructing the incident spectrum may be performed by solving a set of equations, where the measured photocurrent for each photodetector unit is equal to the integral of jo the product of the incident spectrum as a function of wavelength and the pre-calibrated spectral response function as a function of wavelength. The product of the incident spectrum as a function of wavelength and the pre-calibrated spectral response function as a function of wavelength is integrated between the limits of the minimum and maximum wavelengths of operation of the photodetector array.

The method may further comprise generating the reconstructed incident spectrum in dependence upon the calibrated spectral responses and the photocurrent responses by solving a system of linear equations: A x incident spectrum(A)calibrated spectral response 1(A) c12 = photocurrenti (i = 1,2,3..., n) where A,,1 and A,,,,,, determine the photodetector array operational wavelength range and i is the index of the photodetector.

The method may further comprise reconstructing the incident spectrum in dependence or upon the calibrated spectral responses and the photocurrent responses using adaptive regularisation.

The method may further comprise measuring photocurrent data from each photodetector unit. The measurement of photocurrent data from each photodetector unit may be conducted by causing a bias to be applied between at least two electrodes.

The measurement of photocurrent data from each photodetector unit may be conducted by using a p-n junction. -6 -

According to t a third aspect of the invention, there is provided a spectrometer. The spectrometer may comprise at least one photodetector array of the first aspect of the invention and a computer configured to receive signals from the photodetector array 5 and to perform the method of the second aspect of the invention.

The spectrometer may further comprise a lens and/or a prism and/or a diffusor. The spectrometer may have no lens and/or no prism and/or no diffusor.

jo According to a fourth aspect of the present invention there is provided a method of forming a photodetector array. The method comprises providing a substrate with a principle surface. The method further comprises providing a compositionally-graded semiconductor structure having first and second locations on the principle surface of the substrate and providing a mask. The method further comprises disposing a number ri of electrodes connected to and spaced apart along the semiconductor structure between the first location and the second location so as to provide a series of photodetector units along the semiconductor structure and performing passivation on the semiconductor structure.

The passivation may be performed by disposing A1203 by atomic layer deposition (ALD).

The method may further comprise embedding the semiconductor structure in a dielectric.

The method may further comprise removing the oxide layer from the surface of the semiconductor structure and performing passivation before the deposition of the electrodes.

The removal of the oxide layer from the surface of the semiconductor structure and passivation of the semiconductor surface maybe performed by treating the semiconductor structure with nitrogen plasma and then treating the semiconductor structure with ammonium sulphide (NHJ,S solution. -7 -

Brief Description of the Drawings

Certain embodiments of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which: Figure 1 is a real-color photoluminescence (PL) image of a compositionally-graded CdS"Sei,nanowire and corresponding spectra collected from marked representative regions (spot size -5 um; scale bar: 20 pm).

Figure 2 shows a nanowire spectrometer in the form of a packaged chip. A fluorescent micrograph of the spectrometer is also shown (scale bar: to pm).

Figure 3 shows I-V curves measured between two typical nearby electrodes at the red jo (CdSe) end of the nanowire shown in Figure 2, illuminated with different intensities of 490 nm light.

Figure 4 illustrates time response of the same photodetector unit shown in Figure 2 of the pulsed incidence light (490 nm, 0.3 mWcm-2) under 0.5 V bias. Dotted lines indicate to % and 90 % points to the peak value used for calculating the rise (1.5 ms) /5 and fall time (3.5 ms).

Figure 5 shows normalized spectral responses RA) of the constituent units varying continuously along the nanowire shown in Figure 2.

Figure 6 schematically illustrates operation of the nanowire spectrometer shown in Figure 2.

Figure 7A is a simulated spectral response R1(A) and photocurrentL, wherein photocurrent is equal to the red area which indicates the integral responses to the incident light.

Figure 78 illustrates a mathematical description of the relation between F(A), R1(A) and -3 or Figure 7C illustrates a step of spectral reconstruction via solving equation set.

Figure 7D illustrates a reconstructed spectrum of the simulated incident light FA). Figure 8 shows spectrum curves reconstructed from an original 560 nm light with an FWHM of 3 nm (dashed curve), as measured by Thorlabs CCStoo spectrometer (0.3 nm wavelength accuracy).

Figure 9 shows two mixed narrow-band light with peaks separated by 15 nm (dashed curve) are spectrally resolved by two spectrometers with 38 and 30 units.

Figure to shows spectral reconstruction of these two devices cannot reproduce the original two peaks separated by to nm.

Figure 11 shows reconstructed spectra of varying monochromatic light for the 38-unit 35 spectrometer. -8 -

Figure 12 shows reconstructed spectra of broadband light with the 38-unit spectrometer.

Figure 13 shows reconstructed spectra of broadband light with the 38-unit spectrometer.

Figure 14 is a schematic diagram of spectral imaging by the nanowire spectrometer and a photograph of the Cambridge University logo by a CCD camera before and after filtering (bottom panel).

Figure 15 illustrates spectral imaging is conducted by scanning the nanowire spectrometer on the imaging plane in a boustrophedonic (or "serpentine") pattern.

Figure 16A is a schematic diagram of an initial data cube. Each pixel contains a photocurrent value of each photodetector unit and forms an initial data cube. Figure 16B is a schematic diagram of a spectral data cube. The spectral data cube is reconstructed from the initial photocurrent data cube.

Figure 17 illustrates a spectrum of yellow color from the projector. Bandpass filter and spectrometer detectable ranges are indicated in the diagram.

Figure 18 is a series of reconstructed images at various wavelengths. Pixel intensity range of these images is normalized. Top right panel, left is an original photograph of the image. Top right panel, left is a pseudo-colored spectral image converted from the spectra according to CIE color matching functions.

Figure 19 illustrates spectra of A and B points in top right panel of Figure 19, measured by the nanowire spectrometer and a conventional spectrometer.

Figure 20 is a schematic diagram of spectral imaging at microscale by nanowire spectrometer.

Figure 21A is a schematic diagram of an initial photocurrent data cube, where each or layer is shifted because of the different location of the photodetector units.

Figure 21B is a schematic diagram of a spectral data cube reconstructed from the photocurrent overlap region.

Figure 22A Top left: is a schematic diagram of the operation of cell mapping. Top right: is a photograph of the cell mapping apparatus. Bottom: is a micrograph of a naturally pigmented red onion cell surrounded by transparent cells. Scale bar: 50 pm.

Figure 22B shows reconstructed absorption spectra from different parts of the red onion cells.

Figure 23 shows absorption spectral images of onion cells at various wavelengths. Pixel intensity range of these images is normalized.

Figure 24 is a schematic diagram of the source-moving thermal evaporation nanowire growth approach. -9 -

Figure 25 is a real-color photoluminescence image of the nanowire on a Transmission electron microscope (TEM) carbon grid.

Figure 26 is a TEM image of the nanowire shown in Figure 2.

Figure 27 shows panels 131-137: which are high-resolution TEM images and selected-area 5 electron diffraction patterns taken from several representative regions, as marked in Figure 26.

Figure 28 is a scanning electron microscope (SEM) image of one end of the nanowire. Figure 29 illustrates position-dependent composition along the length of the nanowire in Figures 25 to 28.

/o Figure 30 illustrates a comparison of position-dependent bandgap values calculated from both the PL measurements and via Wgard's law using the compositional measurements.

Figure 31 are schematic diagrams of the non-embedded device fabrication process (top panel); a SEM image of the cracked electrodes fabricated by normal lithography methods (bottom panel). Inset is a side-view schematics of these two devices.

Figure 32 are schematic diagram s of the embedded device fabrication process (top panel); a SEM image of the embedded device (bottom panel). Note that the discoloration in the SU-8 area (dot line) is non-significant, and simply due to charging-based image artefacts. Inset is a side-view schematics of these two devices.

Figure 33A is a schematic diagram of apparatus for calibrating the response functions.

Insets are the packaged chip and the zoomed-in image of the nanowire spectrometer. Scale bar: 10 pm Figure 33B is a photograph of nanowire spectrometer calibration apparatus.

Figure 34 shows coefficient of variance of sets of photocurrents measured from the or device before and after A1203 deposition, which are 5% and 2%, respectively.

Figure 35 shows measurement stability of the device with A1203 deposition at different bias voltages.

Figure 36 show photocurrents as a function of light power of the device with A1203 deposition at a bias voltage of 0.5 V. Figure 37 is a schematic diagram of light distribution across the filter-array spectrometer and nanowire spectrometer.

Figure 38 is a schematic diagram of light distribution across the filter-array spectrometer and nanowire spectrometer.

Figure 39 is a schematic diagram of nonuniform light intensity distribution across the filter-array spectrometer with relatively considerable dimensions, which will produce significant distortion in the reconstructed spectra.

-10 -Figure 40 is a schematic diagram of uniform light intensity distribution across the nanowire spectrometer at the micron scale, where spatial intensity non-uniformity can be neglected.

Figure 41 shows a reconstruction of a simulated spectrum with peak wavelength of 565 5 nm by a non-iterative method. Unit number n=38.

Figure 42 illustrates noise of the reconstructed spectrum raises with the increase of the stimulated random error level of each photodetector unit. The spectrum is fully distorted when error is 1 %, which is below the system measurement errors (2 %). Figure 43 illustrates noise of the reconstructed spectrum raises with the increase of the /0 stimulated random error level of each photodetector unit. The spectrum is fully distorted when error is 1%, which is below the system measurement errors (2 %). Figure 44 is a flow diagram of spectrum reconstruction by adaptive regularization. Blue blocks represent input information, orange blocks are processes, green blocks denote output data, and the purple block at bottom is the reconstructed spectrum.

Figure 45A shows a reconstruction of the data used in Figures 41 to 43 by adaptive regularization.

Figure 45B shows reconstructed spectra with simulated random errors of each photodetector at CV= to%. Compared with the ordinary non-iterative methods, the adaptive Tikhonov regularization exhibits high error tolerance.

Figure 45C shows reconstructed spectra with simulated random errors of each photodetector at CV=20%. Compared with the ordinary non-iterative methods, the adaptive Tikhonov regularization exhibits high error tolerance.

Figure 45D shows reconstructed spectra with simulated random errors of each photodetector at CV=30%. Compared with the ordinary non-iterative methods, the or adaptive Tikhonov regularization exhibits high error tolerance.

Figure 45E shows reconstructed spectra with simulated random errors of each photodetector at CV=50%. Compared with the ordinary non-iterative methods, the adaptive Tikhonov regularization exhibits high error tolerance.

Figure 45F shows reconstructed spectra with simulated random errors of each photodetector at CV=80%. Compared with the ordinary non-iterative methods, the adaptive Tikhonov regularization exhibits high error tolerance. Although the spectrum accuracy decreases, the major peak can still be observed even with an error level of 8o%.

Figure 46 is photograph of an example of a nanobelt photodetector array.

Figures 47a to 47g are cross-sectional views of a range of semiconductor structures (for example, photoconductor, photovoltaic) which can be incorporated into a spectrometer.

Figure 48 is a schematic plan view of a photodetector array layout. Electrodes are laid 5 out side by side along the nanostructure and a bias is applied between adjacent electrodes, such that current flowing along the wire.

Figure 49 is a schematic plan view of a photodetector array layout. Electrodes are laid out either side of the nanostructure and a bias is applied between opposite electrodes, such that current is flowing across the wire.

/o Figure 50 is a computer generated model of a possible architecture for a snapshot spectral imager, consisting of repeated pixels, each containing the same spectrometer system, formed of multiple different nanowires, each covering different bandgap (and thus, spectral response) ranges.

Detailed Description of Certain Embodiments

Herein, viability of composition-engineered nanowires 1 is experimentally demonstrated for a new paradigm of ultra-compact computational microspectrometers 2, where the previously distinct elements which separate and detect light have been combined into an individual, micrometer-scale component grown in a single bottom-up process. These semiconductor nanowires 1 are alloyed such that the composition, and thus the spectral response, varies along their structure, as the material at different sections can only absorb photons with energy above its respective bandgap. It is shown that by electronically probing the photocurrent and cross-referencing with a pre-calibrated response function for each of a series of points along the nanowire 1, it is or possible to computationally reconstruct incident light signals. With a sufficient number of points, and through careful optimization of the measurement stability, along with the development of a bespoke algorithm, both monochromatic and broadband spectra can be accurately reconstructed. In this way, the entire active element of the spectrometer 2 is scaled down to a footprint ofjust hundreds of nanometers in diameter and tens of micrometers long, in a system that functions without the need for any complex or dispersive optics. The incorporation of the design into a mapping system allows spectral imaging from centimeter-scale image planes down to lensless, single-cell scale in-situ measurements.

Whilst epitaxial growth of thin-films with wide spatial compositional gradients is fundamentally difficult due to lattice mismatch with the substrate, the nanowire 1 -12 -growth interface is independent of the substrate once nucleated. As such, nanowires 1 can afford an almost arbitrary number of material systems to be alloyed into the same nanostructure through adjusting source vapors during production. This makes this spectrometer 2 design highly versatile; the growth of nanowires 1 with different composition-engineering straightforwardly realizes systems that could operate across any wavelength range from the infrared to ultraviolet.

The active element in the spectrometer 2 is a compositionally-graded semiconducting CdSxSe, nanowire 1, in which one end 3 is composed mainly of elements Cd and S. and /o the other 4 mainly of Cd and Se. This corresponds to a continuous gradient of bandgaps that can span from 1.74 to up to 2.42 eV along their length; Figure 1. The details of the synthesis and characterization of the nanowires are described later. Figure 2 shows the structure of a typical nanowire spectrometer 2. Electron-beam lithography is used to fabricate an array of parallel In/Au electrodes 5 on the compositionally-graded nanowire 1. To achieve stable electrical contacts, the nanowires 1 are treated under nitrogen plasma, followed by immersion in ammonium sulfide solution before electrode 5 deposition. The device 2 is then covered with A1203 via atomic layer deposition, to further improve the stability and consistency of the contacts. The I-V and time response of a typical photodetecting unit 6, between two neighboring electrodes 5 at the red (CdSe) end of the nanowire 1, is shown in Figure 3 and Figure 4 respectively. The unit 6 exhibits a high photoresponsivity of -200 AIAT-' with fast response (-1.5 ms) and recovery time (-3.5 ms). As can be seen in Figures, each of the photodetector units 6 of the spectrometer 2 exhibits a broadband spectral response function, with cutoff wavelengths that vary along the length of the nanowire 1. Details of the spectral response calibration are described later.

The basic operation of the nanowire spectrometer 2 is summarized in Figure 6. During the measurement, incident light represented by an unknown function, F(A), illuminates the spectrometer 2. Due to the small physical dimensions of the nanowire 1, it is considered that the incident light is spatially uniform across the device 2 (see Figures 37 to zjo). The computational strategy is illustrated schematically in Figures 7A to 7C. A data selector 7 scans photocurrent generated between each electrode pair 8, followed by signal processing. The photocurrent data, together with the pre-calibrated response functions are then processed to reconstruct the incident spectrum EGO by solving a system of linear equations: -13 -

AZ

F (A)R i(A) dA = (i = 12,3... n) (1) where A/ and A2 determine the spectrometer's 2 operational wavelength range. It is noted that the reconstruction is only possible for incident light wavelengths within this range. R,(A) is the spectral response of unit i, which is characterized beforehand and used as an input parameter. The photocurrent /i is the integration of F(A)Ri(A) over the wavelength range, as shown in the right panel of Figure 7A. For a spectrometer 2 with n photodetector units 6, there are n sets of equations. Solving these equations by ordinary non-iterative methods is liable to breakdown because the measurement errors in both Ri(A)andi, make the equations ill-posed. In this case, the reconstructed target spectrum is fully distorted at the existing system noise level (see Figures 41 to 43).

Aside from minimizing measurement instabilities through optimizing fabrication, an adaptive Tikhonov regularization scheme is introduced to reduce the influence of the errors during the reconstruction of F(A). The algorithm uses a linear combination of Gaussian basis functions with different amplitudes to fit the target spectrum as /5 illustrated in Figure 7D.

Characterization of two nanowire spectrometers 2 containing 38 and 30 photodetector units 6 is demonstrated through their ability to reconstruct the incident light spectra by this algorithm. The optimal bandwidth of basis functions is calculated by the empirical equation 2(A2 -10/n, which is -8.5 nm and 7 nm for the 30-unit and 38-unit spectrometer 2, respectively. As shown in Figure 8, the bandwidths of the reconstructed spectra of a narrow-band 560 nm light from these two spectrometers 2 are consistent with the bandwidths of their basis functions. Although both configurations can resolve two peaks around 570 nm separated by 15 nm, the device 2 with 38 units 6 reproduces -0 or more spectral information than the device 2 with 30 units 6; Figure 9. For both, reconstruction accuracy breaks down only once separation is decreased to 10 nm; Figure 10. A reduction of the physical gap between the adjacent electrodes 5 or synthesis of longer nanowires 1 with high bandgap gradient uniformity would allow an increase in the number of photodetector units 6, and thus higher spectral resolution.

Figure ii demonstrates a series of reconstructed narrowband spectra across a 130 nm bandwidth using the spectrometer 2 with 38 units 6. It is noted that the spectral range is primarily limited by the material composition used in the nanowires 1. This could be increased by choosing other material compositions such as Zn,Cd,,SySet y (1.74-3.54 eV), In"Gai,N (0.7-3.43 eV) and Si, .Gex (0.66-1.12 eV), allowing the fabrication of -14 -nanowires 1 with bandgaps spanning from UV to infrared. Figure 12 further demonstrates that the nanowire 1 spectrometer 2 is able to reconstruct broad spectra at varying spectral bands. Continuous broadband spectra can also be detected; Figure 13. The optimal detection power for a typical device 2 is between -zo and 1500 RVVcm-2, giving a dynamic range of -75. The response exhibits nonlinear or saturation behavior beyond this region (see Figures 34 to 36). One feasible avenue to increase the dynamic range is to introduce a background, broadband light source as a 'light bias'.

In many applications, such as astronomy, precision agriculture and nanophotonics, ro spectral imaging is in high demand to cross-analyze spectral and spatial information. A point scanning spectral imaging is performed using the 38-unit 6 nanowire spectrometer 2. As shown in Figure 14, a University of Cambridge logo pattern is displayed on a projector screen and focused on the nanowire spectrometer 2 by a lens. The nanowire spectrometer 2 scans across the focal plane, on the centimeter scale; Figure 15. The mapping step depends on the required resolution, which can be equal to or larger than the footprint of the nanowire spectrometer 2. Here a step of 0.3 mm is used. Photocurrent measured at each mapping step is recorded in an initial 3D data cube 9 as shown in Figure 16A. A spectral image 10 (Figure 16B) is extracted from the initial photocurrent data cube 9 by restoring the spectrum information in each pixel. As shown in Figure 17, a bandpass filter is used to filter the spectrum bands beyond the detectable region to avoid distortions in the reconstructed spectrum. Figure 18 shows the spectral images 10 obtained at various wavelengths. The pseudo-colored spectral image 10 in Figure 18 (top right panel) is converted from the spectra according to the International Commission on Illumination (CIE) color matching functions, which is consistent with the photograph with the bandpass filter. In addition, the reconstructed spectra of two points marked in Figure 18 fit well with the spectra measured using a conventional spectrometer; Figure 19.

Furthermore, in-situ micrometer-scale spectral imaging using the nanowire spectrometer 2 is demonstrated, which has long been a great challenge in fields such as cytobiology and biomedicine. The typical footprint of these nanowire spectrometers 2 is between 50-100 pm; as a result, spectral imaging at micrometer scales cannot be achieved by point scanning strategies as used in normal spectral imaging. A shift register strategy is adopted to address this problem by sequencing the photocurrent data and their locations, the principle of which is shown in Figure zo. For example, the photocurrent data of the first unit 6 (cyan) is recorded in register number ito zo, while -15 -the last unit 6 (red) occupies the register number 10 to 30. The spectral image 10 data cube is reconstructed from the overlapping register region of the initial photocurrent data cube 9, as shown in Figures 21A and 21B. The typical photodetector unit 6 size is -1 pm. The scanning step can be any integer multiple, n, of the unit size, depending on 5the required spatial resolution.

Onion cells 11 are used to demonstrate this micro-spectral imaging. As shown in Figure 22A, a red onion cell membrane 12, with a naturally colored cell surrounded by transparent cells 11, is mounted onto the holder and positioned over the nanowire /o spectrometer 2 with a gap of several microns. During measurement, the nanowire spectrometer 2 is moved over the membrane 12 surface by a motorized 2D stage 13 under the illumination of white light through the aperture. Figure 22A shows the optical micrograph of the onion cells ii and the grid of mapping steps. in this case, the step size, n=io units 6, is chosen for higher scanning speed, to avoid the cell membrane 11 drying out during the measurement, but smaller scans should be straightforward with further optimization of the system 14. Figure 22B illustrates the reconstructed absorption spectra of the four points on the onion cells ii, and the spectral images at various wavelengths are shown in Figure 23, revealing that spectral and spatial features are obtained at the cellular level by the nanowire spectrometer 2. Such a nanowire spectral imaging system 14 offers potential for a range of micron-scale in-situ applications, for example, distinguishing multiple fluorescent labelled cells in a microfluidic chip.

In summary, the use of single compositionally-engineered nanowires 1 enables an entire spectroscopy system 14 to be miniaturized down to a scale of tens of micrometers, which could open new opportunities for almost any miniaturized spectroscopic application, including lab-on-a-chip systems, drones, implants, and wearable devices. This proof of concept demonstrates a simple, versatile platform that can be expanded upon through a number of avenues by altering either the hardware or software of the system. For example, widening the spectral range can be straightforwardly achieved by using nanowires 1 of different semiconductor alloys 15. The reconstruction accuracy could be further improved by increasing the number of detector units 6 or by developing an even more sophisticated algorithm. High-speed spectral imaging could also be achieved, through the development of 2-dimensional spectrometer arrays 2. The study offers a practical step forward for other light sensitive -16 -nanomaterials to be directly exploited for customized design of ultra-miniaturized spectroscopy systems 14.

As depicted in Figure 24, the growth setup consists of a quartz tube 16 placed horizontally within a single-zone furnace 17; the two different source materials 15, CdS and CdSe powders contained in alumina boats 18 (2 cm in width, 1 cm in height and 6 cm in length), are positioned at the center and upstream of the heating zone 18, respectively. A constant separation distance d is maintained through the placement of two empty alumina boats 18 between those containing the sources 15. A step motor 20, io via a magnetic attachment 21, is used to drive a quartz push rod 21 within the tube 16 and gradually shift the CdS and CdSe along the furnace 17 during the growth process. The target substrate 23 (silicon wafer) is positioned downstream, at the edge of the furnace 17, and is coated with a 2 nm Au seed layer. To purge oxygen from the tube, and provide a carrier gas for the source vapor, nitrogen is flowed through the system at a rate of 15o sccm. Maintaining a pressure of 300 mbar (30 kPa), the temperature in the center of the tube 16 is raised by 40 per minute, to 830, over a period of 20 minutes. During the growth of the nanowires 1, the composition of the vapor is gradually tuned from CdS, through CdSSe, to CdSe by moving the CdS and CdSe source materials 15 downstream at a speed of 2.5 cm per minute. CdSe is moved from a low temperature zone to a high temperature zone, while the CdS is moved from a high temperature zone to a low temperature zone. As the change in the vapor concentration is reflected in the composition of the newly grown material 15, the S content increases, and Se content decreases, along the nanowires' 1 length. The furnace 17 is finally brought down to room temperature when CdS reaches a position of sufficiently low temperature that no more material 15 is being evaporated and thus the vapor content at the substrate 23 (or material composition of the grown nanowire 1) is pure CdSe.

Figure 25 demonstrates the resultant multicolor emission in a typical as-grown nanowire i via UV light excitation, where the emission color varies from cyan at the CdS end of the wire, through yellow and orange, to red at the CdSe end. The structure of these nanowires 1 is more finely investigated using transmission electron microscopy (TEM). Figures 26 and 27 show the TEM image of the nanowire i and its corresponding high-resolution TEM (HRTEM) images and selected-area electron diffraction (SAED) patterns taken from representative regions along its length. Analysis of these images demonstrates the grown wire's 1 hexagonal wurtzite structure and highly crystalline nature, as shown by the lack of apparent defects or phase segregation in the lattice -17 -profiles, and well-arrayed diffraction spots. The typical diameter of these nanowires is around 300 nm, with lengths in the range of 30-200 um; Figure 28.

To further study the relation between composition and bandgap, the corresponding composition of the nanowire 1 is measured by energy-dispersive spectrometry (EDX) in TEM. As shown in Figure 29, the elemental concentration of Cd element levels off at around 50%, but those for S and Scare complementary to each other along the nanowire 1. The cyan end contains mainly CdS, whilst at the red end the CdSe proportion is higher than that of CdS. For the CdSxSe,,, alloys, the bandgap can be jo represented by the quadratic function of compositional factor x as: Eg(CdS,Sei,)= xEg(CdS)+ (1-x)Eg(CdSe)-x(1-x)b (Si) where Eg(CdS"Se,,), Eg(CdS) and Eg(CdSe) are the bandgap values of CdS"Se,,, CdS and CdSe, respectively. b is the bowing parameter, which is 0.6 eV for CdSSei_i. The band gap values extracted from the measured photoluminescence spectra are consistent with those calculated by applying equation Sr to the measured compositions, as shown in the Figure 30. This result confirms the composition and bandgap gradient along the length of the nanowire i.

Device fabrication In order to fabricate an electrode 5 array 24 on the nanowire 1, two routes are investigated: non-embedded (Figure 31) and embedded (Figure 32). The fabrication of the non-embedded structure follows a standard e-beam lithography process. Nanowires 1 are first transferred from the growth substrate 23 to an intermediary substrate 25 through the contact of sliding one over the other. Individual nanowires 1 are picked up and transferred onto the final Si/SiO, device substrate using an optical fiber probe, heated and drawn to an ultra sharp tip and attached to a micromanipulator stage. Poly(methyl methacrylate) (PMMA) is then spun onto the substrate 25 and patterned by electron beam lithography. The surface treatment before the evaporation of electrodes is crucial to obtain good electronic contact between electrodes 5 and nanowire 1. Two metals as electrodes are studied: Indium (In) and Chromium (Cr). According to the results, the junctions between In (or Cr) electrodes 5 and the bandgapgraded nanowires 1 can be either ohmic or Schottky without the surface treatment. For example, for an untreated device 2, the junctions are usually Schottky on the green end -18 -of the nanowire, while ohmic on the red end. The inconsistency of the junctions between electrodes 5 and the nanowires 1 leads to a difference in the performance between the photodetectors 6 in the final multielectrode 5 nanowire spectrometer 2, which will adversely impact the spectrum reconstruction. Therefore, to make stable ohmic contacts with the electrodes 5, the exposed nanowire 1 is treated by nitrogen plasma for 30s, followed by 5 minutes immersion in diluted (NH4)2S solution (1 part of (NH4)2S to to parts of deionized water). This is followed by electrode 5 deposition (by thermal or electron beam evaporation for In-Au and Cr-Au respectively) and lift-off in acetone. The electrodes 5 are distributed parallel to each other across the nanowire 1 to jo collect the photocurrent generated from each section of the bandgap-graded structure.

The fabrication is completed after the deposition of A1203 via atomic layer deposition (ALD). The A1203 performs as a surface passivation layer and can greatly enhance the measurement stability.

Due to the typical diameters of the nanowires 1 used, electrodes 5 in the multielectrode nanowire 1 spectrometers, are liable to crack or show discontinuity at the interface between nanowires 1 and substrate 25, as shown in Figure 31. The failure of any electrode pair 8 means that the device 2 loses corresponding spectral information, and thus resolution, at the band related to the semiconductor composition at that location.

To achieve reliable connections, a S15-8 spacer layer 26 is used to avoid the abrupt height difference. As shown in Figure 31, the nanowire 1 is coated with SU-8 photoresist, which, when spun at a thickness lower than that of the nanowire 1 diameter, will align itself to the same height as the top of the nanowire 1. Electrodes 5 connect to the nanowire 1 via the SU-8 surface without the problems associated with an or abrupt height difference. The process can be seen in Figure 32. S11-8 photoresist (first layer) and PMMA (second layer) are spun onto the substrate. Due to surface tension, only a thin layer of SU-8 coats the surface of the nanowire (of the order of -to nm, as measured with atomic force microscopy) with a gradual slope either side, down to the main body of resist. Electron-beam lithography is then conducted on the PMMA layer to make the electrode patterns. The exposed SU-8 is partially etched using a nitrogen plasma in a reactive ion etch process to remove the thin layer on top of the nanowire 1, exposing the top surface. In the next stage, the exposed nanowire 1 goes through 3os nitrogen plasma treatment and 5 minutes immersion in diluted (NH4)2S solution, as with the non-embedded device 2. After electrodes deposition and lift-off, the device is covered with A1203 via ALD. The fabricated device 2 is then wire-bonded to a chip -19 -holder for characterizations. According to the studies, the optoelectronic properties of the non-embedded and embedded devices 2 are similar.

Response function characterization The spectral response functions Ri(2i) of each photodetector 6 is predetermined and must be calibrated before the reconstruction. As shown in Figures 33A and 33B, this calibration is carried out using a tunable light (FWHM-2 nm) dispersed from a xenon arc lamp. A diffuser is introduced to the output of the tunable light source to create a uniform light spot. In this case, the light power density is determined by a power meter.

/o An adjustable attentiator is used to equalize the power at each wavelength during the calibration. The photocurrent values of the photodetector unit under each wavelength are saved as the spectral response functions Ri(A), which are in turn used as an input of the reconstruction algorithm.

/5 Reduce system errors System errors can be reduced through surface passivation of the nanowires 1 by atomic layer deposition (ALD) of A1203. Coefficient of variation (CV), the ratio of the standard deviation a to the mean!I, is used to describe the photocurrent measurement error level. As shown in Figure 34, CV in the measurements of device 2 with A1203 deposition decreases because the A1203 minimizes the effect of any photochemistry occurring through the adsorbed gases at the surface. Furthermore, the photocurrents increase after A1203 deposition due to a decrease in the interfacial charge trap density.

The relationship between applied bias voltages and the stability of the current or measured is investigated. As shown in Figure 35, the error level has a significant increase when the bias voltage exceeds o.6 V. This is can be mainly attributed to temperature fluctuations induced by the large photocurrent 0.5 V for the measurement of the device 2 is chosen. Linearity with respect to light intensity is one of the most important parameters of a spectrometer. in addition, linearity strongly affects the measurement errors. The photoresponse linearity of the device is shown in Figure 36. The photocurrent increases nonlinearly with power density under low intensity incident light, while at stronger excitation, the response tends toward a linear relationship. One feasible avenue to achieve a more linear response is to introduce a broadband light source as a 'light bias' and set the measured photocurrents as zero points, representing a baseline for future measurements. Another way is to intentionally fabricate Schottky or p-n junctions on the nanowire to get more linear -20 -power response as ohmically contacted nanowire photodetectors (photoconductors) usually show non-linear behavior with increasing light intensity.

Spectrum reconstruction using adaptive regularization This section describes how the unknown spectrum FOL) is reconstructed pointwise from a finite set of measurement data. This can be expressed as: TAAmax F (A)12 i(A) dA = I (i = 1,2, 3, ... n) (S2) where Ii is the measured photocurrent, Ri(A) is the response function of each unit, Amm /0 and Amax are minimum and maximum wavelength respectively, and n is the total number of measurement units.

One of the most intuitive ways to reconstruct the function F(A) is to use a linear combination of Gaussian basis functions (AA) with constant coefficients a; forj = 1, 2, 75...,m: F (A) P (A) = where q)j(X) is a Gaussian function with a peak at X=X j: 11 ( A - \21 omlWrexP[ 2 a) and the constant o is the control parameter for the width of the Gaussian function with a = (2,/ln2) 8d O. 42478, where 8d is the full width at half maximum (FWHM) of the Gaussian function. The peaks of the Gaussian curves 1./ (j = 1, 2, 3..., m) are generated as a linearly spaced vector between the range of the minimum and maximum wavelength.

ikAmax Amin (k -1) + A",," (k = 1, 2, 3..., m) (S5) m -1 or (S3) (S4) -21 -where m is the total number of basis functions in the Eq. S3. This equation can be substituted into Eq. S2, leading to: (I Amax Ri (A)ctoi(A)chl.) aj = I i (56) i=1 Anini More compactly, represented in matrix form: Aa = c (57) where A is a known m x it matrix with element aii = f Ri (2)(1)1(A)dyl and c = [4, /2, ..., /"ris the known vector of measured photocurrent, which a is the vector of unknow coefficients for the basis functions, given by a = [al, a2, a"1T The objective of the problem is to find the value of each elements in the vector a (i.e. ap with j=1,2,3,...,n), that minimizes the residual norm (i.e. error) e = cliz: mintilAa -clifi a (S8) However, the minimization problem in the form of Eq. S8 is numerically unstable in its nature, since such process is numerically equivalent to the inverse of integration in the Eq. S7, thus amplifying the high-frequency noise signals. To alleviate such instabilities, a regularization factor y, also known as Tikhonov regularization, is introduced into the equation to selectively damp noise signals. Therefore, the minimization problem (Eq. S8) is reformulated as: try = mintilAa - y > 0 (59) a Where Fry = [al, az, slid?' is the optimal basis function coefficients that minimize the reconstruction error.

-22 -The damping coefficient should be carefully selected, in order to balance the requirement of robustness (higher y, more robust), and the resolution (lower y, higher resolution). Instead of manually tweaking the regularization factor, several adaptive parameter-choice methods for y have been proposed to automate the process, such as L-curve method, and Generalized Cross Validation (GCV).

Herein, the GCV method is used to adaptively select the regularization factor. For Tikhonov regularization, the optimal regularization factor? from the cross-validation can be expressed as: Pay c1122 11113,11 [in -trace (A (AT A + 3,21)-1AT)]2 (S10) where m is the number of basis functions, I is the identity matrix and the operator trace sums all element on the main diagonal of a matrix. The workflow of the reconstruction algorithm is shown in Fig. S9.

In addition, the quality of the reconstructed spectrum is also related to the choice of FWHM of the basis function. By incrementally varying this parameter in the Eq. S4, the optimal FWHM of the basis Gaussian function is approximately 8.5 nm and 7 nm for the 30-unit and 38-unit spectrometer 2, respectively.

Modifications It will be appreciated that various modifications may be made to the embodiments hereinbefore described. Such modifications may involve equivalent and other features which are already known in the design, manufacture and use of spectrometers and component parts thereof and which may be used instead of or in addition to features already described herein. Features of one embodiment may be replaced or supplemented by features of another embodiment.

Although claims have been formulated in this application to particular combinations of features, it should be understood that the scope of the disclosure of the present invention also includes any novel features or any novel combination of features disclosed herein either explicitly or implicitly or any generalization thereof, whether or not it relates to the same invention as presently claimed in any claim and whether or -23 -not it mitigates any or all of the same technical problems as does the present invention. The applicants hereby give notice that new claims may be formulated to such features and/or combinations of such features during the prosecution of the present application or of any further application derived therefrom.

Claims (21)

1. -24 -Claims 1. A photodetector array comprising: at least one semiconductor structure having first and second locations, 5 wherein at least a section of the semiconductor structure comprises at least two elements, wherein composition of the semiconductor structure varies between the first location and the second location such that bandgap of the semiconductor structure varies between the first location and the second location; and a number n of electrodes connected to and spaced apart along the /o semiconductor structure between the first location and the second location so as to provide a series of photodetector units along the semiconductor structure, wherein, in response to illumination of the semiconductor structure with light, each photodetector unit generates a respective photocurrent response.
2. 2. The photodetector array of claim 1, wherein the semiconductor structure is a semiconductor nanowire.
3. 3. The photodetector array of claim 1 or 2, wherein the semiconductor structure lies on a surface of a substrate and the photodetector array further comprises a dielectric layer disposed on the substrate arranged to be coplanar with at least a section of the semiconductor structure.
4. 4. The photodetector array of any one of claims 1 to 3, wherein a passivation layer covers at least a portion of at least one surface of the photodetector array.
5. 5. The photodetector array of any one of claims 1 to 4, wherein 71 8 so as to allow reconstruction of an incident spectrum comprising at least two peaks.
6. 6. The photodetector array of any one of claims ito 5, wherein the bandgap increases monotonically between the first and the second locations of the at least one semiconductor structure.
7. 7. The photodetector array of any one of claims ito 5, wherein the at least one semiconductor structure comprises at least one heterostructure.
8. -25 - 8. The photodetector array of any one of claims 1 to 7, wherein the bandgap at the first location of the at least one semiconductor structure is between 0.1 and 2 eV.
9. 9- The photodetector array of any one of claims ito 8 wherein the bandgap at the second location of the at least one semiconductor structure is between 2 and 5 eV.
10. 10. The photodetector array of any one of claims ito 9 wherein the semiconductor structure comprises CdS"Sei ", wherein osxsi and wherein, at the first location, x = xl, and at the second location, x = x2, wherein x1 x2.
11. The photodetector array of any one of claims ito 9 wherein the semiconductor structure comprises ZnxCd,"SySei y, wherein osxsi and wherein, at the first location, x = xl, and at the second location, x = x2, wherein x1 x2 and wherein osysi and wherein, at the first location, y = y1, and at the second location, y = y2, wherein y, y2.
12. 12. The photodetector array of any one of claims ito 9 wherein the semiconductor structure comprises in"Ga,N wherein osxs iand wherein, at the first location, x = x,, and at the second location, x = x2, wherein x, x2.
13. 13. The photodetector array of any one of claims 1 to 9 wherein the semiconductor structure comprises Si,Ge" wherein 0sxs1 and wherein, at the first location, x = x,, and at the second location, x = x2, wherein x, x2.
14. 14. A method of operating a photodetector array comprising a series of photodetector units, the method comprising: receiving calibrated spectral responses for each photodetector unit receiving photocurrent responses for each photodetector unit; reconstructing an incident spectrum in dependence upon the calibrated spectral responses and the photocurrent responses.
15. 15. The method of claim 14, wherein reconstructing the incident spectrum is performed by solving a set of equations, wherein the measured photocurrent for each photodetector unit is equal to: the integral of the product of the incident spectrum as a function of wavelength 35 and the pre-calibrated spectral response function as a function of wavelength, wherein the product of the incident spectrum as a function of wavelength and the pre-calibrated -26 -spectral response function as a function of wavelength is integrated between the limits of the minimum and maximum wavelengths of operation of the photodetector array.
16. 16. The method of claim 14, wherein generating the reconstructed incident spectrum in dependence upon the calibrated spectral responses and the photocurrent responses is performed by solving a system of linear equations: fa'ilnif incident spectrum(A)calibrated spectral responsei(A) dA = photocurrent, (1 = 1,2,3..., where Anita and Amt, determine the photodetector array operational wavelength range and i is the index of the photodetector.
17. 17. The method of any one of claims 14 to 16, wherein reconstructing the incident spectrum in dependence upon the calibrated spectral responses and the photocurrent responses is performed using adaptive regularisation.
18. 18. A spectrometer comprising: at least one photodetector array of any one of claims 1 to 13; and a computer configured to receive signals from the photodetector array and to perform the method of any one of claims 14 to 17.
19. 19. A method of forming a photodetector array, the method comprising; providing a substrate with a principle surface; 23 providing a compositionally-graded semiconductor structure having first and second locations on the principle surface of the substrate; providing a mask; disposing a number 71 of electrodes connected to and spaced apart along the semiconductor structure between the first location and the second location so as to provide a series of photodetector units along the semiconductor structure; performing passivation on the semiconductor structure.
20. 20. The method of claim 19, the method further comprising: embedding the semiconductor structure in a dielectric.
21. 21. The method of claim 19 or 20, the method further comprising removing the oxide layer from the surface of the semiconductor structure and performing passivation before the deposition of the electrodes.