Detecting light in a multispectral world

Narrowband photodetection can be regarded in many respects as the optoelectronic equivalent/extension of our sensorial capability to discriminate colour. Therefore, it is not surprising that an important motive of narrowband photodetection is, in fact, colour detection, i.e. the determination of the composition of visible light (wavelength range λ ≈ 400–750 nm) in relation to a set of primary colours. In virtue of this, narrowband photodetectors operating in the visible range are often alternatively referred to as colour-selective photodetectors ¹ . In further analogy with our visual perception, a large number of colour-selective photodetectors are arranged into arrays (known as colour imagers) with the aim of capturing a two-dimensional colour map of an object or a scene (colour imaging).

The scope of narrowband photodetection is not limited to colour detection in the visible range, nor to applications targeting exclusively a human end user. Looking beyond what the eye can see, narrowband photodetection has been pursued in the near-infrared (NIR) (λ ≈ 750–2000 nm) and in the ultraviolet (UV) (λ ≈ 10–400 nm) ranges. In addition, narrowband photodetection has evolved to provide a finer look at the colours in the visible range and beyond, leading to what is known as multispectral and hyperspectral imaging. In contrast to conventional colour sensing, which relates to a limited set of colour primaries, multispectral/hyperspectral imaging involves a greater number of bands (cf. colours), which populate the wavelength axis more finely than conventional primary colour bands. These bands can cover the visible range, but can also extend, for instance, to the NIR and the UV, as needed by the application at hand. This allows one to capture a large number of two-dimensional maps of a scene, as many as the wavelength bands that the imager can resolve.

Having briefly outlined the general ideas underlying narrowband photodetection, in the remainder of this chapter we will explore in detail its technological significance, thus illustrating the context and motives of organic narrowband photodetectors.

In this section we provide an overview of the target application areas of narrowband photodetection. In particular, we first examine target applications pertaining to the visible range, and then illustrate those relevant to the NIR and UV ranges. This provides the basis for the subsequent discussion of benchmark and organic narrowband photodetector technologies (sections 1.3 and 1.4).

1.2.1. Narrowband photodetection in the visible range

In the traditional sense, narrowband photodetectors for colour acquisition are meant to determine the colour coordinates of a light source within a given colour space. This is most commonly referred to as a set of three additive primary colours, e.g. red, green and blue (RGB), or the derived CIE 1931 XYZ tristimulus values (figure 1.1(a)), which relate to human colour perception [1]. While individually, at the core of devices called tristimulus colorimeters, colour sensors find their most widespread use in image sensors (also known as imagers) (figure 1.1(b)), such as those present in cameras of consumer electronic devices (e.g. digital still cameras, mobile phones) and of specialised equipment (e.g. for video surveillance and medical applications). Their relation to human colour perception requires colour selectivity according to an RGB (or derived) colour map, with spectral bandwidths in the region of 100 nm (see, for instance, CIE standard observer colour matching functions).

Zoom In Zoom Out Reset image size

Over the years, narrowband photodetection in the visible range has significantly expanded its scope to also cover/target applications that move beyond a human end user. In machine vision (also known as computer vision), the end user of the acquired images is a machine (e.g. robots, smart cars, smart appliances), which automatically processes them and identifies key information [2]. Machine vision finds application in industrial inspection (e.g. for quality control on a production line), security (e.g. face recognition for intruder identification, biometrics), military applications (e.g. automatic target recognition), robot vision (e.g. for automated manufacturing, robot-assisted surgery and robot driving) [3, 4], and is relevant to emerging application domains such as the Internet of Things (IoT) and Artificial Intelligence (AI). In view of the fact that a machine is the end user of the captured images, it has been shown that a spectral bandwidth of 100 nm or less is necessary so as to achieve illuminant-invariant colour detection in the visible range (i.e. smaller than that needed for common RGB sensors and imagers) [5, 6]. Apart from machine vision systems aiming at colour recognition, a further trend has emerged that involves machine vision for so-called multispectral sensing/imaging. In contrast to conventional RGB photodetection, multispectral sensing in the visible range involves a greater number of bands (up to approximately 10), which are then characterised by a substantially narrower bandwidth (e.g. figure 1.1(c)). These bands typically cover the visible spectrum, but can also extend to adjacent spectral regions, as needed for the application at hand. For instance, this is relevant to applications that involve satellite image acquisition for land surveying and earth science research [7].

A further group of application areas that have become relevant to visible narrowband photodetectors in recent years involves so-called lab-on-chip systems for biomedical and analytical purposes [13–16]. One such application is flow cytometry (e.g. figure 1.1(d)), a technique used to evaluate physical and chemical properties of an ensemble of cells, thus allowing the detection of microorganisms, biomarkers and diseases. In the context of lab-on-chip systems, the light to be sensed may come from an analyte that inherently emits light or that luminesces under light excitation. Such applications require the use of narrowband photodetectors in the visible (and sometimes in the NIR) range, so as to determine the luminescence within the particular wavelength range(s) of interest. A particularly significant demand in this area is that of lab-on-chip devices that are low cost, miniaturised and disposable, for instance, towards medical applications [17].

Another application domain relevant to narrowband photodetectors concerns optical communications. For the sake of illustration, an application that has emerged in recent years concerns so-called visible light communications (VLC) [18]. In consideration of the ever-growing demands for data transmission and wireless communications, VLC has been envisioned with the aim of overcoming the issues of spectral crowding and bandwidth limitations of existing technologies. Instead of relying on the radiofrequency and microwave regions of the electromagnetic spectrum, VLC uses visible light signals for wireless data transmission. Information is transmitted through optical signals exchanged between a transmitter (in principle, any light-producing object we encounter in our daily lives) and a receiver (individual photodiodes or imagers, such as the ones available in many consumer electronic devices). VLC is regarded as very promising for high-speed communications—e.g. as relevant to the IoT—due to the ubiquity of light-producing objects and the high data transmission rate achievable. One of its implementations would allow multiuser access through the combination of RGB sources and narrowband photodetectors [19, 20]. This motivates the development of highly-sensitive RGB narrowband photodetectors with comparatively fast response (e.g. operating frequency of several tens/hundreds of MHz) [21, 22].

Finally, narrowband photodetectors have also formidable potential for futuristic applications. One such application involves their use in bioelectronics, most notably for prosthetic devices aiming at the restoration of the visual function of patients with retinal damage (e.g. figure 1.1(e)) [23]. Over the past decades, impressive results have been obtained in the development of prosthetic devices that directly interface with the retina to produce visual sensory function [24]. As an example, one such approach relies on an intraocular photodetector array to transduce the incoming illumination into electrical signals, which then stimulate the retinal neurons. While emphasis has been placed on the restoration of basic visual function up to this point, it can be envisaged that narrowband photodetectors may play an important role in the development of implants for colour vision.

1.2.2. Narrowband photodetection in the NIR range

1.2.3. Narrowband photodetection in the UV range

Narrowband photodetection/imaging has been conventionally realised through crystalline inorganic semiconductor technologies, e.g. crystalline silicon (c-Si) and III–V semiconductors. These technologies have successfully delivered for a wide range of applications. Their inherent optoelectronic properties (most importantly, broadband and non-tunable absorption), however, leaves input optical filtering as the only strategy for narrowband photodetection to be achieved. This poses some hard limits on spectral bandwidth and performance, and burdens the complexity [44] of the detector/imager architecture and of the fabrication process. Additionally, conventional inorganic semiconductors rely on fabrication technologies that require expensive and high-temperature fabrication processes.

Crystalline silicon is the dominant technology for narrowband photodetection/imaging in the visible range. Its use is commonplace in colour sensing/imaging, for which it relies on colour separation systems to achieve narrowband response (due to its broadband absorption up to ≈1.0 μm, see figure 1.2(a)). Such systems most often consist of passband absorptive filters. In the case of an imager, the typical arrangement involves a two-dimensional array of broadband c-Si photodetectors covered by a colour filter array (CFA) (figure 1.1(b)) [45]. The CFA appears as a grid of absorptive filters (one for each photodetector), some transmissive in the red, some in the blue, and some in the green, according to a regular pattern (e.g. a red–green–blue–green Bayer filter mosaic) [46]. While the good performance of this scheme has led to its widespread use, it presents a number of drawbacks. Firstly, about two thirds of the incident light are absorbed by the filters and do not contribute to image formation [47, 48]. Additionally, photodetectors sensitive to different colours are at a distance from each other, hence interpolation algorithms are required to reconstruct the colour at each position, leading to aliasing problems. Finally, considering that the colour at each point involves the use of three neighbouring photodetectors, the CFA-based approach inevitably limits the imager resolution. An alternative to the CFA approach consists in employing three different image sensors and a prism. The prim separates the incident light into red, green and blue components, and each is directed onto one of the three image sensors. Therefore, each image sensor is dedicated to the acquisition of an image of a specific colour at the maximum spatial resolution available (as per imager technology). While delivering higher spatial resolution than the CFA-based configuration, this approach is burdened with greater complexity and cost. In summary, the use of colour separation systems to achieve narrowband photodetection brings important limitations and challenges.

Zoom In Zoom Out Reset image size

Shifting our focus to narrowband photodetection in the visible range for lab-on-chip systems, this is most often realised through broadband photomultiplier tubes (PMTs) or avalanche photodiodes [14] (e.g. figure 1.1(d)), again in conjunction with external filtering (absorptive filters or dichroic mirrors). If conducted in the multispectral mode, an array of broadband photodetectors (silicon imagers or multichannel PMTs) have been used in conjunction with a dispersive element (prism or diffraction grating) that separates light of different wavelengths [49]. It is noteworthy that such traditional photodetectors do not generally lend themselves to integration for low-cost, miniaturised and disposable lab-on-chip devices [17] (as demanded by many emerging applications) due to size, cost and the need for input filtering.

The dominant technology for detection in the NIR range (1.0–1.6 μm) is based on indium–gallium–arsenide (InGaAs) [50]. In view of their broadband character (figure 1.2(a)), such detectors are to be used with input filtering so as to achieve narrowband response. With regard to spectroscopic applications, the conventional approach to narrowband photodetection is based on input filtering through bulky diffraction gratings or interferometers [31, 51]. As far as imaging applications, instead, typical implementations involve a two-dimensional array of conventional broadband photodetectors in combination with an input tunable filter, e.g. an acousto-optical filter, a liquid-crystal filter, or a micro-electro-mechanical-system (MEMS) based interferometer [31].

In response to the demand of developing compact and/or handheld NIRS devices, great strides have been made in respect to the miniaturisation of the components needed for spectrally narrowband NIR photodetection. While still resorting to broadband InGaAs photodetectors, progress has been made by replacing conventional bulky dispersive elements with more compact ones (e.g. MEMS interferometers or linearly variable interference filters) [31, 51]. This has brought about significant portability, yet further progress in miniaturisation is still be needed in order to address applications that relate to wearable electronics and biomedicine.

In the UV range, the most conventional approach to narrowband photodetection involves the use of PMTs with solar-blind photocathodes (e.g. made of CsI and Cs₂Te). While such devices manifest excellent UV responsivity and spectral rejection of visible photons, they are based on fragile and bulky components (vacuum tubes), and they require very high driving voltages (several hundreds of volts or higher) [52]. This implies, for instance, that UV-selective PMTs are unsuitable for field applications.

Another conventional approach to UV narrowband photodetection involves input-filtered silicon photodiodes. Indeed, being a semiconductor whose absorption spectrum extends to the NIR range (figure 1.2(a)), silicon requires input filters so as to suppress its response to visible and NIR photons and thus deliver a visible-blind photoresponse. This approach is attractive insofar as it builds on the mature silicon technology, yet the resulting efficiency is generally low (<30%). Additionally, this approach faces important limitations, as input filtering here relies on alkali metal filters—which are expensive, bulky, and prone to degradation [53]—or on photonic crystals—which are demanding in terms of fabrication complexity [54].

More recent approaches to UV narrowband photodetection that have overcome the limitations of long-established technologies involve the use of SiC and of Al_x Ga_1−x N alloys. Their attractiveness resides in their wide-bandgap nature (figure 1.2(a)), which inherently allows filterless UV photodetection. Notwithstanding its suitability for visible-blind photodetection, SiC requires input filtering when tackling applications that demand a photoresponse within a particular UV band [32]. In contrast, the absorption onset of Al_x Ga_1−x N alloys can be tuned between 360 nm and 200 nm through compositional engineering (figure 1.2(a)), thus allowing one to tailor the associated UV photodetector response as demanded by the application at hand [55]. Notwithstanding their merits, SiC and Al_x Ga_1−x N photodetector technologies require high-end deposition processes (e.g. vapour phase epitaxy, low-pressure metal–organic chemical vapour deposition [56, 57]), which result in significant fabrication complexity and comparatively high cost [58].

In recent years we have witnessed a rising demand for narrowband photodetectors that can be used seamlessly at the point-of-need for a wide range of applications, e.g. to enhance visual interaction with the objects and environments of our daily lives (e.g. IoT and AI). This would require sensors that are compact and inexpensive, so as to allow their near-ubiquitous deployment. A particularly significant emerging trend points to having image sensors that are mechanically lightweight and flexible, so as to have them deployed in wearables and the biological environment.

Technologies that require input filtering to achieve narrowband response are inevitably burdened with a comparatively complex fabrication process and higher cost. Additionally, they generally result in devices that are bulky and mechanically rigid. It follows that conventional technologies are generally unable to deliver on the emerging demands for narrowband photodetection. Therefore, a large number of academic and industrial investigators have pursued alternative semiconductor technologies that could surpass these limitations.

Organic semiconductors have emerged as a technology with formidable potential for filterless narrowband photodetection in the visible, NIR and UV ranges. While their properties are discussed in detail in chapter 2, here we briefly note that organic semiconductors inherently allow the tuning of their absorption spectrum (e.g. figure 1.2(b)) by chemical tailoring, with practically limitless possibilities in terms of composition and structure. In particular, a large number of them exhibit narrowband absorption (e.g. figure 1.2(b)). In most cases, such compounds possess an in-band absorption coefficient α approaching or greater than 10⁵ cm⁻¹, thus a layer thicknesses in the 100 nm range is sufficient for their use in photodetectors (cf. c-Si requires micrometre-thick active layers due to its low α ≈ 10⁴ cm⁻¹), pointing to good resilience to optical cross-talk when integrated into imagers [66]. In addition to their soft nature, such thinness allows them to achieve outstanding mechanical flexibility, reliably down to a submillimetre bending radius [67], which points to their ability to meet operational requirements for wearables and biomedical applications.

In addition to their advantageous properties, organic semiconductors are particularly attractive for narrowband photodetectors in view of their potentially low-cost fabrication. While a discussion on the deposition processes required for narrowband organic photodetectors is deferred to chapter 4, here we briefly note that, in many instances, organic semiconductors can be deposited at low temperature (down to room temperature) and from solution [68] (e.g. via coating and printing methods), doing away with expensive, high-temperature and vacuum-based fabrication steps required by conventional photodetector technologies. Additionally, in contrast to other solution-based technologies that have emerged in recent years [69, 70] (e.g. lead-halide perovskites, inorganic nanocrystals and quantum dots), organic semiconductors are generally regarded as biocompatible [23, 71], and in particular they do not contain elements such as heavy metals that give rise to significant toxicity concerns [72, 73]. If processed through vacuum-based methods (e.g. vacuum thermal evaporation), they can be deposited over large areas and at relatively low temperature, thus still providing a scalable technology platform for potentially low-cost narrowband photodetectors and imagers.

These and other aspects concerning organic semiconductors are discussed in much greater detail in the remainder of this book. Here we conclude, however, by noting that it is precisely the combination of the aforementioned properties that have provided the driving force for the investigation and development of organic narrowband photodetectors.

This chapter has introduced the context relevant to narrowband photodetection in the visible, NIR and UV ranges. It has done so by firstly surveying many of the application domains in which narrowband functionality is required. It has shown that these domains are particularly broad in scope. Indeed, going well beyond conventional applications such as colour detection, narrowband photodetection is relevant to optical communications, computer vision, spectroscopic applications, agriculture, industrial monitoring and medicine, etc. A recurrent trend that has emerged in recent years is the push for narrowband photodetectors that are compact, lightweight, portable and low cost, hence potentially allowing their dissemination in a wide range of environments and form factors. Finally, against the backdrop of conventional photodetector technologies, we have illustrated that organic semiconductors are inherently well poised to address the emerging functional and technological demands of narrowband photodetection.