Micrograph image of CMOS chip with bump-bonded micro-LED array. 

Context in source publication

Context 1

... use of fluorescence-based methodologies is at the core of many modern instrumentation technologies, especially in the life sciences. Originally, the interest was in the specific labelling of biological samples for imaging applications, but more recently this has expanded rapidly with the growth of DNA sequencing and micro-array applications. In principle, the equipment needed for such spectroscopic instrumentation has not changed over several decades, in that a narrow wavelength source is required to excite the fluorophore of interest and the resulting fluorescence must pass through an optical device to separate the excitation light from the fluorescence emission, before being detected by a light sensitive instrument and the level of fluorescence determined. However, using modern fabrication and manufacturing techniques, such instrumentation can now be made in complete integrated systems with the potential for volume production. The range of methods for interpreting the fluorescence data has also expanded with the growth of fast electronics, originally destined for the telecommunications and computer markets, now being applied to fluorescence instrumentation. Most commonly used fluorophores are designed to be excited in the near ultraviolet (UV) and blue regions of the optical spectrum. The advent of GaN and other blue emitting semiconductor sources, around 15 years ago [1], has helped stimulate interest in the use of such sources in place of lasers or mercury discharge lamps for excitation in fluorescence-based instrumentation. The advantages of these sources include low cost, reliability, compactness and good wavelength matching to many standard fluorophores. In addition, their ability to produce short ( ≈ 1 ns) pulse lengths offers the opportunity for low cost fluorescence lifetime measurements. It has been appreciated for a long time [2] that the fluorescence decay from a molecule provides a significant level of information about the sample and the surrounding conditions of the fluorophore, from the local viscosity in cell membranes [3] to pH [4]. Most recently interest has been in the rapid growth of Föster resonance energy transfer (FRET) methods, where the spatial separation of two fluorescent molecules significantly affects the fluorescence lifetime of the shorter wavelength ‘donor’ molecule. Through advances in molecular chemistry, FRET based assays and experiments utilizing lifetime measurement as the monitor of the FRET process are emerging as the method of choice for many applications in the life sciences [5]. Several methods of lifetime measurement are now in common use in lifetime instrumentation, but all require the common features of a short pulsed, or high frequency modulated, light source, fast detector and timing electronics. Time correlated single photon counting (TCSPC) is potentially the most accurate method of determining complex decays (multiple lifetimes) and was therefore the method selected to demonstrate the performance of our system. The phase method, in which the phase change in the fluorescence emission relative to the excitation pulse is measured [6], is less sensitive compared with TCSPC and applies more light to the sample which can lead to faster sample decomposition. A slight variation on TCSPC is the use of time gated detectors which provide a slightly less accurate lifetime measurement, but are excellent at providing a method of detecting changes in lifetime [7]. More recently, a variation on the phase method has been developed, whereby the change in output intensity is measured [8]. This method uses a slow detector and thus was not suitable in a system that makes use of rapid detectors. Therefore, we chose the TCSPC method as we believed that this would demonstrate the capability of our novel excitation source and detector. Furthermore, if the system worked in this configuration it should also work in the slightly less demanding phase and gated configurations, though these might be the preferred options in a final instrument configuration. In 1995, Araki and Misawa [9] and in 2005, Davitt et al [10] both reported the use of light-emitting diodes (LEDs) for lifetime measurements; however, their system used a conventional photomultiplier. Although this work demonstrated the capability of LEDs for lifetime measurements with the use of a photomultiplier as the detector, the system was not capable of low cost miniaturization. In this paper, we describe a single chip embodiment for such measurements based on time correlated single photon counting. In our approach, a 16 × 4 array of micro- pixel AlInGaN LEDs emitting at 370 nm [11] has been bonded to an equivalent array of aluminium electrodes made from the top metal of a 0.35 μ m high-voltage CMOS chip. Each electrode has an associated driver circuit capable of switching up to 50 V. Incorporated into each electrode is a single photon avalanche diode (SPAD). The output of the SPAD detectors is processed in real time by a TCSPC module that generates a histogram of the fluorescence decay [12]. Other pertinent features of the CMOS chip include on-chip control of the micro-LED pulse width (1.5–48 ns optical pulse width) and the ability to separately address any micro-LED or SPAD in the array. With this device, we have successfully performed demonstration measurements in a colloidal suspension of quantum dots, achieving lifetimes consistent with manufacturer’s specifications and our own reference measurements taken using a photomultiplier tube detector. This microsystem solution will enable the development of low cost, portable fluorescence lifetime readers for many optical lab-on-a-chip applications such as point-of- care diagnostics equipment and the synthesis and/or read out of DNA micro-arrays. With each element being separately addressable, there is the potential to excite many fluorescent samples in parallel [11]. By integrating the excitation source with a photodetector and on-chip driving electronics, our devices will contribute to the development of lab-on-a-chip (LoC) systems [13]. This includes work by Chodavarapu et al [14], which aims to carry out fluorescence detection using a CMOS-based system which incorporates a detector with signal processing circuitry. Their system does not, however, include an integrated excitation source and uses a discrete LED. Cleary et al [15] demonstrated TCSPC on a micro- scale using integrated optics and microfluidics; however, their system relies on a pulsed diode laser as an excitation source and a discrete SPAD detector. The system presented here offers a greater level of integration, placing the excitation source in the same micro-scale system as the sample and detector. A micrograph image of the CMOS chip with bump-bonded micro-LED array can be seen in figure 1. The micro-LEDs consist of a 16 × 4 array of individually addressable pixels, each pixel having a diameter of 72 μ m on a 100 μ m pitch. The devices were fabricated from ‘standard’ UV LED wafers grown on c -plane sapphire substrates by metal organic chemical vapour deposition and have a peak emission at 370 nm [16]. The array of electrodes was depassivated by pad opening at the foundry to reveal the aluminium top- plate electrodes for bump bonding. A post-processing step of oxygen plasma etching removed the polyimide layer of the chip which has been shown to improve the photon detection probability (PDP) of the underlying SPADs by a factor of 2–5 [17]. An example of the top metal driver plates can be seen in figure 2. An electrical connection between the ...