Counting the options to measure low light levels

05 Mar 2004

Detecting the presence of just a few photons of light is essential for many research and commercial applications. Siân Harris investigates the types of detectors that can help.

From Opto & Laser Europe March 2004

There are many situations in which scientists and engineers need to detect very low levels of light (less than around 1000 photons per second). Spectroscopy, for example, is a valuable tool for many chemists, biologists and materials scientists and relies on exciting a substance and looking at the small amounts of light emitted. The military and meteorologists rely on LIDAR techniques to detect faint reflections from clouds and other objects in the sky. And applications such as quantum communication and cryptography deliberately exploit the properties of single photons.

Researchers in these areas require devices that can detect and count single photons. Such detectors convert photons into electrons and amplify them in order to produce an electrical signal that can be detected easily by conventional electronics. The type of detector used is selected according to the light levels being studied, the wavelength of the light and the background levels present in the measurements.

Photomultipliers The photomultiplier, a device that was first used in the 1950s by astronomers studying single photons of light from distant stars, is one option. Photomultipliers are based around a photocathode, which exploits the photoelectric effect to absorb photons and emit a proportional number of electrons. These electrons are electrostatically accelerated and focused onto a special type of electrode known as a dynode. When an electron hits a dynode it liberates a number of secondary electrons. These electrons are then focused onto the next dynode to release further electrons.

This amplification process is repeated between 10 and 14 times while the secondary electrons released from the final dynode are collected at the photomultiplier's anode. This gain mechanism multiplies the number of electrons by around 10⁶ or 10⁷ times, depending on the device specifications. This increases the output signals so that they can be seen on an oscilloscope.

A variety of different photocathode materials is available and each has a characteristic spectral response. The best choice is usually the one that offers the maximum response in the wavelength region of interest, although other factors such as light levels and dark current should also be considered. Dynode structures also vary, and the most suitable one for you will depend on requirements such as size, gain and timing.

Photomultipliers offer excellent time resolution and are suitable for wavelengths ranging from 1 µm down to around 120 nm. Shorter wavelengths than this could be detected by the photocathode, but radiation below this level is absorbed by the materials used for the windows in the detector. Some specialist photomultipliers operate at longer wavelengths than 1 µm, but these devices are very expensive.

Photomultipliers have a large active area - up to around 20 inches in diameter in some cases. This is an advantage for many applications because it does not require the light to be tightly focused. However, it also has disadvantages because a larger photocathode results in larger dark count rates, which can mask the signal when measuring infrequent events.

Many of the applications of single-photon detectors require a wide dynamic range of anything between a couple of photons per second and 10⁷ photons per second. For lower light levels photomultipliers record electrical pulses. They can cope with up to approximately 10⁸ photoelectrons (the electrons emitted in the photomultiplier) per second in this way.

Photomultipliers have to be corrected for dead time, which is when a new photoelectron cannot be detected because the detector is still busy processing the previous one. However, above 10⁸ photoelectrons per second light can be detected by measuring the current instead of counting pulses, which removes the dead time problem. The number of internal gain steps can also be reduced to expand the range of the device.

Avalanche photodiodes For higher light levels, avalanche photodiodes (APDs) are an ideal choice, provided that the light can be focused onto a small enough spot. When bias voltages are applied to these devices it produces an avalanche effect that multiplies the current, typically by between 100 and 1000 times.

In contrast with the wide detection area of photomultipliers, APDs typically have an active area of around 0.5 mm. This means that they require more tightly focused light sources but are less prone to background, although they tend to require cooling to work well. They may also produce more noise than photomultipliers because they require external amplification. However, in common with photomultipliers, APDs have excellent time resolution.

Two types of APD are available. Silicon-based APDs detect wavelengths between 600 nm and 1 µm and serve the visible and near infrared range. Some research has gone into extending this range down to blue and near ultraviolet wavelengths, but commercial products based on this have yet to appear. As wavelengths become shorter the photons have more energy, and so become easier to detect with photomultipliers.

Longer wavelengths are served by APDs based on indium gallium arsenide. These devices cater for wavelengths between about 1.1 and 1.65 µm.

APDs for single-photon detection Conventional APDs can detect light levels as low as around 1000 photons per second. However, the sensitivity of APDs can be enhanced to enable them to detect single photons by using the so-called Geiger mode. This exploits the behaviour of diodes when the voltage applied to them is too high. In this situation they eventually break down, but they initially sit in a metastable state until a photon is absorbed. This produces a self-sustaining avalanche which is very large and can provide large enough amplification to enable single photon detection. Each time devices based on this effect detect a photon, the avalanche is quenched and the device is reset, ready to detect the next photon.

Both photomultipliers and APDs require a pulse height discriminator to ensure that only pulses above a threshold size are counted.

Electron multiplying CCDs After their early work with photomultipliers, many astronomers turned to charge-coupled devices (CCDs) because they wanted to image areas of the sky, rather than simply count photons.

A CCD is a silicon-based array of imaging sensors. It typically uses tens of thousands or even millions of pixels to collect an image. By contrast, photomultipliers usually have up to 32 sensing elements, and APDs tend not to have more than four. This feature of CCDs means that they offer outstanding spatial resolution compared with the other two systems.

However, photomultipliers and APDs resolve time much better than CCDs. They can provide nanosecond-level resolution, whereas CCDs take between 50 and 500 frames per second to maximize the signal and so can only resolve times down to a few microseconds.

A recent innovation in CCD technology has enabled these devices to amplify image signals on-chip with a low noise-impact ionization process. These electron multiplying (EM) CCDs can provide voltage-controllable gain of up to 1000 times, which removes the limitation of read-out noise from the chip amplifier. The easily controllable gain of the EMCCD gives high-efficiency imaging across a wide range of light levels, from single photon detection to photon fluxes greater than 10¹³ photons/cm²/s.

EMCCDs detect photons from silicon's upper wavelength limit at 1064 nm down to the vacuum ultraviolet and below. To maximize the detection efficiency and for imaging in the ultraviolet, special techniques such as backthinning can help.

To get the best signal-to-noise ratio EMCCDs need to be cooled. This is because silicon naturally releases free electrons when warm and it is important to reduce this dark current as much as possible for low light-level detection. Cryogenic cooling as far as -90 °C is sometime required, but temperatures of -40 or -50 °C are fine for many applications and temperatures of -5 °C are adequate for surveillance applications. Cooling is usually achieved with a bimetal device called a Peltier or TEC cooler, which do not require much power (about as much as a household light bulb) as they only need to cool a small area of silicon.

Final considerations In terms of price, EMCCDs are probably the most expensive option. Normal prices range from around £500 to £2000 (€750-3000) depending on device capability. Photomultipliers that are capable of single-photon counting can be picked up for as little as £100 or £200, although devices for research are more expensive. Silicon APDs can be bought for between £100 and £300, but the compound semiconductor-based infrared APD detectors start from around £1000.

Within each type of detector there is a range of options for features such as the structures or materials of the various components. The best one for a particular application depends on weighing up factors such as wavelength, light levels and temporal and spatial resolution. In the past people needed to build their own detection systems from a range of components, which required specialist expertise. Now this worry has been taken away with a wide range of off-the-shelf light-detection modules that customers can couple straight to their PC and power supply. However, as with any major purchase, it is important to think carefully about what features your application requires. The companies that supply detectors then strongly advise you to speak to their experts before making your final decision, so that you can ensure that you end up with the most appropriate device. In some cases, suppliers may be able to customize devices for particular applications.

Noise and background considerations Although the terms "noise" and "background" are often used interchangeably, they mean different things in the context of single photon detectors. Both are important considerations when choosing a device.

Noise is a statistical process that results from the amplification of the electrons. Every electron is amplified by a slightly different amount. For a photomultiplier, the electrons might ideally be amplified by a factor of exactly 10⁷, for example, but in practice they result in a range of signal sizes. This slight spread occurs with all types of detector but is greater for APDs than for photomultipliers because APDs require more external amplification in addition to the internal gain. Like APDs, CCDs amplify signals on the chip by a factor of around 1000 in a fairly low-noise process. Any further amplification is performed externally.

Background, or dark count or current, results from the electron emissions that occur even when no photons are absorbed. These emissions arise from thermionic effects and the presence of radioisotopes in the material. Cooling any type of detector helps to reduce thermal effects, while the use of low-activity glass can reduce the effect of radioisotopes. Background can be greater in photomultipliers than in APDs because the former have larger detector areas. CCDs have large areas so are very susceptible to dark current at room temperature. However, they are always cooled to reduce this problem.