Photon detectors

The class of photon detectors contains the most important detector types for computer vision. Apart from a few exceptions, such as pho-tographic films, most photon detectors are solid state detectors, which make use of the fact that electrical properties of semiconductors are

4.5 Detecting radiance 99 dramatically altered by the absorption of ultraviolet, visible and in-frared photons.

Intrinsic photoconductors. Photoconductors respond to light by ei-ther changing resistance or conductance of the detector material. In-trinsic photoconductors are the most straightforward way to design a solid state electronic detector. They make use of the inherent electrical property of pure semiconductor materials without additional manipu-lations. At normal temperatures, relatively few electrons will be in the conduction band of a semiconductor, which results in a low electric conductivity of the material. Figure 4.6a illustrates the energy-band diagram for an intrinsic photoconductor.

In order to move from the valence band into the conduction band, an electron must have sufficient energy. By absorbing a photon whose energy is greater than that of the bandgap energy Qg, an electronic bond can be broken and the electron can be lifted into the conduction band, creating an electron/hole pair (Fig.4.6a). Both the electron and the corresponding hole can migrate through the detector material and contribute to the conductivity. If an electric field is maintained across the detector, any absorbed photon results in a small electric current, which can be measured by a high-impedance amplifier.

As thermal excitation contributes to the conductivity in the same way as absorbed radiation, thermal noise will corrupt the signal, espe-cially at high temperatures and low illumination levels. The number of thermally excited electrons follows theBoltzmann distribution:

nt∝exp

−Qg

kBT

(4.32) whereQg,kB, andT are the bandgap energy, the Boltzmann constant, and the absolute temperature, respectively. AsQg becomes smaller, the number of thermally excited charge carriers increases. One way to overcome this problem is to cool the detector down to cryogenic temperatures below 77 K (liquid nitrogen temperature), where thermal excitation is negligible.

The minimum photon energy that can be detected is given be the bandgap energyQg of the detector material. With the photon energy (Eq. (2.2))

ep=hν=hc

λ (4.33)

the maximum detectable wavelengthλc, commonly referred to ascutoff wavelength, is given by

λc= hc

Qg (4.34)

Substituting for the constants, and correcting for units such that wave-lengths are in microns and energy gap in electron volts yields the fol-lowing rule of thumb:

λc[µm]= 1.238

Qg[eV] (4.35)

Intrinsic photoconductor detectors can be made in large arrays and they have good uniformity and high quantum efficiency, typically in the order of 60%. They are the basic components of CCD-arrays (charge coupled devices), which are the most widely used 2-D detectors in the visible, the near infrared, and—to some extent—in the x-ray and ultravi-olet region using special semiconductor compounds. In the infrared re-gion, semiconductors with a small bandgap have to be used. For highly energetic radiation, such as x-rays, the energy exceeds the bandgap of any semiconductor. However, the absorption coefficient of most mate-rials is extremely low at these wavelengths, which makes most sensors almost transparent to shortwave radiation. In order to deposit the en-ergy in the detector, the semiconductor material must contain heavy atoms, which have a higher absorptivity in the x-ray region.

Extrinsic photoconductors. For longer wavelengths toward the in-frared region, it is hard to find suitable intrinsic semiconductor mate-rials with sufficiently small bandgaps. For wavelengths beyond 15µm, materials tend to become unstable and difficulties occur in achieving high uniformity and making good electrical contacts. A solution to this problem is to useextrinsic photoconductors, that is, semiconductors doped with eitherp-type orn-type impurities.

The addition of impurities places available electron states in the pre-viously forbidden gap and allows conductivity to be induced by freeing impurity-based charge carriers. Thus, smaller energy increments are required. As illustrated in Fig.4.6b and c, only the gap between the va-lence band and the impurity level (p-type semiconductors) or the gap between the impurity level and the conduction band (n-type semicon-ductors) has to be overcome by absorption of a photon. In the former case, the conductivity is carried by holes and in the latter case free electrons in the conduction band contribute to the conductivity. The basic operation of extrinsic photoconductors is similar to that of in-trinsic photoconductors, except that the bandgap energyQghas to be replaced by the excitation energyQi (Fig.4.6b and c).

Although extrinsic photoconductors are an elegant way to get long wavelength response, they have some less desirable characteristics:

• Due to the smaller bandgap, extrinsic semiconductors are much more sensitive to thermal noise, which can be inferred from Eq. (4.32), and, therefore, require a much lower operating temperature than do intrinsic photoconductors.

4.5 Detecting radiance 101

+

+

-- hν

hν

n-type depletion p-type

Figure 4.7:Band diagram of thep-njunction in a photovoltaic detector (photo-diode). In the p-type material, photogenerated electrons diffuse into the deple-tion region and are swept into the n-type region by the electric field. The same process occurs in the n-type material, except the roles of the holes and electrons are reversed.

• Extrinsic photoconductors have a quantum efficiency that is sub-stantially smaller than that of intrinsic materials (30% compared to 60% in average). This results from the fact that the impurities are necessarily more sparse than the host material, which leads to a smaller optical absorption cross section.

• The electrical conductivity of extrinsic materials differs fundamen-tally from that of intrinsic materials. In intrinsic photoconductors, electron/hole pairs are generated by the excitation process, both contributing to the charge transport (Fig.4.6a). In extrinsic photo-conductors, individual charge carriers are generated whose comple-mentary charge resides in an ionized atom, which remains immobile in the crystal structure and cannot carry current (Fig.4.6a and b).

As the number of semiconductor atoms always outnumbers the im-purity atoms, the intrinsic effect dominates in both types of extrinsic material at high temperatures (where all impurity charge carriers are thermally excited) and for wavelengths smaller than the cutoff wave-length of the intrinsic material. To reduce the response from intrinsic conduction, all wavelengths below the anticipated long-wave radiation have to be blocked by spectral filters.

Photodiodes (photovoltaic detectors). A photovoltaic detector ac-tively generates a voltage or current from incident electromagnetic ra-diation. The most common realization is based on a junction between two oppositely doped zones (p-njunction) in a semiconductor mate-rial. As this setup acts as a diode, this type of detector is also called photodiode.

Photodiodes allow large resistance and simultaneously high pho-toconductive gain within a small volume to be obtained. The n-type material has a surplus (and thep-type material has a deficiency) of elec-trons compared to the crystal bond of the semiconductor material. In

the adjacent region of both oppositely doped zones, electrons migrate from the n- to the p-region acceptor atoms and holes migrate from thep- to then-region donors, if thermal excitation frees them. Within the contact region all bonds are complete and the material is depleted of potential charge carriers. This results in a high resistance of this re-gion, as opposed to the relatively high conductivity of thep- andn-type material. As the charge carriers diffuse, a voltage is established across the depletion region, called thecontact potential, which opposes the diffusion of additional electrons. The net result is a permanent equi-librium voltage across thep-njunction. The resulting bandstructure across the contact zone is shown in Fig.4.7.

When photons of energies greater than the forbidden gap energy are absorbed in, or close to ap-njunction of a photodiode, the resulting electron/hole pairs are pulled by the electric field of the contact poten-tial across thep-njunction. Electrons are swept from thep-region into then-region, and holes in the opposite direction (Fig.4.7). As the charge carriers are spatially separated across the detector, a resulting voltage can be measured. If then- and thep-type region are connected, a small current will flow between both regions. This phenomenon is called the photovoltaic effect.

Because photodiodes operate through intrinsic rather than extrin-sic absorption, they can achieve a high quantum efficiency in small vol-umes. Photodiodes can be constructed in large arrays of many thou-sands of pixels. They are the most commonly used detectors in 1-6-µm region [3] (e. g., InSb infrared focal plane arrays) and are also used in the visible and near ultraviolet.

Photoemissive detectors. Photoemissive detectors operate with ex-ternal photoelectric emission. The excited electron physically leaves the detector material and moves to the detecting anode. Figure4.8a il-lustrates the principal setup. A conduction electron is produced in the photocathode by absorption of a photon with an energy greater than the intrinsic bandgap of the detector material. This electron diffuses through the detector material until it reaches the surface. At the sur-face of the photocathode it might escape into the vacuum. Using an electric field between the photocathode and the anode helps to acceler-ate the electron into the vacuum, where it is driven towards the anode and counted as current. Suitable photocathode materials must have the following properties:

• high-absorption coefficient for photons

• long mean-free path for the electron in the cathode material (low transport losses of electrons migrating to the surface of the cathode)

• low electron affinity, that is, low barrier inhibiting the electron emis-sion

4.5 Detecting radiance 103

a

Vsupp

vacuum

photo-electron

photocathode anode

hν

-V_out

b

photomultiplier tube

-microchannel

-Figure 4.8:Photoemissive detectors. aDetection process for a vacuum photo-diode;blight amplification by a microchannel (top) and a photomultiplier tube (bottom).

The simple vacuum photodiode (Fig.4.8a) can be improved by elec-tron multipliers, increasing the number of elecelec-trons contributing to the output current for each detected photon. A commonly used photoemis-sive detector is thephotomultiplier, illustrated in Fig.4.8b. It consists of a vacuum tube including several intermediate anodes. Each anode, called adynode, is given a voltage higher than the previous one. The geometrical arrangement is such that emitted electrons are accelerated towards the next adjacent dynode. If the voltage difference is high enough, each photoelectron leaving a dynode gets fast enough to eject multiple electrons from the next dynode upon impact. This process is repeated until the avalanche of electrons finally reaches the anode.

The voltages required for operation are provided by a single supply, divided by a chain of resistors. The photocathode is held at a large negative voltage in the order of several thousand volts relative to the anode.

Photomultipliers are large devices, restricted mainly to single de-tectors. A different form of electron multipliers, which is of practi-cal relevance for computer vision, are made from thin tubes of lead-oxide glass. These microchannels have diameters of 8-45µm and a length-to-diameter ratio of about 40[3], and are suitable for integra-tion into small-scale detector arrays. Microchannel plates are arrays of approximately one million channel electron multipliers, fused into solid wafers [6]. Figure 4.8b illustrates the principal mechanism of a single microchannel. The microchannel wall consists of three layers:

an emitting layer; a conducting layer; and bulk glass. The conductive layer has a high resistance and allows a large voltage to be maintained across the ends of the tube. Electrons that enter the tube are

acceler-ated along the tube until they collide with the wall. The inner surface layer, called the emitting layer, is made from PbO, which acts as an elec-tron multiplier. Upon impact, the accelerated elecelec-trons create multiple secondary electrons that are accelerated by the voltage along the tube until they strike the walls again and produce more free electrons. This operation is comparable to a continuous dynode chain and the gains are nearly as large as those of photomultipliers.

Microchannel plates are used in modern light intensifying cameras, suitable for low-illumination applications, such as fluorescence imaging and night vision devices.