Solid-State Electronic Devices: Optoelectronic Devices [part 1]

Home | Articles | Forum | Glossary | Books

So far we have primarily concentrated on electronic devices. There is also a wide variety of very interesting and useful device functions involving the interaction of photons with semiconductors. These devices provide the optical sources and detectors that allow broadband telecommunications and data transmission over optical fibers. This important area of device applications is called optoelectronics. In this section we will discuss devices that detect photons and those that emit photons. Devices that convert optical energy into electrical energy include photodiodes and solar cells. Emitters of pho tons include incoherent sources such as light-emitting diodes (LEDs) and coherent sources in the form of lasers.


1. Understand solar cells

2. Study photodetectors such as APDs

3. Study incoherent light sources (LEDs) and coherent light sources (lasers)

Earlier, we saw that bulk semiconductor samples can be used as photoconductors by providing a change in conductivity proportional to an optical generation rate. Often, junction devices can be used to improve the speed of response and the sensitivity of detectors of optical or high-energy radiation. Two-terminal devices designed to respond to photon absorption are called photodiodes. Some photodiodes have extremely high sensitivity and response speed. Since modern electronics often involves optical as well as electrical signals, photodiodes serve important functions as electronic devices. In this section, we shall investigate the response of p-n junctions to optical generation of electron-hole pairs (EHPs) and discuss a few typical photodiode detector structures. We shall also consider the very important use of junctions as solar cells, which convert absorbed optical energy into useful electrical power.

1. Photodiodes

FIG. 1 Optical generation of carriers in a p-n junction: (a) absorption of light by the device; (b) current Iop resulting from EHP generation within a diffusion length of the junction on the n side; (c) I-V characteristics of an illuminated junction.

1.1 Current and voltage in an illuminated junction

In section 5 we identified the current due to drift of minority carriers across a junction as a generation current. In particular, carriers generated within the depletion region W are separated by the junction field, electrons being collected in the n region and holes in the p region. Also, minority carriers generated thermally within a diffusion length of each side of the junction diffuse to the depletion region and are swept to the other side by the electric field. If the junction is uniformly illuminated by photons with hv / Eg, an added generation rate gop (EHP/cm^3 - s) participates in this current (FIG. 1). The number of holes created per second within a diffusion length of the transition region on the n side is ALpgop. Similarly ALngop electrons are generated per second within Ln of xp0 and AWgop carriers are generated within W. The resulting current due to the collection of these optically gener ated carriers by the junction is

Iop = qAgop(Lp + Ln + W) (eqn. 1)

If we call the thermally generated current described in Eq. ( 5-37b)

I th,we can add the optical generation of Eq. (eqn. 1) to find the total reverse current with illumination. Since this current is directed from n to p, the diode equation

[Eq. ( 5-36)]


(eqn. 2)

Thus the I-V curve is lowered by an amount proportional to the genera tion rate (FIG. 1c). This equation can be considered in two parts--the current described by the usual diode equation and the current due to optical generation.

When the device is short circuited (V = 0), the terms from the diode equation cancel in Eq. (eqn. 2), as expected. However, there is a short-circuit current from n to p equal to Iop. Thus the I-V characteristics of FIG. 1c cross the I-axis at negative values proportional to gop. When there is an open circuit across the device, I = 0 and the voltage V = Voc is

(eqn. 3a)

For the special case of a symmetrical junction, pn = np and tp = tn,we can rewrite Eq. (eqn. 3a) in terms of the thermal generation rate pn/tn = gth and the optical generation rate gop. Neglecting generation within W:

(eqn. 3b)

Actually, the term gth = pn/tn represents the equilibrium thermal generation-recombination rate. As the minority carrier concentration is increased by optical generation of EHPs, the lifetime tn becomes shorter, and pn/tn becomes larger (pn is fixed, for a given Nd and T). Therefore, Voc cannot increase indefinitely with increased generation rate; in fact, the limit on Voc is the equilibrium contact potential V0 (FIG. 2). This result is to be expected, since the contact potential is the maximum forward bias that can appear across a junction. The appearance of a forward voltage across an illuminated junction is known as the photovoltaic effect.

Depending on the intended application, the photodiode of FIG. 1 can be operated in either the third or fourth quarters of its I-V characteristic. As FIG. 3 illustrates, power is delivered to the device from the external circuit when the current and junction voltage are both positive or both negative (first or third quadrants). In the fourth quadrant, however, the junction volt age is positive and the current is negative. In this case power is delivered from the junction to the external circuit (notice that in the fourth quadrant the current flows from the negative side of V to the positive side, as in a battery).

FIG. 2 Effects of illumination on the open-circuit voltage of a junction: (a) junction at equilibrium; (b) appearance of a voltage Voc with illumination.

FIG. 3 Operation of an illuminated junction in the various quadrants of its I-V characteristic; in (a) and (b), power is delivered to the device by the external circuit; in (c) the device delivers power to the load.

If power is to be extracted from the device, the fourth quadrant is used; on the other hand, in applications as a photodetector we usually reverse bias the junction and operate it in the third quadrant. We shall investigate these applications more closely in the discussion to follow.

1.2 Solar cells

FIG. 4 (a) Solar cell arrays attached to the International Space Station. The solar array wings measure 74 m tip to tip. Each wing contains 32,800 solar cells and can produce 62 kW of power for the station. (Courtesy of NASA.) (b) 72MW terrestrial solar plant in Italy. (MEMC/Sun Edison.)

FIG. 5 Configuration of a solar cell: (a) enlarged view of the planar junction; (b) top view, showing metal contact "fingers."

Since power can be delivered to an external circuit by an illuminated junction, it is possible to convert solar energy into electrical energy. If we consider the fourth quadrant of FIG. 3c, it appears doubtful that much power can be delivered by an individual device. The voltage is restricted to values less than the contact potential, which in turn is generally less than the band gap voltage Eg /q. For Si the voltage Voc is less than about 1 V. The current generated depends on the illuminated area, but typically Iop is in the range 10-100 mA for a junction with an area of about 1 cm^2. However, if many such devices are used, the resulting power can be significant. In fact, arrays of p-n junction solar cells are currently used to supply electrical power for many space satellites. Solar cells can supply power for the electronic equipment aboard a satellite over a long period, which is a distinct advantage over batteries. The array of junctions can be distributed over the surface of the satellite or can be contained in solar cell "paddles" attached to the main body of the satellite (FIG. 4a). A 72MW solar cell plant in Italy is shown in FIG. 4b.

To utilize a maximum amount of available optical energy, it is necessary to design a solar cell with a large area junction located near the surface of the device (FIG. 5). The planar junction is formed by diffusion or ion implantation, and the surface is coated with appropriate materials to reduce reflection and to decrease surface recombination. Many compromises must be made in solar cell design. In the device shown in FIG. 5, for example, the junction depth d must be less than Lp in the n material to allow holes generated near the surface to diffuse to the junction before they recombine; similarly, the thickness of the p region must be such that electrons generated in this region can diffuse to the junction before recombination takes place. This requirement implies a proper match between the electron diffusion length Ln, the thickness of the p region, and the mean optical penetration depth 1/a. It is desirable to have a large contact potential V0 to obtain a large photovoltage, and therefore heavy doping is indicated; on the other hand, long lifetimes are desirable and these are reduced by doping too heavily. It is important that the series resistance of the device be very small so that power is not lost to heat due to ohmic losses in the device itself. A series resistance of only a few ohms can seriously reduce the output power of a solar cell (Exc. 7). Since the area is large, the resistance of the p-type body of the device can be made small. However, contacts to the thin n region require special design. If this region is contacted at the edge, current must flow along the thin n region to the contact, resulting in a large series resistance. To prevent this effect, the contact can be distributed over the n surface by providing small contact fingers as in FIG. 5b. These narrow contacts serve to reduce the series resistance without interfering appreciably with the incoming light.

FIG. 6 I-V characteristics of an illuminated solar cell. The maximum power rectangle is shaded.

FIG. 6 shows the fourth-quadrant portion of a solar cell characteristic, with I r plotted upward for the convenience of illustration. The open-circuit voltage Voc and the short-circuit current Isc are determined for a given light level by the cell properties. The maximum power delivered to a load by this solar cell occurs when the product VI r is a maximum. Calling these values of voltage and current Vm and Im, we can see that the maximum delivered power illustrated by the shaded rectangle in FIG. 6 is less than the IscVoc product. The ratio ImVm/IscVoc is called the fill factor, and is a figure of merit for solar cell design.

Applications of solar cells are not restricted to outer space. It is possible to obtain useful power from the sun in terrestrial applications using solar cells, even though the solar intensity is reduced by the atmosphere. Current worldwide total power generation capability is ~15 TW and corresponds to an annual energy usage of ~500 quads (1 quad = 10^15 , or a quadrillion, BTU), with an annual increase of 1-2,. Of this, ~80, comes from fossil fuels (oil, natural gas, and coal), which will be exhausted in several hundred years. In addition, CO2 emissions and global warming have spurred a renewed interest in "green" sources of energy such as photovoltaics, of which the current installed capacity worldwide is about 100 GW. About 1 kW/m^2 is available in a particularly sunny location (translating to about 600 TW potentially available worldwide), but not all of this solar power can be converted to electricity. Much of the photon flux is at energies less than the cell band gap and is not absorbed. High-energy photons are strongly absorbed, and the resulting EHPs may recombine at the surface. A well-made single-crystal Si cell can have about 25, efficiency for solar energy conversion, providing ~250 W/m^2 of electrical power under full illumination. This is a modest amount of power per unit solar cell area, considering the cost and effort involved in fabricating a large area of Si cells. Amorphous Si thin-film solar cells can be made more cheaply, but have lower efficiencies (~10,) because of the defects in the material.

The cost and scalability of photovoltaic technology are of paramount importance for terrestrial applications. Currently, it costs only about 3 cents per kWh for electricity generation from fossil fuels, but about 10 times that amount from amorphous Si solar cells, and the time to recover the investment in photovoltaics is about four years. In terms of scalability, at 10, cell efficiency, approximately 3, of the land area would have to be covered with solar cells to meet US energy needs, which would, of course, create other environmental problems. One approach to obtaining more power per cell is to focus considerable light onto the cell with the use of mirrors. Although Si cells lose efficiency at the resulting high temperatures, GaAs and related compounds can be used at 100°C or higher. In such solar concentrator systems, more effort and expense can be put into the solar cell fabrication, since fewer cells are required. For example, a GaAs-AlGaAs heterojunction cell provides good conversion efficiency and operates at the elevated temperatures that are common in solar concentrator systems. In such systems, the area required is shifted from the cell to the concentrator.

1.3 Photodetectors

When the photodiode is operated in the third quadrant of its I-V characteristic (FIG. 3b), the current is essentially independent of voltage but is proportional to the optical generation rate. Such a device provides a useful means of measuring illumination levels or of converting time-varying optical signals into electrical signals.

FIG. 7 Schematic representation of a p-i-n photodiode.

In most optical detection applications the detector's speed of response, or bandwidth, is critical. For example, if the photodiode is to respond to a series of light pulses 1 ns apart, the photogenerated minority carriers must diffuse to the junction and be swept across to the other side in a time much less than 1 ns. The carrier diffusion step in this process is time consuming and should be eliminated if possible. Therefore, it is desirable that the width of the depletion region W be large enough so that most of the photons are absorbed within W rather than in the neutral p and n regions. When an EHP is created in the depletion region, the electric field sweeps the electron to the n side and the hole to the p side. Since this carrier drift occurs in a very short time, the response of the photodiode can be quite fast. When the carriers are generated primarily within the depletion layer W, the detector is called a depletion layer photodiode. Obviously, it is desirable to dope at least one side of the junction lightly so that W can be made large. The appropriate width for W is chosen as a compromise between sensitivity and speed of response. If W is wide, most of the incident photons will be absorbed in the depletion region, leading to a high sensitivity. Also, a wide W results in a small junction capacitance, thereby reducing the RC time constant of the detector circuit. On the other hand, W must not be so wide that the time required for drift of photogenerated carriers out of the depletion region is excessive, leading to low bandwidth.

One convenient method of controlling the width of the depletion region is to build a p-i-n photodetector (FIG. 7). The "i" region need not be truly intrinsic, as long as the resistivity is high. It can be grown epitaxially on the n-type substrate, and the p region can be obtained by implantation. When this device is reverse biased, the applied voltage appears almost entirely across the i region. If the carrier lifetime within the i region is long compared with the drift time, most of the photogenerated carriers will be collected by the n and p regions. An important figure of merit for a photodetector is the external quantum efficiency hQ, defined as the number of carriers that are collected for every photon impinging on the detector. For a photocurrent density Jop, we collect Jop/q carriers per unit area per second. For an incident optical power density of Pop, the number of photons shining on the detector per unit area per second is Pop/hv. Therefore,

ηQ = (Jop/q)/(Pop/hv) (eqn. 4)

For a photodiode that has no current gain, the maximum hQ is unity. If low-level optical signals are to be detected, it is often desirable to operate the photodiode in the avalanche region of its characteristic. In this mode, each photogenerated carrier results in a significant change in the current because of avalanche multiplication, leading to gain and external quantum efficiencies of greater than 100,. Avalanche photodiodes (APDs) are useful as detectors in fiber-optic systems.

The type of photodiode described here is sensitive to photons with energies near the band gap energy (intrinsic detectors). If h n is less than Eg, the photons will not be absorbed; on the other hand, if the photons are much more energetic than Eg, they will be absorbed very near the surface, where the recombination rate is high. Therefore, it is necessary to choose a photodiode material with a band gap corresponding to a particular region of the spectrum. Detectors sensitive to longer wavelengths can be designed such that photons can excite electrons into or out of impurity levels (extrinsic detectors). However, the sensitivity of such extrinsic detectors is much less than intrinsic detectors, where EHPs are generated by excitation across the band gap.

By using lattice-matched multilayers of compound semiconductors, the band gap of the absorbing region can be tailored to match the wavelength of light being detected. Wider band gap material can then be used as a window through which the light is transmitted to the absorbing region (FIG. 8). For example, we saw that InGaAs with an In mole fraction of 53, can be grown epitaxially on InP with excellent lattice-matching. This composition of InGaAs has a band gap of about 0.75 eV, which is sensitive to a useful wavelength for fiber-optic systems (1.55 mm), as we shall see in Section 2.2. In making a photodiode using InGaAs as the active material, it is possible to bring the light through the wider band gap In0.52Al0.48As (also lattice-matched to InP), thus greatly reducing surface recombination effects. In the case of APDs requiring narrow-band-gap material, it is often advantageous to absorb the light in the narrow-gap semiconductor (e.g., InGaAs) and transport the resulting carriers to a junction made in wider band gap material (e.g., InAlAs), where the avalanche multiplication takes place at high fields. Such a separation of the absorption and multiplication APDs avoids the excessive leakage currents typical of reverse-biased junctions in narrow-gap materials. In this particular structure, there is also a doped InAlAs charge layer, leading to a SACM APD, which helps optimize (decrease) the electric field further between the multiplication and absorption regions. In certain APDs, one can grade the alloy composition between the high-band-gap multiplication regions and the lower-band-gap absorption region to avoid any band-edge discontinuities, which can trap photogenerated carriers. The photocurrent and dark current both increase with bias because of avalanche multiplication (FIG. 8b). One obviously wants to maximize the difference delta_I between the photocurrent Ip and the dark current Id. The ratio of delta_I at different voltages to that at a low reference voltage is defined as the gain of the APD.

FIG. 8 Use of multilayer hetero-junctions to enhance the photodiode operation: (a) an APD in which light near 1.55 mm is absorbed in a narrow-band-gap material (InGaAs, Eg = 0.75 eV) after passing through a wider-gap material (InP and InAlAs); electrons are swept to an InAlAs junction, where avalanche multiplication takes place. The i regions are lightly doped; (b) photocurrent, dark current, and gain increasing as a function of bias because of avalanche multiplication; (c) typical gain-bandwidth characteristics of such a SACM APD. [After X. Zheng, J. Hsu, J. Hurst, X. Li, S. Wang, X. Sun, A. Holmes, J. Campbell, A. Huntington, and L. Coldren, IEEE J. Quant. Elec., 40(8), pp. 1068-1073, Aug. 2004.]

1.4 Gain, bandwidth, and Signal-to-Noise Ratio of Photodetectors

In optical communication systems, the sensitivity (which depends on gain) of the photodetector and its response time (bandwidth) are of critical importance. Typical gain-bandwidth characteristics of such a SACM APD are shown in FIG. 8c and are limited by the transit time of carriers through the structure. Unfortunately, designs that increase gain tend to decrease bandwidth and vice versa. It is common to express the gain-bandwidth product as a figure of merit for detectors. In a p-i-n diode, there is no gain mechanism, since at most one EHP is collected by the junction for each photon absorbed. Thus, the gain is essentially unity, and the gain-bandwidth product is determined by the bandwidth, or frequency response. In a p-i-n, the response time is dependent on the width of the depletion region.

Another important property of detectors is the signal-to-noise (S/N) ratio, which is the amount of usable information compared with the background noise in the detector. In the case of photoconductors, the major source of noise is random thermal motion of the carriers, leading to fluctuations in the dark current (called Johnson noise). The noise current increases with temperature (~kT) and with the conductance of the material in the dark. Therefore, the photoconductor noise at a given temperature can be reduced by increasing the dark resistance. Yet another source of noise at low frequencies is 1/f or flicker noise, due to carrier trapping and de-trapping at defects.

In a p-i-n diode, the dark current is smaller and the dark resistance much higher than in a photoconductor, and the main source of noise is random thermal generation-recombination of EHPs within this region (called shot noise). The shot noise is ultimately due to the quantization of the charge of electrons and holes. The noise in a p-i-n device is considerably lower than that in a photoconductor, as well as in APDs, for reasons described next.

Avalanche photodiodes have the advantage of providing gain through the avalanche multiplication effect. The disadvantage is increased noise relative to the p-i-n, due to random fluctuations in the avalanche process. This noise is reduced if the impact ionization in the high-field region is due to only one type of carrier, since more fluctuations in the ionization process occur when both electrons and holes participate. In Si, the ability of electrons to create EHPs in an impact ionization event is much higher than for holes. Therefore, Si APDs can be operated with high gain and relatively low noise. Unfortunately, Si APDs cannot be used for most fiber-optic trans mission because Si is transparent at the wavelengths of low loss and low dispersion (l = 1.55 and 1.3 mm) for optical fibers. For these longer wave lengths, In0.53Ga0.47As has become the material of choice. However, the ionization rates of electrons and holes in most compound semiconductors are comparable, which degrades their noise and frequency response relative to Si APDs.

The various sources of noise determine the S/N ratio in a photodetector. One quantifies it as a noise-equivalent power (NEP), which is the minimum detectable signal that would produce the same rms output as the noise. The detectivity of the photodetector is then defined as D = 1/ NEP.

The NEP depends on the area of the photodetector as well as the band width. The specific detectivity, D*, is then defined as that for a detector of unit area and a bandwidth of 1 Hz. Clearly, once the bandwidth requirements are met, it is desirable to choose a photodetector with the highest D*.

Another approach that has demonstrated excellent performance in terms of both high sensitivity and bandwidth utilizes a waveguide structure (FIG. 9). Here, unlike the situation with the p-i-n photodiode (FIG. 7) or the APD (FIG. 8), the light strikes the photodiode perpendicular to the current transport. The advantage is that now the absorption region can be made quite long, along the photon path, leading to high sensitivity, while at the same time the photogenerated carriers have to traverse a short distance in the perpendicular direction, leading to short transit times and high bandwidth.

FIG. 9 Schematic of a waveguide photodiode. The photons are strongly absorbed in the narrow-band-gap InGaAs region A, and carriers are multiplied by the avalanche process in region M. The charge region C helps optimize the electric field profile between A and M.

2. Light-Emitting Diodes

When carriers are injected across a forward-biased junction, the current is usually accounted for by recombination in the transition region and in the neutral regions near the junction. In a semiconductor with an indirect band gap, such as Si or Ge, the recombination releases heat to the lattice. On the other hand, in a material characterized by direct recombination, considerable light may be given off from the junction under forward bias. This effect, called injection electroluminescence, provides an important application of diodes as generators of light. The use of LEDs in digital displays is well known. There are also other important applications, in traffic and automotive signals and in illumination. Another important device making use of radiative recombination in a forward-biased p-n junction is the semiconductor laser. As we shall see in Section 4, lasers emit coherent light in much narrower wave length bands than LEDs, with more collimation (directionality), and are very useful for fiber-optic communication systems, as described in Section 2.2.

For LEDs, the frequency (color) of the photon is governed by the band gap of the semiconductor as given by the Planck relation, hv = Eg, which, in appropriate units, can be expressed as Eg (eV) = 1.24 / lambda (um). A very important metric of an LED is the external quantum efficiency eta_ext, which is defined as the light output divided by the electrical input power:

eta_ext = (Internal radiative efficiency) * (Extraction efficiency) (eqn. 5)

The internal efficiency is a function of the quality of the material and the structure and composition of the layer. Defects in the material will clearly lead to nonradiative recombination. However, even if the internal efficiency is high, not all emitted photons are extracted from the LED. The emitted photons from an LED have a wide angular distribution, unlike those in a laser. For example, if an LED had a planar surface, the photons incident on the semiconductor-air interface at angles greater than a critical angle would undergo total internal reflection and ultimately be lost via absorption within the semiconductor. Hence, typically, LEDs are made with a dome-type encapsulation, which acts as a lens so that more of the photons can be extracted.

2.1 Light-emitting materials

The band gaps of various binary compound semiconductors are illustrated relative to the spectrum. There is a wide variation in band gaps and, therefore, in available photon energies, extending from the ultraviolet (GaN, 3.4 eV) into the infrared (InSb, 0.18 eV). In fact, by utilizing ternary and quaternary compounds the number of available energies can be increased significantly (see Figs. 1-13 and 3-6). A good example of the variation in photon energy obtainable from the compound semiconductors is the ternary alloy gallium arsenide-phosphide, which is illustrated in FIG. 10. When the percentage of As is reduced and P is increased in this material, the resulting band gap varies from the direct 1. 43-eV gap of GaAs (infrared) to the indirect 2. 26-eV gap of GaP (green). The band gap of GaAs1-xPx varies almost linearly with x until the 0.45 composition is reached, and electron-hole recombination is direct over this range. The most common alloy composition used in red LED displays is x ? 0.4. For this composition the band gap is direct, since the G minimum (at k = 0) is the lowest part of the conduction band. This results in efficient radiative recombination, and the emitted photons (~1.9 eV) are in the red portion of the spectrum.

For GaAs1-xPx with P concentrations above 45 percent, the band gap is due to the indirect X minimum. Radiative recombination in such indirect materials is generally unlikely, because electrons in the conduction band have different momentum from holes in the valence band (see Fig. 3-5). Interestingly, however, indirect GaAs1-xPx (including GaP, x = 1) doped with nitrogen can be used in LEDs with light output in the yellow to green portions of the spectrum. This is possible because the nitrogen impurity binds an electron very tightly. This confinement in real space (?x) means that the electron momentum is spread out in momentum space ?p by the Heisenberg uncertainty principle. As a result, the momentum conservation rules, which generally prevent radiative recombination in indirect materials, are circumvented. Thus nitrogen doping of GaAs1-xPx is not only useful technologically, but also provides an interesting and practical illustration of the uncertainty principle.

FIG. 10 Conduction band energies as a function of alloy composition for GaAs1-xPx.

In many applications light from a laser or an LED need not be visible to the eye. Infrared emitters such as GaAs, InP, and mixed alloys of these compounds are particularly well suited to fiber-optical communication systems or TV remote controls. For example, a laser or LED can be used in conjunction with a photodiode or other photosensitive device to transmit information optically between locations. By varying the current through the diode, the light output can be modulated such that analog or digital information appears in the optical signal directed at the detector. Alternatively, the information may be introduced between the source and detector. For example, a semiconductor laser-photodetector arrangement can be used in a compact disc or DVD system for reading digital information from the spinning disc. A light emitter and a photodiode form an optoelectronic pair, which provides complete electrical isolation between input and output, since the only link between the two devices is optical. In an optoelectronic isolator, both devices may be mounted on a ceramic substrate and packaged together to form a unit that passes information while maintaining isolation.

FIG. 11 Improvement of luminous intensity of LEDs over time.

In view of the broad range of applications requiring semiconductor lasers and LEDs with visible and infrared wavelengths, the wide variety of available III-V materials is extremely useful. In addition to the AlGaAs and GaAsP systems shown in Fig. 3-6 and FIG. 10, the InAlGaP system is useful for red, yellow, and orange wavelengths, and AlGaInN is a strong emitter in the blue and green. Why is there so much interest in short wave length emitters such as blue-green LEDs? As shown in FIG. 11, moder ate efficiency (~10 lumens/watt) red, green, and yellow-green LEDs have existed for a long time in the GaAsP system, based on concepts such as N isoelectronic doping. More recently, in the mid-1990s, much higher brightness and efficiency (~30 lumens/watt) red-orange-yellow LEDs based on the InAlGaP system have been developed. The higher-band-gap InAlGaN system has direct band gaps over the entire alloy composition range and hence offers highly efficient light emission in the blue and green part of the spectrum. A major goal of the optoelectronics community has been the achievement of high-efficiency red, green, and blue emitters, because those colors are the three primary colors of the spectrum. By combining these color LEDs (or in conjunction with suitable phosphors), one can form intense white light sources (~500 lumens) with luminous efficiencies ~2 times greater than those of conventional incandescent light bulbs. Since general-purpose illumination typically requires kilolumens of light, it may be practical to combine a few of these high-brightness LEDs for that purpose. They also have much longer lifetimes (5, 000 -100,000 hours versus 2,000 hours for conventional lightbulbs) and much higher energy efficiencies. The costs of these LEDs have been reduced considerably in recent years, making them more competitive. Continued application of these LEDs in lighting can lead to a significant reduction in global energy demand. Arrays of red, green, and blue emitters are being used in outdoor displays and TV screens, as well as in automotive taillights and signal lights. Red, yellow, and green LEDs are being used for traffic lights because they have a much higher reliability and lifetime than conventional lightbulbs, and they save energy.

For many years an efficient blue emitter was elusive, because materials with wide band gap also tend to require crystal growth at high temperatures. Therefore, they are difficult to grow on convenient substrates. An excellent candidate in terms of band gap is GaN, which has Eg corresponding to the ultraviolet. Unfortunately, deposition of GaN by MOCVD or other methods requires a substrate that can withstand high temperatures. A convenient substrate is sapphire, which has a melting point above 2000 C.

However, there is considerable lattice mismatch between sapphire and GaN, and epitaxial GaN films suffer from dislocations that thread from the sapphire-GaN interface through the film. These threading dislocations can serve as non-radiative EHP recombination sites where carriers recombine with their energy given up as heat rather than light. This is not good news for an LED material, and for many years GaN films were inefficient light emit ters. As sometimes happens (too bad it's not more often), nature provides a way around a problem that seems insurmountable. In this case it turns out that addition of In during the growth of the GaN film provides a way to avoid catastrophic loss of carriers at dislocations. Since In is not completely miscible in GaN, In-rich clusters form in InGaN which have a smaller band gap than the rest of the film. Therefore, carriers diffusing through the InGaN can fall into the potential well of the lower band gap cluster and recombine radiatively. If the density of such low-bandgap clusters significantly outnumbers the density of threading dislocations, many of the electron-hole pairs can survive to recombine giving off light. This discovery revolutionized the search for wide-bandgap materials to allow light emission in the blue and violet spectra. White light can be obtained by combining red, green, and blue emitters or by coating a blue or violet emitter with phosphors that can reemit various wavelengths to achieve white light. In recent years these white LED have been engineered to be extremely efficient, allowing the application of such high brightness LEDs in lighting applications.

One problem that is faced with LEDs is that many of the photons that are generated cannot be extracted from the device if it is a planar structure. The reason is that the semiconductor has a higher refractive index than the surrounding (air) medium, and for a light ray impinging on the surface at greater than the critical angle, it undergoes total internal reflection, and is re-absorbed. Hence, it is important to texture the surface of the LEDs to create micro-lenses such that a larger fraction of the light rays can be extracted. There are also important packaging challenges to dissipate heat

from the devices by creating proper heat sinks. Therefore, considerable device engineering has been applied to the design of these devices in recent years, and the results have been impressive in terms of lumens out for a given power into the LED. Finally, after many years of work by crystal growers, device designers and packaging engineers, LEDs have become widespread in lighting as well as in displays.

2.2 Fiber-optic communications

The transmission of optical signals from source to detector can be greatly enhanced if an optical fiber is placed between the light source and the detector. An optical fiber is essentially a "light pipe" or waveguide for optical frequencies. The fiber is typically drawn from a boule of glass to a diameter of ~25 um. The fine glass fiber is relatively flexible and can be used to guide optical signals over distances of kilometers without the necessity of perfect alignment between source and detector. This significantly increases the applications of optical communication in areas such as telephone and data transmission.

One type of optical fiber has an outer layer of very pure fused silica (SiO2), with a core of germanium doped glass having a higher index of refraction (FIG. 12a).

Such a step-index fiber maintains the light beam primarily in the central core with little loss at the surface. The light is transmitted along the length of the fiber by internal reflection at the step in the refractive index.

FIG. 12 Two examples of multimode fibers: (a) a step index having a core with a slightly larger refractive index n; (b) a graded index having, in this case, a parabolic grading of n in the core. The figure illustrates the cross section (left) of the fiber, its index-of-refraction profile (center), and typical mode patterns (right). 1The index of refraction (or refractive index) n compares the velocity of light v in the material with its velocity c in a vacuum: n = c/v. Thus, if n1 > n2 in FIG. 12a, the velocity of light is greater in material 2 than in 1. The value of n varies somewhat with the wavelength of light.

Losses in the fiber at a given wavelength can be described by an attenuation coefficient A [similar to the absorption coefficient of Eq. ( 4-3)]. The intensity of the signal at a distance x along the fiber is then related to the starting intensity by the usual expression.

(eqn. 6)

The attenuation is not the same for all wavelengths, however, and it is therefore important to choose a signal wavelength carefully. A plot of A vs. l for a typical silica glass fiber is shown in FIG. 13. It is clear that dips in A near 1.3 and 1.55 mm provide "windows" in the attenuation, which can be exploited to reduce the degradation of signals. The overall decrease in attenuation with increasing wavelength is due to the reduced scattering from small random inhomogeneities which result in fluctuations of the refractive index on a scale comparable to the wavelength. This type of attenuation, called Rayleigh scattering, decreases with the fourth power of wavelength. This effect is observed at sunrise and sunset, when attenuation of short wavelength blue and green light results in red and orange sunlight. Obviously, Rayleigh scattering encourages operation at long wave lengths in fiber-optic systems. However, a competing process of infrared absorption dominates for wavelengths longer than about 1.7 um, due to vibrational excitation of the atoms making up the glass. Therefore, a useful minimum in absorption for silica fibers occurs at about 1.55 um, where epitaxial layers in the (In, Ga) (As, P) system can be grown lattice-matched to InP substrates .

FIG. 13 Typical plot of attenuation coefficient A vs. wavelength l for a fused silica optical fiber. Peaks are due primarily to OH-impurities.

Another consideration in choice of operating wavelength is the pulse dispersion, or spreading of data pulses as they propagate down the fiber. This effect can be caused by the wavelength dependence of the refractive index, causing different optical frequencies to travel down the fiber with slightly different velocities. This effect, called chromatic dispersion, is much less pronounced at the 1.3 um window in FIG. 13. Another cause of dispersion is the fact that different modes propagate with different path lengths (FIG. 12a). This type of dispersion can be reduced by grading the refractive index of the core (FIG. 12b) such that various modes are continually refocused, reducing the differences in path lengths.

The light source in a fiber-optic system today is a laser, because the light is made up of essentially a single frequency and allows a very large information bandwidth. (Semiconductor lasers suitable for fiber-optic communications will be discussed in Section 4.) In early optoelectronic systems for fiber optics, it was most convenient to use the well-established GaAs-AlGaAs system for making lasers and LEDs. These light sources are highly efficient, and good detectors can be made using Si p-i-n or APDs. However, these sources operate in the wavelength range near 0.9 um, where the attenuation is greater than that for longer wavelengths. Modern systems, there fore, operate near the 1. 3-and 1.55 om minima shown in FIG. 13. At these wavelengths, lasers can be made by using InGaAs or InGaAsP grown on InP, and detectors can be made of the same materials (see FIG. 8).

Multimode fibers are larger (~25 um in diameter) than single-mode fibers (~5 um) and transmit a coherent laser beam. By forming numerous optical fibers into a bundle, with an appropriate jacket for mechanical strength, an enormous amount of information can be transmitted over long distances.

Depending upon the losses in the fibers, repeater stations may be required periodically along the path. Thus, many photodetectors and laser sources are required in a fiber-optic system. Semiconductor device development, including appropriate binary, ternary, and quaternary compounds for both emitters and detectors, is therefore crucial to the successful implementation of such optical communications systems. A schematic of a fiber-optic transmission system is shown in FIG. 14.

[Transmission rates of 40 Gbit/s are becoming standard, but even higher rates of 400 Gbit/s have been achieved using ultradense-wavelength-division multiplexing (UDWDM). This approach uses slightly different wavelengths or colors to carry different channels of information along the same fiber. As a convenient calibration of the 40 G bit/s rate, it is worth noting that the human eye is able to transmit about 1 G bit/s to the brain.]

FIG. 14 Schematic of a fiber-optic communication system illustrating the transmission of analog signals, such as those in telephony or TV. After the signal is converted to a digital electrical signal, it is used to modulate the laser light output as light pulses that are transmitted down the fiber, with periodic amplification of the signal with repeaters to compensate for fiber loss. Switching circuits route signals appropriately. After the optical signal is transformed to an electrical output by a photodetector and a low-noise preamplifier (LNA), it is converted to an analog signal, once distortions of the digital signals due to propagation through the fiber have been corrected with the use of a "regenerator." NEXT>>

Top of Page

PREV.   NEXT   Guide Index HOME