William J. Wadsworth 1994
The diode laser is rapidly becoming an indispensable device in modern life. It is present in the domestic CD player, in optical communications, laser printers, laser pointers, and of course in the laboratory. The attraction of diode lasers over all other types of laser is their extreme compactness, and ease of operation. The diodes themselves have sub-millimetre dimensions, and their high efficiency means that they have small power supplies. For example a 10 mW diode laser might require 100 mW of electrical power, easily provided by a battery, whereas a gas laser such as a He-Ne laser would require 10 W, and thus must be mains powered. At the high power end the differences are even more stark. A 40 W copper vapour laser, one of the most efficient gas lasers, needs 5 kW of electrical power. This means using a three phase supply. By contrast a 40 W diode array might require 200 W of electrical power, less than an average computer supply. Thus the high efficiency not only reduces the energy costs of running the lasers, but also dramatically reduces the size, cost, and complexity of the associated power supplies.
Diode lasers are also very user friendly, being readily packaged as devices with one button control, and they may have lifetimes of thousands of hours in low stress applications. At present the uses of laser diodes are limited by their low maximum powers, their susceptibility to optical and electrical damage, and the limited range of wavelengths available. Improvements are continually being made on all three points.
Towards the middle of this century, stimulated emission was first used to create coherent radiation in the microwave region of the spectrum. This Microwave Amplification by the Stimulated Emission of Radiation ( MASER ) was soon seen to be a very important development as a source of electromagnetic radiation. It was not for some years, however, that optical MASERs, later called LASERs, were theoretically predicted[1] then experimentally realised. The first laser, built in 1960 by Maiman[2] at Hughes aircraft company was a ruby laser, employing coiled flashlamps for the pump energy source. Diode lasers followed in November 1962, but it was not until the early seventies that robust commercial devices were possible, and a further ten years elapsed before the diode laser finally showed its potential in mass markets. This slow development followed by blossoming into a mass use device, contrasts sharply with the development of gas and solid state lasers. These relatively rapidly found their way into research laboratories as a wide range of experiments were designed which took advantage of the unique properties of LASER radiation. However even after thirty years of development these lasers are still primarily a tool for research, though other applications, for example in medicine and machining, are growing. The unique advantages of a diode laser, small size and high efficiency, have opened up the mass market in domestic goods, and spurred researchers on through the many years of development.
In making a laser the first requirement is that the medium must emit light. So before any attempt to build a diode laser, the Light Emitting Diode (LED) should be demonstrated. It is necessary first to examine the mechanisms by which a semiconductor diode junction may emit light.
The simple picture of the action of a p-n junction semiconductor diode is shown in Fig 3.1.
Figure 3.1 The p-n junction
In equilibrium, the charge carriers, excess electrons and holes, are confined to the n- and p-type regions respectively by a potential barrier across the junction. When the junction is forward biased by placing an external potential across it, this barrier is reduced and charge carriers are able to diffuse across the junction. There are then both holes and excess electrons in the junction region and recombination can occur. This allows more charge carriers to move into the junction region to replace those lost in recombination, and an electrical current flows. Recombination is clearly a process in which energy is released as the excess electrons fall down across the band gap to fill the holes. It might be expected that this energy would be emitted as a photon, giving a light emitting diode. This process does occur in some semiconductors such as gallium arsenide (GaAs), but in others such as silicon, the energy is lost as heat.
The reason for the very different behaviour of silicon and GaAs is that the recombination must conserve both energy and momentum. Fig 3.2 shows schematically the energy-momentum relations for electrons and holes in silicon and gallium arsenide (momentum is expressed as a wavenumber)
Figure 3.2 Indirect and Direct gap semiconductors
In equilibrium, the charge carriers occupy their lowest energy states, electrons at the bottom of the conduction band, and holes at the top of the valence band. In silicon these states do not have the same momentum. Therefore if a recombination is to result in the emission of a photon, which has little momentum, a quantum of lattice vibration (a phonon) must also be created to carry away the excess momentum. This is known as an indirect process and such semiconductors are known as INDIRECT GAP semiconductors. The two particle process is not favoured, and recombinations in indirect gap semiconductors usually occur by thermal or collisional processes. The silicon chips in a computer do not glow, they just get warm.
GaAs however is a DIRECT GAP semiconductor. The minima of electron and hole energies occur at the same momentum. Thus recombination can result in a photon alone. LED action was first observed in GaAs in 1952. The light is emitted in a narrow range of wavelengths which is determined by the size of the band gap of the semiconductor. The red, green and yellow LEDs available today are made using semiconductor compounds with different band gaps, for example GaAs, GaAlAs, AlInGaP, GaAsP.
The LED must satisfy two more conditions before it is possible to make a laser diode. Firstly stimulated emission must be able to dominate over absorbtion and spontaneous emission, and secondly there must be an optical cavity. Fortunately both are relatively simple to achieve. The population inversion required for stimulated emission to dominate is realised by increasing the current through the diode. This increases the density of electron-hole pairs in the junction region, and thus creates an inversion. The optical cavity is made by cleaving the crystal along two parallel crystal planes, perpendicular to the junction plane. This creates optically flat and parallel surfaces. They do not need to be coated to act as laser mirrors as the high refractive index of GaAs gives the GaAs-air interface a reflectivity of 35% which is found to give sufficient feedback to sustain laser action. The structure of the basic diode laser is shown in fig. 3.3.
Figure 3.3 The basic diode laser (from Schawlow 1963)
The simple devices described in the previous section had very high threshold currents of 50,000 to 100,000 A/cm2. Even with the sub-millimetre dimensions typical of these devices this still translates to device currents of up to 50 A. The associated heating of devices run at these very high currents required cooling in liquid nitrogen, and even then only pulsed operation of the lasers was possible. To understand this and to see how the threshold current can be reduced we need to discuss what the threshold current is. It is the current required to give a sufficient population inversion for the round trip gain in the cavity to be greater than the round trip loss. The gain may be increased by improving the flatness of the junction. In a rough junction any particular part of the laser beam will only pass through the gain region of the junction for a fraction of the length of the crystal. The round trip gain will clearly be greater in a flat junction where the beam can be in the gain region for the full length of the crystal. The junction flatness was improved by the development of new growth techniques for the junctions. Instead of making the junction by diffusing acceptor impurities (such as tin) into an n-type crystal to make the top part p-type, the n-type substrate was polished flat and then a p-type layer grown on top from a melt of p-type material. This epitaxial method gives a junction almost as flat as the initial polished surface. Losses from the cavity occur as photons pass out of the cavity, both as a useful beam and from scattering from dislocations in the crystal. More careful growth techniques improved the crystal quality, reducing dislocations, and along with flat junctions this reduced thresholds to 25,000 to 30,000 A/cm2 and device currents to about 10 A. Though this was a considerable improvement it was still a large current and further improvements were clearly necessary.
A diode laser made from a single material such as GaAs has some fundamental limitations[4]. Firstly, carriers can diffuse out of the junction region, creating a wide gain region which reduces the round trip gain along any particular path through the crystal. Secondly there is considerable loss of light by spontaneous emission perpendicular to the cavity axis. This is particularly important at threshold as it is the spontaneous photons travelling along the cavity axis and being reflected back into the gain region which initiate the laser action. Diodes made up from layers of different semiconductors, known as heterostructures, can overcome these natural limits by confining the light and the charge carriers to a defined region of the diode. The principle of the confinement is shown in fig 3.4. This shows the two types of heterojunction used. In each case the junction is between two semiconductors with differing band gaps. In the left-hand diagram, both are p-type. This means that in equilibrium it is the valence band which is equalised across the junction, leaving a step in the conduction band. The level valence band does not impede the majority carriers, the holes, but the step in the conduction band presents a barrier to any electrons diffusing into this region from a conventional n-p junction to the left. The right-hand diagram shows an n-p heterojunction. Here there is a large step in the valence band even under forward bias, which presents a barrier to the diffusion of holes out of the junction region. Fig 3.5 shows how this may be used in single and double heterostructures to confine the carriers to a small region which is defined, not by the nature of the materials, but by the thickness the layers are grown to. This gives us the ability to define a narrow lasing region, about 200nm wide, in which the gain is concentrated. As well as carrier confinement, the heterostructures achieve light confinement due to the different refractive indices of the different semiconductors.
Figure 3.4 Heterojunctions (from Panish 1971)
These techniques dramatically reduce the threshold currents to 1000 A/cm2, with device currents below 1 A[5,6]. This allows the diodes to be run continuously at room temperature. Improved heat sinking, and thus improved performance, may be achieved by placing the electrodes as a narrow stripe on a wider diode. The heat generated in the diode may then be conducted sideways out of the lasing region as well as up and down. A typical diode of this type reported in 1972 was capable of producing 120 mW of light, continuous wave, at 900nm. The efficiency was only 7%, which is far from the theoretical maximum for a diode laser, but already two orders of magnitude better than a typical gas or solid state laser.
Figure 3.5 Heterostructures (from Panish 1971)
Once reliable devices had been developed which could operate continuously at room temperature they had to be modified to suit the many applications which had been proposed. The first major use envisaged was in optical communications. This is a low power application but requires a narrow bandwidth close to the transmission peak of optical fibres at 1.55 mm. Small size is also important. A diode laser for use in a compact disc player is not required to have a particularly narrow bandwidth but it must have good spatial beam quality and be of short wavelength so that it can be focused to a small diffraction limited spot. Diode lasers usea in recent applications in the pumping of solid state lasers are required to deliver very high power of 1 to 100 W, but with little requirement for either narrow bandwidth or good beam quality.
A free running diode laser has a bandwidth of 1 to 10 nm. This is owing to the distribution of the holes and excess electrons in the levels of the valence and conduction bands. If one surface of the diode is antireflection coated, then an external end mirror may be used and all types of dispersive elements inserted to reduce the laser bandwidth. This type of external cavity is the same as would be used in a solid state or dye laser. A development of this is the Cleaved Coupled Cavity, or C3 laser[7,8]. The design is shown in fig 4.1. The tuning mechanism is essentially the same as an intracavity etalon, but with important practical differences: i ) The etalon is a second laser, so the gain region can still fill the entire cavity, which keeps the threshold low. ii ) The cavities have similar lengths, so only there will be only one mode overlap in the gain profile of the diode, and no further wavelength selection is needed. iii ) The device is manufactured from a single crystal, which is then cleaved in the middle. The bottom electrode is left intact so the two halves are kept in perfect alignment automatically.
Figure 4.1 The CleavedCoupled Cavity laser (from Tsang 1984)
The laser may be tuned across the gain profile of the diode by using the relationship between current and refractive index in a diode below lasing threshold. Current 2 can be set above threshold to cause that section to lase. A small current 1 may then be applied to the other section. The magnitude of current 1 will then determine the refractive index of that section and hence the mode spacing. Changing current 1 will tune the comb of modes from section 1 across those of section 2 and give overlap on a different wavelength.
The line narrowing and tuning mechanisms can only give wavelengths close to the band gap energy of the semiconductor used for the laser. GaAs emits at about 900nm; for the 1.55 mm needed for communications, or for 670nm visible lasers, other semiconductor compounds must be found with the required band gap energies. There are many III-V semiconductors which are suitable for the production of diode lasers which have extended the range of commercial devices now available to wavelengths from 600nm to 1.6 mm.
Single diode devices have a power limit of the order of hundreds of milliwatts. At higher output powers the high currents required cause unwelcome heating. For really high power applications many individual diodes are stacked in arrays. Lenses placed in front of the stack then combine their outputs into a single beam. As the laser cavities are independent of each other the spatial and spectral quality of such stacks is poor. However it is amply sufficient for pumping solid state lasers.
Making laser diodes in the blue region of the spectrum from wide band gap semiconductors is proving to be very difficult. Again the main problem arises from high threshold currents and the associated heating. In wide band gap diodes the heating comes mainly from ohmic resistance at the interfaces between the wide band gap material and either metal electrodes or narrow band gap substrates. The fact that the heating is localised worsens its effect. There has still been some success, and (Zn,Cd)Se / Zn(S,Se) heterostructures have been demonstrated[9] with 15 mW pulsed at 500nm at room temperature and continuous operation at 77 K.
Possibly a more promising route to short wavelengths is nonlinear frequency conversion of red diodes or diode-pumped solid state lasers. Electricity to blue light efficiencies of 1-10% have been realised by these methods[10,11], which is small when compared with efficiencies of 40% in red diodes, but is still much better than the alternatives at these wavelengths.
For optical communication between and within computers it is desirable to integrate processing circuitry and laser diodes onto a single chip. This would greatly increase the speed of communications between and within computers and also reduce the total size of a system by reducing the number of chips. The problem here is that it is cheapest and simplest to fabricate electronic circuitry from silicon, but diode lasers must be made from gallium arsenide. This presents an engineering problem, though not an insurmountable one.
[1] A.L. Schawlow and C.H. Townes Physical Review 112,p1940 ( 1958 )
[2]T.H. Maiman Journal of the Optical society of America 50,p1134 ( 1960 )
[3] A.L.Schawlow Scientific American 209,p34 ( 1963 )
[4] M.B. Panish and I. Haysashi Scientific American 225,p32 ( 1971 )
[5] M.B. Panish, I. Haysashi and S. Sumski Applied Physics Letters 16,p326 ( 1970 )
[6] Zh.I. Alferov, V.M. Andreev, V.I. Korol'kov, E.L.Portnoi and D.N. Tret'yakov Soviet Physics Semiconductors 2,p1289 ( 1969 )
[7] W.T. Tsang, N.A. Olsson, R.A. Logan, R.A. Linke, B.L. Kasper, I.P. Kaminow Journal of the Optical society of America B-Optical Physics 1,p448 ( 1984 )
[8] W.T.Tsang Scientific American 251,p127 ( 1984 )
[9] H. Jeon M. Hagerott, J. Ding, A.V. Nurmikko, D.C. Grillo, W. Xie, M. Kobayashi and R.L. Gunshor Optics Letters 18,p125 ( 1993 )
[10] L.Y. Liu, M. Oka, W. Wiechmann and S. Kubota Optics Letters 19,p189 ( 1994 )
[11] W.J. Kozlovsky and W. Length Optics Letters 19,p195 ( 1994 )
By William wadsworth (1994)