Introduction

The use of electromagnetic wave theory and geometrical optics to describe propagation of light is satisfactory when there is no exchange of energy between radiation and matter. All observable characteristics – direction of propagation, polarization, diffraction, interference, the energy of radiation, and so forth – are accurately described by one of these so-called classical theories. However, with the exception of absorption by lossy media (eqn. 4.36) or the effects of dispersion (eqn. 4.22), classical theories fail to give reasons for many phenomena that have practical consequences in modern technology and that involve interaction between radiation and matter. Even the partially successful descriptions of dispersion or loss cannot explain the existence of the multitude of spectrally distinct absorption lines; cannot predict the wavelengths for absorption or the wavelengths for anomalous dispersion; cannot account for all the absorbed energy; and, most importantly, cannot predict or describe the effect of optical gain, which is closely related to absorption. Thus, the principles of operation of many important devices of modern optics such as lasers, photodiodes, electronic cameras, and television screens cannot be explained by classical theories. Even the emission by the sun or an incandescent lamp are beyond the scope of these theories. It is evident therefore that a new, modern theory is necessary – one that is either more complete than, or simply complements, the classical theories.

Introduction

In the introduction we saw that explaining the concept of light may require more than just one theory. However, before we begin our journey through the disciplines of optics, we must identify the physical parameters needed to quantify the phenomena that are associated with light. But even before that, we should recognize that the phenomenon we call light is only a part of the broader phenomenon of radiation. If we consider radiation to be

then light may be defined (American National Standard 1986) as

(This definition is sometimes extended to include ultraviolet and infrared radiation.) Note that, by the present definition, radiation that is spectrally detectable to the eye will be called light even if it is too faint to be seen. Although these definitions include terms (such as electromagnetic waves and spectra) that are yet to be explained, it is evident that light represents a subcategory of radiation. Thus, once the physical quantities that specify radiation are defined they may also be used for the quantification of light. Furthermore, the measurement of radiation – or radiometry – does not depend on how we define what is detectable by the eye. Therefore, radiometry is more objective than photometry, which is the process of measuring light.

Introduction

Spectroscopic techniques are an excellent tool for thermodynamic measurements. Typically they do not require a direct contact with the tested medium and are therefore considered nonintrusive. This is a significant advantage over measurements that require physical probes, particularly when the environment at the point of measurement is hostile owing to such extreme conditions as high temperature, aggressive chemical reaction, or fast flows. In addition, many spectroscopic techniques can be adapted to image the two- or three-dimensional distribution of select properties. With the advent of electronic cameras, images can be digitized and stored almost instantaneously by computers. Exposure times of as short as 5 ns are possible with the use of CCD (charge-couple device) intensified cameras. Recording rates of up to 12 kHz can be achieved with image converter cameras, and storage of hundreds of images is possible on most desktop computers. These device characteristics enable such fast phenomena as explosions or detonation waves to be recorded electronically; depending on the selected spectroscopic technique, images of the distribution of thermodynamic parameters of such events can be obtained.

Spectroscopic techniques can be classified in several ways. A simplistic but useful classification includes only two groups: passive techniques, which depend on spontaneous emission by the tested object; and active techniques, which require an excitation of the tested object to induce detectable emission.

Over the last two decades, such new technologies as laser-aided material processing, optical communication, holography, and optical measurement techniques have advanced from laboratory curiosities to commercially available engineering tools. Faced with these developments, engineers must now apply in their practice many concepts of physics that in the past were considered outside the boundaries of classical engineering. Advanced levels of electromagnetic theory and of quantum mechanics are now required to answer fundamental questions about interference, diffraction, or polarization of light and their applications. For example: How does radiation interact with gases, liquids, and solids? How does one obtain optical gain? When should a laser be used, and when would an ordinary light source suffice? How can a laser beam be produced and focused? Answers to these questions are essential for selecting an optical technique for measuring the properties of gas, liquid, or solid phases, or when designing a laser system for material processing, surgery, communication, or entertainment. Although most engineering students attend two or three undergraduate courses in physics, they seldom acquire the proficiency required to fully understand the intricacies of modern optics or laser applications. Similarly, midlevel engineers who obtained their formal education before many new techniques were developed may find an increasing gap between the knowledge acquired in undergraduate physics classes and the requirements of professional practice.

Introduction

The presence of metastable defects in device materials can have a variety of effects, some of them slight and some very important; but because the emphasis in this book is material phenomena, device effects are only summarized here. There are cases in which optically (or electronically) induced changes in defects degrade device performance, and some other cases in which similar processes speed the annealing of existing degradation. Either way, the inherent properties of metastability make its effects more prominent at lower temperatures, with the metastable-barrier energy determining the maximum temperature at which significant effects are observed. Since the energy gap of the host material limits the size of this barrier, Si devices operating at room temperature (Eg= 1.1 eV) are not much affected; GaAs (Eg = 1.4 eV) devices are affected, generally at lower temperatures; and compounds with larger gaps show the greatest effects. By far the greatest device impact is in photovoltaic cells of a-Si:H, whose degradation by bright light has been discussed at length in Chapters 4 and 5 and is the motivation for a great deal of the research reviewed there. There are effects in devices using crystalline III–V materials, but their impacts are less pervasive, and in other devices of a-Si:H they are not critical. For stable deep-level defects, the most prominent effect is as lifetime “killers” in minority-carrier devices, and indeed that is also the dominant effect of metastable defects in a-Si:H. In majority-carrier devices or semiconductor lasers other effects are more important.

Amorphous Semiconductors and Their Defects

For the reader unfamiliar with amorphous semiconductors, we present a brief summary of some of the essential principles needed as a context for the treatment of metastable defects in hydrogenated amorphous Si (a-Si:H). For more detailed discussions of properties of amorphous solids the reader is referred to books by Elliott (1983) and Zallen (1983); for theory of tetrahedral semiconductors to that by Overhof and Thomas (1989); and for broad descriptions of a-Si:H to a four-part set edited by Pankove (1984), a two-volume set edited by Fritzsche (1989), and a book by Street (1991b). For additional historical background there is the landmark book of Mott and Davis (1979).

It is essential to recognize from the start that, although amorphous materials lack the long-range order of crystals, atoms in amorphous structures do not have random locations as occur in a gas. Rather, they retain the same shortrange order that characterizes the local atomic relationships of a crystal of the same material; in this respect amorphous and liquid semiconductors have much in common. When some liquids are cooled sufficiently to solidify, they may form a glassy state, which is defined by such high viscosity as to provide structural rigidity (i.e., shear strength) although crystallization does not occur.

Most semiconductors, when cooled from the melt, form crystallites rather than glasses; but other more rapid cooling techniques can force them to become amorphous solids, although these are then limited to thin films.

CdS and CdSe

Descriptions of exotic effects as the result of photoexcitation of II–VI materials, particularly CdS and CdSe, can be found extending back to when Boer, Borchardt, and Borchardt (1954) reported temperature-dependent, slow decreases of photoconductivity with time of photoexcitation of CdS crystals. In this section we review some examples of these effects and the kinds of models that have been suggested. It must be admitted that genuinely authenticated models do not exist for many of these early photoinduced defect effects, and so we primarily call attention to the suggestions that investigators have made.

Apparently related effects were found by measurements of thermally stimulated conductivity depending on the temperature of photoexcitation before making the measurement, and in the effect of photoexcitation on impure CdS, particularly on CdS:Cu crystals showing both a decrease in bulk photoconductivity of the material and a reversible degradation of the junction collection properties of a CuxS/CdS heterojunction. The phenomenon of persistent photoconductivity has also been observed. In many cases decreases of luminescence with time under photoexcitation have been observed, paralleling the decrease in photoconductivity. A photoinduced defect interaction model for some of these kinds of effects has been described by Tscholl (1968).

Decrease in Photoconductivity with Time in CdS and CdSe

The experimental results at 100 °C of the original experiment of Boer and coworkers (Boer 1954, 1990 [p. 1101]; Boer, Borchardt, and Oberlander 1959) are shown in Figure 3.1.

Introduction

Among defects in semiconductors the DX center is unique in the extent and detail of its studies and in the advanced present state of its understanding. It (or similar centers) has been observed in n-type III–V compounds doped with a variety of species of atoms and in some II–VI compounds; it may also be related to the dominant defect in a-Si:H. The donor-related DX center in AlGaAs alloys (the most widely studied) can thus serve as a model that can usefully be invoked in the study of other semiconductor defects; the EL2 defect in GaAs – a native defect in undoped material now thought to be an AsGa antisite defect (an As atom on a site that normally would have a Ga atom) – has some similar properties, and is discussed in Section 2.6. We show later in this chapter that this model provides a new and convincing explanation for a major limitation on the ability to dope some III–V materials. This limitation has caused even more serious impediments to the applications of II–VI materials, and (as discussed in Chapter 3) the DX model is useful there too. Present understandings followed nearly twenty-five years of efforts to elucidate metastability in III–V compounds, and the literature is large; we therefore present here only a brief summary of the early work, which is described well in reviews by Lang (1992), Langer (1980), and Mooney (1990, 1991), who also discussed the impacts of such deep donors on the properties of devices using these materials.

A number of observations of anomalous effects have been reported for these centers, with the consequent conclusion that the familiar one-electron picture of deep-level centers is inadequate.

Over the past few decades a fascinating revolution has taken place with regard to the way that defects in semiconductors and their properties are regarded. Traditional understandings of a variety of effects have used models of isolated point defects with fixed properties, operating independently in the matrix of the solid, and describable in terms of one-electron energy bands and defect levels. More recently we have come to realize that this classic picture of a simple defect with fixed properties may be the exception – that in real semiconductors there is a wide variety of interactions between defects with deep energy levels and the lattice that leads to variations in local atomic configurations comprising the defect, and in some cases to complexes between defects. The central distinguishing feature of most of the defect processes of interest is that they inextricably couple the electronic system and local structural configurations. Consequently, processes involving these defects cannot be fully described by either electronic transitions or defect states alone.

Some of these new configurations are relatively stable, but many others are not: They can be brought into metastable existence in the solid by the action of photoexcitation, other high-energy excitation, injection of excess carriers, and/or thermal energy, and they can be removed once again by thermal or optical anneal that returns them to their ground state. The metastability of their existence and the radical changes in electronic properties that can exist between their ground state and metastable state provide interesting new possibilities for semiconductors, but often they also provide mechanisms for instability and degradation that handicap the applications of semiconductors.

In this chapter we consider examples of photoinduced defect processes in a number of other amorphous semiconductors for comparison with the results previously described. These examples are drawn from research on amorphous AlGaAs, compensated amorphous silicon, amorphous germanium, alloys of amorphous silicon and germanium, amorphous silicon nitride, and finally the amorphous chalcogenides. General references for these subject areas are Disordered Semiconductors (Kastner, Thomas, and Ovshinsky 1987), the International Conference on Amorphous Semiconductors (1993), and the review paper “Photoinduced effects and metastability in amorphous semiconductors and insulators” (Shimakawa, Kolobov, and Elliott, in press).

DX Centers in Amorphous AlGaAs Films

A comparison of DX-center effects in crystalline and amorphous Si-doped A10.34Ga0.66As has been reported by Lin, Dissanayake, and Jiang (1993). Two questions are treated: How are the relaxation properties of the DX center affected by changes from crystalline to amorphous? What is the connection between the DX type of defect in crystalline and amorphous semiconductors?

Below 250 K, DX centers in amorphous AlGaAs exhibit persistent photoconductivity behavior, such as is characteristic of their behavior in crystalline materials. The decay of PPC can be described by a stretched exponential with decay time constant τ and decay exponent β, such that τ decreases with increasing temperature. A comparison of the temperature dependence of τ for amorphous and crystalline material is shown in Figure 6.1, and a similar comparison of the temperature dependence of β is shown in Figure 6.2. The values of r are larger for the crystalline than for the amorphous material, but the temperature dependence is approximately the same for both.

Kinetics of Generation and Annealing of Photoinduced Defects

In efforts to elucidate the nature and origin of the metastable defects induced by light in a-Si:H (Staebler and Wronski 1977), there have been many studies of the kinetics of their generation and annealing. Although such studies do not establish any microscopic models, they can provide significant guides as to which models may be acceptable. In the interpretation of such studies it is useful to be mindful of the various other ways of producing dangling-bond defects with similar properties. In particular, the observation that such defects can also be produced in the dark by passage of forward current in a p–i–n device (Staebler et al. 1981) has led to the general belief that the defects are not produced by the light directly, but rather by the energy released when excess carriers recombine (Staebler et al. 1981) or are captured (Crandall 1991) at a localized center. Another significant observation is that light of photon energies down to and below 1 eV can cause this degradation even though the band gap is ≈ 1.75 eV.

The earliest observations of photoinduced degradation of lightly doped a-Si:H – the Staebler–Wronski effect – found that both the dark conductivity and photoconductivity decreased as a result of light exposure (see Fig. 1.2), and that these effects could be annealed away at about 150 °C for a few hours (Staebler and Wronski 1977). Thus there were two kinetic terms evident for the rate of change of the defect density N: a positive, defect-generating term that must contain the light intensity, and a negative, thermal annealing term.

Types of Defects

This discussion of defects in semiconductors deals with those having significant photoelectronic interactions. The word defect is used here as a shorthand for “imperfection.” It may therefore include any departure from the ideal periodic lattice of a crystal or, in the case of amorphous materials, any departure from an ideal continuous random network. These defects may take a variety of forms:

1. native point defects, such as isolated vacancies, interstitials, or antisite atoms of the host crystal;

2. point defects associated with the presence of isolated impurity atoms, in either substitutional or interstitial positions;

3. defect complexes formed by the spatial correlations between different point defects, such as donor–acceptor or impurity–vacancy pairs;

4. line defects, such as dislocations;

5. defects associated with grain boundaries in a polycrystalline material; and

6. defects associated with the existence of a surface or interface.

In keeping with the thrust of this book, some defects (e.g., phonons, dislocations, and interfaces or surfaces) are not treated independently.

General Effects of Defects on Electronic Properties

Any of the above defects can play a variety of electronically active roles that affect the electrical and optical properties of a semiconductor. Some of the traditionally accepted roles can be summarized as follows.

Donor or Acceptor Fundamentally, a donor is a defect that is neutral when electron-occupied, or positive when unoccupied; an acceptor is a defect that is negative when electron-occupied, or neutral when unoccupied.

Introduction

Optical frequencies are so high that a small relative bandwidth is a very large bandwidth in absolute terms. This fact is behind the success of optical short pulse generation. The first successful generation of short optical pulses via modelocking started in the 60s with Nd:glass. Figure 1.1 gives the history of the advances in short pulse generation. Steady progress in generating shorter and shorter pulses was achieved with dye lasers, principally because these were CW lasers that had much more predictable modelocking behavior than Q-switched lasers. After subpicosecond pulse generation was achieved with dye lasers, it was soon realized that the short pulses produced by the lasers were much shorter than the relaxation times of the dyes used for the gain and absorption media. It was recognized that the pulses were generated through combined action of the saturable absorber dye that, by saturating, opened the ‘shutter’ for the transmission of the pulse, and the gain, by saturating, closed the ‘shutter.’ This was an important discovery since there exist very few saturable absorbers with subpicosecond relaxation times appropriate for intracavity operation. The shortest pulses achieved were 27 fs in duration, and after spectral spreading in a fiber and subsequent compression 6 fs was achieved, still the record today.

In 1978 the first modelocking of semiconductor lasers was achieved. They are of prime importance in communications applications and are limited to about 100 fs pulse duration, a pulse width adequate for communications for many years to come.