Perspectives1

Introduction

1. Perspectives on Electronic and Photonic Materials

Tim Smeeton, Colin Humphreys

Electronic and photonic materials have a tremen-dous impact on the modern world. They includea wide range of material classes and are developedthrough a deeply interdisciplinary combination ofphysics, chemistry, materials science, and engi-neering. In this introductory chapter, we give someperspectives on this exciting and ever-changingfield. We give an example of the tremendous in-tegration of different materials used in today’sconsumer products, and then take a historicallook at the development of some key semicon-ductor materials and devices from inception totoday. Focusing in particular on the developmentof the transistor and integrated circuit and someof the key electronic and photonic applications ofcompound semiconductors, we take advantage of

1.1 Tremendous Integration ....................... 2

1.2 The Silicon Age..................................... 31.2.1 The Transistor and Early Semiconductor

Materials Development.......................... 31.2.2 The Integrated Circuit ............................ 5

1.3 The Compound Semiconductors ............ 71.3.1 High-Speed Electronics ......................... 81.3.2 Light Emitting Devices ........................... 91.3.3 The III-Nitrides ..................................... 11

References ..................................................... 14

the long-distance view to point out some unifyingthemes across the wide portfolio ofmaterials whileappreciating their unique features.

It can be easy to forget how remarkable electronic andphotonic materials are. Take the light emitting diode(LED) as an example: An electric current is passedthrough a tiny stack of layers of slightly different mate-rials and brilliant colored light is emitted. Of course, theLED is carefully designed, and there are powerful the-ories to explain the behavior, but that should not detractfrom the initial moment of wonderment that it works atall. Or that a room can be lit by photons generated in theactive region of an LED which includes just a few thou-sand cubic micrometers of material – about the samevolume as a grain of flour.

We all literally see the output from LEDs everyday, from television screens, vehicle lights, or lightingluminaires. Most other electronic and photonic materi-als and devices are less conspicuous to our senses buttogether they have played an indisputable role in defin-ing the way people live, work, and communicate in thetwenty-first century: from microprocessors containingbillions of transistors which provide immense com-putational power to laser diodes and radiofrequencytransceivers which enable trans-global and wirelesscommunication. These complex devices are only avail-able today thanks to the work of countless researchersover the past century who have identified, designed andunderstood a vast array of materials, produced them

with extraordinary purity, quality and economy, anddeployed them in device designs which harness theirpower.

The detailed chapters in this handbook providecomprehensive introductions to the huge range of tech-nical fields which electronic and photonic materialsnow occupy, written by world experts. In this intro-ductory chapter, we have the luxury of stepping backfrom the details of these complex fields and reviewingthe whole, in search of some perspective. First, we takea look at the complex mixture of electronic and pho-tonic materials which are in use today and reflect onsome common themes. Next, the majority of the chapteris dedicated to a review of the development of a hand-ful of the most important semiconductor materials anddevices since the early breakthroughs of the 1940s: thebirth of electronics in germanium and silicon; the in-tegrated circuit (IC); some of the key electronic andoptoelectronic uses of III–V semiconductors; and therecent rise of III-nitride materials. These reviews serveas a tribute to breakthroughs of the last 75 years, hope-fully providing some perspective of how the science andtechnology of these materials has come to its currentstate and perhaps providing some encouragement andinspiration to those who are striving to develop materi-als today.

© Springer International Publishing AG 2017S. Kasap, P. Capper (Eds.), Springer Handbook of Electronic and Photonic Materials, DOI 10.1007/978-3-319-48933-9_1

Introduction

2 Introduction

1.1 Tremendous Integration

A starting point for perspective is to appreciate the com-plex mixture of materials and devices that are routinelyintegrated together. For a case study, we need look nofurther than something which is probably within touch-ing distance as you read this sentence: a smartphone.One of the defining technologies of the twenty-first cen-tury, it has been enabled by an astonishing convergenceof electronic and photonic materials and devices. Notwo models are quite the same, but let us teardowna hypothetical handset and its supporting infrastruc-ture, laying forward references to the detailed chaptersin this book as we go. As a user, we see text, im-ages and video from a high-resolution liquid crystaldisplay (LCD) in which blue light emitted from in-dium gallium nitride (InGaN) LEDs (Chaps. 31, 35,and 40) excites a rare-earth-doped yttrium aluminumgarnet (YAG) phosphor (Chap. 38) to produce whitelight which, in turn, is color-filtered into red, green,and blue components and modulated at the pixel levelby liquid crystal cells (Chap. 36) controlled by sili-con or indium gallium zinc oxide (IGZO) thin filmtransistors (TFTs) (Chap. 44). We have touch controlthanks to capacitive sensors which use an opticallytransparent and electrically conducting oxide such asindium tin oxide (ITO) (Chap. 58). Low power con-sumption multicore processors, RAM, and flash mem-ories – all silicon – provide the computing power andstorage (Chap. 21). The device is powered by a highenergy-density lithium ion battery (Chap. 11). Wire-less connectivity is provided between gallium arsenide(GaAs) based transistor radio transceivers on the hand-set and base stations equipped with silicon, siliconcarbide (SiC), or aluminum gallium nitride (AlGaN)transistor amplifiers. Beyond that, data communica-tion – which gives the smartphone its purpose – is viasilica (SiO2) optical fibers (Chap. 41) using indiumphosphide (InP)-based laser diodes, optical receivers,switches and amplifiers (Chap. 35) to data centers basedon high-performance silicon processors. And we cango further still: The data centers consume electricalpower distributed through more power transistors (sil-icon, SiC and AlGaN) and perhaps originally generatedby crystalline silicon, amorphous silicon, copper in-dium gallium selenide (CIGS) or cadmium telluride(CdTe) photovoltaic cells (Chap. 43).

The list of core materials is already well into dou-ble figures, but we have still only scratched the surface.If we consider just the families of semiconductor com-ponents in a little more detail, these include multiplealloy compositions (Chap. 30) with optimized combina-tions of impurity doping (Chap. 2). Not even to mentiondozens – probably hundreds – of different electrode, in-

terconnect, dielectric and packaging (Chaps. 27, 29, and53) materials.

We could go on, but surely the point is made:A remarkably complex combination of materials anddevices is involved. It is a phenomenal achievement oftechnology and business – we could argue a wonderof the modern world – that this dense concentration ofdifferent materials and technologies can be brought to-gether with such synergy in a reliable, mass-producedproduct weighing less than 150 g that can be sold ata price accessible to billions of people. It is often theproduct design or the software of these devices whichsteal the limelight in marketing, but it is the elec-tronic and photonic materials which are the unassumingheroes!

Electronic and photonic materials include almost alltypes and forms of material:

� Inorganic compounds (semiconductors, dielectrics)� Organic molecules (liquid crystals, semiconduc-tors)� Ionic crystals (batteries), polymers (packaging,lithography masks)� Metals/alloys (electrodes, interconnect wiring)� Single crystal� Polycrystalline� Amorphous� Thin film� Nanostructured� Two-dimensional.

Despite this breadth, there is at least one strongunifying theme across all: the need for extraordinaryprecision in the design and fabrication of the materials.These materials demand combinations of high purities,low defect densities, and precise chemical compositionthat probably exceed the requirements of any other areaof technology. These requirements have been met bya unique convergence of physics, chemistry, materialsscience and engineering which brings together the coretheoretical understanding with the means for fabrica-tion.

Two aspects, in particular, are vital for deliveringthe necessary precision: exact material fabricationmethods and detailed understanding of material struc-ture and properties. In many cases, the ultimate deviceperformance is determined by how far we fall shortof fabricating the perfect material. If the number ofcrystalline defects is too high, or if there are too manyimpurities, or if an interface is too rough, or if . . . ,a device may fail. Minute deviations from a targetspecification can have catastrophic effects. Taking LEDor laser diode fabrication as an example, a change

Perspectives on Electronic and Photonic Materials 1.2 The Silicon Age 3Introd

uction

in composition of material or doping accounting tomuch less than 1% of a layer just a few nanometersthick can be the difference between a viable deviceand a costly failure. In nearly all device fabricationsteps, from bulk growth methods for fabrication ofsubstrates with near-perfect quality, to thin film growthmethods to deposit precise material compositions withatomic-scale precision, to processing technology topattern device features at the nanometer scale, andthrough to packaging and integration methods, materialproduction techniques are the lifeblood of the field.The precision of fabrication techniques needs to beat least matched by the precision with which we candetermine the structure of the resulting materials. Thismeans that the structural and chemical characterizationtools and techniques are an inseparable aspect of thestory of all electronic and photonic materials. Accurateunderstanding of the quality (or otherwise) of thematerials that we make – obtained from microscopy orother tools – has been a continuous driver or enablerfor progress in material and device development.

When the material fabrication expertise and under-standing of structure and properties are combined withdeep understanding of the physics of device design,the result is the remarkable devices that can enrich ourlives.

So, today’s electronic and photonic materials spana tremendous range of fields and disciplines. Howdid this myriad of technologies successfully emergeand grow? Why have the materials that are in usetoday become dominant over others? In the nextsections, we will give some very selective answersto these questions, hoping to provide a perspectiveof how some technologies have reached the cur-rent state. In particular, we have chosen to focuson a few sectors of semiconductor development toexplain some of the early development of what be-came mainstream technologies. Often the key mes-sages are learnt in these early years – before a tech-nical field fragments into highly specialized sec-tors – and may provide the most valuable perspec-tive.

1.2 The Silicon Age

1.2.1 The Transistor and EarlySemiconductor Materials Development

Electronics is the control of electrons to produce usefulproperties; electronic materials are the media in whichthis manipulation takes place. The foundations for to-day’s electronics were laid in 1947 – 50 years afterJ.J. Thompson had discovered the electron – with thefirst demonstration of the semiconductor transistor ef-fect. Building on quantum theories to explain semicon-ductor behavior from the 1930s [1.1, 2] and improve-ments in purity of candidate semiconductor materials,John Bardeen and Walter Brattain used germanium tobuild and demonstrate the first semiconductor triode.Later, to be named the point contact transistor to reflectits transresistive properties, this success at Bell Labo-ratories was obtained just a few years after a researchgroup led by William Shockley was established to fo-cus on understanding semiconducting materials. It wasto earn Brattain, Bardeen, and Shockley the 1956 NobelPrize for Physics.

The first point-contact transistor was based aroundthree contacts onto an n-doped germanium block: Whena small current passed between the base and emitter, anamplified current would flow between the collector andemitter [1.3]. The emitter and collector contacts neededto be located very close to one another (50�250�m),and this was achieved by evaporating gold onto the

corner of a plastic triangle, cutting the film with a ra-zor blade, and touching this onto the germanium –the two isolated strips of gold serving as the two con-tacts [1.4]. At about 1 cm in height, based on relativelyimpure polycrystalline germanium and adopting a dif-ferent principle of operation, the device bears barelyany resemblance to today’s IC electronics components.Nonetheless, it was the first implementation of a solid-state device capable of modulating (necessary for sig-nal amplification in communications) and switching(needed for logic operations in computing) an electriccurrent. In a world whose electronics were deliveredby the thermionic vacuum tube, the transistor was im-mediately identified as a component which could beemployed as an amplifier, oscillator, and for other pur-poses for which vacuum tubes are ordinarily used [1.3].

In spite of this, after the public announcement of theinvention at the end of June 1948, the response of boththe popular and technical press was somewhat muted.It was after all still little more than a laboratory curios-ity [1.5] and ultimately point-contact transistors werenever suited to mass production. The individual devicesdiffered significantly in characteristics, the noise levelsin amplification were high and they were rapidly to besuperseded by improved transistor types.

A huge range of transistor designs have been intro-duced from the late 1940s to now. These successivegenerations either drew upon or served as a catalyst

Introduction

4 Introduction

for a range of innovations in semiconductor materialsprocessing and understanding. There are many fascinat-ing differences in device design but from a materialsscience point of view the three most striking differ-ences between the first point-contact transistor and themajority of electronics in use today are the choice ofsemiconductor, the purity of this material, and its crys-talline quality. Many of the key electronic materialstechnologies of today derive from the developments inthese fields in the very early years of the semiconductorindustry.

Both germanium and silicon had been producedwith increasing purity throughout the 1940s [1.6]. Prin-cipally because of germanium’s lower melting tem-perature (937 ıC compared with 1415 ıC) and lowerchemical reactivity, its preparation had always provedeasier and was therefore favored for the early devicemanufacture such as the first transistor. However, theproperties of silicon make it a much more attractivechoice for solid-state devices. Although germanium isexpensive and rare, silicon is, after oxygen, the secondmost abundant element. Silicon has a higher break-down field and a greater power handling ability; itssemiconductor band gap (1:1 eV at 300 K) is substan-tially higher than that of germanium (0:7 eV), so silicondevices are able to operate over a greater range of tem-peratures without intrinsic conductivity interfering withperformance.

These two materials competed with one another indevice applications until the introduction of novel dop-ing techniques in the mid-1950s. Previously, p- andn-doping had been achieved by the addition of dopantimpurities to the semiconductor melt during solidi-fication. A far more flexible technique involved thediffusion of dopants from the vapor phase into the solidsemiconductor surface [1.7]. It became possible to dopewith a degree of two-dimensional precision when itwas discovered that silicon’s oxide served as an ef-fective mask to dopant atoms and that a photoresistcould be used to control the etching away of the ox-ide [1.8, 9]. Successful diffusion masks could not befound for germanium and it was soon abandoned formainstream device manufacturing. Dopant diffusion ofthis sort has since been superseded by the implantationof high-energy ions which affords greater control andversatility.

Shockley was always aware that the material of thelate 1940s was nothing like pure enough to make re-liable high performance commercial devices. Quantummechanics suggested that to make a high quality tran-sistor out of the materials, it was necessary to reducethe impurity level to about one part in 1010. This wasa far higher degree of purity than existed in any knownmaterial. However, William Pfann, who worked at Bell

Laboratories, came up with the solution. He inventeda technique called zone-refining to solve this problem,and showed that repeated zone refining of germaniumand silicon reduced the impurities to the level required.The work of Pfann is not widely known but was a criti-cal piece of materials science that enabled the practicaldevelopment of the transistor [1.10, 11].

At a similar time, great progress was being madein reducing the crystalline defect density of semicon-ducting materials. Following initial hostility by someof the major researchers in the field, it was rapidlyaccepted that transistor devices should adopt singlecrystalline material [1.12]. Extended single crystals ofgermanium several centimeters long and up to 2 cmin diameter [1.12, 13] and later similar silicon crys-tals [1.14] were produced using the Czochralski tech-nique of pulling a seed crystal from a high puritymelt [1.15]. The majority of material in use today isstill derived from this route. To produce silicon withthe very lowest impurity concentration, an alternativemethod called float zoning was developed where a poly-crystalline rod was converted to a single crystal bythe passage of a surface tension confined molten zonealong its length [1.16–18]. No crucible is required inthe process, so there are fewer sources of impurity con-tamination. Float zoning is used to manufacture someof the purest material in current use [1.19]. The earlyCzochralski material contained dislocation densities of105�106 cm�2 but by the start of the 1960s dislo-cation free material was obtained [1.20–23]. Initially,most wafers were on the silicon (111) plane, whichwas easiest to grow, cut, and polish [1.24]. For field-effect devices, which are discussed below, use of the(100) plane was found to offer preferable properties,so this was introduced in the same decade. The impu-rity concentration in dislocation-free silicon has beencontinually reduced up to the present day and waferdiameters have increased almost linearly (though accel-erating somewhat in recent years) from about 10 mm inthe early 1960s to the dinner plate 300 mm which dom-inates today [1.19] and with 450 mm tentatively in thepipeline within the next decade. These improvementsrepresent one of the major achievements in the growthand processing of semiconductor materials.

A series of generations of transistors followed inrapid succession after Brattain and Bardeen’s first tri-umph. Here, we only mention a few of the majordesigns whose production has traits in common withtechnology today. Early in 1948, Shockley developeda detailed formulation of the theory of p–n junctionsthat concluded with the conception of the junction tran-sistor [1.25, 26]. This involved a thin n-doped base layersandwiched between p-doped emitter and collector lay-ers (or vice versa). This p–n–p (n–p–n) structure is the

Perspectives on Electronic and Photonic Materials 1.2 The Silicon Age 5Introd

uction

simplest form of the bipolar transistor (so-called be-cause of its use of both positive and negative chargecarriers), a technology that remains important in ana-logue and high-speed digital ICs today. In April 1950,by successively adding arsenic and gallium (n- andp-type dopants, respectively) impurities to the melt,n–p–n junction structures with the required p-layerthickness (� 25�m) were formed from single crystalgermanium. When contacts were applied to the threeregions, the devices behaved much as expected fromShockley’s theory [1.25, 27]. Growth of junction tran-sistors in silicon occurred shortly afterwards and theyentered production by Texas Instruments in 1954 [1.12].

By the later years of the 1950s, the diffusion dopingtechnique was used to improve the transistor’s speed re-sponse by reducing the thickness of the base layer in thediffused base transistor [1.28]. This began the trend ofmanufacturing a device in situ on a substrate material,so, in a sense, it was the foundation for all subsequentmicroelectronic structures. Soon afterwards, epitaxialgrowth techniques were introduced (what would todaybe described as vapor phase epitaxy (VPE)) which havesince become central to both silicon and compoundsemiconductor technology. Gas phase precursors werereacted to produce very high quality and lightly dopedcrystalline silicon on heavily doped substrate wafers toform epitaxial diffused transistors. Even though the col-lector contact was made through the thickness of thewafer, the use of highly doped (low resistance) wafersreduced the series resistance and therefore increased thefrequency response [1.29].

For some years, the highest performance deviceswere manufactured using the so-called mesa pro-cess where the emitter and diffused base were raisedabove the collector using selective etching of the sili-con [1.25]. The planar process (which was at the heartof nearly all device production until the last few years)was subsequently developed, in which the p–n junc-tions were all formed inside the substrate using oxidemasking and diffusion from the surface. This resultedin a flat surface to which contacts could be made us-ing a patterned evaporated film [1.30]. This processingtechnique was combined with some exciting thoughts atthe end of the 1950s and led to the application of tran-sistor devices and other components in a way which wasto transform the world, that is, the IC.

1.2.2 The Integrated Circuit

With the benefit of hindsight, the IC concept is quitesimple. The problem faced by the electronics industryin the 1950s was the increasing difficulty of physicallyfitting into a small device all of the discrete electroniccomponents (transistors, diodes, resistors, and capaci-

tors), and then connecting them together. It was clearthat this problem would eventually limit the complexity,reliability, and speed of circuits which could be cre-ated. Transistors and diodes were manufactured fromsemiconductors but resistors and capacitors were bestformed from alternative materials. Even though theywould not deliver the levels of performance achievablefrom the traditional materials, functioning capacitorsand resistors could be manufactured from semiconduc-tors, so, in principle, all of the components of a circuitcould be prepared on a single block of the semicon-ducting material. This reasoning had been proposedby Englishman G.W.A. Dummer at a conference in1952 [1.31], but small-scale attempts to realize circuitshad failed, largely because they were based on connect-ing together layers in grown-junction transistors [1.32].In 1958, however, Jack Kilby successfully built a simpleoscillator and flip–flop logic circuits from componentsformed in situ on a germanium block and intercon-nected to produce circuits. He received the Nobel Prizein 2000 for his part in the invention of the IC.

Kilby’s circuits were the first built on a single semi-conductor block, but by far the majority of the circuit’ssize was taken up by the wires connecting together thecomponents. Robert Noyce developed a truly IC in theform that it was later to be manufactured. While Kilbyhad used the mesa technique with external wiring,Noyce applied the planar technique to form transistorson silicon and photolithographically defined gold oraluminum interconnects. This was more suited to batchprocessing in production and was necessary for circuitswith large numbers of components.

In the 1960s, a new transistor design – the metal ox-ide semiconductor field effect transistor (MOSFET) –was introduced that replaced the bipolar device usedin the first chips [1.33]. In this device, a gate was de-posited onto a thin insulating silicon oxide layer onthe silicon. The application of a voltage to the gateresulted in an inversion layer in the silicon below theoxide, thereby, modifying the conducting channel be-tween source and drain contacts. This structure wasa p-MOS device (current transfer between the collec-tor and emitter was by hole conduction) grown on (111)silicon using an aluminum gate. Earlier attempts at sucha device had failed because of trapped impurities andcharges in the gate oxide – this new structure had re-duced the density of these to below tolerable levelsbut the device still could not compete with the bipolartransistors of its time [1.24]. By 1967, however, (100)silicon (which offered lower densities of states at theSi/SiO2 interface) was used together with a polycrys-talline silicon gate to construct a more effective andmore easily processed device with advantages over thebipolar transistor. In the early 1970s, the n-MOS device,

Introduction

6 Introduction

1994 1997 1999 2001Time of PC purchase

120 Å

60 Å

100 MHz

40 Å

300 MHz

500 MHz1GHz20 Å

Fig. 1.1 The relentless march to zero thickness. HRTEMimages of SiO2 gate oxide thickness used in personalcomputers before high-k gate oxides were introduced. (Af-ter [1.34], with permission from Cambridge UniversityPress)

which was even less tolerant of the positive gate oxidecharges, was realized thanks to much improved clean-liness in the production environment. With conductionoccurring by the transfer of electrons rather than holes,these were capable of faster operation than similar p-MOS structures (the mobility of electrons in silicon isabout three times that of holes). By the 1980s, these twodevices, were combined in the complementary MOS(CMOS) device which afforded much lower power con-sumption and simplified circuit design [1.35].

The MOSFET is still a common structure in mi-croelectronics today. Of course, now it can be muchsmaller, with transistor dimensions well below 100 nm,enabling more transistors within a single chip. Thesmaller dimensions are achieved through improvementsin optical lithographic patterning methods. The min-imum dimension of components which can be litho-graphically patterned on an IC is ultimately limitedby the wavelength of radiation used in the processand this has continually been decreased over the pastfew decades. In the late 1980s, wavelengths of 365 nmwere employed; by the late 1990s, 248 nm were com-mon; and today, 193 nm is being used; sometimescoupled with double-patterning processes to extend tothe smallest dimensions. Research into extreme ultravi-olet lithography (e.g., at 13:5 nm wavelength) continuesfor potential use in future generation processes.

Providing good performance from transistors withsmaller dimensions required significant advances inmaterials. For example: better control of doping; use ofcopper interconnects instead of aluminum; and low-kdielectric insulating layers between interconnects in-stead of silicon dioxide. Perhaps, the most disruptivemodification was the replacement of the native siliconoxide as the gate insulator by high-k dielectric material

100

101

102

103

104

105

106

107

108

109

1955 1965 1975 1985 1995 2005 2015Year of introduction

Number of transistors

(Moore, 1965)

40048008

8080

8086 286

386 DX 486 DX

ItaniumItanium 2

Pentium 4Pentium III

Pentium IIPentium

Core i7

Fig. 1.2 The realization of Moore’s law by commercial In-tel processors: the logarithmic increase in the number oftransistors in each processor chip

such as hafnium oxide (Chap. 27). Smaller MOSFETdimensions had required ever thinner silicon gate oxidethicknesses – as shown clearly in Fig. 1.1 – eventuallyculminating in gate insulator layers just a few atomsthick; sufficiently thin that electron quantum mechan-ical tunneling through it became a significant powerdrain. Use of high-k material such as hafnium ox-ide, which has a significantly higher dielectric constantthan silicon oxide (k � 22, compared with k � 3:9),enabled a return to thicker, tunneling-resistant, gate in-sulators for devices launched in the early 2000s. Thisimprovement required huge commitment in researchand development, including development and produc-tion scale-up of new methods to deposit sufficientlyperfect high-k material (e.g., atomic layer deposition,ALD).

A most recent development in transistor technol-ogy has seen the abandonment of the traditional pla-nar geometry and the first widespread use of three-dimensional (3-D) transistor geometries to further re-duce leakage problems associated with the extremeproximity of source and drain. For example, the fin fieldeffect transistor (FinFET), where the transistor channelis formed in a several nanometers wide ridge that pro-trudes above the planar substrate surface and is coveredon its three exposed edges by gate material.

The development in complexity and performance ofsilicon devices, significantly due to materials scienceprogress, is unparalleled in the history of technology.Never before could improvements be measured in termsof a logarithmic scale for such a sustained period. Thisis often seen as the embodiment of Moore’s law. Not-ing a doubling of the number of components fitted ontoICs each year between 1959 and 1965, Moore predictedthat this rate of progress would continue until at least

Perspectives on Electronic and Photonic Materials 1.3 The Compound Semiconductors 7Introd

uction

10 years later [1.36]. From the early 1970s, a modifiedprediction of doubling the number of components ev-ery couple of years has been sustained to the currentday. Since the goals for innovation have often been de-fined assuming the continuation of the trend, it shouldperhaps be viewed more as a self-fulfiling prophecy.A huge variety of statistics relating to the silicon mi-

croelectronics industry follow a logarithmically scaledimprovement from the late 1960s to the current day:the number of transistors shipped per year (increas-ing); average transistor price (decreasing); and numberof transistors on a single chip (increasing) are exam-ples [1.37]. The final member of this list is plotted inFig. 1.2.

1.3 The Compound Semiconductors

It has been said that silicon is to electronics what steelis to mechanical engineering [1.38]. Steel is very ef-fectively used for most of the world’s construction butthere are some tasks which it is incapable of perform-ing and others for which alternative structural materialsare better suited. In the same way, there are some cru-cial applications – such as optoelectronics and very highspeed electronics – that silicon cannot usually deliverbut which a wide range of compound semiconductorsare better equipped to perform.

Silicon’s band gap is indirect (an electron–hole re-combination across the band gap must be accompaniedby an interaction with a phonon in the lattice) thatseverely limits the potential efficiency of light emis-sion from the material. Many of the important com-pound semiconductors, such as the alloys AlxGa1�xAs,InxGa1�xN, AlyGa1�yN, and AlxInyGa1�x�yP, exhibitdirect band gaps (no phonon interaction is required),so they can efficiently emit brilliant light in LEDsand laser diodes. Furthermore, in these alloy systems,where the band gap can be adjusted by changing thecomposition, there is a means of selecting the energyreleased when an electron and hole recombine acrossthe gap and therefore controlling the wavelength ofthe photons emitted. From the AlxInyGa1�x�yAs andAlyInxGa1�x�yN alloy systems, there is, in principle,a continuous range of direct band gaps from deep inthe infrared (InAs; �D 3:5�m) to far into the ultravi-olet (AlN; �D 200 nm). The semiconductor band gapsof these materials and the corresponding photon wave-lengths are put into context with the visible spectrum inFig. 1.3.

Compounds are also very useful in high speed elec-tronics applications. One of the determining factors inthe speed of a transistor is the velocity of the charge car-riers in the semiconductor. In GaAs, the electron driftvelocity is much higher than in silicon, so its transistorsare able to operate at significantly higher frequencies.The electron velocity in InAs is higher still. Further-more, in the same way that silicon was preferred togermanium, devices manufactured using semiconduc-tors such as GaN, which have much wider band gaps

than silicon (3:4 eV compared with 1:1 eV), are capableof operating in much higher temperature environments.

Apart from these advantageous properties of com-pound semiconductors, the use of different alloy com-positions, or totally different semiconductors, in a sin-gle device introduces entirely new possibilities. In sil-icon, most device action is achieved by little morethan careful control of dopant impurity concentrations.In structures containing thin layers of semiconductorswith different band gaps (heterostructures), there is thepotential to control more fundamental parameters suchas the band gap width, mobilities, and effective massesof the carriers [1.38]. In these structures, importantnew features become available which can be used bythe device designer to tailor specific desired proper-

3.5

3.0

2.5

2.0

1.5

1.0

0.5

6.0

5.54.0

200210

220

350

400

500

600

8001000

1.55 μm

3.2 3.4 3.6 5.4 5.6 5.8 6.0

Ultra-violet

Energy (eV) Wavelength (nm)

Hexagonal latticeconstant (a) (Å)

Cubic latticeconstant (a)(Å)

Direct band gapIndirect band gapDirect band gap alloyIndirect band gap alloy

ZnSe

AlAsGaP

Si

Ge

InP

Al0.48In0.52As

In0.53Ga0.47As InAsInN

SiC

GaN

AlN

Violet

Blue

Green

Yellow

Red

Infrared GaAs

Fig. 1.3 Parameter perspective: band gaps and lattice parametersof selected semiconductors. The important wavelengths for opticalstorage (CD, DVD, and Blu-Ray) and the 1:55�m used for efficientdata transmission through optical fibers are labeled

Introduction

8 Introduction

ties. Hebert Kroemer and Zhores Alferov shared theNobel prize in 2000 for developing semiconductor het-erostructures used in high-speed- and optoelectronics.

We will mainly consider the compounds formed be-tween elements in group III of the periodic table andthose in group V (the III–V semiconductors); princi-pally those based around GaAs and InP which weredeveloped over much of the last 40 years, and GaNand its related alloys which have been most heavilystudied only during the last two decades. Other fami-lies are given less attention here, though they also haveimportant applications (e.g., the II–VI materials in op-toelectronic applications). It can be hazardous to try andconsider the compound semiconductors as a single sub-ject. Though lessons can be learnt from the materialsscience of one of the compounds and transferred to an-other, each material is unique and must be consideredon its own (i. e., of course, the purpose of the special-ized chapters which follow in this handbook).

It is worth repeating that the power of the compoundsemiconductors lies in their use as the constituent lay-ers in heterostructures. The principal contribution fromchemistry and materials science to enable successfuldevices has been in the manufacture of high-qualitybulk single crystal substrates and the development oftechniques to reliably and accurately produce real lay-ered structures on these substrates from the plans drawnup by a device theorist. In contrast to silicon, the com-pound semiconductors include volatile components, soencapsulation has been required for the synthesis oflow-defect InP and GaAs substrates such as in the liq-uid encapsulated Czochralski technique [1.39, 40]. Thesize and crystalline quality of these substrates lag someway behind those available in silicon. Crucial to thecommercialization of electronic and optoelectronic het-erostructures were the improvements over the last fewdecades in the control of epitaxial growth available tothe crystal grower. The first successful heterostructureswere manufactured using deposition onto a substratefrom the liquid phase (liquid phase epitaxy (LPE)) –a beautifully simple technology but with severe limi-tations [1.38]. However, the real heterostructure revo-lution had to wait for the 1970s and the introductionof molecular beam epitaxy (MBE) and metallorganicchemical vapor deposition (MOCVD) – also known asmetallorganic vapor phase epitaxy (MOVPE) providedthat the deposition is epitaxial.

MBE growth occurs in an ultrahigh vacuum withthe atoms emitted from effusion cells forming beamswhich impinge upon and form compounds at the sub-strate surface. It derives from pioneering work at thestart of the 1970s [1.41]. MOCVD relies on chemicalreactions occurring on the substrate involving metallor-ganic vapor phase precursors and also stems from initial

work at this time [1.42]. In contrast to LPE, these twotechniques permit the combination of a wide range ofdifferent semiconductors in a single structure and of-fer a high degree of control over the local composition,in some cases on an atomic layer scale. The success-ful heterostructure devices of the late 1970s and 1980swould not have been achievable without these two toolsand they still dominate III–V device production and re-search today.

1.3.1 High-Speed Electronics

The advantages of the III–V materials over siliconfor use in transistors capable of operating at highfrequencies were identified early in the semiconduc-tor revolution [1.43]. Shockley’s first patent for p–njunction transistors had included the proposal to usea wide-gap emitter layer to improve performance, andin the 1950s, Kroemer presented a theoretical de-sign for a heterostructure transistor [1.44]. Some yearslater, the structure of a GaAs metal semiconductorFET (MESFET) was proposed and realized soon af-terwards [1.45, 46]. In these devices, a Schottky barriersurface potential was used to modulate the conductivityof the GaAs channel. One of the earliest applications ofthe III–Vs was the low-noise amplifiers in microwavereceivers that offered substantial improvements relativeto the silicon bipolar transistors of the time. The deviceswere later used to demonstrate subnanosecond switch-ing in monolithic digital ICs [1.47]. Today, they formthe core of the highest speed digital circuits and areused in high speed electronics in microwave radar sys-tems and wireless communications which incorporatemonolithic ICs.

For at least 30 years, there have been repeatedattempts to replicate the MOSFET, the dominant tran-sistor form in silicon ICs, on GaAs material. Theseattempts have been frustrated by the difficulty of re-producibly forming a high quality stoichiometric oxideon GaAs. In direct analogy with the initial failure ofconstructing working n-MOSFETs on silicon, the GaAsdevices have consistently been inoperable because ofpoor quality gate oxides with a high density of surfacestates at the GaAs–insulator interface [1.40]. One of theresearch efforts focused on realizing this device was,however, to be diverted and resulted in the discoveryof probably the most important III–V electronic device,that is, the high electron mobility transistor (HEMT).

The background to this invention lies in the beau-tiful concept of modulation doping of semiconductorswhich was first demonstrated in 1978 [1.48]. One of thetenets of undergraduate semiconductor courses is thedemonstration that as the dopant density in a semicon-ductor increases, the mobility of the carriers is reduced,

Perspectives on Electronic and Photonic Materials 1.3 The Compound Semiconductors 9Introd

uction

because the carriers are scattered more by the ionizeddopants. It was found that in a multilayer of repeating n-AlGaAs layers and undoped GaAs layers, the electronssupplied by donor atoms in the AlGaAs, moved into theadjacent potential wells of the lower-band gap GaAslayers. In the GaAs these suffered from substantiallyless ionized impurity scattering and therefore demon-strated enhanced mobility.

While working in a group attempting to createGaAs MOSFETs (and seemingly despairing at thetask [1.49]), Takashi Mimura heard of these results andconceived of a field effect transistor where the con-ducting channel exploited the high mobility associatedwith a modulation doped structure. In essence, a dopedAlGaAs layer was formed above the undoped GaAschannel of the transistor. Donated carriers gathered inthe GaAs immediately below the interface where theydid not suffer from as much ionized impurity scatteringand so their mobility would approach that of an ultra-pure bulk semiconductor. The current was conductedfrom the source to the drain by these high mobility car-riers and so the devices were able to operate in higherfrequency applications [1.49]. Realization of the struc-ture required a very abrupt interface between the GaAsand AlGaAs and was considered beyond the capabilityof MOCVD of the time [1.49]. However, following theadvances made in MBE procedures during the 1970s,the structure was achieved by that technique withina few months of the original conception [1.49, 50]. Thefirst operational HEMT chips were produced on De-cember 24, 1980: By pleasing coincidence this was theanniversary of Brattain and Bardeen’s demonstration oftheir point-contact resistor to the management of BellLabs in 1947. Structures based on the same principle asMimura’s device were realized in France very shortlyafterwards [1.51].

The commercialization of the HEMT became sig-nificant in the late 1980s, thanks to broadcasting satel-lite receivers. The improved performance of the devicescompared with the existing technology allowed thesatellite parabolic dish size to be reduced by at leasta factor of two. Structures similar to these have sinceplayed a crucial role in the massive expansion in mo-bile telephones.

The evolution in HEMT structures since the early1980s is a fine example of how fundamental compoundsemiconductor properties have been exploited as thetechnology has become available to realize new de-vice designs. The electron mobility in InAs is muchhigher than that in GaAs and rises as the indiumcontent in InxGa1�xAs is increased [1.52]. The intro-duction of an InGaAs (as opposed to GaAs) channelto the HEMT structure, to create the so-called pseudo-morphic HEMT (pHEMT), resulted both in increased

electron mobility and a higher density of carriers gath-ering from the doped AlGaAs layer (because of thelarger difference in energy between the conductionband minima of InGaAs and AlGaAs than that be-tween GaAs and AlGaAs) [1.39]. The indium contentand thickness of the channel are limited by the lat-tice mismatch with the GaAs (Fig. 1.3). If either isincreased too much, then misfit dislocations are formedwithin the channel. The restriction is reduced by grow-ing lattice matched structures on InP, rather than GaAs,substrates. Al0:48In0:52As and In0:53Ga0:47As are bothlattice matched to the InP (Fig. 1.3) and their con-duction band minimum energies are well separated,so that in the InGaAs below the interface betweenthe two compounds a high density of electrons witha very high mobility is formed. Compared to thepHEMTs, these InP-based HEMTs exhibited signifi-cant improvements, were shown to exhibit gain at over200 GHz, and became established as the leading tran-sistor for millimeter-wave low noise applications suchas radar [1.39].

1.3.2 Light Emitting Devices

LEDs and laser diodes exploit the direct band gap semi-conductors to efficiently convert an electric current intophotons of light. Work on light emission from semicon-ductor diodes was carried out in the early decades ofthe twentieth century [1.53], but the start of the mod-ern era of semiconductor optoelectronics traces fromthe demonstration of lasing and LED behavior from p–n junctions in GaAs [1.54, 55] and GaAs1�xPx [1.56].The efficiency of these LEDs was low and the lasers hadlarge threshold currents and only operated at low tem-peratures. A year later, in 1963, Kroemer and Alferovindependently proposed the concept of the double het-erostructure (DH) laser [1.57, 58]. In the DH device,a narrow band gap material was to be sandwiched be-tween layers with a wider gap so that there would besome degree of confinement of carriers in the activelayer. By the end of the decade, DH devices had beenconstructed that exhibited continuous lasing at roomtemperature [1.59, 60]. Alferov’s laser was grown byLPE on a GaAs substrate with a 0:5�m GaAs activelayer confined between 3�m of Al0:25Ga0:75A on ei-ther side. The launch of the compact disk in 1982 sawthis type of device, or at least its offspring, becomingtaken for granted in the households of the world.

One of the major challenges in materials selectionfor heterostructure manufacture has always been avoid-ing the formation of misfit dislocations to relieve thestrain associated with lattice parameter mismatch be-tween the layers. AlxGa1�xAs exhibits a direct bandgap for x< 0:45 and the early success and sustained

Introduction

10 Introduction

dominance of the AlGaAs/GaAs system derive signif-icantly from the very close coincidence of the AlAsand GaAs lattice parameters (5:661 Å and 5:653 Å;Fig. 1.3). This allows relatively thick layers of AlGaAswith reasonably high aluminum content to be grownlattice matched onto GaAs substrates with no misfitdislocation formation. The use of the quaternary al-loy solid solution InxGa1�xAsyP1�y was also suggestedin 1970 [1.59] to offer the independent control of lat-tice parameters and band gaps. Quaternaries based onthree group III elements have since proved very pow-erful tools for lattice matching within heterostructures..AlxGa1�x/0:5In0:5P was found to be almost perfectlylattice matched to GaAs and additionally have a verysimilar thermal expansion coefficient (which is impor-tant to avoid strain evolution when cooling after growthof heteroepitaxial layers at high temperatures). By vary-ing x in this compound, direct band gaps correspondingto light between red and green could be created [1.61].Lasers based on this alloy grown by MOCVD area common choice for the red wavelengths (650 nm)used in DVD reading (Fig. 1.3).

Obtaining lattice matching is not so crucial for lay-ers thinner than the critical thickness for dislocationproduction and can be less of an issue these days be-cause of probably the most important development inthe history of optoelectronic devices, that is, the intro-duction of the quantum well. In some way a quantumwell structure is an evolution of the double heterostruc-ture but with a very much thinner active layer. It is thechosen design for most solid-state light emitting de-vices today. With the accurate control available fromMBE or MOCVD, and following from some earlywork on superlattices [1.62], very thin layers of care-fully controlled composition could be deposited withinheterostructure superlattice stacks. It became possi-ble to grow GaAs layers much less than 10 nm thickwithin AlGaAs–GaAs heterostructures. The carriers inthe GaAs were found to exhibit quantum mechani-cal confinement within the one-dimensional potentialwell [1.63, 64]. Lasing from GaAs/Al0:2Ga0:8As quan-tum wells was reported the following year [1.65], butit was a few years before the emission matched thatachievable from DH lasers of the time [1.66] and thequantum well laser was further advanced to signifi-cantly outperform the competition by researchers in the1980s [1.67].

The introduction of heterostructures with layerthicknesses on the nanometer or atomic scale repre-sents the final stage in scaling down of these devices.Similarly, Brattain and Bardeen’s centimeter-sized tran-sistor has evolved into today’s microprocessors withfar submicron FETs whose gate oxide thicknesses aremeasured in Angstroms. Throughout this evolution,

materials characterization techniques (Chaps. 17–20)have contributed heavily to the progress in our under-standing of electronic materials and these techniquesdeserve a brief detour here. As the dimensions havebeen reduced over the decades, the cross-sectional im-ages of device structures published in the literaturehave progressed from optical microscopy [1.28], toscanning electron microscopy (SEM) images [1.66], totoday’s high- or atomic-resolution transmission elec-tron microscopy (TEM) analysis of ultrathin layers. Thechallenge of obtaining relevant and representative datafrom the smaller volume of material, which is avail-able in higher magnification analysis, has been metduring the past decade or so by widespread adoptionof highly site specific specimen preparation techniquesbased around focused ion beam (FIB) milling. Foreach new material family, understanding of defects andmeasurement of their densities (e.g., by TEM and x-ray topography) have contributed to improvements inquality. Huge improvements in x-ray optics have seenhigh-resolution x-ray diffraction techniques develop tobecome a cornerstone of heterostructure research andproduction quality control [1.68]. Scanning-probe tech-niques such as scanning tunneling and atomic force mi-croscopy have become crucial to the understanding ofMBE and MOCVD growth. Chemically sensitive tech-niques such as secondary ion mass spectroscopy andRutherford backscattering have improved to provide in-formation on doping concentrations and compositionsin layered structures with excellent depth resolution.Chemically sensitive measurements have truly reachedthe atomic scale in the past decade through the ex-ploitation of new generations of aberration-correctedand monochromated transmission electron microscopesand three-dimensional atom probe microscopy (3DAP).The disruptive improvement of spatial resolution en-abled by the new generation TEMs continues to yieldinvaluable insight into atomic-scale material, interface,and device structures, especially when combined withchemically sensitive techniques such as electron energyloss spectroscopy (EELS). Meanwhile, the applicationof 3DAP to semiconducting and insulating materialsand devices provides, for the first time, the holy grail of3-D, atom-by-atom chemically resolved maps. By wayof example, we show an atomic-scale chemical mapof the critical layers in a III-nitride laser diode struc-ture in Fig. 1.4, revealing crucial information on thedistribution of alloy and dopant atoms. The materialscharacterization process remains a critical element ofelectronic and photonic materials research.

A primary commercial driver for semiconductorlaser diode research has been to produce more effec-tive emitters of infrared wavelengths for transmissionof data along optic fibers. Devices based on InP sub-

Perspectives on Electronic and Photonic Materials 1.3 The Compound Semiconductors 11Introd

uction

3.5 nm In0.18Ga0.82N QW

60 40 20 0 –20 –40 –60

100

200

300

400

y (nm) x (nm)

z (nm)

235 nm GaN

20 nm GaN

55 nm GaN

5 nm Al0.2Ga0.8N

100 nm In0.02Ga0.98N

InAl

Fig. 1.4 Reconstructed atom map of the active region ofa III-nitride laser diode structure determined by 3-D atomprobe microscopy (3DAP). Indium atoms are displayed asblack (25% displayed); aluminum atoms are displayed asgray (25% displayed); other atoms are not displayed. (Af-ter [1.69], with permission from AIP Publishing)

strates have proven to be extremely effective because ofits fortuitous lattice parameter match with other III–Valloys which have band gaps corresponding to the low-absorption windows in optic fibers. While remaininglattice matched to InP, the InxGa1�xAsyP1�y quaternarycan exhibit band gaps corresponding to infrared wave-lengths of 1:3 and 1:55�m at which conventional opticfibers absorb the least of the radiation (the absoluteminimum is for 1:55�m). Room temperature contin-uous lasing of 1:1�m radiation was demonstrated fromthe material in 1976 [1.70], and InP-based lasers andphotodiodes have played a key role in the optical com-munications industry since the 1980s [1.39].

A second major driver for laser diode research sincethe 1970s was to produce laser components for read-ing or writing optical storage media. We have alreadymentioned the AlGaAs infrared (�D 780 nm) emittersused to read compact discs (CD) and the AlGaInP red(�D 650 nm) lasers in DVD devices (Fig. 1.3). Aslasers emitting the shorter wavelength became avail-

able, the optical disc’s surface pits (through whichbits of data are stored) could be made smaller andthe storage density increased. The pursuit of shorterwavelength laser diodes for higher density optical stor-age supported extensive development of wide bandgap II–VI compounds, principally ZnSe, for their po-tential in green (�� 520 nm) and blue (�� 440 nm)wavelengths. Laser operation in this part of the vis-ible spectrum proved difficult to realize [1.71] and,although a blue/green laser was demonstrated in theearly 1990s following improvements in the p-dopingof ZnSe [1.72], the devices were prone to rapid dete-rioration during operation and were never commercial-ized. Ultimately, the winning laser diode technologyfor the third generation of optical storage laser wasbased on III-nitride materials following breakthroughsin the 1990s that are introduced in the next section.Blue/violet (�D 405 nm) emitting InxGa1�xN-basedlaser diodes enabled Blu-ray (BD), the third and cur-rent generation of optical storage. CD, DVD, and BDtechnologies were introduced in 1982, 1995, and 2006,respectively, providing data storage per layer of approx-imately 0:7 GB, 4:7 GB, and 25 GB; a growth of datacapacity five times per decade.

It is likely that BD also marks the final phase ofthe consumer optical disk era as network-based datadistribution and storage replace the widespread use ofoptical discs. This is a good time, therefore, to note thatfor over 30 years the rapid advances in home entertain-ment (music, films, and console games) and computing(software distribution and data storage) were enabled inpart by laser diode materials and device technology. Thetechnology race for optical storage laser diodes leavesa legacy of extraordinary laser component technologiesthat inevitably evolve to bring value to new sectors, forexample, the imminent growth of laser diode-based im-age projectors and automotive headlights. Of course,the growth in network data distribution which hascaused the consumer optical disk era to peak is enabledby the laser diodes used in network data communi-cation, a wonderful example of the ebb and flow oftechnology dominance.

1.3.3 The III-Nitrides

In the final section of this chapter, we focus on theIII-nitride materials, such as GaN, InxGa1�xN, andAlyGa1�yN and their devices, which have been subjectto some of the most widespread research and develop-ment among all electronic and photonic materials sincethe mid-1990s. III-nitride LEDs and laser diodes haverevolutionized the lighting, display, and optical storageindustries during the past decade. Electronic devicesbased on these materials are expected to have a strong

Introduction

12 Introduction

impact on electrical power distribution and electric ve-hicles in the next.

The relevance of the InxGa1�xN alloy for visiblelight emitting devices is clear from Fig. 1.3. The InNand GaN direct band gaps correspond to wavelengthsstraddling the visible spectrum, and the alloy poten-tially offers access to all points in-between. The earlycommercially successful InGaN light emitters weremarketed by Nichia Chemical Industries following theresearch work of Shuji Nakamura who demonstratedthe first blue InGaN DH LEDs [1.73], blue InGaNquantum well LEDs, and laser diodes soon after [1.74].Nakamura, together with other pioneers in III-nitridegrowth, Isamu Akasaki, and Hiroshi Amano receivedthe 2014 Nobel Prize for Physics for their contribu-tions to blue LED technology. Meanwhile, the wideband gap and other advantageous properties of GaNand AlyGa1�yN alloys provide new opportunities forhigh-frequency and power transistors for high voltageswitching and deep ultraviolet light emitters.

The development of III-nitride materials and de-vices has much in common with the early researchof other III–V systems. For example, MOCVD andMBE growth technologies were both successfully usedfor III-nitride device growth (MOCVD has becomethe preferred technology for mass manufacturing oflight emitting devices); both electronic and optoelec-tronic device applications were explored from an earlystage; and one of the obstacles limiting early devicedevelopment was achieving sufficiently high p-type car-rier densities. However, in some ways III-nitrides pre-sented new challenges: Although conventional III–Vsall share the same cubic crystallographic structure, thenitrides most readily form in a hexagonal allotrope thathas strong polar and piezoelectric properties; the vastmajority of III-nitride device development and man-ufacturing has used heteroepitaxial growth on highlylattice-mismatched substrates leading to devices withvery high defect densities, in contrast to a reliance onhomoepitaxy or closely lattice matched substrates inpreviously successful III–V systems where much lowerdefect densities are standard; and many III-nitride de-vices have been found to be astonishingly more tolerantto the presence of these crystalline defects than otherIII–Vs. We will focus briefly on two fascinating as-pects of the story of the III-nitrides today: the variety ofsubstrate materials in mainstream use, and the remark-able – and still not fully understood – properties of theInxGa1�xN alloy.

First, the plurality of substrate materials in use forLEDs. Synthesis of large bulk GaN crystals – the ob-vious substrate choice from a technical perspective – ismuch more difficult than for other III–Vs such as GaAsor InP. Methods including ammonothermal growth and

hydride VPE are technically viable routes to producehigh quality GaN substrates today. However, in theearly stages of III-nitride development, there were nolarge area GaN substrates available and heteroepitaxialgrowth of GaN on foreign substrates was the only op-tion for commercial LED production. Even today, GaNsubstrate costs are very high owing to combinationsof high temperature, high pressures, slow growth rates,and expensive precursor materials in the manufacturingprocess.

A consequence of the initial necessity of heteroepi-taxy, and ongoing commercial barriers to the use ofGaN substrates, is that a total of four substrate materialsare in use for commercial production of InGaN-basedblue LEDs: sapphire (’-Al2O3), silicon carbide, silicon,and GaN. Substrate choice is not purely based on tech-nical factors – commercial considerations are strongdrivers – but in the context of conventional compoundsemiconductors, it is extraordinary that four completelydifferent substrates can be used to produce LEDs withapproximately competitive device performance (say,competitive within a factor of two).

Current LED production is dominated by use of sap-phire substrates, with silicon carbide substrates a distantsecond and silicon and GaN substrates just entering themarket in recent years. Sapphire is by no means an idealchoice. First, it is electrically insulating which meansthat electrical contacts cannot be made to the devicethrough the substrate material, thus, requiring LED de-signs with both n- and p-contacts on the same surface.Second, the refractive indices of sapphire and GaN aresignificantly different, which lead to low efficiency ofextraction of light from the LED chip unless specialmeasures are applied, such as the removal of the sub-strate or 3-D patterning of the sapphire surface priorto GaN growth. Third, and perhaps most importantly,sapphire has a large lattice mismatch of approximately16% with GaN [1.75]. This large lattice mismatch isrelieved by the formation of misfit dislocations at theGaN/Al2O3 interface which lead to threading dislo-cations propagating through the GaN into the activelayers of the devices. The key discovery for reducingthe defect densities to tolerable levels was the use ofbuffer layers at the interface with the sapphire [1.75].Dislocation densities of approximately 109 cm�2 wereadequate for the first commercial LEDs and countlessrefinements to buffer layer growth have led to typ-ical threading dislocation densities of approximately107�108 cm�2 in devices today. In principle, siliconcarbide has advantages over sapphire: it can be elec-trically conducting, the substrate can be readily etchedto form chip shapes that improve light extraction ef-ficiency, and it has a smaller lattice mismatch withGaN. However, despite these significant differences,

Perspectives on Electronic and Photonic Materials 1.3 The Compound Semiconductors 13Introd

uction

devices grown on sapphire and silicon carbide have re-mained broadly competitive with one another. Cost andavailability considerations led to sapphire becoming themainstream substrate choice for LEDs during the early2000s.

More recently, silicon and GaN have emerged as al-ternative substrate choices, for quite distinct reasons.Silicon substrates are pursued as a route to reduce LEDmanufacturing costs compared with incumbent produc-tion on sapphire. A key driver is the lower cost ofsilicon wafers, especially in large wafer formats (atleast 150 mm diameter) which can be processed byotherwise unused and fully depreciated wafer process-ing lines previously established for silicon electronicsprocessing. In addition to the lower substrate cost,GaN-on-silicon enables cost reductions to be made inprocessing, packaging, and phosphors. For example,GaN-on-silicon facilitates chip scale packaging (CSP),chip scale optics (CSO), and die level integration. CSPhas been used for silicon electronics for many years andthis technology can now be used for GaN-on-silicon.CSP also reduces the phosphor usage, gives improvedthermal performance and is a route to wafer scalepackaging (WSP). The main technical challenges to de-livering viable device performance and yield are largelattice mismatch (again) and large difference in ther-mal expansion coefficient compared with GaN, whichleads to cracking and bowing during cool down aftergrowth unless careful strain engineering is used duringgrowth [1.76]. In contrast, GaN substrates are substan-tially more expensive than sapphire, and only viable insmaller formats (typically 50�75 mm diameter). How-ever, LEDs grown on GaN substrates benefit from muchlower threading dislocation densities and consequentlydeliver some of the best device efficiencies, especiallyfor small chips operated at high drive currents [1.77].The next few years will reveal if silicon substrates orGaN substrates can substantially displace the incum-bent sapphire and silicon carbide-based production.

Throughout the history of InGaN LED research,there has been ongoing fascination and controversy re-garding the precise mechanism of light emission fromInGaN quantum wells. The original driver behind thiswas the extraordinarily high light emission efficiencyof InGaN quantum wells, despite the presence of dislo-cation densities as high as 109 cm�2. Obtaining similarlight emission efficiencies in other III–V materials typ-ically required dislocation densities to be several ordersof magnitude lower.

For a few years, a popular explanation for the highefficiency of light emission from InGaN quantum wellswas that strong fluctuations in the indium concentrationin the quantum wells created low-energy quantum-dotlike sites that localized electrons and holes and re-

duced the deleterious impact of dislocations on the lightemission process. This explanation was based primar-ily on TEM analysis of InGaN quantum wells whichindicated nanometer-scale regions with greatly higherindium fraction than the remainder of the quantumwell – so called indium clusters. Many attempts weremade to control and optimize this clustering process toincrease light emission efficiency. However, since wedemonstrated that the apparent indium concentrationfluctuations could be a misleading artifact of TEM anal-ysis [1.78, 79], and that high light emission efficiencycould be obtained from InGaN quantum wells with ran-dom indium distribution [1.80], this theory is no longerwidely supported. The origin of localization is now sup-posed to be due to much subtler features of the InGaNquantum wells.

Recently, the controversy concerning InGaN lightemission has evolved to also explaining the phe-nomenon of efficiency droop in InGaN LEDs. Drooprefers to the observed reduction in efficiency of state-of-the-art LEDs when they are operated at high currentdensities. There is great pressure to understand and re-duce droop because of the potential cost benefit fromusing smaller LED chips with higher operating currents(i. e., high current densities). The subject of dozens ofhigh profile publications and review articles over thelast 10 years, many theories have been vigorously pro-posed for lower efficiencies at high current densities:increased Auger recombination of electrons and holes;overflow of electrons from the LED active region; thedelocalization of carriers; and a complex role of large-scale crystalline defects.

Although these important technical debates are stilllargely unresolved, the global impact of InGaN LEDscannot be disputed. This is a good example of how com-mercial success of electronic and photonic devices caneasily occur before full technical understanding of thematerials is established. Twelve years ago, in the intro-duction to the 1st Edition of this handbook, we wrote ofthe potential for III-nitride LEDs to replace traditionallight sources such as incandescent light bulbs, but notedthe need to reduce costs and increase the output powerfrom individual LEDs to reach this goal. These com-mercial and technical barriers were emphatically over-come and the potential is now being realized. Produc-tion costs have been driven down by massive economiesof scale during epitaxial growth, improvements in yield,cheaper device processing, and improved device perfor-mance which mean fewer, smaller LEDs are requiredfor a given application. Tremendous improvements inLED performance – in particular, higher efficiency andhigh power light output – have been delivered throughcountless improvements in all aspects of the LEDtechnology, including the core semiconductor material

Introduction

14 Introduction

fabrication. As a consequence, InGaN blue- or violet-emitting LEDs, combined with phosphors that producewhite light emission, are now cornerstone technologiesin domestic, retail, industrial and outdoor lighting, vehi-cle headlights, and LCD backlighting. These solid-statetechnologies provide large energy savings comparedwith conventional light sources. Every day for the fore-seeable future, we will directly sense – through oureyes – the light generated by these devices. Next gen-eration LED lighting may be even more efficient, byomitting phosphors and generating white light by mix-ing red, yellow, green, and blue LEDs. Such lightingwould be color tunable and could mimic sunlight in-doors. There is increasing evidence that such optimizedlighting would be good for our health, increase produc-tivity at work, and even improve examination results inschools.

In this introductory chapter, we have glimpsed justone sector of the electronic and photonic materials land-scape, focusing on just a few semiconductor families.We have given examples of the complex interdepen-

dence and balance of power between different mate-rials and technologies: the astonishing integration ofmaterials which we take for granted in a budget smart-phone; the switch from germanium to silicon in theearly years of the transistor; the dominance and re-lentless performance improvements in silicon ICs overseveral decades; the laser-diode enabled growth of con-sumer optical storage to its peak; and the co-existenceof dramatically different approaches to make LEDswhich are revolutionizing the way we light the world.There is a complex fusion of science and engineer-ing excellence, ingenious breakthroughs, serendipity,and commercial considerations. But underpinning it allis the need to fundamentally understand and preciselycontrol these fascinating materials. And this is all themore exciting because the pace of progress is so high,and the influence of the materials is so great, thanks tothe electronic and photonic materials being the enablingtechnology in so many aspects of modern life. The de-tailed chapters in the rest of this book are a valuableresource to support further advances in this vital field.

References

1.1 A.H. Wilson: Proc. R. Soc. Lond. A 133, 458 (1931)1.2 A.H. Wilson: Proc. R. Soc. Lond. A 134, 277 (1931)1.3 J. Bardeen, W.H. Brattain: Phys. Rev. 74, 230 (1948)1.4 J. Bardeen: Nobel Lecture, Physics, 1942–1962 (El-

sevier, Amsterdam 1956), http://www.nobelprize.org/nobel_prizes/physics/laureates/1956/shockley-lecture.html

1.5 E. Braun, S. MacDonald: Revolution in Miniature, 2ndedn. (Cambridge Univ. Press, Cambridge 1982)

1.6 W.H. Brattain: Nobel Lecture, Physics, 1942–1962(Elsevier, Amsterdam 1956), http://www.nobelprize.org/nobel_prizes/physics/laureates/1956/brattain-lecture.html

1.7 C.S. Fuller: Phys. Rev. (Ser. 2) 86, 136 (1952)1.8 I. Derick, C.J. Frosh: US Patent, 2802760 (1955)1.9 J. Andrus, W.L. Bond: US Patent, 3122817 (1957)1.10 W.G. Pfann: Trans. Am. Inst. Mech. Eng. 194, 747

(1952)1.11 W.G. Pfann: Zone Melting (Wiley, New York 1958)1.12 G.K. Teal: IEEE Trans. Electron Devices 23, 621 (1976)1.13 G.K. Teal, J.B. Little: Phys. Rev. (Ser. 2) 78, 647 (1950)1.14 G.K. Teal, E. Buehler: Phys. Rev. 87, 190 (1952)1.15 J. Czochralski: Z. Phys. Chem. 92, 219 (1917)1.16 P.H. Keck, M.J.E. Golay: Phys. Rev. 89, 1297 (1953)1.17 H.C. Theurer: US Patent 3 060 123 (1952)1.18 H.C. Theurer: Trans. Am. Inst. Mech. Eng. 206, 1316

(1956)1.19 K.A. Jackson (Ed.): Silicon Devices (Wiley-VCH, Wein-

heim 1998)1.20 W.C. Dash: J. Appl. Phys. 29, 736 (1958)1.21 W.C. Dash: J. Appl. Phys. 30, 459 (1959)1.22 W.C. Dash: J. Appl. Phys. 31, 736 (1960)1.23 G. Zeigler: Z. Naturforsch. 16a, 219 (1961)

1.24 M. Grayson (Ed.): Encyclopedia of SemiconductorTechnology (Wiley, New York 1984) p. 734

1.25 I.M. Ross: Bell Labs Tech. J. 2(4), 3 (1997)1.26 W. Shockley: Bell Syst. Tech. J. 28(4), 435 (1949)1.27 W. Shockley, M. Sparks, G.K. Teal: Phys. Rev. 83, 151

(1951)1.28 M. Tanenbaum, D.E. Thomas: Bell Syst. Tech. J. 35, 1

(1956)1.29 C.M. Melliar-Smith, D.E. Haggan, W.W. Troutman:

Bell Labs Tech. J. 2(4), 15 (1997)1.30 J.A. Oerni: IRE Trans. Electron Devices 7, 178 (1960)1.31 J.S. Kilby: IEEE Trans. Electron Devices 23, 648 (1976)1.32 J.S. Kilby: Nobel Lectures in Physics: 1996–2000

(Imperial College, London 2000), http://www.nobelprize.org/nobel_prizes/physics/laureates/2000/kilby-lecture.html

1.33 D. Kahng, M.M. Atalla: Silicon-silicon dioxide fieldinduced surface devices, Proc. Solid State Res. Conf.,Pittsburgh (1960)

1.34 F.H. Baumann, C.P. Chang, J.L. Grazul, A. Kamgar,C.T. Liu, D.A. Muller: C4.1, Mat. Res. Soc. Symp. 611,1 (2000)

1.35 J.T. Clemens: Bell Labs Tech. J. 2(4), 76 (1997)1.36 G.E. Moore: Electronics 38(8) (1965)1.37 G.E. Moore: IEEE Int. 1, 20 (2003)1.38 H. Kroemer: Nobel Lectures in Physics: 1996–2000

(Imperial College, London 2000), http://www.nobelprize.org/nobel_prizes/physics/laureates/2000/kroemer-lecture.html

1.39 O. Wada, H. Hasegawa (Eds.): InP-Based Materialsand Devices (Wiley, New York 1999)

1.40 C.Y. Chang, F. Kai: GaAs High-Speed Devices (Wiley,New York 1994)

Perspectives on Electronic and Photonic Materials References 15Introd

uction

1.41 A.Y. Cho: J. Vac. Sci. Technol. 8, S31 (1971)1.42 H.M. Manasevit: Appl. Phys. Lett. 12, 156 (1968)1.43 H.J. Welker: IEEE Trans. Electron Devices 23, 664 (1976)1.44 H. Kroemer: RCA Review 18, 332 (1957)1.45 C.A. Mead: Proc. IEEE 54, 307 (1966)1.46 W.W. Hooper, W.I. Lehrer: Proc. IEEE 55, 1237 (1967)1.47 R. van Tuyl, C. Liechti: IEEE Spectrum 14(3), 41 (1977)1.48 R. Dingle, H.L. Störmer, A.C. Gossard, W. Wiegmann:

Appl. Phys. Lett. 33, 665 (1978)1.49 T. Mimura, S. Hiyamizu, T. Fujii, K. Nanbu: Jpn.

J. Appl. Phys. 19, L225 (1980)1.50 T. Mimura: IEEE Trans. Microw. Theory Tech. 50, 780

(2002)1.51 D. Delagebeaudeuf, P. Delescluse, P. Etienne, M. Lav-

iron, M. Chaplart, N.T. Linh: Electron. Lett. 16, 667(1980)

1.52 D. Chattopadhyay, S.K. Sutradhar, B.R. Nag: J. Phys. C14, 891 (1981)

1.53 E.E. Loebner: IEEE Trans. Electron Devices 23, 675(1976)

1.54 R.N. Hall, G.E. Fenner, J.D. Kingsley, T.J. Soltys,R.O. Carlson: Phys. Rev. Lett. 9, 366 (1962)

1.55 M.I. Nathan, W.P. Dumke, G. Burns, F.H. Dill Jr.,G. Lasher: Appl. Phys. Lett. 1, 62 (1962)

1.56 N. Holonyak, M.B. Panish, P.W. Foy, S. Sumski: Appl.Phys. Lett. 1, 82 (1962)

1.57 H. Kroemer: Proc. IEEE 51, 1782 (1963)1.58 Z.I. Alferov: Nobel Lectures in Physics: 1996–2000

(Imperial College Press, London 2000), http://www.nobel.se/physics/laureates/2000/alferov-lecture.html

1.59 Z.I. Alferov, V.M. Andreev, D.Z. Garbuzov,Y.V. Zhilyaev, E.P. Morozov, E.L. Portnoi, V.G. Trofim:Fiz. Tekh. Poluprovodn. 4, 1826 (1970)

1.60 I. Hayashi, G.E. Fenner, J.D. Kingsley, T.J. Soltys,R.O. Carlson: Appl. Phys. Lett. 17, 109 (1970)

1.61 A. Žukauskas: Introduction to Solid-State Lighting(Wiley, New York 2002)

1.62 L. Esaki, R. Tsu: IBM J. Res. Dev. 14, 61 (1970)1.63 L.L. Chang, L. Esaki, R. Tsu: Appl. Phys. Lett. 24, 593

(1974)1.64 R. Dingle, W. Wiegmann, C.H. Henry: Phys. Rev. Lett.

33, 827 (1974)1.65 J.P. van der Ziel, R. Dingle, R.C. Miller, W. Wiegmann,

W.A. Nordland: Appl. Phys. Lett. 26, 463 (1975)1.66 R. Dupuis, P.D. Dapkus, U. Holonyak Jr., E.A. Reack,

R. Chin: Appl. Phys. Lett. 32, 295 (1978)1.67 Z.I. Alferov: Semiconductors 32, 1 (1998)1.68 D.K. Bowen, B.K. Tanner: High Resolution X-ray

Diffractometry and Topography (Taylor Francis, Lon-don 1998)

1.69 S.E. Bennett, T.M. Smeeton, D.W. Saxey,G.D.W. Smith, S.E. Hooper, J. Heffernan,C.J. Humphreys, R.A. Oliver: J. Appl. Phys. 111,053508 (2012)

1.70 J.J. Hsieh, J.A. Rossi, J.P. Donneley: Appl. Phys. Lett.28, 709 (1976)

1.71 H.E. Ruda (Ed.): Widegap II–VI Compounds for Opto-Electronic Applications (Chapman Hall, London 1992)

1.72 M.A. Haase, J. Qiu, J.M. DePuydt, H. Cheng: Appl.Phys. Lett. 59, 1272 (1991)

1.73 S. Nakamura, T. Mukai, M. Senoh: Appl. Phys. Lett.64, 1687 (1994)

1.74 S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa,T. Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto:Jpn. J. Appl. Phys. 35, L74 (1996)

1.75 S. Nakamura, G. Fasol: The Blue Laser Diode (Springer,Berlin 1997)

1.76 D. Zhu, D.J. Wallis, C.J. Humphreys: Rep. Prog. Phys.76, 106501 (2013)

1.77 C.A. Hurni: Appl. Phys. Lett. 106, 031101 (2015)1.78 T.M. Smeeton, M.J. Kappers, J.S. Barnard, M.E. Vick-

ers, C.J. Humphreys: Appl. Phys. Lett. 83, 5419 (2003)1.79 T.M. Smeeton, C.J. Humphreys, J.S. Barnard,

M.J. Kappers: J. Mater. Sci. 41, 2729 (2006)1.80 M.J. Galtrey: J. Appl. Phys. 104, 013524 (2008)