Baca A.G., Ashby C.I.H. Fabrication of GaAs Devices

Подождите немного. Документ загружается.

Introduction to GaAs devices

automotive lighting and signage. Some of these applications are

enabled by the high reliability of solid-state devices compared to

traditional light sources (think trafﬁc lights). Others are enabled

because of the need for a single primary colour, which is very

inefﬁcient to produce with a white light source and a ﬁlter, and

the light becomes too hot. Other applications are enabled by

extending the battery life due to higher efﬁciency as well as the

brightness of LEDs (such as portable electronics). Most other

lighting applications are extremely cost sensitive and will adopt

solid-state lighting solutions only when their cost comes down

further. Ultimately, LEDs offer the potential to save billions of

dollars in electricity if theoretical efﬁciencies can be practically

attained.

Another photonic device is the edge-emitting laser. Edge-

emitting lasers have applications as diverse as telecommunic-

ations, data storage, materials processing, medical surgeries,

spectroscopy and sensors. In telecommunications, the laser is used

to transmit encoded data along an optical ﬁbre to carry telephone

calls along with digital data, initially for long-haul communica-

tions in the mid 1980s, but recently for all areas of extremely high

trafﬁc or bandwidth such as metropolitan areas. Many telephone

calls and other sources of data are combined electronically (multi-

plexed) and sent along an optical ﬁbre using the output of a single

laser.

The telecommunication industry is highly competitive and the

GaAs (or other compound semiconductor) devices must enable

a lower overall system cost in order to be widely adopted. Fibre-

optic lines are more costly to put in place compared to twisted-pair

copper lines (the legacy of voice technology), but they are cost

effective if enough trafﬁc is present. Their capacity is also more

easily expanded by sending extra wavelengths (more lasers per

ﬁbre) of light down the same ﬁbre at the same time by a technique

called wavelength division multiplexing (WDM). To minimise

cost, one should minimise the use of electronic repeaters as much

as possible.

When sending light down an optical ﬁbre two things happen that

cause a loss of the signal. One is that some of the light gets lost

along each length of ﬁbre, mainly by absorption. The other prob-

lem arises because different wavelengths travel at different speeds

(dispersion) through the optical ﬁbre medium. This effect causes

spatial separation of the different wavelengths along the length of

the ﬁbre that will eventually result in loss of the ability to decodethe

signal. Minimising both of these problems is largely the respons-

ibility of the optical ﬁbre makers, who try to reduce impurities

in the ﬁbre and engineer it for minimum dispersion. By sending

light at the minimum in dispersion (1.3 μm), or at the minimum of

Introduction to GaAs devices

loss (1.55 μm), maximum transmission distance can be obtained.

GaAs-based lasers do not reach such long wavelengths (up to

1.1 μm is possible using InGaAs quantum wells strain-matched to

GaAs) and initially had greater reliability problems, so the closely

related InP-based lasers became the mainstay of ﬁbre-optic com-

munications. However, GaAs-based lasers acquired an important

role in ﬁbre-optic communications in the 1990s in all-optical

repeaters.

When the optical signal gets weak enough, it must be ampliﬁed

or regenerated. In the early days, this meant converting the light

to an electrical signal, regenerating (reshaping pulses to their ori-

ginal form) and amplifying the signal, and driving a laser (or in

very-high-frequency systems, driving a laser modulator) to again

transmit the signal down the optical ﬁbre. In the 1990s, a means of

optical ampliﬁcation was introduced where the weak optical signal

was sent through a small section of Er-doped ﬁbre along with a CW

laser source to pump the Er-doped ﬁbre (through Er absorption at

1400 or 1100 nm). After passing through the Er-doped ﬁbre, the

ampliﬁed signal is reintroduced into the standard ﬁbre-optic line.

InP-based lasers can be used to amplify at 1400 nm or GaAs-based

lasers (with InGaAs active regions) can be used at 1100 nm. The

GaAs-based lasers offer lower noise and have garnered a signiﬁc-

ant part of this market. Er-doped ﬁbre ampliﬁers greatly reduced

the cost of transmitting data over optical ﬁbres and enabled the

adoption of WDM techniques to greatly expand the capacity of

ﬁbre-optic systems without having to dig up streets and lay new

cable lines.

Another large category of laser use is in materials processing in

industrial applications [21]. These applications include precision

cutting, welding, shaping and cleaning. The precision nature of

the cutting as well as the ﬂexibility in the laser placement have

led to expanding usage of lasers in cutting and welding. Although

the majority of these applications employ other high-power lasers

such as CO

lasers, semiconductor lasers have gained a small part

of this market as well and will continue to ﬁnd new applications

as the maximum output power levels of GaAs lasers are increased

over time. One notable advance in solid-state lasers such as yttrium

aluminium garnet (YAG) lasers has been the use of semiconductor

diode lasers to pump the YAG laser rod. YAGlaserswere ﬁrst intro-

duced with ﬂashlamps for optical pumping. Since incorporating

GaAs semiconductor lasers, they have become smaller and more

reliable (ﬂashlamps were notoriously short-lived). These diode-

pumped solid-state lasers, which now operate in excess of 100 W

and should continue to increase in power output with time, have

gradually captured a signiﬁcant part of the laser cutting and related

markets.

Introduction to GaAs devices

GaAs edge-emitting lasers are also used in large quantities

for CD/DVD players and data storage. Read-only applications

have been around for nearly 20 years and use GaAs lasers at

wavelengths 830 and 780 nm. Once reliable output power levels

were pushed beyond 30 mW, read/write applications became feas-

ible. Read/write applications initially used the 830 and 780 nm

lasers and then later 650–680 nm to increase data storage densities.

These applications will soon use blue GaN-based lasers.

The vertical-cavity surface-emitting laser (VCSEL) is another

important photonic device. Since edge-emitting lasers emit paral-

lel to the plane of the wafer along a cleaved edge, they are generally

not testable until they have been packaged. This unfortunate real-

ity makes edge-emitting lasers very costly to produce if the yields

are not high because package and assembly costs are incurred in

discarded parts. VCSELs, on the other hand, are testable at wafer

level, since lasing can be characterised normal to the plane of the

wafer. Bad dice are marked and discarded without incurring the

cost of assembly and packaging. VCSEL processing is also sim-

pler and cheaper than that for edge-emitting lasers. VCSELs have

other desirable properties as well. They can be made into arrays,

effectively building up speed (bandwidth) through combining par-

allel paths rather than building single, serial, high speed devices.

Short-wavelength VCSELs (usually less than 1100 nm) are a less

mature technology than edge emitters, but they are attractive for

certain short-length data transmission applications. VCSELs may

soon be suitable for replacing edge-emitting lasers in long-haul

applications through progress in VCSEL research at 1.3 μm and

longer wavelengths.

VCSELs are predominantly GaAs-based devices at present. In

order to build high-reﬂectivity mirrors above and below the active

region, one needs lattice-matched materials with high contrast in

the index of refraction. This allows high-reﬂectivity mirrors (dis-

tributed Bragg reﬂectors) to be made from a large number of pairs

of alternating planes made from the high- and low-index ﬁlms. For

GaAs-based devices, GaAs, AlAs and their alloys provide both a

lattice match and a high-index contrast, but these are lacking for

devices lattice matched to InP.

One of the early VCSEL applications that established their com-

mercial viability was as the ﬁbre channel standard for storage

area networks (SAN), a method of allowing greater scalabil-

ity in data storage by the addition of storage elements directly

to a network rather than to particular servers. Data is trans-

ferred between the storage network elements by ﬁbre-optic lines.

More recently, VCSELs are being adopted for all kinds of

short-to-intermediate-range computer and network connectivity

applications.

Introduction to GaAs devices

Other photonic devices are also used in communications applic-

ations as well as a myriad of niche applications. These devices

include photonic waveguides and related circuits, photodiodes and

optical modulators.

1.5.2 Electronic device applications

GaAs electronic devices have historically served in a myriad of

applications as well, but these have been winnowed down in

recent years to a few very-high-volume applications. The applica-

tions of greatest importance today are cellular phone technology,

ﬁbre-optic communications, electronic test equipment and military

applications.

The cell phone market comprises the cell phone handset as well

as equipment infrastructure segments. Users with handsets ori-

ginate calls within a calling area and the handset communicates

with basestations built by the wireless service provider. An eco-

nomic tradeoff exists in basestation placement versus the power

requirements in the handset and its call quality. Obviously, fewer

basestations built further apart will result in a less costly network,

at least in the initial buildout. However, this scenario will result

in greater requirements on the power-transmitting capability of

the cell phone, which create a variety of difﬁculties, not the least

of which is that its battery is small and discharges quickly. Cell

phone OEMs like to differentiate their handsets on the basis of

premium features and factors such as small size and length of

talk time between battery charges. The cell phone power ampliﬁer

(PA) is the most power-hungry component in the handset because

it is required to transmit up to several miles. GaAs technology

is very competitive in meeting the technical requirements for cell

phone PAs and has achieved a large market share in this tech-

nology by meeting cost targets for these parts (not always easy

with demanding OEMs!). InP-based technology, in contrast, can

achieve higher power-added efﬁciency PAs than GaAs, but it is not

yet able to meet cost targets. Si and SiGe bipolar technology are

generally less desirable because of lower power-added efﬁciency

and are only attractive in the lowest cost segment of the market.

Apart from the PA, GaAs devices can be used in other parts

of a wireless communications system, but they have less com-

pelling advantages and must compete based on cost. All digital

functions have always been more suitable for Si CMOS (e.g.

digital signal processors). Even among other mixed-signal com-

ponents, there are serious competitive pressures. Functions such

as switches, receivers and mixers can often be implemented with

better technical speciﬁcations in GaAs, but the differences com-

pared with Si may not be compelling enough to gain design wins.

Introduction to GaAs devices

Besides cost advantages, Si offers increasing (mixed-signal) integ-

ration and even promotes the long-term idea of a system on a

chip (digital/mixed-signal/analogue). GaAs competes by trying

to integrate other functions such as switches into modules. For

example, there are three worldwide standards for cell phones, each

with a different frequency and requiring different PAs. One recent

trend is to make cell phones that can be used on all three types of

network (called tri-band phones) and integrate all three types of

PA with GaAs devices that switch between each band.

Cell phone applications represent the most prominent but not

the only type of application of wireless communications systems.

The frequency spectrum is allocated for different purposes by the

Federal Communications Commission (FCC) in the United States

and other regulatory bodies in other countries. Cell phone com-

munications have been allocated frequency bands in the 1–2 GHz

range for different wireless protocols that are implemented by

wireless communications providers. We will discuss some of the

applications that have speciﬁc portions of the spectrum allocated

(examples are from the United States; other countries have dif-

ferent ways of allocating the frequency spectrum) as examples

of other addressable markets. One important frequency band is

at 2.4 GHz. Below a certain power level that does not interfere

with the licensed use, anyone can use this part of the spectrum.

Wireless industry groups have developed standards for high-speed,

local-area data transmission. These standards have resulted in

the adoption of wireless networking cards for computers that

are now widely used for wireless broadband home networks and

wireless “hot spots” at certain commercial locations such as air-

ports, hotels and coffee shops. The same frequency band is used

for a “bluetooth” standard that allows almost any instrument or

appliance to receive and send wireless communications. These

wireless technologies are similar in many ways to cell phone tech-

nology, with one main difference being the much lower power

levels at which they transmit. Other examples of commercial wire-

less spectrum allocations are the wireless local-area-network band

near 5 GHz, and the wireless point-to-point band near 30 GHz.

The former is similar to the 2.4 GHz wireless data transmission,

while the latter was conceived as an alternative way of provid-

ing telecommunications carriers “last mile” access to the end user

and spurring competition among incumbent wireline telephone

companies. Not all of these technologies have developed or will

develop in the way they were envisioned when the spectrum was

initially allocated. These technologies represent opportunities for

GaAs technologies, particularly with co-integration of multiple

applications (the cell phone-PDA-satellite radio-computer), but

they are more cost-sensitive than cell phone PA technology.

Introduction to GaAs devices

transmitter

optical

amplifier

receiver

FIGURE 1.6 A schematic of a ﬁbre-optic communications system.

The cell phone infrastructure market uses GaAs PAs for slightly

different reasons than the handset market. Here, linearity is the

key. In order to process as many calls as possible within a given

frequency band, each call is assigned a certain part of the frequency

spectrum. The details of how this is done and by what modulation

scheme is beyond the scope of this book, but sufﬁce it to say that

the speciﬁcations of the PA become more stringent. The base-

station broadcasts calls within its area. It is important that any

distortion in the signal (such as intermodulation distortion) not

be interpreted as data from another frequency band. In order to

keep intermodulation distortion to an acceptable level, extremely

high linearity is required of the power ampliﬁers. Again, GaAs

devices compete well in terms of price and technical speciﬁcations,

but a silicon device called LDMOS also performs well. Future

competition from GaN PAs is also likely.

Both of these wireless technology segments are highly com-

petitive. GaAs technology, although superior in many technical

respects, is continually challenged by Si, SiGe and Si LDMOS

technologies, both in terms of pricing and in attempts to close the

technology gap.

Fibre-optic communications utilise GaAs devices for some of

the components used in high-bit-rate electronics. A ﬁbre-optic

system schematic is shown in FIGURE 1.6. A communication

signal is originated in the transmitter and converted from an elec-

trical signal to a stream of bits through a series of electronic circuits

ending with the high-speed multiplexer. The stream of bits is sent

(within the transmitter) to the laser driver, which modulates the

laser. In cases where the laser is not fast enough to be modu-

lated directly, a modulator driver is used. The modulated light

travels down an optical ﬁbre until the signal becomes attenuated

and needs ampliﬁcation or regeneration(seeSection 1.5.1 for ﬁbre-

optic ampliﬁcation). After a few rounds of ampliﬁcation without

regeneration, the signal must be regenerated. The signal exits the

ﬁbre and is passed into a photodiode of the receiver. It is then amp-

liﬁed with a transimpedance ampliﬁer. After the transimpedance

ampliﬁer, the signal will be reconstructed with a clock and data

recovery circuit. If the signal is at its ﬁnal destination, it will then

go to a demultiplexer and be converted from a stream of bits to the

electrical signal representing the communication. If it is not at its

Introduction to GaAs devices

ﬁnal destination, it will go to a laser driver or modulator driver to be

converted back to light and sent on its way down the optical ﬁbre.

Fibre-optic communications systems were ﬁrst ﬁelded in the

1980s for terrestrial long-haul and transoceanic communications.

They were then gradually adopted for many other applications

where high capacity or high bandwidth was required. InP-based

devices have long been used for the lasers and photodiodes. Until

recently, the highest bit rate optical communications systems used

GaAs devices for multiplexers, demultiplexers, transimpedance

ampliﬁers, clock and data recovery circuits, and laser (and mod-

ulator) drivers. Si ICs were used for other digital functions and

lower-bit-rate systems. However, SiGe bipolar and even CMOS

circuits are now readily available to build most of these circuits

even in the high-bit-rate systems and GaAs devices have been

relegated to a minor role in all but laser and modulator driver

circuits. The driver circuits appear likely to remain the domain of

GaAs-based or InP-based circuits in high-bit-rate systems because

they require higher operating voltages than are achievable with

CMOS or SiGe when clock rates are at 10 Gbit/s or above. Like-

wise, many other applications of high-speed digital logic circuits

were once the domain of GaAs technology (even up to 100 k gates

of logic). Now the speeds of Si technology have caught up with

GaAs. Future, higher speed applications may evolve towards InP

devices.

Military applications utilise a wide range of GaAs circuits at a

wide range of frequencies. These include radars, communications

systems, electronic warfare (EW) circuits and others. Because mil-

itary systems operate worldwide and on irregular assignments,

they do not deal with a well-deﬁned spectrum allocation the

way commercial communications systems do. Therefore, milit-

ary systems design and build their radar systems around a broad

range of frequencies from below 1 GHz at the low end to low

mm-wave frequencies (near 30 GHz) at the high end. Radar and

EW applications have higher bandwidth requirements than wire-

less communications systems. They are also qualitatively different

since radars are an imaging technology. The wide-ranging frequen-

cies, high bandwidth requirements and challenging performance

speciﬁcations have led to widespread adoption of analogue and

microwave GaAs and InP-based technologies in preference to

Si devices. This trend is likely to remain for the foreseeable

future.

Other applications of GaAs (or InP) devices include collision-

avoidance radar, other commercial radars, space solar cells,

radiation-hard electronics, space satellite uplinks and downlinks,

space communications systems, ultra-low-noise receivers and

certain niche communications components such as oscillators.

Introduction to GaAs devices

REFERENCES

[1] S. Mahajan, K.S. Sree Harsa [Principles of Growth and Processing

of Semiconductors (WCB McGraw-Hill, Boston, MA, USA, 1999)]

[2] M. Madau [Fundamentals of Microfabrication (CRC Press, Boca Raton,

FL, USA, 1997)]

[3] S.K. Ghandi [VLSI Fabrication Principles: Silicon and Gallium Arsenide,

Second Edition (John Wiley and Sons, New York, 1994)]

[4] S. Wolf, R.N. Tauber [Silicon Processing for the VLSI Era, Volume 1 –

Process Technology (Lattice Press, Sunset Beach, CA, USA, 1986)]

[5] S.J. Pearton, C.R. Abernathy, F. Ren [Topics in Growth and Device

Processing of III-V Semiconductors (World Scientiﬁc, Singapore, 1996)]

[6] R. Williams [Modern GaAs Processing Methods, Second Edition (Artech

House, Boston, 1990)]

[7] H.C. Casey Jr., M.B. Panish [Heterostructure Lasers (Academic Press,

London, 1978)]

[8] S.M. Sze [Semiconductor Devices: Physics and Technology (John Wiley

and Sons, New York, 1985)]

[9] W. Lui [Fundamentals of III-V Devices: HBTs, MESFETs, and HFETs/

HEMTs (John Wiley and Sons, New York, 1999)]

[10] S.L. Chuang [Physics of Optoelectronic Devices (John Wiley and Sons,

New York, 1995)]

[11] S.M. Sze (Ed.) [High-Speed Semiconductor Devices (John Wiley and

Sons, New York, 1990)]

[12] M. Shur [GaAs Devices and Circuits (Plenum Press, New York, 1987)]

[13] C.Y. Chang, F. Kai [GaAs High-Speed Devices: Physics, Technology, and

Circuit Applications (John Wiley and Sons, New York, 1994)]

[14] L.A. Coldren, S.W. Corozine [Diode Lasers and Photonic Integrated

Circuits (John Wiley and Sons, New York, 1995)]

[15] J.T. Verdeyen [Laser Electronics (Prentice-Hall, New Jersey, 1981)]

[16] W.R. Wisseman, W.R. Frensley [in VLSI Electronics Microstruc-

ture Science vol.11, GaAs Microelectronics Eds. N.G. Einspruch,

W.R. Wissman (Academic Press, Orlando, FL, USA, 1985) p.1]

[17] G.P. Agrawal, N.K. Dutta [Semiconductor Lasers (Kluwer Academic

Publishers, 1993)]

[18] Z.I. Alferov [Semiconductors (USA) vol.32 (1998) p.1]

[19] Z.I. Alferov et al. [Lov. Phys. Solid Atate vol.9 (1967) p.208 ]

[20] A. Bergh, G. Craford, A. Duggal, R. Haitz [Phys. Today (USA) vol.54

(2001) p.42]

[21] D.F. Welch [IEEE J. Sel. Top. Quantum. Electron. (USA) vol.6 (2000)

p.1470]

Chapter 2

Semiconductor properties, growth,

characterisation and processing techniques

Chapter scope p.21

Semiconductor properties p.21

Energy levels and band structure in

semiconductors p.22

Charged carriers in

semiconductors p.27

Carrier transport and continuity

equations p.32

Bulk crystal growth p.33

Methods of crystal growth p.34

Substrate properties and device

requirements p.36

Epitaxy p.38

Molecular beam epitaxy p.40

Metal-organic chemical vapour

deposition p.43

Material characterisation p.46

Light-based techniques p.47

Electron-beam techniques p.49

Ionic techniques p.53

Electrical characterisation p.56

Processing techniques p.60

Back end processing and

analysis p.60

Backside processing, die

separation and packaging p.61

Reliability p.64

Failure analysis p.66

Conclusion p.66

References p.67

2.1 CHAPTER SCOPE

The purpose of this chapter is to provide a concise overview of a

lot of different subject matter that forms the basis of understanding

semiconductors, semiconductor devices and start-to-end product

realisation using them. Much of this material is considered a pre-

requisite to understanding devices in traditional learning settings.

To master the prerequisite material the reader will need formal

courses or in-depth texts for self study. The authors recognise that

not all readers may have become well schooled in these subjects

prior to learning GaAs processing techniques, in part because of

their breadth of scope and of the theoretical character. Hopefully,

these readers will beneﬁt at this time from simpliﬁed explanations

of a number of broad subject areas and from the references to

sources with more complete coverage. These “layman’s” intro-

ductions should help readers place the processing techniques and

the necessary multidisciplinary topics quickly into proper context

within the more fundamental subjects.

The basics of semiconductors and their properties [1–5]

most useful in semiconductor devices are given in Section 2.2.

Section 2.3 introduces the basic techniques for GaAs crystal

growth. Section 2.4 presents the main methods of epitaxial growth,

metal-organic chemical vapour deposition and molecular beam

epitaxy. Section 2.5 introduces some of the more important

semiconductor characterisation techniques. In Section 2.6, some

processing techniques that are common to many of the later

chapters are presented. Specialised processing techniques and

equipment are introduced in later chapters as the need arises.

2.2 SEMICONDUCTOR PROPERTIES

The electrical properties of solids allow their classiﬁcation as con-

ductors, insulators or semiconductors. The electrical properties of