Baca A.G., Ashby C.I.H. Fabrication of GaAs Devices
Подождите немного. Документ загружается.
Introduction to GaAs devices
electronic technology. The device-oriented books describe ideal
operation and do not treat the impact of specific processing choices
on device operation [7–17]. The process-oriented books address
some impacts of processing choices but do not adequately relate
these to subtleties in device performance and reliability. Especially
lacking in any of these books is an adequate treatment of the crit-
ical importance of surfaces and interfaces in GaAs devices and
their correlation with device performance.
This book attempts to address all aspects of GaAs pro-
cessing that deal with GaAs free surfaces and with interfaces
between GaAs and metal contacts or dielectrics. These subjects
are intimately involved in virtually all aspects of device pro-
cessing. We have attempted to provide both fundamental perspect-
ive and practical advice in the areas of cleaning and passivation
(Chapter 3), wet etching and photolithography (Chapter 4), dry
etching (Chapter 5) and GaAs-metal contacts (Chapters 6 and 7).
Chapters 8 and 9 discuss device performance for HBTs and FETs
and how this relates to processing choices. Chapter 10 deals with
special processing issues such as wet oxidation that are especially
important in optoelectronic devices. Chapter 2 provides a variety
of background material about physical concepts, semiconductor
materials and characterisation techniques so that the student does
not have to consult multiple texts to gain perspective while using
this book.
Our main goal is to provide both fundamental and practical
information to aid both the beginning and the practising engineer in
relating device performance, degradation and non-idealities with
processing choices made during device fabrication. Above all,
we have attempted to make this an “apprenticeship in a book”
by presenting as much practical information as possible that we
have learned through many years of processing of GaAs devices
ourselves. To the extent that some of these goals are achieved,
another book on GaAs processing is worthwhile.
1.2 GaAs MATERIALS
Soon after the discovery of the transistor in 1947, gallium arsen-
ide and many other semiconductors were assessed as candidate
materials for electronic devices. GaAs became recognised as a
material with a large electron velocity that might be suitable for
high-speed electronics. It took several decades before this “mater-
ial of the future” made a significant commercial entry in the 1970s,
much to the frustration and delayed ambitions of many of its
practitioners. Now, GaAs is the basis of a several billion dollar
worldwideindustryfor high-frequencyandhigh-speed electronics,
2
Introduction to GaAs devices
which places it at a few percent of the overall semiconductor mar-
ket. Although by market share alone GaAs may not seem to deserve
much attention as a semiconductor material, its importance lies in
the applications it enables. As seen in Section 1.5.2, the applic-
ations that use GaAs electronics are some of the most exciting
around.
Although GaAs materials properties are more easily understood
after assimilating some of the topics in Chapter 2, a brief con-
trast of GaAs to other semiconductors is instructive here. One
important classification is that of elemental versus compound
semiconductors. The main elemental semiconductors are Si and
Ge. These semiconductors are generally easier to grow, purify
and process than compound semiconductors such as GaAs, so
they were used for devices at an earlier date and continue to have
important material-related advantages. Si, the dominant semicon-
ductor commercially, is more defect free, has better mechanical
properties, and is available as larger diameter wafers. Its commer-
cial dominance in electronics is usually ascribed to two important
advantages. The first is its unparalled surfacepassivationwith SiO
2
that leads to unequalled transistors based on surface inversion,
i.e. the MOSFET (metal-oxide-semiconductor field effect tran-
sistor). The second is the unrivalled small size and high packing
density of its transistors and other integrated circuit components
(e.g. trench capacitors). Part of this second factor is the sheer
investment in tools for high-density, high-volume manufactur-
ing that is, in principle, available to compound semiconductor
technologies as well (but not in practice due to smaller wafers
and smaller addressable markets necessary to pay for the capital
investments). The foundational part of the high-density advantage
for Si-based electronics is due to its superior material properties.
High-density devices require small lateral and vertical dimensions
in the device as well as low power density per device. Virtually
every important dimension is easier to reduce in Si devices because
of the unique advantagesof its oxide and the processing advantages
of an elemental semiconductor (i.e. shallower and more flexible
doping).
Compound semiconductors find niches in certain applications
due to other material properties. The most important material prop-
erties are the high mobility (carrier speed in response to an electric
field), light-emission capability, and a large variety of bandgap-
engineering approaches available. These properties derive in large
part from the band structure of semiconductors (Section 2.2.1). In
particular, GaAs has a direct bandgap (Section 2.2.1) that permits
high-efficiency light emission as well as a conduction-band struc-
ture that leads to fast electron conduction. GaAs electronics find
niches in small-sized circuits that require high speed or low power
3
Introduction to GaAs devices
consumption. Soon after the semiconductor laser was discovered
in 1962, the light emitting properties of GaAs gained a new import-
ance and ultimately became a large commercial attribute. GaAs-
based lasers and LEDs share a broad range of important applica-
tions with several other compound semiconductors. Interestingly,
Si, the dominant electronic semiconductor, does not have a direct
bandgap nor does it address light-emitting applications (except as
part of the transmitting and receiving electronic functions).
1.3 TYPES OF GaAs DEVICES
GaAs devices are categorised broadly by whether they are
electronic devices or light-based (photonic) devices. This cat-
egorisation is due to differences in specialisation ranging from
understanding the physics of the device operation to the con-
siderable differences in measurement techniques. Processing of
photonic devices is remarkably similar to electronic device pro-
cessing and many of the topics in this book are equally applied to
both categories of devices.
In this section, the types of GaAs devices will be briefly defined
and introduced for the purpose of referring to the types of applica-
tions of GaAs devices. This introduction will not include details of
the physical principles or the operation of the devices, the subject
of later chapters. Readers without a prior familiarity with semi-
conductor devices may find this chapter more useful after reading
the rest of the book.
1.3.1 Electronic devices
Electronic devices are of three basic types: bipolar transistors, field
effect transistors and diodes. All three types have important com-
mercial applications. The purpose of this introduction is to provide
a level of understanding useful for a discussion of applications of
GaAs devices (Section 1.5).
emitter - n
base - p
collector - n
FIGURE 1.1 A schematic of a
bipolar transistor with electrical
contacts to the emitter, base and
collector.
GaAs applications use heterojunction bipolar transistors
(HBTs), field effect transistors (FETs) and diodes. The bipolar
transistor is illustrated in FIGURE 1.1. The bipolar transistor can
be considered to be two pn junction diodes connected together with
a common middle region. The npn bipolar transistor is the most
common and uses a common p-region. As shown in FIGURE 1.1,
the bipolar transistor has an emitter, a base region and a collector
region, and electrical contacts to each. A heterojunction transistor
uses emitter material with larger bandgap (Section 2.2.1) than the
base material. In an npn transistor, electrons flow from the emit-
ter to the collector regions, passing through the base. The base
4
Introduction to GaAs devices
is made very thin so that few electrons recombine in that region.
The electron flow is modulated by an electrical signal applied to
the base-emitter junction. A small change in the signal applied to
this junction leads to a large current change at the collector. This
amplification of the signal is what makes transistors have more
uses than diodes. Details of the operation of bipolar transistors are
found in Section 9.2.
Field effect transistors are illustrated in FIGURE 1.2. Like the
bipolar transistor, the FET is a three-terminal device with metal
contacts for the source, the gate and the drain. In the field effect
transistor, electrons flow from the source to the drain when a drain
voltage is applied. A voltage signal to the gate can cause a change
in the drain current resulting in amplification. As with the bipolar
transistor, amplification of an input signal occurs for a properly
designed device.
drain
source
gate
channel
FIGURE 1.2 A schematic of a
field effect transistor with electrical
contacts to the source, gate and
drain.
1.3.2 Photonic devices
Fourbroad categoriesof photonic deviceswill be introduced: light-
emitting diodes (LEDs), laser diodes (LDs), photodetectors and
waveguides. As for electronic devices, a rudimentary introduction
of device types is useful for a discussion of applications of GaAs
devices (Section 1.5).
n-region
p-region
FIGURE 1.3 A schematic of a
light-emitting diode with electrical
contacts to the n- and p-regions.
Light-emitting diodes produce light after passing current
through a pn junction in a semiconductor with a direct bandgap.
A cross-section of an LED is shown in FIGURE 1.3. Electrical cur-
rent flows between a top contact and a bottom contact, shown here
at the bottom of the semiconductor. “Bottom” contacts may also
be made to the front side of the wafer (but below the top contact).
The structure of FIGURE 1.3 is generally rotated about the axis for
a circular geometry, but other shapes are also possible. An applied
electrical bias sends electrons from the negative region to the posit-
ive region and holes (Section 2.2.2) in the opposite direction. LEDs
emit photons when electrons and holes in the same area of the semi-
conductor recombine and release the energy of recombination as
a photon of light. The top contact is shown as a cross-section of
opposite ends of a ring which conducts current from the edge of
the cavity (central area between top contacts) to the bottom con-
tact (and spreading in the process). Sometimes instead of circular
ring contacts, planar transparent contacts are used. Because of the
band structure of the semiconductor (Section 2.2.1), the light is
generally monochromatic. GaAs-based LEDs generally emit in
the red and yellow regions of the spectrum. Light is emitted in all
directions in the LED. Many of the photons are absorbed by the
semiconductor or by the package and never contribute to the light
output. Researchers and applications engineers spend much effort
5
Introduction to GaAs devices
trying to find ways to get as much light as possible out of the LED
in order to make the devices more efficient and cost effective.
Laser diodes also have pn junction structures. As in the LED,
photons are generated by passing the current through a pn junc-
tion. However, one establishes an optical cavity by creating mirrors
in the semiconductor so that photons can make multiple passes
through the cavity to enable the amplification required for lasing.
An edge-emitting laser, as illustrated in FIGURE 1.4, will have
cleaved edges to form the mirrors as reflecting edges of the semi-
conductor. A vertical-cavity laser, which is similar in cross-section
to the LED of FIGURE 1.3, will have superlattice mirrors above
and below the pn junction (Section 10.5.1 and FIGURE 10.13).
Like the LED, laser diodes create electron-hole pairs by injecting
minority carriers into an active region where they can recombine
and emit light. Like any laser, the light is very directional and can
have very narrow bandwidths for the emitted light. The wavelength
of the light can be tuned by changing the bandgap or energy levels
in the material of the active region.
n-region
p-region
FIGURE 1.4 A schematic of an
edge-emitting laser with electrical
contacts to the n- and p-regions.
One type of photodetector, the p-i-n diode, may be considered
to be an LED operated in reverse. However, instead of an abrupt
pn junction, an undoped (or i for intrinsic) region with reasonable
thickness is used for photon absorption. Some fraction of photons
with energy above the bandgap of the i-region entering the detector
are absorbed by the semiconductor and create electron-hole pairs.
The electric field across the reverse-biased junction sweeps the
charge out of the i-region to permit its detection.
1.4 A BRIEF HISTORY OF GaAs DEVICES
A brief history of GaAs devices will be presented here so that the
reader can place the devices of interest within a historical context.
For readers completely unfamiliar with GaAs devices, much of
the historical context may be better understood by referring to
this section as other material in the book is digested. The history
of GaAs electronic devices is presented first (because electronic
devices were researched earlier), followed by the history of GaAs
photonic devices.
1.4.1 History of GaAs electronic devices
The transistor was invented in 1947 at Bell Laboratories. This tran-
sistor was called a point contact transistor but operated as a bipolar
transistor. This and other early transistors used a germanium semi-
conductor. This invention opened the door for the exploration
of many other semiconductor materials. Quickly germanium and
6
Introduction to GaAs devices
silicon became the dominant semiconductors of the 1950s, with
Si becoming dominant in the 1960s.
The field effect transistor was invented in 1952. However,
most of the early transistors of any importance in the 1950s and
1960s were still bipolar transistors. Not surprisingly, the ini-
tial interest in GaAs devices centred on bipolar transistors that
would be superior to germanium (and later silicon) transistors
for high-frequency applications. The first GaAs bipolar transistor
with RF performance superior to that of silicon bipolar transist-
ors and approaching that of germanium devices was reported in
1961. Large efforts supported by US government funding aimed at
developing GaAs bipolar transistors with superior high-frequency
performance. Technological limitations forced these early devices
to be based on homojunction analogues to Si and Ge devices, often
with diffusion as the means of creating the pn junctions in bipolar
devices. By 1980 a GaAs bipolar transistor with an operating fre-
quency greater than 1 GHz (unity current gain, Section 8.2.1) had
been achieved using ion implantation. These early attempts at
making high-performance GaAs BJTs were not very successful
and competitive GaAs bipolar transistors would have to wait for
the development of heterostructure devices.
The field effect transistor was first analysed in the early 1950s
by Shockley. In 1966, the first GaAs FETs were developed, both
a metal-semiconductor FET (MESFET) using a Schottky barrier
gate (Section 8.2.1) and a junction field effect transistor (JFET)
with a Zn-diffused region for a pn junction gate. The first MESFET
with GHz capability was demonstrated a year later. By 1970,
MESFETs with 30 GHz performance (unity current gain) had
been demonstrated with 1 μm gates. The 1970s saw a series of
demonstrations of GaAs FET use in a variety of high-frequency
analogue circuit applications such as low-noise amplifiers and
power amplifiers. JFETs based initially on diffusion and later on
ion implantation for pn junction gates never became as widely
used as MESFETs because the pn junction gate has a parasitic
capacitance that limits its performance. The majority of early
MESFET and bipolar structures were grown by vapour phase
epitaxy (VPE), which offered better active layer thickness control
and higher surface quality.
In addition to GaAs transistors, there were two important
microwave diode technologies discovered in the 1960s: the Gunn
and the IMPATT (IMPact Avalanche and Transit Time) diodes.
Apart from challenging physicists to explain the source of their
microwave oscillations, these devices still play an important role
as very-high-frequency power sources.
Two key ingredients of high-performance GaAs electronics are
the semi-insulating substrate for low parasitics and the use of
7
Introduction to GaAs devices
heterostructures in all GaAs devices. The first GaAs crystals grown
by the Czochralski (CZ) method were produced in 1956. The
liquid encapsulated Czochralski (LEC) technique (Section 2.3)
was first applied to GaAs in 1965. High-purity crystal growth
has been enabled by advances in this technique. By the early
1980s, round substrates grown by LEC in 2-inch and 3-inch sizes
were commercially available. Prior to that GaAs crystals had odd
shapes that were not conducive to the development of high-volume
commercial applications.
A historically important development for semi-insulating sub-
strates was the use of Cr doping, first reported in the early 1960s.
Today Cr is no longer needed to produce semi-insulating crystals
(Section 2.3), but in the early days of poor purity and stoi-
chiometry control, it enabled high-frequency devices that would
not otherwise have been possible.
Bringing performance to higher levels in the 1980s and bey-
ond depended primarily on two factors, dimensional scaling and
the emergence of new devices based on heterostructures. Dimen-
sional scaling depended primarily on the development of better
processing techniques and better lithography tools, which largely
leveraged the investments made in Si technology. Devices based
on heterostructures had been proposed long before the materials
growth techniques were able to produce these structures.
In the 1960s, liquid phase epitaxy and vapour phase epitaxy
were the main methods of creating heterostructures. Neither
offered the type of control needed for thin active layers in hetero-
structure transistors. Dimensional control of thin AlGaAs and
GaAs heterostructures was demonstrated by Cho in 1971 using
a newly developed technique of MBE (Section 2.4). By 1975,
high-quality electronic and optical devices were demonstrated.
The history of the heterostructure bipolar transistor is nearly as
old as the transistor itself. In 1948, Shockley outlined the advant-
age of incorporating a heterostructure into a bipolar transistor.
In 1957, Kroemer formulated the basic HBT theory. The first
microwave HBT was reported in 1972. By 1982, the HBT was
used in digital integrated circuits. By 1990, GaAs HBT operation
was pushing past 100 GHz.
E
C
E
V
E
F
2DEG
FIGURE 1.5 An illustration of a
two-dimensional electron gas.
The high electron mobility transistor (HEMT), was first demon-
strated in 1979 by Mimura and co-workers. It was preceded
by the engineering of superlattice heterostructures of GaAs and
AlGaAs by Dingle and co-workers and a single quantum-well
“two-dimensional electron gas” (2DEG) structure. The 2DEG
structure (FIGURE 1.5) uses electron dopant impurities placed
in the higher bandgap material (AlGaAs). The band alignments of
the adjacent GaAs and AlGaAs shifted the free electrons from the
AlGaAs into the GaAs. Hence, electron mobility was improved
8
Introduction to GaAs devices
by transport in this pure, high-quality layer and the separation
between impurity donors and the electron gave rise to the name
“two-dimensional electron gas”.
The next development in HEMT devices was the incorpor-
ation of a strained layer of InGaAs in the channel, termed a
pseudomorphic HEMT (PHEMT). The higher transport of this
strained layer pushed frequency performance towards 100 GHz
by the late 1980s, although GaAs ceded the performance advant-
age to InP-based HEMTs by then. In spite of higher frequency
operation for InP-based devices, GaAs devices’ favourable com-
bination of speed and breakdown characteristics led to the dom-
inance of PHEMTs and HBTs in power amplifier applications
(Section 1.5.2).
1.4.2 History of GaAs photonic devices
Prior to the invention of the solid-state laser in 1962, light emission
from semiconductors was little more than a curiosity. Instantly, the
potential for directed light sources of all types was recognised and
laser research in semiconductors took off. Many early advances
in GaAs lasers came out of the Ioffe Institute, Bell Laborator-
ies and IBM laboratories [7,18]. GaAs homojunction lasers were
demonstrated early on but suffered a number of problems relat-
ing to the lack of confinement. The double heterostructure laser
was conceived theoretically well before the experimental means
of achieving such a device was perfected. The double heterostruc-
ture (DH) was designed to inject electrons and holes into an active
region while the potential barriers of the heterostructures serve to
inhibit escape of the injected carriers and to enhance population
inversion. The heterostructures are also engineered with a differ-
ent dielectric constant to confine the light in the active region and
prevent optical losses. Such a structure proved to be far superior to
homojunction lasers, but was initially thought to be a paper curios-
ity. Little was known about the interface quality between candidate
heterostructure materials and widespread skepticism existed about
whether the ideal heterojunction with a non-defective interface
could be fabricated [18]. Many early researchers considered the
search for the ideal heterojunction very unlikely to succeed and
few groups at that time attempted it.
Since GaAs was known from early physics and electronic invest-
igations to show a favourable combination of properties, the search
for a DH laser centred initially on GaAs for an active region. Its
direct bandgap, low effective mass, high mobility at the conduc-
tion band minimum and sharp optical absorption were among its
favourable properties. The search for compatible heterostructure
materials was initially focused on GaP and AlAs. AlAs, while
9
Introduction to GaAs devices
having a compatible lattice constant, was quickly found to be
chemically unstable. Work then centred on GaAsP and the first
double heterostructure GaAs laser was fabricated in 1966 [19].
However, the difficulty of achieving a lattice matched condition
led to a search for other materials, and shortly thereafter AlGaAs
was found to be a lattice matched material that exhibited the chem-
ical stability that AlAs lacked. The first CW stripe laser using
AlGaAs/GaAs was demonstrated in 1970. It exhibited a lower
lasing threshold than earlier approaches. Thresholds were to con-
tinue improving throughout the 1970s. At this point the evolution
of laser materials research turned to quaternaries. They offered
the flexibility of independently varying the lattice constant and the
bandgap.
These early laser developments were carried out mainly with
liquid-phase epitaxy as the means of material growth. LPE pro-
duced heterostructures with adequate quality interfaces but had
many shortcomings that would limit the further development of
GaAs lasers. The lack of uniformity of the layers caused current
crowding and optical self-focusing. Also, LPE is incapable of
growing thin layers. The thick active regions grown by LPE caused
high losses in the cavity. In the 1970s, two new growth tech-
niques began to have a large impact on GaAs devices. Molecular
beam epitaxy (MBE) and metal organic chemical vapour depos-
ition (MOCVD) brought the ability to grow semiconductor films
to monolayer accuracy with high purity. This control allowed
the growth of thin, uniform active regions which ultimately led
to quantum-well active regions. Quantum-well laser diodes were
made with reduced threshold current and higher gain.
1.5 APPLICATIONS OF GaAs DEVICES
GaAs applications fall into two categories: those based on its light
emitting properties (due to its direct bandgap – see Section 2.2.1)
and those based on its transport properties. A complete discus-
sion of all applications will not be possible, but we will present a
wide enough range of applications to give a good overview. In the
process, we will focus on the more established, high-volume, com-
mercial technologies that represent real, not potential sales. In this
section, the order of presenting electronic and photonic applic-
ations is reversed from the history of the devices (Section 1.4)
because of their history; photonic devices were commercial-
ised earlier than electronic devices while electronic devices were
researched earlier than photonic devices.
Each of these categories may contain consumer, enterprise and
military segments. In the consumer segment, it is understood that
10
Introduction to GaAs devices
products must meet technical requirements, but only become truly
successful by offering low cost. Every OEM (original equipment
manufacturer) knows the competitive pressures of a consumer
product all too well and plans to maintain profit margins by con-
tinually reducing manufacturing costs and receiving regular cost
reductions from their suppliers. The enterprise segment, that which
mainly consists of products for large companies, is also cost sens-
itive, but not nearly so much as the consumer segment, since the
lack of extremely high volume will not necessarily attract the same
range of potential competitors. The least competitive (butstillcom-
petitive) segment of all is the military segment, since successful
product introductions in this segment often require lengthy product
development cycles and long-term customer-supplier relationships
that are not easily upset by new entrants offering better prices. Each
business category offers advantages and disadvantages for sup-
pliers of GaAs devices. The consumer and enterprise categories
offer high growth rates but are extremely competitive. The milit-
ary category offers a stable pricing environment, but relatively low
volume, and it is a business that can act as a buffer during economic
downturns.
1.5.1 Photonic device applications
The biggest application of GaAs devices continues to be LEDs.
LEDs had humble beginnings but have since followed a version
of Moore’s law for several decades (roughly ten times improve-
ment per decade). Light output per watt, light output per lamp
and cost per unit of light have shown a consistent exponential
rate of improvement over three decades [20], although in some
cases with more jumps and pauses than with Si technology. Cumu-
lative improvements have reached 3–4 orders of magnitude over
the first three decades since the introduction of red LEDs and
similar rates over shorter periods for other colours. With con-
tinued innovation, LEDs can continue producing such gains for
another decade or two. Practitioners of the LED art talk about
a revolution in lighting that is under way and has three com-
ponents. First is the recent addition of blue and green colours,
which had proven elusive in traditional III–V materials but have
recently come of age in the mid 1990s in GaN-based materials.
Second is the recent introduction of white LEDs, which either
combine three primary colours or use a near-UV LED with a
phosphor. Third is continuing steady improvement in cost struc-
ture, luminous efficiency and total light output per device. Since
the early 2000s, LEDs have become the preferred component
for traffic lighting applications, portable electronic displays (cell
phones, personal desk assistants, etc.), backlighting applications,
11