See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/229003908

GaAs-based high-frequency and high-speed devices

Article · February 2003

CITATIONS READS

15 1,101

2 authors:

Masaaki Kuzuhara Shinichi Tanaka

University of Fukui Shibaura Institute of Technology
248 PUBLICATIONS 3,556 CITATIONS 26 PUBLICATIONS 218 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Masaaki Kuzuhara on 11 February 2016.

The user has requested enhancement of the downloaded file.

GaAs-based high-frequency and
high-speed devices
Masaaki KUZUHARA and Shinichi TANAKA

The explosive growth of new communi- The first is the inherent superiority of the The second factor is the solid progress
cation technologies such as the internet and electronic properties of GaAs materials. For that has been made by industrial process
mobile phones has expanded the applica- the fabrication of high-speed electronic devic- technology. Process technology can be broad-
tions of high-speed electronic devices based es, these advantages include the following: 1) ly classified into crystal growth technology
on compound semiconductors. In particular, high electron mobility, 2) high electron drift and material processing technology. With re-
the development of structures with supe- velocity, 3) stable Schottky barrier height at gard to the former, it is impossible to ignore
rior electronic properties and advanced mass- room temperature, 4) the existence of ther- the progress made in the use of larger wafer
production process technology has resulted mally stable semi-insulating substrates, and 5) sizes on GaAs single-crystal substrates. The
in high-speed GaAs-based transistors that the ability to form a variety of heterojunctions standard diameter of GaAs wafers is currently
are making an important contribution to the on these materials. changing from four to six inches. To accom-
improved performance of the latest mobile Advantages 1) and 2) are the basic pre- modate this increased substrate size, epitaxial
phone handsets and base stations. In this pa- requisites for high-speed and high-frequency growth techniques are also being developed
per we describe how the technology behind operation. In particular, in the latest devices for processing multiple 6-inch wafers simul-
such devices has developed so far, and out- for low-voltage mobile terminal applications, taneously, and it is also becoming possible to
line the way in which GaAs-based high-speed a competitive edge can often be obtained by control layer thicknesses to within 1 nm.
electronic devices are currently implemented. reducing the parasitic resistance as much as On the other hand, in material processing
Finally we review the prospects of cultivating possible. As a result, a high electron mobility technology, dry-etching techniques are being
new markets for such devices in the future. is especially important. developed with superior reproducibility and
Keywords: gallium arsenide transistor, Advantage 3) may seem rather mundane uniformity and with low damage by com-
high-power device, metal-semi- at first, but without a stable Schottky bar- bining anisotropic etching with the selective
conductor field-effect transis- rier height it would have been impossible to etching properties of heterojunction materi-
tor, heterojunction field-effect contemplate the development of the modern als. The use of heterojunctions is important,
transistor, heterojunction bipo- field-effect transistor (FET). Advantage 4)— not only for obtaining new functions from
lar transistor, mobile communi- the existence of thermally stable semi-insulat- their physical properties but also because
cation, microwave, millimeter- ing substrates—is a property that can be tak- they provide a means of improving the di-
wave en for granted with GaAs. However, silicon (Si) mensional controllability of crystal processing.
substrates do not exhibit this property, and The secret to success in the mass production
1. Introduction at present it is only possible to provide them of high-performance devices is hidden in the
High-speed electronic devices based on with equivalent characteristics through the properties of heterojunctions.
compound semiconductors are used in an use of advanced process technologies such as In this paper we will describe the tran-
ever-widening range of fields, such as ampli- SIMOX (Separation by Implanted OXygen). In sitions that have so far been made in the
fier devices in high-frequency radio equip- this respect it seems that GaAs materials are technology and the current status of its
ment (e.g., mobile communication terminals) far better suited to high-frequency applica- implementation, with particular reference
and switching elements in high-speed digital tions. Advantage 5)—the ability to form a va- to mobile communication applications and
circuits. Among such devices, the best cost/ riety of heterojunctions—provides designers GaAs-based high-speed electronic devices.
performance ratios are achieved by GaAs- with a greater degree of latitude in the types We will also describe the latest techniques
based transistors, which are establishing a and combinations of materials that can be that are currently under development. We
firm position in the market for mass-pro- used to configure devices. This facilitates the will then describe the development trends in
duced devices that help to meet the needs of development of GaAs device families tailored device technology for third-generation (3G)
the modern telecommunication society. It is to the requirements of a diverse range of mobile phones—for which services were
worthwhile studying how these devices have products with different functions. Heterojunc- started in 2001—by focusing separately on
managed to attain such a status. There are at tions that can typically be formed on GaAs their application to mobile terminals and base
least two contributing factors, which are as substrates include AlGaAs/GaAs and InGaP/ stations. Finally, we will describe some other
follows: GaAs. potential applications for this technology be-

Photonic and Wireless Devices Research Laboratories, NEC Corporation (2-9-1 Seiran, Otsu 520-0833)
e-mail : [email protected]

4 JSAP International No.7 (January 2003)

sides mobile communications. threshold voltage may
result in a gate bias of 0.6 4

Electron Density (1018cm-3)

2. Application to Mobile Phone -3V or less, which is in- (a) uniformly doped
Vgs=0.5V
Terminals convenient to prepare 0.4 3

Energy (eV)
2.1 Field-effect transistors when operating with a
0.2 2
2.1.1 Double-doped/double hetero struc- 3 V supply. Also, even
tures when operating at 3 V,
0.0 1
Over the last few years, mobile phone a gate breakdown volt-
terminals have evolved into pocket-sized de- age of at least 10 V is -0.2 0
vices weighing no more than 100 g and oc- generally required for
cupying no more than 100 cm3. This level of stable operation, and 0.6 4

Electron Density (1018cm-3)

miniaturization has been achieved by reduc- conventional MES struc-
(b) pulse-doped
0.4 Vgs=0.5V 3

Energy (eV)
ing the voltage of the battery power sources. tures are limited in this
Until the mid-1990s, mobile phones were respect.
0.2 2
operated either at 5.8 V (five nickel-cadmium On the other hand,
(NiCd) or nickel monohydride (NiMH) cells) or w i t h h e t e ro j u n c t i o n 0.0 1
at 4.6 V (four of these cells instead of five), FETs it is possible to in-
but due to the wide availability of lithium ion crease the drain current -0.2 0
secondary cells and the development of pow- without causing much 0 20 40 60 80 100
er transistors capable of operating with a 3 V change to the threshold
Distance from Surface (nm)
supply, the use of 3 V power sources has now voltage or gate break-
become the de facto industry standard. Also, down voltage. The opti-
Fig. 1: The distribution of conduction band energy and electron concentration
since transistors have been developed with mal structure for meet- across double-doped/double heterojunctions. (a) Uniformly doped structure. (b)
Pulse-doped structure. In the uniformly doped structure, it can be seen that elec-
structures that make them suitable for high- ing this requirement is
trons spread out into the AlGaAs electron supply layer.
efficiency operation, subsequent technical a double-doped/double
reforms have made it possible to achieve sub- heterostructure.1,2) This
stantial improvements to the talk time and structure is characterized by having a channel structure, and a saturation-drain voltage (knee
standby time. layer of GaAs or InGaAs sandwiched between voltage) about 0.10–0.15 V lower. These ben-
But to produce power transistors capable layers of n-AlGaAs situated above and be- efits are due to the use of an undoped Al-
of delivering output signals of the order of low it. By striking a design balance between GaAs Schottky layer at the surface, whereby 1)
one watt, the switch to low-voltage 3 V pow- the upper and lower doped layers, the gate the injection of electrons from the gate metal
er sources means that these devices must be breakdown voltage and the maximum drain to the semiconductor is suppressed, and 2)
operated with higher currents than before. current can be controlled independently. the parasitic parallel conduction of electrons
The Japanese domestic technical standard for Figures 1(a) and 1(b) show the conduc- in the AlGaAs layer (parasitic MESFET effect)
personal digital cellular communication (PDC) tion band energy distribution and electron is suppressed.
requires mobile phone terminals to produce density distribution in two types of double- When an FET with a pulse-doped struc-
a microwave output power of at least 1 W. doped/double heterostructure. In the uni- ture was used for power amplification (AB-
According to simple AC theory, the current formly doped structure (Fig. 1(a)), a gate class operation) at 950 MHz with a 3 V pow-
drawn by the transistor is about 1.3 A for a electrode directly contacts with an n-AlGaAs er supply, it was found to have a saturated
3 V bias if the transistor’s internal resistance Schottky layer doped uniformly with impu- output power of 1.4 W, a power-added effi-
(on-resistance) is ignored. In practice, since rities (e.g., n =1.0×10 18 cm −3, thickness 34 ciency of 60%, and a linear gain of 12.7 dB.3)
the presence of on-resistance further reduces nm). On the other hand, in the pulse-doped This report was the first to demonstrate the
the amplifier’s power conversion efficiency, structure (Fig. 1(b)), there is a thin layer of benefits of using double-doped/double het-
the required current is even higher. n-AlGaAs doped with a high concentration erojunction FETs as power amplifier transis-
With the aim of achieving optimal op- of n-type impurities (e.g., n=4.5×1018 cm−3, tors that operate efficiently at low voltages.
eration with a 3 V supply, we first looked into thickness 9 nm) below an undoped AlGaAs In 1993, analog mobile phone systems
increasing the maximum drain current den- Schottky layer (e.g., thickness 26 nm). began to be replaced by digital systems that
sity per millimeter of gate width in the FETs. 2.1.2 Depletion-type heterojunction FETs are sensitive to waveform distortion. To sat-
For this purpose it is effective to increase the The results of evaluating the DC and isfy Japan’s domestic PDC technical standard
electron density present in the FET channel. high-frequency characteristics of prototypes with a 3 V power supply, it is thus necessary
However, in a conventional MES (metal–semi- of the abovementioned two types of hetero- to achieve an output of 1 W in the linear op-
conductor) FET structure, increasing the con- junction FET (gate length 1.1 µm) were pre- erating region of the FET, as opposed to ana-
centration of impurities in the channel layer sented in an earlier report.3) log phones which used amplification in the
causes a large negative shift in the threshold With a pulse-doped structure it was pos- saturated region. Also, to extend the phone’
voltage and a sharp drop in the gate break- sible to obtain a gate breakdown voltage 3 s talk time it is also important to improve the
down voltage. A large negative shift of the –4 V higher than that of a uniformly doped power conversion efficiency (i.e., the power-

JSAP International No.7 (January 2003) 5

 ������ ���� ����� ������ ���� �����
������
������������ ������
�������� ��������������
������������ ������������
�������������� ��������������������
�������� ��������������
�������������� ��������������������
���������������������� ����������������������
�������������� ��������������������
�������� ��������������
������ ������

�������������� ��������������

(a) (a)
���
��� ���������������� ������� ������������
�������������� ��� ���� �������
���
������������

������������
������
������ ������������
��� ������������
��� ������
�������
���
��� ������
���� ���� ������
���� ���
������
����
� �� �� �� �� ��� ��� � �� �� �� ��
(b) �������������������������� (b) ���������������������������

Fig. 2: A heterojunction FET with a multilayer cap structure. (a) Cross-sectional struc- Fig. 3: A heterojunction FET with an Al 0.5Ga0.5As barrier layer. (a) Cross-sectional
ture. (b) Conduction band energy distribution under the ohmic electrode. The use of structure. (b) Conduction band energy distribution under the gate electrode. The in-
a multilayer cap structure reduces the potential barrier at the n-AlGaAs layer near the troduction of a barrier layer suppresses the injection of electrons from the channel
surface. towards the gate electrode.

added efficiency), and to reduce the FET’s negative threshold voltages. A drawback of a two-dimensional electron gas is formed in
knee voltage (i.e., reduce the on resistance). this type of device is that two power sup- the GaAs layers in contact with both sides of
Furthermore, it should be possible to mass plies—one positive, the other negative—are the abovementioned heavily-doped n-AlGaAs
produce these devices with high yield, and needed to achieve amplification. Since Si- layer, there is a smaller potential barrier at the
with small chip dimensions. MOSFETs and bipolar transistors (described n-AlGaAs layer that also serves as an etching
To satisfy these requirements, the gate below) can operate with a single power sup- auto-stop layer. This effect makes it possible
length was reduced to 0.8 µm and a two- ply, the fact that twin positive and negative to reduce the contact resistance along the
stage recess embedded structure was em- power supplies are only needed for hetero- path from the heavily-doped n-GaAs surface
ployed for the gate electrode (see Figs. 2 and junction FETs is a distinct disadvantage. To layer to the InGaAs channel layer.
3 below).4) As a result, the drain current den- reduce the number of circuit components Tests were performed on an experimen-
sity was increased to 600 mA/mm (about 1.5 and develop products that are more compact tal enhancement-type heterojunction FET
–2 times higher than a conventional device), and less expensive, it is desirable to develop with a gate length of 0.5 µm and a multilayer
and the on-resistance was reduced to 2.3 Ω enhancement-type heterojunction FETs that cap structure.5) When a threshold voltage of
·mm. Selective dry etching of AlGaAs and have a positive threshold voltage and can be +0.20 V was selected, a large maximum drain
GaAs was used to form the two-stage recess, operated with a single power supply. How- current density of 390 mA/mm was achieved,
and the variation of recess etching depth ever, enhancement-type FETs have generally and the device also exhibited a large gate
across the wafer was kept to a minimum by suffered from reduced maximum drain cur- breakdown voltage of 17 V. By employing a
using the former as an etching stop layer. rent and increased on resistance, leading to multilayer cap structure, the parasitic resis-
The large-signal characteristics of this im- reduced power density and efficiency. tance could be reduced by 0.1 Ω·mm and it
proved FET were measured with a 950 MHz To address the abovementioned problems was possible to obtain an on-resistance of 1.6
QPSK (quadrature phase shift keying) digitally of enhancement-type FETs, a new multilayer Ω·mm.
modulated signal.4) In a miniature chip with a cap structure was introduced.5) Figure 2(a) For an FET with a gate width of 24 mm,
gate width of 7 mm, a power-added efficien- shows the cross-sectional structure of this FET, as a result of evaluating the power charac-
cy of 56.3% was achieved with an output and Fig. 2(b) shows the conduction band en- teristics with a single 3.5 V power supply, an
power of 1.23 W when operated at 3.4 V. ergy distribution below the ohmic electrodes. output power of 1.0 W and a power-added
2.1.3 Enhancement-type heterojunction In this structure, an undoped GaAs layer and efficiency of 65.5% were achieved—enough
FETs a heavily-doped n-AlGaAs layer are deposited to satisfy the PDC technical standard. In this
The heterojunction FETs introduced so in turn below the heavily-doped n-GaAs layer device, it was also possible to cut off the
far have all been depletion-type devices with in contact with the ohmic electrodes. Since drain current to just 2 µA/mm just by setting

6 JSAP International No.7 (January 2003)

the gate voltage to 0 V. Consequently, there with a single 3.5 V supply: output power 460 duced device that brings important benefits,
was no need to turn off the drain supply by mW, power-added efficiency 58.9%, associ- especially to applications such as power am-
providing a separate switch circuit. In oth- ated gain 12.6 dB. Moreover, the gate leak- plifier elements in mobile phone terminals.
er words, use of an enhancement-type FET age current during peak output was only 0.8 µ Factors that have led to the growth in popu-
makes it possible to eliminate not only the A, thus verifying the effect of introducing the larity of these devices include the fact that
negative power supply circuits but also the barrier layer. they help to make mobile terminals more
drain bias switching circuits from the mobile In a W-CDMA scheme, the transmitted compact and light-weighted because they
phone terminals. power strength of the terminal side is con- can be driven by a single positive power
In an enhancement-type FET, as the dif- trolled over a wide range according to the source, and the fact that they are suitable for
ference between the threshold voltage and distance from the base station. When the communication applications involving digital
the gate forward on voltage (VF) decreases, amplifier is biased under class A or class B modulation schemes where linear operation
it becomes harder to maintain a large maxi- operation, the efficiency generally becomes is required because they have comparatively
mum drain current. Also, during power op- worse as the output power is reduced from low distortion when operating with a large
eration a forward gate current flows more the saturated output point. output signal.
readily, increasing the likelihood of the ampli- A method proposed for avoiding this Although AlGaAs has for a long time
fier operation becoming unstable. According- problem involves using a DC-DC voltage con- been used as a wide bandgap emitter materi-
ly, it becomes necessary to increase VF. verter to dynamically control the drain bias al, the fact that InGaP and GaAs allow selec-
Figure 3(a) shows the cross-sectional voltage V ds.7) It has been reported that this tive etching has recently resulted in a prefer-
structure of an enhancement-type hetero- method allows the power-added efficiency ence for the latter, which is more suitable for
junction FET developed for this purpose, and of the amplifier to be increased from 8% (Vds mass production. It is also apparent that the
Fig. 3(b) shows the conduction band energy = 3.5 V) to 21% (Vds = 1.0 V) when the out- characteristics of the InGaP/GaAs heterojunc-
distribution under the gate electrode.6) This put power is 20 mW. Here, it was possible tion interface substantially improve the reli-
differs from a conventional structure in that to achieve normal amplifier operation even ability of HBTs.9)
an Al 0.5Ga 0.5As barrier layer (0.5 nm thick) when V ds was reduced to 1.0 V because a There are two factors that can be said to
with a large bandgap energy is inserted be- double-doped/double heterojunction FET has contribute to the improved reliability associ-
tween the InGaAs channel layer and the up- a sufficiently low knee voltage. In practice, ated with introducing an InGaP/GaAs hetero-
per n-AlGaAs electron supply layer. By em- in a heterojunction FET with a gate width of junction. First, since the density of defect lev-
ploying this new structure, it is possible to 28 mm where the on-resistance is reduced to els at the InGaP/GaAs interface is lower than
achieve an equivalent increase of V F while 1.3 Ω·mm, an output power of 1.0 W and a in the case of an AlGaAs/GaAs interface, it
avoiding the injection of electrons traveling power-added efficiency of 59% have been has the effect of reducing the recombination
from the channel toward the gate electrode achieved with a low bias voltage of 1.0 V.8) current arising from defect levels that con-
when a positive gate bias is applied. tributes to the device lifetime. Secondly, since
Figures 4(a) and (b) show the drain cur- 2.2 Bipolar transistors the difference in bandgap energy that occurs
rent vs. voltage characteristic and the gate The heterojunction bipolar transistor between the InGaP and GaAs (about 0.5 eV)
forward characteristic of an enhancement- (HBT) is widely recognized as a mass-pro- mostly appears as a discontinuity in the va-
type FET having an Al0.5Ga0.5As barrier layer
(gate length 0.5 µm). The maximum drain
current density was 370 mA/mm, and the
threshold voltage was +0.17 V. The value of
��� ��
VF determined from the gate forward current
���������
vs. voltage characteristic is 0.9 V, which repre-
���������������������

��������������������

��������
sents an improvement of 0.2 V compared to ����� ��������������������
��� ��
the device with no Al0.5Ga0.5As barrier layer.
For an enhancement-type FET with a �����
�����
high VF of this sort (gate width 16 mm), the ��� �� ��������������������

input/output power characteristics were eval- �����

uated with an input signal consisting of a 1.95 ��� �

GHz digital phase-modulated signal diffused �����
at a chip rate of 3.86 Mcps according to a
����� �
W-CDMA (wideband code division multiple �
� � � � � � � � ��� ��� ��� ��� ��� ���
access) scheme. By adjusting the input and ����������������� ����������������
output matching circuits to maximize effi- (a) (b)
ciency while suppressing power leakage to an
adjacent channel (separated by 5 MHz) to -35
Fig. 4: (a) Drain current vs. voltage characteristics and (b) gate forward characteristics of an enhancement-type het-
dBc or less, the following favorable results
erojunction FET with an Al0.5Ga0.5As barrier layer. The introduction of a barrier layer improves the forward on voltage
were obtained when operating the device by about 0.2 V.

JSAP International No.7 (January 2003) 7

lence band, it has the effect of suppressing The advantages of a drift base HBT in- materials and heterojunctions incorporating
the recombination current that arises from clude: 1) a larger average drift velocity for these materials. The ability of HBTs to oper-
the reverse injection of holes in the base layer electrons in a given accelerating field due ate from a single power source and their high
into the emitter. to the ability to maintain a large conduction power densities are significant advantages
It has recently been estimated that an band energy difference between the and L over FET devices, but there are still several
InGaP/GaAs HBT with a junction temperature valleys, and 2) the absence of Al exposed at problems that need to be addressed, such as
of 135°C has a very long lifetime of 3.6×108 the surface where it can interfere with the the existence of offset voltages in the collec-
hours, 10) making HBTs essential for mobile formation of external base layer electrodes. tor characteristics, achieving uniform-temper-
terminal applications where high reliability is Adding In to the GaAs is preferable in ature operation across a chip, dealing with
required. that it makes the bandgap in the base layer thermal runaway, and coping with surges
Studies are being made of how to im- narrower and reduces the operating thresh- and load fluctuations. It will be a while yet
prove the materials used in GaAs-based old voltage. However, since lattice mismatch before these problems are fully addressed. In
HBTs, not only to make them easier to mass problems may arise, there is a limit to the the market for mobile terminal products, it is
produce and more reliable, but also to en- amount of In that can be added. Attention worthwhile predicting how the market share
hance their performance. As in the case of has thus recently turned to GaInNAs compo- of FETs and HBTs will change in the future,
heterojunction FETs, a wider range of device sitions where small quantities of nitrogen are but for HBTs it is vital to have a comprehen-
possibilities can be made available if the lat- added to the InGaAs, resulting in powerful sive debate at least including the problems
tice matching constraints can be relaxed. An materials that can be used to make HBTs that mentioned above.
example of this is the drift base HBT, which operate at lower voltages while maintain-
uses a layer of InxGa1−xAs with a ramped com- ing lattice matching with the GaAs. Figure 3. Application to Mobile Phone
position (see Fig. 5). This HBT structure can 6 shows the relationship between the band Base Stations
improve the current gain blocking frequency positions in GaAs and GaInNAs. A problem 3.1 Current status of microwave
without producing dislocations or defects, to be addressed with this technique is that transistors for base station transmit-
even when a strained layer is introduced into when GaInNAs and GaAs form a heterojunc- ters
the device.11) tion, most of the difference in bandgap ener- The main purpose of a mobile phone
gy appears as an energy discontinuity on the base station is to relay voice and data com-
conduction band side.12) munications over radio channels while up-
The abovementioned result dating the positional information relating to
shows that a GaInNAs base struc- mobile terminals that are constantly on the
�������� ture is ideal for pnp transistors, move. According to figures gathered in Sep-
but is not suited for npn transis- tember 2001, some 65 million mobile termi-
E� tors as it is because it produces nal contracts have been registered in Japan,
�������� a potential barrier to electrons. and the number of base stations that have
An example has been report- been built across the country to accommo-
ed in which a GaAs/GaInNAs date these users is already in excess of 33
E� ����x����x�� chirp superlattice is employed thousand.14)
�x������������� ������
to achieve excellent npn opera- To accommodate the recent explosive
����x����x�� tion.13) In this example, an arti- growth in the number of mobile phone ter-
�x������������� ficial ramped composition layer minals, the number of communication chan-
is formed by the superlattice, nels needs to be increased. For this purpose,
Fig. 5: Energy band diagram of a drift base HBT incorporating a layer
of InxGa1−xAs with a ramped composition.
and the small bandgap of the a transition is under way towards third-gen-
Ga 0.89 In 0.11 N 0.02 As 0.98 base layer eration (3G) mobile phone systems that use
(0.98 eV) is effectively linked to bandwidth more efficiently. At the same time,
the reduction of the threshold to increase the volume of traffic that can be
voltage. However, there are still carried in each base station area, increas-
Ec problems associated with the ing the output power of microwave signals
crystal quality and doping con- transmitted from the base station antennas is
GaAs centration in npn HBT base mate- also an important issue. In particular, in cur-
(1.43 eV) GaInNAs rials, and further remedial studies rent systems where a single amplifier is used
(0.98 eV) should be carried out to achieve to amplify multiple carrier waves at the same
a high current gain combined time, it is necessary to cope with instanta-
Ev with a low base resistance. neous peak outputs approximately an order
As described above, GaAs- of magnitude larger than the average output
based HBTs are expected to open power so that the transmitter amplifier be-
Fig. 6: Relationship between the band positions in GaAs and GaInNAs. up new possibilities in the future comes never saturated. This is referred to as
Most of the difference in bandgap energy appears as an energy discon-
tinuity at the conduction band side.
through the introduction of new operating the power amplifier with back-off,

8 JSAP International No.7 (January 2003)

 square of the current, this makes it harder to
���������������������������������������������������� achieve energy savings at the base stations.
������ As a result, the use of higher drain currents
��� �������
is no longer an ideal solution for achieving
���
higher output power levels.
����������������

���
������
������� 3.2 Heterojunction FETs with field
���
��� plates
������ An attempt has been made to improve
������� ���
the operating voltage of GaAs FETs to 20 V or
���
������
��� more, which is twice the normal voltage for
������ ������� ���
������� ��� such devices.16) This involved the use of a het-
���
�� erojunction FET with a new structure where
�� a field plate (FP) connected to the gate elec-
� trode is situated on the surface passivation
���� ���� ���� ���� ���� ���� film between the gate and drain.
������������� The cross-sectional structure of this FP-
FET is shown in Fig. 8. The gate is formed
Fig. 7: The year-on-year growth in the performance of high output GaAs FETs developed for mobile on an n-AlGaAs surface layer, and controls
phone base stations.
the concentration of electrons in the n-GaAs
channel layer below it. The FP electrode plays
and allows high-quality digital radio commu- vices. According to the latest data, there are two primary roles. One is to moderate the
nication to be achieved with a low bit error reports of 240 W products being available on electric-field strength at the drain end of the
rate. For these reasons, there is currently a the market, and 300 W devices have already gate when the gate is subjected to a large
strong need for increased output power from been produced in the laboratory.15) negative bias when operating with large AC
the microwave transistors used in base sta- Power supply voltages have remained signals. It achieves this by depleting the chan-
tion transmitters. fixed at 10–12 V throughout this period, so nel layer underneath the negatively charged
FETs based on GaAs and Si semicon- the output power levels attained so far have FP electrode, and as a result the gate break-
ductors have been widely used in the trans- all been achieved by increasing the drain cur- down voltage is substantially improved. The
mission power amplifiers of mobile phone rent. Although this can be done by using second role is to facilitate the flow of large
base stations. The former have better high- larger transistor chips or greater numbers of AC drain currents, since the depletion layer in
frequency characteristics but have a serious chips, there are a number of accompanying the channel under the FP electrode is instan-
drawback with respect to high voltage opera- problems that are exacerbated by this ap- taneously reduced when the gate bias swings
tion. On the other hand, although the latter proach. in the positive direction. This has the effect
can be successfully operated at voltages as First, there are limits to how large the of improving the linearity of the input/output
high as 20–30 V, it is hard to achieve favor- chips can be made. The package dimensions power characteristics and allowing a large
able high-frequency characteristics. In terms are also limited, so it will become increas- saturated output power to be obtained.
of cost, Si-based FETs are much less expen- ingly difficult to carry on increasing the chip An experimental heterojunction FET
sive, but 3G mobile phones—which are ex- size in the future. Another problem is that in- with an FP electrode (gate length 1.0 µm, FP
pected to become widespread in the future— creased output current
use the 2 GHz band which is higher than is associated with re-
the frequency of conventional systems. Ac- duced input impedance.
cordingly, the current status is that the better Since matching circuits
high-frequency characteristics of GaAs FETs with a large impedance Passivation Film Field Plate
Gate
give them the advantage in terms of overall conversion ratio add to Source Drain
performance. the complexity, this ag-
n-GaAs
At the product level, let’s see how in- gravates problems such undoped GaAs
n-AlGaAs
creases have so far been made to the output as increased reflection
n-GaAs
power of high-power GaAs FETs. Figure 7 loss and bandwidth re-
shows the year-on-year growth in the perfor- duction. Finally, there is buffer
mance of high-power GaAs FETs developed the problem that larger S.I. GaAs sub.
for mobile phone base stations. Since 1997, power supply circuits
when the 100 W barrier was broken for the generate more heat.
first time by mounting four FET chips in a sin- Since the losses in a
Fig. 8: Cross-sectional structure of a new type of heterojunction FET incorporat-
gle package, there has been steady growth power supply circuit ing a field plate. The field plate is connected to the same potential as the gate
in the output power available from such de- rise in proportion to the electrode.

JSAP International No.7 (January 2003) 9

length 0.5 µm) has exhibited a peak drain at high voltages. 60 GHz band has been configured by flip-
current of 300 mA/mm and a gate break- Although these results were obtained chip mounting a coplanar MMIC (monolithic
down voltage of 45 V. Figure 9(a) shows a with an experimental FP-FET with a GaAs microwave integrated circuit) consisting of
photograph of an FET chip with a multi-cell channel, the fabrication of an FP-FET with an 0.15 µm gate GaAs heterojunction FETs onto
structure designed for high power applica- InGaP channel was also recently reported.17) a multilayer glass ceramic substrate.19)
tions, and Fig. 9(b) shows a photograph of Such devices will allow base stations to oper- Figure 10 shows a photograph of the
an FET amplifier with a partial matching cir- ate at even higher voltages. transmitter and receiver modules. The trans-
cuit for push-pull operation that incorporates mitter module incorporates three types of
four of these chips. Each FET chip has a total 4. The Progress Toward New MMIC—a 30/60 GHz frequency multiplier, a
of 108 gate fingers in a parallel arrangement, Applications modulator and a transmitter amplifier—and
and the overall gate width is 86.4 mm. The As the internet continues to grow, the the receiver module incorporates two types of
chip measures 1.5 mm × 4.4 mm. data transfer speeds that networks are re- MMIC—a low-noise amplifier and a demodu-
The output power of the push-pull am- quired to support are increasing rapidly. For lator. This 60 GHz band transceiver module
plifier was evaluated at 2.1 GHz, and was wireless networks, a peak transfer speed of was used to build a prototype two-way radio
found to increase linearly as the drain volt- 54 Mbit/s has been achieved in 5 GHz band transceiver conforming to IEEE 1394, and was
age increased. At a drain voltage of 22 V, this wireless systems such as HiperLAN/2 and able to operate at 500 Mbit/s with an ASK
device delivered an output power of 230 W IEEE802.11a, which are currently being stan- (amplitude modulation) scheme.
with a power-added efficiency of 42% and a dardized.18) However, cable-based systems are Oscillators are essential components of
linear gain of 11 dB. A W-CDMA signal was progressing at an even faster rate. Accord- millimeter-wave systems of this sort. HBTs are
input to this push-pull amplifier, and the leak- ingly, even in wireless systems it is expected suitable for oscillator applications because
age into a neighboring channel separated by that gigabit wireless communications will ap- these devices have a vertical configuration.
5 MHz was measured. As a result, the power pear for applications such as real-time video This means that they are affected little by 1/f
leakage at an output power of 40 W (8 dB streaming and high-speed downloading. noise sources located at the semiconductor
back-off point) was found to be −35 dBc or Since August 2000, the 60 GHz band surfaces, and as a result they exhibit signifi-
less, with a 25% power-added efficiency. The (59–66 GHz) has been released for the op- cantly low phase noise (fluctuation of the os-
distortion characteristics at the back-off point eration of license-free wireless systems, and cillator frequency)—which is one of the most
were improved by about 4 dB due to the in- it is strongly anticipated that this band will be important factors determining the perfor-
creased drain voltage, and it was verified used for wireless systems operating at 100 mance of an oscillator. Figure 11 shows the
that this device exhibits not only high output Mbit/s and above. For example, an experi- surface appearance of an experimental 38
power but also low distortion when operated mental miniature transceiver module for the GHz band MMIC oscillator chip which uses

Transmitter Module
MMIC
Filter
MMIC
Receiver
Module
(a) ��������������

Antenna
Port

Fig.10: Photograph of miniature 60 GHz band transceiver modules. The transmitter module
incorporates three types of MMIC—a frequency multiplier, a modulator and a transmitter am-
plifier. The receiver module incorporates two types of MMIC—a low-noise amplifier and a de-

(b) ������������ modulator—and a bandpass planar filter.

Fig. 9: Heterojunction FETs with field plates. (a) Photograph of a chip with a
multi-cell structure consisting of a heterojunction FET chip with field plates (chip
dimensions: 1.5 mm × 4.4 mm). (b) Surface photograph of an amplifier with a
partial matching circuit that incorporates four FET chips in the same package.

10 JSAP International No.7 (January 2003)

 References
���������������� �������������� 1 ) K. Inoue and H. Sakaki: Jpn. J. Appl. Phys. 23,
L61 (1984).
2 ) K. Hikosaka, Y. Hirachi, T. Mimura and M. Abe:

��� IEEE Electron Devices Lett. EDL-6, 341 (1985).

3 ) N. Iwata, K. Inosako and M. Kuzuhara: IEEE
MTT-S Int. Microwave Symp. Dig., Atlanta,
1993, p. 1465.
4 ) N. Iwata, M. Tomita, K. Yamaguchi, H. Oika-
wa and M. Kuzuhara: Tech. Dig. IEEE GaAs IC
������ Symp., Orlando, 1996, p. 119.
5 ) Y. Bito and N. Iwata: Inst. Phys. Conf. Ser. No.
162, p. 237 (1999).
6 ) Y. Bito, T. Kato and N. Iwata: IEEE Trans. Elec-
tron Devices 48, 1503 (2001).
7 ) T. B. Nishimura, N. Iwata and G. Hau: IEEE
MTT-S Int. Microwave Symp. Dig., Anaheim,
1999, p. 1091.
8 ) T. Kato, Y. Bito and N. Iwata: IEICE Trans.
Electron. E84-C, 249 (2001).
������������ 9 ) T. Takahashi, S. Sasa, A. Kawano, T. Iwai and
Fig. 11: A 38 GHz MMIC oscillator. An HBT with a graded composition InGaAs base structure is used T. Fujii: Tech. Dig. IEEE Int. Electron Devices
as the active element. When detuned by 1 MHz, this oscillator has a favorable phase noise of −114 Meet., San Francisco, 1994, p. 191.
dBc/Hz. 10) M. Tanomura, Y. Miyoshi, M. Kawanaka, T.
Nishimura, F. Harima, H. Sato, H. Takahashi,
S. Yoshida, K. Higashi, N. Iwata and H. Shi-
an HBT with an InGaAs graded composition high-speed wireless internet access technolo- mawaki: IEICE Electronics Soc. Conf. 2, p. 59
(2000).
base structure. gies to complement ADSL (Asymmetric Digi-
11) H. Shimawaki, Y. Amamiya, N. Furuhata and
tal Subscriber Line) and FTTH (Fiber To The K. Honjo: IEEE Trans. Electron Devices 42,
5. Conclusion Home). GaAs will probably play an important 1735 (1995).
Although GaAs is a semiconductor material role in the wireless devices used for these ap- 12) H. P. Xin, C. W. Tu and M. Gena: Appl. Phys.
with some favorable electronic properties, this plications. Lett. 36, 1416 (1999).
13) R. J. Welty, H. P. Xin, K. Mochizuki, C. W.
does not mean that it has outstanding quali- As mentioned in this paper, GaAs devices
Tu and P. M. Asbeck: Device Research Conf.
ties. Its mobility is inferior to that of InGaAs, are also being used for the initial prototypes Dig., p. 145 (2000).
and its peak drift velocity is less than that of InP. of 60 GHz band millimeter-wave wireless sys- 14) http://www.hokuriku-bt.go.jp/data/
Its bandgap energy is larger than that of Si, but tems. There is no doubt that GaAs—which musen/musen.html
much smaller than those of SiC or GaN. How- has the longest history of all group III-V semi- 15) K. Ebihara, K. Inoue, H. Haematsu, F. Yamaki,
H. Takahashi and J. Fukaya: IEEE MTT-S Int.
ever, its overall performance makes it one of conductors—will continue to be used as the
Microwave Symp. Dig., Phoenix, 2001, p.
the best candidates for use in high-frequency central material in high-speed transistors 649.
devices. The successful use of GaAs in power across a wide range of frequency bands while 16) K. Matsunaga, K. Ishikura, I. Takenaka, W.
amplifier devices for mobile phone terminals is keeping other new semiconductor materials Contrata, A. Wakejima, K. Ota, M. Kanamori
also due to its overall superiority over other ma- at bay. and M. Kuzuhara: Tech. Dig. IEEE Int. Electron
Devices Meet., San Francisco, 2000, p. 393.
terials rather than the overwhelming benefits of
17) A. Wakejima, K. Ota, K. Matsunaga, W.
any particular characteristic. Hopefully this pa- Acknowledgements Contrata and M. Kuzuhara: IEEE RF-IC Symp.
per has helped to convey this situation to some We obtained most of the research re- Dig., Phoenix, 2001, p. 151.
extent. sults introduced in this paper through joint 18) Nikkei Electronics: No. 803, p. 57 (August 27,
Although there is not enough space for studies with Naotaka Iwata, Keiko Yamagu- 2001).
19) K. Maruhashi, M. Ito and K. Ohata: Proc. Eu-
this article to offer many specific examples of chi, Takeshi Nishimura, Yasunori Bito, Wal-
ropean Microwave Conf., London, 2001, Vol.
GaAs-based high-speed electronic devices, ter Contrata, Takehiko Kato and Gary Hau— 1, p. 411
there will be an abundance of new applica- thank you all very much. We would also like 20) K. Hosoya, S. Tanaka, Y. Amamiya, T. Niwa, H.
tions for such devices in the future. It has to thank Masaki Ogawa and Masashi Mizuta Shimawaki and K. Honjo: IEEE MTT-S Int. Mi-
been decided that specific frequency bands for their helpful discussions. crowave Symp. Dig., Boston, 2000, p. 47.
(Accepted for publication: November 21,
between 20 and 40 GHz will be used for
2001)

Masaaki Kuzuhara (D. Eng.) Shinichi Tanaka (D. Eng.)

Completed a Master’s course in engineering at Kyoto University in 1981. Since joining NEC Completed a Master’s course in engineering at the University of Tokyo in 1986. Since joining
Corporation, he has been involved in the R&D of ion implantation in compound semicon- NEC Corporation he has been involved in the R&D of heterojunction bipolar transistors and
ductors, Monte Carlo simulation of compound semiconductors, compound semiconductor MMICs. From 1992 to 1993 he was a guest researcher at Purdue University.
heterojunction devices and MMICs. From 1987 to 1988 he was a guest researcher at the
University of Illinois.

JSAP International No.7 (January 2003) 11

View publication stats