Chapter 4

Semiconductor Applications
Katashi Masumoto
The Research Institute for Electric and Magnetic Materials,
2-1-1 Yagiyama Minami, Taihaku-ku, Sendai 982-0807, Japan

Akinori Katsui
School of High-Technology for Human Welfare, Department of Materials Science and Technology,
Tokai University, 317 Nishino, Numazu, Shizuoka 401-0321, Japan

Takashi Matsuoka
NTT Basic Research Laboratories, Nippon Telegraph and Telegraph Corporation, 3-1 Wakamiya
Morinosato, Atsugi, Kanagawa 243-0198, Japan

1. Introduction

Generally, semiconducting compounds have a diamond

structure or one of its derivatives as a basis (see
Parthe, Chapter 14 of Volume 1). This structure has
two types of positions: the f.c.c. positions and the b.c.c.
positions of half the octants. Figure 1 shows the related
zinc blende structure, wurtzite structure, and chalcopyrite
structure. The zinc blende structure is the one in which
cations lie in the f.c.c. positions and anions lie in the b.c.c.
positions with equivalent coordination. A ternary
compound with a diamond structure may be obtained
by substituting only cations or by substituting both
cations and vacancies into the tetrahedral sites and
octahedral sites of the f.c.c. lattice. Figure 2 is
a diagram showing an example of the induction
to higher-order semiconducting compounds with the
diamond structure. In this arrangement, each of
the compounds must have a positive valence of
four on average (Pamplin, 1964). This requirement
also applies to all of the ternary to multinary
compounds having a diamond or similar structure
described in this chapter. Thus Figure 2 shows that
Ge, GaAs, ZnFeAs2, and CuGaGe2As4 are all possible
adamantane compounds.

Many of the semiconducting compounds show a

variety of optical phenomena, such as luminescence,
laser beam emission, and nonlinear optical effects, none
of which is present in pure silicon. Many of these
compounds have charge carriers with higher electron
mobilities than silicon. Furthermore, there is a wide
selection of materials from the same groups of
compounds in constructing ternary and quaternary solid
solutions of the diamond structure. Therefore, it is
possible to change the fundamental properties of a
semiconductor, such as the forbidden band gap,
mobility, and lattice constant, over a usefully wide
range. The result is a flexibility in material design for
a variety of applications. There is more and more cutthroat competition in the development of compound
semiconductors.
Table 1 as an example shows a comparison of
materials used in some important compound semiconductor light-emitting diodes (LEDs). Many of the
materials are compounds of two or more kinds of
elements. A much higher level of material preparation
technology is needed to construct useful devices from
these compound semiconductors than is needed for an
elemental semiconductor. The quality of presently
available commercial materials is not yet sufficiently

Magnetic, Electrical and Optical Applications of Intermetallic Compounds. Edited by J. H. Westbrook and R. L. Fleischer 1995,
2000 John Wiley & Sons Ltd

Cation
Anion

Group I or III atom

Group III or I atom
Group Vl atom (anion)

Figure 1. Crystal structures of (a) zinc blende (B3, cF8), (b)

wurtzite (B4, hP4) and (c) chalcopyrite (El 1 , tI16)

high for obtaining the required properties as semiconductors, although their prices are extremely high.
Yet there is more and more demand for compound
semiconductor devices in all kinds of industries;
therefore, material preparation technology is advancing
steadily into the first stages of commercial production.
The III-V semiconducting compounds are in the most
advanced stage of research and development, as a result
of long-term studies of their physicochemical and
electronic properties. They are now the most promising
electronic materials. Particularly advanced materials in

this field are optoelectronic semiconductors based on

GaP, GaAs, and InP and magnetoelectronic semiconductors based on InSb and InAs. In the future,
higher functionality and improved properties of these
compounds will be needed.
There has been rapid and extensive progress in the
technologies of preparing semiconducting compounds
and manufacturing applied electronic devices. Recent
developments in the techniques of crystal growth via the
vapor phase have been remarkable. Particularly,
molecular beam epitaxy (MBE) and metal organic
vapor-phase epitaxy (MOVPE) make it possible to stack
one layer of atoms onto another. Such deposition of
different semiconductors, having different widths of
their forbidden bands, serves to form an artificial
* super lattice'* having a width potentially equivalent to,
or less than, the de Broglie wavelength. It is further
possible, by repeated deposition of dissimilar lamellae,
to make a modulated lamellar * super lattice' with a
controlled and varied potential period. This technique
for depositing very thin atomic layers makes it possible
to attain heterojunctions between structures of different
lattice constants. Thus, we now have more degrees of
freedom in the combination of semiconducting materials
to form a new type of 'superlattice', namely the 'strained
superlattice'. One new proposal is that, without a
periodic change in potential but with a periodic change
in another physical property (for example, the effective
mass), it might be possible to form both a quantum well
and a 'superlattice'.
Using advanced techniques of crystal growth and
materials processing, we can change the spatial distribution of impurity atoms at short (atomic scale)
intervals, even in a semiconducting material with a
forbidden band gap. This leads to a change in the spatial
distribution of electric charges, and hence to the
formation of a periodic change in potential to produce
a 'superlattice'. It has also been found possible to form
buried layers of compounds, e.g. semiconducting
CoSi2, by high-dose ion implantation followed by
annealing. Silicide layers formed by this mesotaxy
process are single crystal, aligned with the substrate,
with smooth, abrupt interfaces and excellent electrical
characteristics (Celler and White, 1992).
In summary, starting with the conventional technique
for forming a simple heterojunction, the advanced
technology for producing a quantum well and many

Quotes are used here to distinguish this type of material from the
superlattice of the metallurgist, an ordered array of atoms formed from
a disordered binary (or higher) solid-solution alloy.

e/a

16

32

e
4

8
16
Figure 2. Induction to binary, ternary, and quaternary semiconducting compounds with diamond structure (e = number of valence
electrons for the atoms, e/a = the electron to atom ratio, n = vacancy)

kinds of 'superlattices' has resulted in a concept of new

physical effects unattainable in bulk crystals. We can
call this a new physics. We are about to make the
concept a reality.
Section 2 deals with semiconducting compounds of
Groups H-VI, HI-V, IV-IV, IV-VI and V-VI,
transition metal silicides, and ternary systems. It

describes examples of such compounds, their characteristics, and recent related research and development.
Section 3 presents examples and recent applied studies
of electronic devices. Section 4 covers optical devices,
and Section 5 other devices including magnetoelectric
devices. Finally, Section 6 summarizes these
developments and considers future problems.

Table 1. Comparison of compound semiconductor materials for emitting devices

External quantum efficiency (0Io)
Compound
HI-V

Epitaxial layer

Substrate
GaP

GaAs

GaP:Zn, O
GaPrN
GaP
GaP
GaAs 0 - 1 5 P 0 ^N
GaAsO35Po.65:N
GaAs
GaAs 0 6 P 0 4
Ga

ii-vi

IV-IV
1-IH-VI2
IV-VI

GaAlAs
GaAs
InP
Sapphire
ZnSe
GaAs
ZnS
GaP
SiC
GaAs, GaP
PbSe, PbTe

'Commercially available.

O.65 A 1 O.35 A s

Ga065Al035As(DH)
GaAlInP
GalnAsP/InP
GaN
ZnSe
ZnSe, ZnS
ZnS
SiC
CuAlS2, CuGaS2
PbSnSe, PbSnTe

Device
LED (red)"
LED (yellow)0
LED (green)*
LED (yellow)*
LED (yellow)0
LED (orange)0
LED (0.8-0.9/*m)
LED (red)
LED, LD (red, ~0.9/tm)
LED (red)
LD (-0.6 ^m)
LD, LED (1.3-1.6 jim)
LED (blue)0
LED (blue)0
LED, LD (blue)
LED (blue)0
LED (blue)
LED (blue)
LED (visible)
LD (>10|im)

Commercially
available

Maximum
observed

-4
0.3
0.15
0.1
0.12
0.3

15
0.7
0.2

0.2
-3
-15

0.2
7
21

3.2

9.1
1.3

0.25
0.65

0.04
0.05

2. Compound Semiconductors for Device

Applications
'

2 A Group H-VI Semiconducting Compounds

Group II-VT semiconductors are binary compounds and
their solid solutions, consisting of Group HB elements
(Zn, Cd, Hg) and Group VIB elements (O, S, Se, Te),
e.g. ZnSe, ZnS and (Hg, Cd)Te, are typical.
2.1.1 ZnSe and ZnS
The forbidden band gaps Eg of Zn chalcogenides are
2.26 eV for ZnTe, 2.67 eV for ZnSe and 3.7 eV for ZnS,
all at 300 K. Their mutual solid solutions exist over all
compositions, and thus Eg of Zn chalcogenides can
change continuously from the yellow region to the violet
region. Attention has especially centered around ZnSe
and ZnS as materials for light-emitting device applications in the blue region.
ZnSe and ZnS possess a phase transition between the
zinc blende type and wurtzite type. The transition temperatures are 1698 K and 1297 K for ZnSe and ZnS,
respectively (Klakov et al., 1976).
ZnS has a high equilibrium dissociation pressure
(3.7 x 102IcPa) at the melting point Tm (2103 K). The
Tm (1793 K) and dissociation pressure (53 kPa) of ZnSe
are not as high. However, the vapor pressures of Zn,
S and Se are exceedingly high in comparison with the
compounds. Consequently, it is difficult to grow large,
high-quality single crystals of ZnSe and ZnS from the
melt. The vapor-growth method has been intensively
studied to avoid these problems. High-purity ZnSe single
crystals including few twin boundaries have been grown
by means of the sublimation method (Fujita et al.,
1979).
As-grown crystals are high-resistance n-type, and an
anneal in the melt or under the vapor pressure of Zn
is essential for low resistivity. Epitaxial thin-film growth
has been achieved using liquid-phase epitaxy (LPE),
vapor-phase epitaxy (VPE), MOVPE and MBE. Among
these alternatives, the MOVPE and MBE methods have
been intensively examined because of the nonequilibrium
character of film growth. Since ZnSe and ZnS wafers
for epitaxial growth are not commercially available,
GaAs (mismatch of 0.27% for ZnSe) and GaP (0.77%
for ZnS) are very frequently employed as substrates. The
epitaxial growth temperature is 350 to -500 0 C.
Undoped films are highly resistive (specific resistivity
p above 106Q cm), and the photoluminescence (PL)
spectra generally show strong near-band edge emission
and a weak self-activated emission.

Formation of a p-n junction leads to realization of

high injection efficiency in LED applications, and hence
conduction-type conversion is needed. However, control
of conduction type by means of impurity doping has
proven unachievable in Group II-VI semiconductors,
except for CdTe. The reason still remains unclarified.
For n-type films, iodine is useful as a donor species,
and the carrier concentration can be controlled in the
range 1016-1019cm~3 (Shibata et al., 1988). The resistivities of films doped with 1 x 1019cm~3 iodine atoms
decreased to as little as 1.3 x 10 ~3 U cm at room temperature. These doped films showed a strong blue PL
at 300 K. For p-type films, nitrogen (Ohki et al., 1988),
phosphorus (Bhargava, 1982), lithium (Nishizawa, 1986;
Yasuda et al., 1988; Cheng et al., 1988) and oxygen
(Akimoto et al., 1989) have been examined as acceptor
species. Recently, a nitrogen-doping technique using a
high-frequency plasma cell was established. Furthermore, a radio-frequency plasma gas cell was installed
into an MBE chamber and low-resistance p-type films
were grown on a GaAs substrate by plasma-assisted
doping with nitrogen (Park et al., 1990; Ohkawa, 1990).
Still further, a laser diode (LD) with a (Zn, Cd)Se/ZnSe
quantum-well structure has been manufactured, and
laser emission with a wavelength of 494 nm at 300 K was
observed by electric current injection (Jeon et al., 1991;
Haase^a/., 1991).
2.1.2 (Hg, Cd)Te
The narrow band gap semiconductor (Hg, Cd)Te is a
solid solution of the Group II-VI semiconductors HgTe
and CdTe. The band gap changes continuously from
HgTe (-0.3eV at OK) to CdTe (1.6 eV at OK).
The terminal compound CdTe generally shows no
phase transition. As-grown crystals, grown from a melt
with a Te-rich composition, proved to be p-type and
to contain a high concentration of Cd vacancies
(Triboulet and Marfaing, 1977). By annealing under Cd
vapor pressure, the crystals become low-resistance ntype (Triboulet, 1981).
Epitaxial film growth has also been studied. CdTe,
(Cd, Zn)Te and (Cd, Mn)Te have been used as substrates. Undoped (Hg, Cd)Te films formed by the LPE
method contained 10 16 -10 17 cm 3 Hg vacancies and
showed p-type conduction. By annealing under Hg
vapor pressure, the films were converted to n-type
conduction. Both conduction types could also be
produced by impurity doping.
(Hg, Cd)Te has been developed as a material for an
infrared detector in the range 1-20 ^m. Furthermore,
(Hg, Cd)Te has high electron mobility and forms an

insulating native oxide of high quality in the film form;

it is therefore highly useful for electrical applications.
A 'superlattice' of modulated CdTe/HgTe and
ZnTe/HgTe structure is highly desirable for high-speed
devices, such as Gunn diodes and hot-electron transistors. In an (Hg, Cd)Te/HgTe/(Hg, Cd)Te structure
with a double barrier type, single quantum well, a
negative resistance attributed to a resonance tunnel
effect was found (Reed et al., 1987). The peak to valley
ratio was 1.4:1, comparable with that of AlGaAs/GaAs.
2.2 Group III- V Semiconducting Compounds
Group IH-V semiconductors are binary compounds and
their solid solutions consist of Group IHB elements (B,
Al, Ga, In) and Group VB elements (N, P, As, Sb).
Table 2 shows some fundamental characteristics of
Group III-V semiconductors. Practical interest has been
most intensively centered on GaAs, InP, and the solid
solutions (Al, Ga)As, (In, Ga)As, and (In, Ga) (As, P).
GaAs and InP melt congruently; however, they have
high dissociation pressures. Hence, pressures above
atmospheric pressure are required over the melt to
maintain a stoichiometric composition during the growth
process. The pressures are 100 and 2.7xl0 3 kPa for
GaAs (Tm = 1511 K) and InP (Tm = 1335 K), respectively.
These crystals usually contain high concentrations of
compositional defects. Thus, growth of perfect single
crystals of large diameter has been investigated by means
of horizontal Bridgman (HB), liquid-encapsulated
Czochralski (LEC), and vertical-gradient freezing (VGF)
methods. Point defects, diffusion mechanisms, and
superlattice disordering in GaAs-based materials have
been reviewed in depth by Tan et al. (1991). Cr, oxygen
and native point defects form deep acceptor levels in
GaAs (Hobgood et al., 1982; Martin et al., 1980).

In InP, transition metal elements such as Fe (Mizuno

and Watanabe, 1975; Kamada et al., 1984) and Ti
(Iseler, 1986; Katsui, 1988) form deep levels. The deep
levels compensate the residual carriers electrically, and
as a result the crystals become semi-insulating with p
above 5 X 106 Q cm.
The semi-insulating crystals are useful as substrates
for optoelectronic integrated circuits (OEICs), capable
of reducing floating capacitance and providing electrical
isolation for devices.
Recent practical interest has been greatly centered on
OEICs. Furthermore, marked advancement in the
MOVPE and MBE methods and in the sophistication
of small-scale, high-resolution processing techniques
have enabled producers to propose various multifunctional or high-performance devices. A voluminous
and excellent secondary literature exists dealing with
Group III-V semiconductors and their applications (for
example, Willardson and Beer, 1975-1992).
2.3 Group IV-IV Semiconducting Compounds
SiC, one of the Group IV-IV semiconductors, possesses
crystal polymorphism (Marshall et al.., 1974). Hexagonal
Qf-SiC is an indirect transition type semiconductor
having a band gap Eg of about 3.OeV at 300K. It is
useful as a material for LEDs in the blue region. Lowresistance n-type and p-type films are grown by doping
with nitrogen and aluminium, respectively. The blue
emission is probably due to a donor-acceptor (D-A)
emission in the n-type layer.
High-quality a-SiC single crystals have recently been
grown by an improved Reigh method and a sublimation
method (Tairov and Tsvetkov, 1981). A p-n junction
has been formed in these crystals by a rotational dipping
method (Ikeda et al., 1979). First, a p-type layer was

Table 2. Fundamental properties of Group III-V compound semiconductors

Properties

GaAs

InP

InAs

InSb

AlN

BN (cubic)

E% (eV) at 300 K
Transition type
Recombination coefficient

1.43
Direct

1.35
Direct

0.36
Direct

0.18
Direct

5.9
Direct

8.0
Indirect

1.26X10 9

8.5XlO" 11

4.58XlO- 11

4500
4.5 XlO4

33 000
1.5 XlO5

80 000
IxIO6

14 (/*h)

0.2 (500 K)

1.14X10 1 2
12.5

2.35 XlO" 12
18.7

5xl012

12.1
0.68
1343
540

0.273
1215
380

0.166
798
220

3.2
-2723
1200

(Cm 2 S 1 )
7.2IxIO- 1 0
Hall mobility /xe (cm2 V 1 s"1)
300K
8500
77 K
Surface recombination velocity
(cms" 1 )
2 XlO7
Dielectric constant (C/N)
~2.7xl012
Relative permittivity e/e0
12.9
Thermal conductivity
1
1
(W cm" deg" )
0.455
Melting point (K)
1511
2
Hardness (kg mm )
750

IxIO3

13
>2973
7300-10 000

grown on the crystals by immersion in an Si melt doped

with Al, and then an n-type layer was formed by
introducing nitrogen gas into the melt.
LEDs with a central wavelength of 475 nm, an
external quantum efficiency of 2x 10 ~4, a maximum
luminous intensity of 12 mcd at an operational current
of 20 mA and continuous room-temperature operation
times above 5000h have been developed (Koga et al.,
1985).
For an SiC LED, since the band gap of a-SiC is larger
than the energy of photoemission, reabsorption of
emitted light in the LED interior is negligibly small.
In cubic (3-SiC9 undoped n-type films with a carrier
concentration of 1016-1017cm~3 have been grown on
a two inch diameter Si wafer by means of chemical
vapor deposition (CVD).
Just as with cx-SiC, /3-SiC is chemically and structurally
stable to high temperatures and is highly thermally
conductive. Diodes and transistors operable at high
temperatures and high electric power have been studied.
*

2.4 Group IV-VI Semiconducting Compounds

Table 3 shows some fundamental properties of Group
IV-VI semiconductors. (Pb, Sn)Te compounds which
form a complete solid-solution series of the Group IVVI semiconductors PbTe and SnTe are direct transition
type semiconductors.
The crystal structure of (Pb, Sn)Te is f.c.c, NaCltype Bl (cF8). The band gap of (SnxPb1 _JTe at 77 K
decreases linearly with increasing x from 0.22 eV for the
terminal compound PbTe, becomes zero at x=0.4, and
again increases to 0.27 eV for the other terminal
compound SnTe at x= 1 (Paker and Johnson, 1981).
As-grown, undoped (Pb, Sn)Te crystals are p-type,
owing to high concentrations of Pb and/or Sn vacancies.

The n-type crystals are obtained by annealing as-grown

crystals in a metal-rich atmosphere, so that the Pb
and/or Sn vacancies are compensated. Furthermore, Pb
and/or Sn interstitials may be generated (Paker and
Johnson, 1981). The n-type conversion is also achieved
by doping with Sb, Cd or In. Tl is used as a dopant
for p-type crystals (Antcliffe and Wrobel, 1970).
(Pb, Sn)Te is mechanically soft, similar to aluminum
metal. Plastic deformation readily occurs and low-angle
boundaries are easily formed. Efforts have been made
to grow bulk single crystals without low-angle
boundaries (Kinoshita and Sugii, 1983).
Recent studies have been made of epitaxial thin-film
growth by means of LPE using a Te solvent and MBE.
(Pb, Sn)Te is promising as a material for variable
wavelength LDs and for photodetectors for the wavelength region 5-30 /*m. Although (Pb, Sn)Te has the
same band gap as the Group H-VI semiconductor (Hg,
Cd)Te, (Pb, Sn)Te has a larger effective electron mass
and a smaller energy loss owing to nonradiative
recombination; therefore, it is more useful for device
applications.
LDs with homojunctions and double hetero junctions
(DH) have been studied (Tomasett and Fonstad, 1974;
Horikoshi et al., 1982). In DH-type structures, the laser
oscillation temperature is below about 80 K.
A 'quantum well' in semiconductors signifies a
quantum-mechanical well of a certain potential energy,
and within it carriers, electrons or holes, are twodimensionally confined. A single quantum well is the
most fundamental structure, and consists of an
extremely thin well layer sandwiched between barrier
layers. A quantum well may also be formed such that
a light beam is confined in it, as well as electrons. Then,
their confinement becomes highly effective in a multiple
quantum-well (MQW) structure, and therefore the upper

Table 3. Fundamental properties of Group IV-VI compound semiconductors and semimetals

Properties
Lattice constant (nm)

PbS

PbSe

PbTe

SnS

SnSe

SnTe

0.59362

0.6147

0.6461

59.9

62.98

8.15

8.25

a-0.599, b = 0.434,
c=l. 120
5.08

6.18

6.51

20
0.14
1.6

18-20
0.15
2.0
428
32.6
30
1190
0.22
0.22

0.19

0.18

21
0.18
20

Density (g cm " 3 )
7.59
Volume dilatation coefficient
(1(T 6 K- 1 )
20
Specific heat ( J g 1 K 1 )
0.14
1
1
Thermal conductivity (W m K" ) 2.5
Permittivity e
Vickers hardness (kg mm 2)
Melting point (K)
Band gap (eV) at 77 K
Effective mass

1349
0.30
0.2

1076
0.22
0.27

1154
-1.08*

"In these semimetals a negative band gap implies an overlap of the conduction and valence bands.

1133
-0.9
0.15

35.9
62
1079
-0.27
0.256

temperature limit can be expected to be raised in LDs

provided with the MQW structure. Pulse oscillation at
204 K was demonstrated in an MQW laser with a
quantum-well layer of (Pb, Sn)Te and a barrier layer
of Pb(Se, Te), which was capable of emitting at a wavelength of 6 fim (Shinohara et aL, 1985). However, even
in a simple DH-type laser of wavelength 3.27/xm
structured by (Pb, Cd) (S, Se)/PbS/(Pb, Cd) (S, Se),
pulse oscillation operation at 200K was observed
(Koguchi et al.9 1987).
2.5 Group V-VI Semiconducting Compounds
Bi2Se3, Bi2Te3 and Sb2Te3 compounds have a layered
structure with a rhombohedral unit cell (C33, hR5). The
physical and mechanical properties of these compounds
are very anisotropic arising from the hexagonal
symmetry of the crystal. Solid solutions of these narrow
band gap semiconductors have been utilized for thermoelectric devices (Lovett, 1977; Uemura and Nishida,
1988; see also Vedernikov, Chapter 20 in this volume).
The figure of merit Z for evaluating thermoelectric
materials is given by a2o/x, where a is the thermoelectromotive force, a is the electrical conductivity and
x is the thermal conductivity. Some n-type and p-type
thermoelectric materials are prepared with Bi-rich and
Sb-rich solid solutions, respectively. The optimum Z
value is attained by controlling the electron and hole
concentrations with halogens and tellurium, respectively,
as dopants. Commercial thermoelectric materials are
commonly made with a uniaxial solidification technique
in which the a axis is grown as the preferred direction
since it has better thermoelectric properties than the caxis (Uemura and Nishida, 1988). The Z value of the
material oriented along the a axis is about
3.OxIO -3 K" 1 . In order to increase the mechanical
strength, however, the Z value of most commercial
materials is suppressed to 2.0-2.6XlO -3 K" 1 by
allowing some misorientation of the a axis.
Since Bi2Te3 crystals with the layered structure show
pronounced mechanical cleavage on the c plane,
the uniaxially solidified polycrystalline samples and the
single crystals are readily broken along the solidification
direction and the c plane. It was found that the mechanical strength of Bi2Te3 can be improved by powder
metallurgy techniques without spoiling the thermoelectric
properties. Despite the fact that the thermoelectric power
and electrical resistivity of the sintered materials are
varied, optimization is again realized by controlling the
composition of the alloying element and the dopant and
by increasing the phonon thermal resistivity owing to
grain boundary scattering. Recently, anistropic sintered

materials of Bi2Te3 have been fabricated by the hotpressing technique (Ohsugi et al.9 1989). A maximum
figure of merit for n-type Bi2Te2 85Se015 of up to
3.7XlO - 3 K 1 , perpendicular to the pressing direction,
has been achieved, which is much higher than that of
single crystals (Kaibe, 1989; Kaibe et al., 1989).
These sintered materials are quite suitable for very
small thermomodule applications, for example, for
building cooling modules for optical communication
devices and equipment for semiconductor processing, etc.
2.6 Transition Metal Silicide Semiconductors
Transition metal silicides, which have been investigated
mainly as refractory materials, have interesting physical
and chemical properties owing to the continuous
changes in the valences of the transition metal atoms.
CrSi2, MnSi2_x (0.273^ x^0.250) and /3-FeSi2 are
semiconductors developed as thermoelectric conversion
materials for use at temperatures above 1200K (see
Vedernikov, Chapter 20 in this volume). Their heatresistant properties are attractive, even though the
thermoelectric power of these materials is not as large
as that of the chalcogenide thermoelectric materials, and
their electrical conductivities are comparatively high.
Of all the silicides, a degenerate semiconductor of ptype CrSi2 with a hole concentration of 1020cm~3 has
superior heat resistance and is used for a highperformance thermoelectric leg (Nishida, 1972). Also,
nonstoichiometric MnSi2 _x is a p-type, degenerate
semiconductor (Nishida, 1972; Kawasumi et al.9 1981).
It is noted that /3-FeSi2 becomes p-type by doping with
Mn and/or Al (Nishida, 1973) and n-type with Co
atoms. These materials have found use in highperformance, high temperature applications (Ware and
McNeill, 1964).
On the other hand, the thermoelectric performance
of the semimetal CoSi (its conduction and valence bands
overlap by about 0.02 eV) is inferior to that of the
silicides mentioned above; it is an n-type material with
a comparatively large thermoelectric power (Asanabe
et al., 1964). When this material is combined with ptype CrSi2, one can obtain a CoSi-CrSi2 thermocouple. As with the previously mentioned silicides,
including FeSi2, mass production technology is established (Tokushima et al.t 1972; Uemura and Nishida,
1988), and developmental research on these materials
is now in progress for thermoelectric generator
applications. Since the electric power of these thermocouples is sufficient for power supplies for safety devices
for gas and petroleum combustion instruments and so
on, various ideas have been proposed in this field.

Note that since these materials are useful for

application at higher temperatures with material costs
much lower than those of the chalcogenides, special
applications are possible, such as high-temperature
controllers for combustion and chemical reaction
instruments or furnaces and for thermoelectric generators
that utilize waste heat.
Mesotaxially formed, buried layers of silicides have
also been investigated. Such patterned layers of CoSi2
have been used as gates for permeable-base transistors,
and have been proposed for application as buried
collector contacts for high-speed bipolar transistors, as
ground planes for microstrip lines, and as mirrors in
GaAs-Si modulators. In addition to CoSi2, Ti, Fe, Cr,
Ni, and Y silicides have also been formed by mesotaxial
implantation (Celler and White, 1992).
2.7 Ternary Semiconducting Compounds
As ternary compounds, Group H 2 -III-VI 4 (defect
chalcopyrite type), Group I 3 -V-VI 4 (famatinite), and
Group I 2 -II-VII 4 (stannite) are known (see Parthe,
Chapter 14 in Volume 1). Chalcopyrite-type compounds
have been most intensively studied of all ternary semiconducting compounds. They are expressed by the
molecular formulae 1-HI-VI2 and H-IV-V2, and have
the same tetragonal structure with c/a 2 as chalcopyrite
CuFeS2 (El1, tI16) (Shay and Wernick, 1975). Highquality single crystals of 1 cm3 have been grown by the
traveling heater method (Sugiyama et al., 1989) and by
the gradient freezing method (Ando and Katsui, 1981).
The principal features of chalcopyrite-type compounds
are birefringence and a large, nonlinear optical constant.
Typical compounds are AgGaS2, AgGaSe2, ZnGeP2
and CdGeAs2. These compounds with a wide band gap
and large lattice distortion show a degeneracy of the
refractive index at a specified wavelength (Lotspeich,
1979). Recently, optical filters were prepared utilizing
this property (Horinaka et al., 1989).
Chalcopyrite-type semiconductors are direct transition
type. Among them, CuAlS2 and CuGaS2, belonging to
the 1-HI-VI2 compounds, have Eg above 2eV and
thus show potential for use in light-emitting devices in
the visible region. The conduction type of as-grown
crystals is usually highly resistive p-type, the resistivity
of which can be decreased by annealing under a
constituent chalcogen atmosphere.
High-quality, impurity-doped, heteroepitaxial films
have been studied by means of MOVPE using cyclopentadienyl(triethylphosphine)copper(I) (CpCuTEP) as
a Cu source material (Hara et al., 1987) in a closed-tube
halide VPE method with chloride sources and MBE.

A site
B site
Anion

Figure 3. Spinel structure H l 1 (cF56) (unit cell of A sites).

The heavy lines define a subcell which is one fourth of the unit
cell of A sites

Recent interest has focused on CuInSe2 as a material

for a high conversion efficiency, thin-film solar cell
without optical degradation. CuInSe2 has an Eg of
about 1.0 eV at room temperature and a high absorption
coefficient, above 1 x 105 cm"1. The efficiency of solar
cells with polycrystalline CuInSe2 films, produced mainly
by means of selenization and vacuum evaporation
methods, exceeds 14% (Michell et al., 1988; Basol and
Kapur, 1990). Development of CuInSe2 solid solutions
and tandem structures with amorphous Si are in
progress to realize conversion efficiencies above 20%.
2.8 Ferromagnetic Semiconducting Compounds
There are many kinds of magnetic semiconductors; most
are binary and ternary compounds. The attention of
many researchers has focused on a group of ferromagnetic chromium chalcogenide-based ternary compounds represented by MCr2X4, where M stands for a
nonmagnetic metal, such as Cd or Hg, and X is S or
Se. They, like ferrite, are of spinel structure (Hl 1 ,
cF56), as shown in Figure 3. The portion of the structure
defined by the thick solid lines corresponds to one fourth
of the unit cell in the spinel structure. All M ions lie
at A sites, each of which is a lattice point at the center
of a tetrahedron of anions. The Cr ions all lie at B sites,
each of which is a lattice point at the center of an
octahedron of anions. Neighboring pairs of Cr ions are
ferromagnetically bonded.
Table 4 lists the properties of major chromium chalcogenide-based ternary compounds. Compounds having Cu
atoms at A sites are metallic conductors and not ferromagnetic; compounds having Zn atoms at A sites are
antiferromagnetic. Although these compounds are not
ferromagnetic, they are listed in Table 4 because a comparison of ferromagnetic materials with nonferromagnetic

Table 4. Properties of chromium chalcogenide based ternary compounds

Lattice
constant
Compound
CdCr2O4

a (A)
U
(lA = 0.1nm) parameter0
8.596

Magnetic
moment
(4.2 K)foeper
molecule)

0.396

CdCr2S4
CdCr2Se4
HgCr2S4
HgCr2Se4
MnCr2S4
FeCr2S4
CoCr2S4
CuCr2S4
CuCr2Se4
CuCr2Te4
ZnCr2O4
ZnCr2S4

10.239
10.745
10.244
10.743
10.108
9.989
9.916
9.822
10.337
11.137
8.328
9.983

0.375
0.383
0.392
0.390
0.387
0.385
0.382
0.381
0.380
0.379
0.387
0.385

ZnCr2Se4

10.443

0.384

5.55
5.94
5.35
5.64
1.27
1.6
2.4
4.58
4.94
4.93

Curie point
Tc
(K)
Antiferromagnetic,
Neel point TN = 9 K
86
121
36
110
80
192
235
420
460
365
T N = 16 K
TN=18K
TN = 2 3 K

Asymptotic
Curie point

Resistivity

(room temperature)
(ficm)

(K)
-83
+ 152
+ 204
+ 137
+ 192
-12+10
-234+14
-390 40
+ 390
+ 465
+ 400
-392
+ 18
+ 115

The u parameter gives the lattice site of the anion (ideally u = 0.375, see Figure 3).

~10
-104

~10 9
~10 3
~1(T 4

~io- 4
~10~ 4

Magnetic
type
Antiferromagnetic
Ferromagnetic
Ferromagnetic
Ferromagnetic
Ferromagnetic
Ferrimagnetic
Ferrimagnetic
Ferrimagnetic
Ferromagnetic
Ferromagnetic
Ferromagnetic
Antiferromagnetic
Antiferromagnetic

Conduction
mode

Semiconductor
Semiconductor
Semiconductor
Semiconductor
Semiconductor
Semiconductor
Semiconductor
Metallic
Metallic
Metallic
Semiconductor
Semiconductor
Semiconductor

materials, both having the same crystal structure, is of

significance in understanding the chromium chalcogenide-based magnetic semiconductors.
With the above exceptions, chromium chalcogenide
ternary compounds are both ferromagnetic and semiconducting, and they show unique conducting behavior
as general magnetic materials. These compounds have
been the objects of numerous studies on their electrical,
magnetic, and optical properties. A combination of
these properties seems promising for new electronic
devices and is of worldwide interest from the viewpoints
of physicochemical qualities and applications. Ferromagnetic semiconductors containing Cd or Hg atoms
at the A sites can be doped with impurities to exhibit
p-type or n-type conduction. These materials have
charge carriers with the largest mobilities among all the
magnetic materials. An appropriate chemical treatment
makes the magnetic anistropy of these compounds less
significant and forces them to have a narrow microwave
absorption width. Thus, we may be able to find new
fields for their application.
CdCr2Se4 is a material that has been studied as a
thin film for p-n junction elements (Wen et al., 1968).
HgCr2Se4 and CdCr2Se4 are promising materials for
tunable semiconductor lasers for the mid-infrared region
(Masumoto and Koguchi, 1976, 1982; Koguchi and
Masumoto, 1978; Takahashi et al., 1983). The wavelength of their laser oscillations changes with ambient
temperature and applied magnetic field; the rate of
change is about 10 times that of IV-VI semiconducting
compounds. For example, the optical absorption edge
of HgCr2Se4 is 0.83 eV (1.49 /on) at 300K; it sharply
decreases with decreasing temperature (a phenomenon
known as the red shift), reaching 0.28 eV (4.43 fim) at
4.2 K (Harbeke et al., 1968; Lehmann and Emmenegger,
1969). At the Curie temperature (about 120 K), the edge
is further decreased by 10% with the application of a
10 kG magnetic field (Lehmann and Emmenegger, 1969).

3. Electrical Applications

These characteristics make Si highly suitable for

integrated electronic applications. Although Si at present
is certainly in the very large scale integration (VLSI) age,
attempts to realize with compound semiconductors
better integrated circuits (ICs) than those with Si have
been increasingly made.
In materials for high-speed device applications, high
electron mobility and high saturation carrier velocity
are required. Table 5 shows a comparison of characteristics among high-mobility Group IH-V semiconductors.
The narrower the band gap and the smaller the effective
mass, the higher is the electron mobility. Saturation
velocity has the same trend as the mobility. The mobility
of carriers is affected by various scattering mechanisms,
and generally scattering by optical phonons is predominant in compound semiconductors.
GaAs, InP, InSb and InAs show high electron
mobilities and high saturation velocities relative to Si.
They therefore are suitable for high-speed device
applications. In GaAs, semi-insulating crystals are
commercially available, and thus low power consumption is expected, as well as high-speed operation.
Both InSb and InAs have higher electron mobility than
GaAs. However, conduction in both of these semiconductors, owing to their band gaps at 300 K, is intrinsic
and, hence, low temperatures are needed for operation
with a high margin of reliability.
There are two conceptions in the further development
of high-speed integrated circuits using compound semiconductors. One direction aims at speeds above those
of Si ICs, and the other direction is toward Si/VLSI
devices resistant to radiation damage and with low
power consumption for space development applications.
In the latter case speed is sacrificed. In the former, as
well as in GaAs metal Schottky field effect transistors
(MESFETs), the IC being at an LSI level, superhighspeed devices such as the high electron mobility
transistor (HEMT), the heterojunction bipolar transistor
(HBT), the permeable-base transistor (PBT), the static
inductance transistor (SIT), and the metal insulator
semiconductor field effect transistor (MISFET) are
objectives.

3.1 Materials with High Electron Mobility

Si has the following characteristics: (i) its electron
mobility is relatively high; (ii) high-purity and highperfection single-crystal wafers up to eight inches in
diameter are commercially available; (iii) insulating
SiO2 is quite compatible with Si and is highly useful for
diffusion masking and passivation; (iv) its strength is
high; and (v) its Eg is moderate for integration with
other materials.

3.2 GaAs MESFET ICs

The driving force for GaAs FET IC developments is
the exploitation of electron mobilities several times
larger than that of Si by direct circuit integration
on a semi-insulating substrate using ion implantation and extension of recently developed, ultrahighresolution processing techniques to Group IH-V
semiconductors.

Table 5. Property comparison of high electron mobility semiconductors

Property
_ (eV) 300 K
77 K
Lattice constant (nm)
Electron affinity (eV)
Electron mass0 m/m0
me

InSb

InAs

GaSb

InP

GaAs

Ge

Si

0.18
0.228
0.64787
4.59

0.36
0.41
0.6058
4.9

0.67
0.80
0.6095
4.06

1.35
1.40
0.58687
4.38

1.43
1.47
0.5653
4.07

0.665
0.7343
0.5658
4.13

1.107
1.1532
0.54198
4.01

0.024
0.41

0.041
0.26

0.0810
0.56

0.0667
0.71

0.0815
0.32

0.1905
0.50

0.0139
0.32
Hall mobility at 300 K (cm2 V- 1 s"1)
80 000
450
/*H
IxIO6
Zi6(Cm2V"1 s"1) at 77 K
Saturation drift velocity vm
5.2 XlO7
(cms1)
5.6 XlO2
Saturation field Em (V c m 1 )
Intrinsic carrier concentration
2 XlO16
nx ( c m 3 ) 300K
77 K
3.8 XlO9
m

33 000
450
1.5 XlO5

4000
1420

4500
150
4.5 XlO4

8500
435

3800
1800
3.6 XlO4

1450
500
2.4 XlO4

3.2 XlO7
4 XlO3

<6xlO6

2.3 XlO7
1.2X104

1.7-2.OxIO 7
5.2 XlO3

6 XlO6
4 XlO3

IXlO 7
2 XlO4

1 x 1015
3.5 XlO3

2.7 XlO12

1.4 XlO7

1.2 XlO7

1.8X1013

1.2 XlO10

Currently under development are MESFETs, metal

oxide semiconductor (MOS) FETs, junction-type FETs
and so on, the FET being a fundamental device for
GaAs ICs. Because of the underdeveloped planar techniques for p-n junctions, the leading device type is the
MESFET.
Studies of GaAs ICs started in 1970, at which time
the Si LSI was developed. By 1974, the normally-on
GaAs IC was proposed. In 1977 and 1978, integrated
circuits and normally-off ICs were produced,
respectively. Since then, remarkable progress in the
degrees of integration and speed achieved has been made
yearly.
A MESFET is a transistor with a structure which is
formed from a Schottky gate electrode and source and
drain ohmic electrodes on an n-type semiconductor
layer. A metal-semiconductor ohmic contact is used as
a gate electrode in the MESFET, instead of a p-n
junction. The operational function is such that the gate
input signal changes the spread of the Schottky contact
depletion layer, and thus changes the thickness of the

channel layer, which provides the conductance between

the source and drain. Figure 4 shows the two typical
structures of MESFETs. There are normally-on
elements (depletion type) and normally-off elements
(enhancement type) in fundamental circuit constructions
of MESFETs. In the normally-on type, the current is
cut off by applying a negative electric field. On the other
hand, in the normally-off type, the source and drain
pinch off at the built-in voltage and are turned 'on' by
a positive voltage.
Although normally-on types operate at very high
speed, the power consumption is large, the circuit is
complex, the occupied area is large, and hence they are
unsuitable for integration. Normally-off types have
simple circuit construction and small power consumption, and thus are suitable for high integration.
However, normally-off types possess several problems
still to be solved: the manufacturing conditions are
severe, the logic amplitude is small, the limitation on
the pinch-off voltage is strict and the switching speed is
rather low. Developments of GaAs FET ICs are making
(b) Normally off type

(a) Normally on type

Depletion
layer
Drain
Gate
Source

Source

Depletion
layer
Drain
Gate

Layer of
n-type
Semi-insulating
GaAs

Figure 4. Fundamental structures of the two MESFET types

Semi-insulating
GaSa

steady progress through a reiterative process of experimental production and evaluation.

A serious problem on the performance side of the
GaAs MESFET LSI is that, compared with the Si
LSI, the speed is not as high as expected. Furthermore,
there is a problem of low yield. Uniformity of characteristics is important for LSI, and hence reliable achievement of a high-quality, semi-insulating GaAs substrate
is necessary.
3.3 New High-speed Devices
Superhigh-speed devices based on new operational
principles have been proposed as a result of marked
progress with heteroj unction techniques and highresolution processing techniques. Here, HEMTs, HBTs,
PBTs, SITs and MISFETs are introduced.
3.3.1 HEMTs
A modulated, doped GaAs/AlGaAs heteroj unction is
a junction formed from undoped GaAs and AlGaAs
heavily doped with an n-type impurity. In single heterojunctions and in a 'superlattice' using modulated,
doped GaAs/AlGaAs, the mobility, especially at low
temperatures, is greatly increased compared to GaAs

of a nearly equal carrier concentration (Dingle et al.,

1978).
Since in such a junction as GaAs/Al0 3Ga0 7As the
electron affinity of GaAs (iig = 4.07 eV) is larger than
that of Al03Ga07As ("g = 3.75 eV), electrons supplied
from the donor in n-type AlGaAs flow to the GaAs side
and are accumulated in quite a narrow region of the
heteroj unction, which then forms a two-dimensional
electron gas (Dingle et al., 1978). Conduction electrons
are spatially isolated from the donor and are less scattered
by ionized and neutral impurities; therefore, the mobility
is exceedingly high (Hiyamizu and Mimura, 1981).
The HEMT is a superhigh-speed device utilizing a
single, modulation-doped heterojunction (Mimura et al.y
1980). Fundamental to the HEMT is a heterojunction
gate. Figure 5 shows the structure and schematic band
diagram near the gate of the MESFET and the heterojunction gate FET. In the MESFET, the channel layer
must have impurity concentrations of about 1016cm~3
in order to switch the channel and to obtain the required
channel current. Consequently, subsequent impurity
scattering caps the electron mobility at the channel. On
the other hand, in heterojunction types impurity
concentrations in the channel are below 1015 cm~3. A
heterojunction gate with high impurity concentrations
is constructed by forming a layer with a larger band gap

(a) MESFET

(b) Heterojunction gate FET

Heteroepitaxial film (large n)

Source

Drain

Gate

Source

Gate

Drain

Epitaxial
film
Semi-insulating GaAs

Channel layer

Semi-insulating GaAs

Channel layer

Metal

Substrate

Figure 5. Comparison of MESFET and heterojunction gate FET (upper, structures; lower, schematic band diagrams near the gates)

Next Page

and smaller electron affinity than the channel. A

sufficient supply of the required electrons is obtained
from the high-concentration layer.
In the HEMT structure, the low-field electron
mobility is 8 x 103 cm2 V" ] s"l at 300 K, twice that of
the GaAs MESFET, and the saturation velocity is
1.5-1.9XlO7CmS"1; hence the current gain cutoff
frequency/T in switching is high. For example,/T was
80 GHz at a gate length of 0.25 /mi, twice that of a GaAs
MESFET with the same gate length (Abe et al., 1987).
As materials for HEMTs, AlGaAs/InGaAs/GaAs
(Chao et al., 1987) and InAlAs/InGaAs/InP (Henderson,
1987) have also been studied.
3.3.2 HBTs
The concept of the HBT is defined in Shockley's patent
(Shockley, 1948). With the advancement of heterojunction technology and ion implantation technology,
devices have since been experimentally produced and
their performance confirmed.
The features of the HBT are a high degree of freedom
in structural design and high speed, because movements
of electrons and holes are separately controlled
(Kroemer, 1982).
Injection of minority carriers from base to emitter
is so restricted, compared to homojunction bipolar
transistors, as to allow doping with high concentrations
of carriers. Thus, base resistance, base thickness and
conductance between base and collector decrease, and
the design of transistors with high speed and excellent
high-frequency characteristics can be realized.
Recently, n-p-n-type (Beneking and Su, 1981) and
p-n-p-type (Su et a/., 1981) bipolar transistors of homojunction or heterojunction types using GaAs with
AlGaAs have been examined. Also, an HBT with an
/ T of 105 GHz and a maximum oscillation frequency
/MAX of 55 GHz was developed through improvements
in device structures and processing (Ishibashi, 1987).
A review article dealing with recent HBT studies has
been published (Sugeta and Ishibashi, 1990).
3.3.3 PBTs and SITs
In both PBT (Bozler and Alley, 1980) and SIT
(Nishizawa et al., 1975) devices there is a vertical FET,
as is expected for superhigh-frequency operation.
The PBT has the same structure as a base-embedded
SIT (Alley et al., 1980; Vojak et al., 1983), and consists
of a grating made from a thin film of tungsten which
is embedded in an epitaxial GaAs layer. The electrons
flow from emitter to collector through slits in the metal

film. The tungsten grating is used as a gate, and a gate

length of about 0.1 /mi is easily attainable. In a
theoretical comparison of PBT, MESFET and HEMT
devices with very short gate lengths, an /MAX above
500 GHz was projected in PBTs for an effective gate
length of 0.15^m (Das, 1987).
The structure of an SIT is fundamentally the same
as that for a PBT but the dimensions and the impurity
concentrations of the active layer differ from those of
the PBT. In an SIT, the distance between source and
gate is extremely short, and the structure has a
sufficiently small resistance Rs between source and gate
that the product of Rs and mutual conductance Gm is
< < 1 (Sugeta, 1986). Furthermore, the impurity concentration is so low that the whole active layer is
depleted at the gate voltage.
The SIT is also useful for high electric power
applications (Nishizawa, 1986).
3.3.4 MISFETs
The electron saturation drift velocity of InP is 1.5 times
as large as that of GaAs. Theoretical (Frey and Wada,
1979) and experimental studies of high-speed InP FETs
have been made.
To realize an MESFET in InP is difficult, owing to
the small potential barrier of the Schottky junctions in
the n-type layer. In an InP MIS-type structure, an
electron accumulation layer and an electron inversion
layer can be easily formed at the insulating film-InP
interface (LiIe et al., 1978), and hence InP is suitable
for MISFET applications. This situation is quite in
contrast to the case of GaAs (Meiners, 1978).
The largest factor in the realization of high-speed InP
MISFETs is the formation of a high-quality gateinsulating film. Enhancement-type MISFETs using
SiO2, Si3N4, P3N5 or anodic Al2O3 as the gate-insulating
film have been investigated. The channel layer was
directly formed on a semi-insulating substrate
surface in InP, and an effective mobility of about
3000Cm2V-1S"1 was attained for an Al2O3/native
oxide double-gate structure.
A problem common to InP MISFETs is the drift of
the drain current. Recently, drain current drift below
10% of the initial value was attained (Hasegawa,
1984). From this result, it was speculated that the drift
is probably owing to interface levels being distributed
energetically and spatially below a conduction band.
Furthermore, enhancement-type MISFETs using
InGaAs have been studied and an effective mobility of
1500cm2 V - 1 S" 1 was obtained using plasma oxidation
of Al2O3 as the gate film.