Handbook of Compound Semiconductors. Growth, Processing, Characterization, and Devices

918 Pages·1996·17.565 MB·English

by  Paul H. Holloway and Gary E. McGuire (Eds.)

Preface
This book is a state-of-the-art reference on the growth and process-
ing of compound semiconductors. The leading experts in the important
growth techniques, processing steps, characterization methods, packaging,
and devices have contributed their knowledge. While the scope of the book
is compound semiconductors, there are so many different semiconductors
being studied and used that complete coverage of all materials is impossible
in boonoek . Therefore the emphasis in this book si on gallium-arsenide- and
indium-phosphide-based materials. Several other III-V and some II-VI
compound semiconductors are discussed where they provide particular
insight or illustrate specific properties and/or processes.
Chapters in the book provide a complete overview of the technolo-
gies necessary to grow bulk single-crystal substrates, and grow hetero- and
homoepitaxial films using moleculabre am epitaxy (MBE) omre tal-organic
chemical vapor deposition (MOCVD). Technologies necessary to process
compound semiconductors into test structures and devices are covered,
including electrical contacts, dielectric isolation, interface passivation, ion
implantation, wet darnyd etching, and rapid thermal processing. Techniques
to characterize the materials and devices using electrons, ions, and photons,
are described. While the emphasis of the book si on materials growth and
processing, the technologies are placed in perspective by a review of the
important electronic and optoelectronic devices, and epitaxial lift-off, and
other device packaging issues.
vii
viii Preface
With this complete coverage of the critical topics, we believe the
book will be a valuable reference for persons currently performing research
on compound semiconductors. It will also be an excellent reference for
advanced graduate courses in materials science, electrical engineering, and
applied physics. In our judgement, the authors of the chapters have provided
exceptionally comprehensive, authoritative work, and for this we are ex-
tremely grateful.
September, 1995
GainesviUe, Florida Paul .H Holloway
Research Triangle Park, Gary E. McGuire
North Carolina
Contributors
Shin-ichi Akai Sidney I. Ingrey
Sumitomo Electric Industries Ltd. Bell Northern Research
Itami, Hyogo, Japan Ottawa, Ontario, Canada
Kambiz Aiavi Nan Marie Jokerst
Department of lacirtcelE gnireenignE Microelectronics Research Center
University of Texas at Arlington Georgia Institute of Technology
Arlington, TX Atlanta, GA
Scott A. Chambers Kevin S. Jones
Molecular Science Research Center Department of Materials Science &
Pacific Northwest Laboratory gnireenignE
Richland, WA University of Florida
Gainesville, FL
Eric Y. Chan
Boeing Company Avishay Katz
Seattle, WA Standard Motor Products, Inc.
Long Island City, NY
Stephen W. Downey
AT&T Bell Laboratories Richard Y. Koyama
Murray Hill, NJ TriQuint Semiconductor
Beaverton, OR
Paul H. Holloway
Department of Materials Science & Derek L. Lile
Engineering tnemtrapeD of Electrical gnireenignE
University of Florida Colorado State University
Gainesville, FL Fort CollinCsO,
/x
x Contributors
Vinod Malhotra Brian J. Skromme
tnemtrapeD of lacirtcelE gnireenignE tnemtrapeD of lacirtcelE gnireenignE
University of Hawaii Arizona State University
Honolulu, HI Tempe, AZ
Gary E. McGuire Eberhard Veuhoff
Microelectronics Center of North snemeiS AG
Carolina Munich, Germany
Research Triangle Park, NC
Carl W. Wilmsen
Stephen J. Pearton tnemtrapeD of lacirtcelE gnireenignE
AT&T Bell Laboratories Colorado State University
Murray HiNlJl , Fort Collins, OC
Rajendra Singh Masamichi Yokogawa
tnemtrapeD of lacirtcelE gnireenignE Electric Sumitomo Industries Ltd.
Clemson University Itami, Hyogo, Japan
Clemson, SC
NOTICE
To the best of our knowledge the information in this publication is
accurate; however the Publisher does not assume any responsibility
or liability for the accuracy or completeness of, or consequences
arising from, such information. This book is intended for informational
purposes only. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use by the Publisher.
Final determination of the suitability of any information or product
for use contemplated by any user, and the manner of that use, is the
sole responsibility of the user. We recommend that anyone intending
to rely on any recommendation of materials or procedures mentioned
in this publication should satisfy himself as to such suitability, and
that he can meet all applicable safety and health standards.
1
Bulk Crystal Growth
Shin-ichi Akai and Masamichi Yokogawa
INTRODUCTION
III-V compound semiconductors are widely used as substrates for
optical devices such as LED's and laser diodes and for electronic devices
such as FET's, HEMT's, HBT's and IC's. These applications are becom-
ing key elements ni an advanced information society. In this chapter, our
focus is on GaAs and InP and recent advancements in their crystal growth
technology.
Horizontal Bridgman (liB) and Liquid Encapsulated Czochralski
(LEC) are two representative growth methods of III-V compound crystals.
The HB method is favorable for reducing the dislocation density and is,
therefore, used ni providing substrates for optical devices. The LEC method
is advantageous for increasing the crystal diameter and is, therefore, used in
providing substrates for electronic devices. Incorporating the advantages of
both growth methods, practical applications of early hot wall crystal growth
techniques[ll such as Vertical Gradient Freeze ,12[)FGV( Vertical Bridgman
13[)BV( and Liquid Encapsulated Vertical Bridgman (LEVB)Ial have re-
centlbye en tried. The basic objectives ni the development of crystal growth
technology are larger crystals (diameter and length), reduction of crystal
defects (such as dislocation) and higher purity. In addition, stoichiometry
control is important ni the case of compound crystals.
Compound srotcudnocimeS
1.0 REDUCTION OF DISLOCATION DENSITY
Dislocations affect device performance. Therefore, the reduction of
dislocation density is a very important issue in crystal growth technology.
Various techniques to achieve low dislocation density are summarized in
Fig. .1 Among them are the reduction of thermal stress (or reduction of
temperature gradient) and the increase of critical resolved shear stress. It
has bccn ixWroper by Matsumoto et al. ]5[ that a roof-shaped thermal baffle
installed on the top of a heat shield alleviated very strong gaseous convec-
tion in a chamber, thus reducing the temperature gradient in the B203 layer
to 30 - 60~ This gradient is about one third of the temperature
gradient in the conventional LEC method. The radiation effect becomes
more significant in LEC-grown crystals having a high melting point. In a
conventional single-zone heater puller, the heat outflow from the surface of
growing crystals is large because of the heat radiation toward the cold
chamber wall. Consequently, the temperature gradient at the solid-liquid
interface bex~mes large. On the other hand, in a puller with a multi-zone
heater, the heat inflow to the surface of a growing crystal is large because of
the heat radiation from the after-heater and consequently, the temperature
gradient can be reduced. ]6[ A low temperature gradient means that the
Improved hot zone
Reduction of
, Multizone heater
thermal stress
' ' ('Reduction of )-~
FEC (Fully Encapsulated Czochralski)
\ temperature gradient
Ambient gas
Prevention of
As (or P) dissociation
Control of yrtemoihciots
Increase J Melt composition
of CRSS
Isoelectronic doping
, In, B (for GaAs)
Ga, As (for InP)
erugiF .1 suoiraV sehcaorppa ot eht noitcuder of noitacolsid ni ytisned V-IH dnuopmoc
.slatsyr crotcudnocimes
Bulk Crystal Growth 3
solidified crystal stays at elevated temperature longer and therefore, the
group V element tends to dissociate from the crystal surface. LEC growth
with a low temperature gradient requires a special measure to suppress the
dissociation of group V elements. Crystal growth in an arsenic (or phospho-
rus) atmosphere ~IH7[ and latsyrc growth with lluf noitaluspacne by B203 11I[
haveb een .detpmetta The yrtcmoihciots of eht latsyrc also stceffa eht
dislocation density. ]41[']21[
The effect of impurity doping on the dislocation density and dislo-
cation mobility in semiconductor crystal is well-known. In order to obtain a
semi-insulating crystal for IC applications, an isoeleetronic impurity must
be used. For GaAs, these dopants are N, AI, In, Sb, B and ]SB-lSlt.p
However, to date, only indium doping has been effective for obtaining low-
dislocation, semi-insulating CraAs crystals of an industrially useful size
(that is, with a diameter larger than 2 inches and a lot size greater than 50
wafers).[ 112I-]91 Doping with indium, together with the VM-FEC (Vertical
Magnetic field applied, Fully Encapsulated Czochralski) method, has en-
abled the growth of dislocation-free and striation-free GaAs crystals. [lq A
4" diameter dislocation-free GaAs crystal has been obtained using this
method. 122[ While In doping yields materials of high quality, it has the
disadvantages of smallleort sizes and higher wafer costs in comparison with
conventional undoped crystals. This is due to the fact that a cellular
structure is generated during crystal growth, thus greatly reducing material
yield. This phenomenon occurs when the In concentration in the melt
exceeds a critical value. The melt then goes to a compositional supercooled
state. In order to retard the generation of this cellular structure, two
techniques have been tried. The first method is the G/R technique ]32[ in
which the temperature gradient (G) and the growth rate (R) are simulta-
neously controlled, and the second is a double crucible method 142[ in which
the segregation coefficient of In in sAm.C is artificially adjusted to unity.
The In metdhoopdi ng is also effective in HB sAarC crystal growth. 1~[
Because the HB method has a lower temperature gradient compared to the
LEC method, the In concentration necessary for reducing the dislocation
density is relatively small. Horizontal Zone Melting fflZM), which can
alter an impurity segregation coefficient, has been tried in order to obtain a
uniform impurity distribution along the growth ]62t.noitcerid
Isoelectronic impurities for InP are Ga, As and Sb. An impurity
segregation coefficient is not generally unity, and therefore, the impurity
concentration changes along the growth direction. Accordingly, co-doping
of two kinds of impurities, where the segregation coefficient of one impurity
dnuopmoC srotcudnocimeS
is smaller than unity and that of the other impurity is larger than unity, has
been trim to obtain a uniformd istribution of total impurity concentration in
InP crystal.[ 16
2.0 HB GaAs
The HB method is extensively employed as a low dislocation
density crystaglr owth technique. This is because reduction of thtee mpera-
ture gradient is relatively easy and precise control of the stoichiometry is
possible in comparison with the LEC method. Basic approaches to achieve
a low dislocation density have already been described in the previous
section. This section addresses the methods of cooling a solidified crystal
and the solid-liquid interface epahs in order to obtain very low dislocation
density CraAs crystals.
2.1 Cooling Methods
Figure 2 shows a schematic diagram of the BIF method. An entire
ingot is kept in the 2 T zone after solidification at the solid-liquid interface
and then gradually cooled to room temperature. A temperature distribution
is generated in the crystal during this cooling process, and ~rdingly,
thermal stress is induceA in the crystal. Since CraAs is more susceptible to
dislocation's generation and propagation at low temperature by lower
thermal stress in comparison with i, S optimization of the cooling process is
critical for the reduction of dislocation density. Figure 3 shows the influ-
ences of the cooling process and crystal diameter on the dislocation density.
When the cooling rate is - 5 10~ in the temperature range between 900~
and 1,200~ the etch pit density (EPD) of 2" diameter crystals can be
reduced to lower than 5,000 cm .2- However, when the cooling rate is 01 ~
100~ in the same temperature range, the EPD increases to about twice
this value.
2.2 Influence of Solid-Liquid Interface Shape
It has been found that the generation of lineage, which is a collection
of dislocations, is related to the shape of the solid-liquid interface in LEC ]72[
and GF (Gradient Freeze) growth. ]s2[ We have investigated the relation
between the solid-liquid interface shape and the dislocation density by doing
Bulk Crystal Growth 5
AB etching[ ]92 on Si-doped HB GaAs crystals. These crystals were grown
along the 11[ l] crystallographic direction, and sample wafers werteh en cut
along the 12[ l] direction, which is parallel to a free surface ofthe ingot. The
results are shown in Fig. 4. The shape of the solid-liquid interface of case
(a) is nearly flat and the EPD is quite low across the wafer. On the other
hand, the solid-liquid interface shape of case (b) is not synunetric from right
to left and shows a strong concavity toward the melt. The EPD in the
corresponding region is high. Figure 5 shows the EPD maps of a 2"
diameter Si-doped HB GaAs crystal grown in a thermal environment, which
makes the solid-liquid interface shape symmetric and fiat. It shows that the
entire ingot has an EPD of less than 400 em .2-
_~- i 250 ,~T (1240---- ! 245~ )
~.- t238 - " "
~u 1200 2T (1100...1220oc)
-~ t 001
<
UJ
a.
Ts (6Q0----620"C)
LU \
Quartz Boat
Figure 2. Schematic diagram of three-temperature zone 3]]-1 (Horizontal Bridgman)
crystal growth method.
dnuopmoC srotcudnocimeS
)a(
10 4
4"
lE
E
L)
o , t-
n 3"~
w A ~ -
A
103 A "2 #,
I I I I I ,I i
40 60 80 001
DIAMETER OF WAFER [mm]
Figure 3. Cooling rate dcpcndcncc of EPD of HB grown CmAs crystal. Cooling rate in the
temperature range between 900~ and 1,200~ is (a) 10 ~- 100~ and (b) 5 -- 10~
(a) (21]') plane
o.'o+x( )~m
0 O ~ 0 O ! O
~sqOOOO
Ave 110cm ~"
Max 700cm 2"
(b) (2~T) plane
(XlO2cm.2)
/o o, 2,,,oN
X l~176176176176 /
middle----.~x \~, t o o (cid:12)9 o o , , +,mz-.~t,,,m. ,,=J/ .
kko + o o ,o.,,,.-1/
\\ + 2 t o +o't3ua/
+o.tto,,,
Ave e 960cm
Max 5,200cm 2"
Figure 4. Relationship between solid-liquid interface shape and dislocation density.
Solid-liquid interface shape is (a) flat, and (b) not symmetric from fight to left.