 |
Description  |
|
|
TECHNICAL FIELD
The present invention relates to methods for forming monocrystalline III-V
arsenide-nitride semiconductor materials, especially those materials
having a high bandgap (greater than 2.0 eV). More particularly it relates
to direct bandgap light emitting materials, preferably those with cubic
crystal lattices, which can be used to make devices such as semiconductor
lasers that emit light with frequencies in a range that more than span the
visible spectra.
BACKGROUND ART
Semiconductor lasers that emit light in the long wavelength visible (red)
and near infrared spectra have been known for many years. To date,
however, it has been difficult to produce such lasers that emit light in
the short wavelength, blue region of the visible spectra. A wide variety
of applications await such "blue" lasers, should they become commercially
available.
Certain II-VI semiconductor compounds such as zinc selenide sulphide
(ZnSeS) have been considered promising candidates for blue lasers. In an
article entitled "Blue-green laser diodes", Applied Physics Letters, v.
59, Sep. 9, 1991, M. A. Haase et al. describe achieving lasing action at a
wavelength of 490 nm with a ZnSe based structure under pulsed current
injection at 77 K. More recent advances in this material system using a
remote plasma source for the introduction of radical nitrogen (N) atoms
have resulted in the demonstration of pulsed laser operation at room
temperature at a wavelength of 480 nm.
The ZnSeS based structures have several limitations, however, due to
conduction band alignment and the activation energy of defects. The
confinement energy of the cladding layers limits efficient operation of
the ZnSeS system to a wavelength of 530 nm, as the electron confinement is
too small for efficient operation of laser diodes at 480 nm or shorter
wavelengths. The low activation energy of defects in the crystal lattice
of such a material limits the processing and growth temperature of the
structure to no more than a few hundred degrees Celsius.
The ZnSeS materials are further complicated by the problems associated with
growth on a gallium arsenide (GaAs) substrate. The interface between the
ZnSeS and GaAs results in a cross doping of the compounds. The Ga and As
act as dopants in the ZnSeS and the Zn, Se and S act as dopants in the
GaAs. As a result the interface between the two compounds becomes quite
resistive. Laser diodes fabricated with this interface require operating
voltages in excess of 20 V. The heat dissipated from this is also high,
preventing continuous wave (CW) operation at room temperature.
An alternative to the II-VI compounds for blue light emission are the III-V
semiconductor compounds, such as AlGaInN. The advantages of the III-V
material systems are several. First, they include materials having large
direct bandgaps ranging from 2 eV to 6 eV. Second, large energy
differences between the valleys of the lowest direct bandgaps and those of
the lowest indirect bandgaps exist. Third, electrons and holes can be
satisfactorily confined in heterostructures, including quantum well
structures, due to large confinement energies. Fourth, production of these
materials is compatible with metal organic chemical vapor deposition
(MOCVD) growth reactors. Fifth, the materials typically have low
resistance to n-type and p-type doping. Finally, the III-V materials can
avoid the cross doping and defect creation and propagation problems that
plague the II-VI materials.
AlGaInN materials have so far been grown only on mismatched substrates such
as Si, SiC, single crystal Al.sub.2 O.sub.3 and MgO. Mismatched crystal
lattices tend to have defects that absorb light, lowering the efficiency
of light generation and reducing lasing potential. Defects also tend to
propagate through crystals, lowering the useful life of the crystals.
Another difficulty is that AlGaInN type materials typically have a native
wurtzite, or generally hexagonal, crystal lattice, which has few
convenient cleavage planes to form mirrored facets for Fabry-Perot
reflection and which cannot be easily grown to exactly layered depths for
quantum wells or other cladding confined structures. For laser diodes, it
is desirable that the heterostructure material have a generally cubic
zinc-blende crystal lattice. While in AlGaInN materials the cubic
structure is metastable at typical production temperatures, and thus not
impossible to form in principle, to actually produce such a cubic crystal
lattice material by epitaxial growth techniques appears to require a
substrate that is not only lattice matched to the AlGaInN type material
but also has a cubic lattice that induces the growth of a cubic substrate
lattice of the AlGaInN type material layers. Substrates which have
commonly been used for group III-V high bandgap semiconductors, such as
basal plane {0001} sapphire (Al.sub.2 O.sub.3) and .alpha.-SiC, have the
disadvantage of producing the less desirable wurtzite crystal lattice
structure.
In U.S. Pat. No. 5,146,465, Kahn et al. describe growing layers of AlGaN
with alternating concentrations of Al and Ga on an AlN buffer layer which,
in turn, was grown on an Al.sub.2 O.sub.3 substrate. Kahn et al. appear to
overcome some of the difficulties inherent in the wurtzite lattice by
polishing walls of the device and adding mirrors. However, cleaved facets,
when available, are inherently better mirror surfaces than etched or
polished surfaces because of their lower defect density. In U.S. Pat. Nos.
5,173,751 and 5,274,251, Ota et al. take advantage of the lattice constant
of .alpha.-ZnO, which falls between that of several III-V nitride
materials, allowing mixing of those materials in proportions calculated to
match the substrate lattice of ZnO, to form AlGaInN or AlGaNP layers. The
resultant crystal lattice appears to be wurtzite in form.
In U.S. Pat. No. 4,862,471, Pankove describes the growth on a gallium
phosphide (GaP) substrate of gallium nitride (GaN). Indium nitride (InN)
or aluminum nitride (AlN) layers are similarly grown to form a quantum
well light emitting device. Similarly, U.S. Pat. No. 5,076,860 to Ohba et
al. teaches a compound semiconductor material of GaAlBNP with a zincblende
(cubic) crystal lattice, grown on a GaP substrate. Ohba et al. describe
several different materials, including a GaAlN semiconductor grown on a BP
substrate and the growth of group III-V materials having ordered bonds but
non-lattice matched crystals. U.S. Pat. No. 5,042,043 to Hatano et al.
describes a semiconductor laser formed from alternately stacking BP and
GaAlN layers to form Ga.sub.x Al.sub.y B.sub.1-x-y N.sub.z P.sub.1-z
material on a GaP substrate. In all of these cases, the use of GaP as a
substrate results in a severe lattice mismatch with the nitride material
layers (about 20% mismatch for GaN). In an article entitled
"High-Efficiency Aluminum Indium Gallium Phosphide Light-Emitting Diodes,"
Hewlett-Packard Journal, August 1993, pp. 6-14, R. M. Fletcher et al.
describe gallium arsenide phosphide doped with nitrogen (GaAsP:N) and
gallium phosphide doped with nitrogen (GaP:N).
In an article entitled "Luminescence quenching and the formation of the
GaP.sub.1-x N.sub.x alloy in GaP with increasing nitrogen content",
Applied Physics Letters, Vol. 60, No. 20, May 18, 1992, pp. 2540-2542, J.
N. Baillargeon et al. teach N doping in GaP of up to 7.6% using molecular
beam epitaxy (MBE), and note that increasing nitrogen content tends to
shift emission spectra lower above a certain nitrogen concentration.
Similarly, X. Liu et al. describe observing a red shift in emission
spectra from GaP:N as nitrogen (N) concentration is increased in "Band gap
Bowing in GaP.sub.1-x N.sub.x alloys", Applied Physics Letters, Vol. 63,
No. 2, Jul. 12, 1993, pp. 206-210. Difficulties are encountered when mixed
compounds having significant amounts of both nitrogen and other group V
elements are attempted. Miyoshi et al. describe a miscibility gap for
growth of GaP.sub.1-x N.sub.x for x.gtoreq.0.04 in an article entitled
"Metalorganic vapor phase epitaxy of GaP.sub.1-x N.sub.x alloys on GaP",
Applied Physics Letters, Vol. 63, No. 25, Dec. 20, 1993, pp. 3506-3508.
Of primary concern to reliable visible laser diode operation is the
optimization of the substrate and growth buffer layers. To form a
semiconductor material having desirable lacing properties, it is
advantageous to grow such a material on a lattice matched substrate in
order to avoid the promulgation of defects that absorb light. It is also
desirable, whenever possible, that the substrate promote growth in the
semiconductor material of a cubic crystal lattice that facilitates the
relatively easy formation of cleaved reflective facets for defining
resonant laser cavities.
SUMMARY OF THE INVENTION
The present invention involves formation of monocrystalline III-V compound
semiconductor materials having at least nitrogen and arsenic at group V
lattice sites of the crystal material. Other group V elements, such as
phosphorus and antimony, can also be present at group V lattice sites in
addition to the nitrogen and arsenic. The group III atomic species can be
any combination of boron, aluminum, gallium and indium.
The exact composition of a particular material to be produced, that is, the
relative concentration of each of the group III and group V elements in
the III-V compound, is generally selected so as to substantially lattice
match with the selected substrate's growing surface. Some lattice strain
due to lattice mismatch greater that 1% can be accommodated provided the
mismatched layer is sufficiently thin to avoid formation of lattice
defects. In addition, for light emitting devices, such as laser diodes and
LEDs, the desired emission wavelength determines the required
semiconductor bandgap for the material and hence plays a major role in the
choice of material composition and of a suitable substrate. For emission
wavelengths shorter than about 620 nm, a bandgap greater the 2 eV is
required. The III-V compound arsenide-nitride materials of the present
invention are characterized by a large bandgap bowing parameter which must
be taken into account when matching up the desired semiconductor bandgap
with a possible material composition. GaAs.sub.x N.sub.1-x material, where
x.ltoreq.0.10, has a bandgap greater than 2.0 eV. Laser diodes further
require that the selected material be a direct bandgap material, that is,
one where the energy of the lowest direct bandgap is below that of the
lowest indirect bandgap, while LEDs are capable of operating using either
direct or indirect bandgap materials. The GaAs.sub.x N.sub.1-x noted above
is a direct bandgap material. The addition of other group III and group V
elements, such as boron, aluminum, indium, phosphorus and indium, to the
basic GaAsN composition allows somewhat independent adjustment of the
lattice constant and bandgap, allowing different materials with different
emission wavelengths to be lattice matched to the same substrate. For
LIDAR systems, laser diodes using strained InGaAs.sub.1-y N.sub.y active
regions, where y.ltoreq.0.04, can produce light emissions in the 2.0 .mu.m
to 2.5 .mu.m range. It is also possible to produce electronic devices,
such as transistors, which are capable of high temperature operation,
using the high bandgap semiconductor material of the present invention.
The monocrystalline material of the present invention if formed by
epitaxial growth on a monocrystalline substrate. Depending on the choice
of substrate material and the orientation of the substrate's crystal
lattice with respect to the substrate's surface, that is, the choice of
crystal lattice plane for the growing surface, the resulting III-V
compound arsenide-nitride material layers may have either a cubic
zincblende crystal structure or a hexagonal wurtzite crystal structure.
InGaAlAs materials have a native zincblende structure, but with the
appropriate choice of substrate can be induced to grow in the metastable
wurtzite structure. InGaAsAlN materials, on the other hand, have a native
wurtzite structure, but with the appropriate choice of substrate can be
induced to grow in the metastable zincblende structure. Substrates can be
selected from the group consisting of Al.sub.2 O.sub.3 (sapphire) using
either basal plane or r-plane growing surfaces, diamond, Si, Ge, SiC in
both wurtzite (.alpha.) and zincblende (.beta.) forms, InN, GaP, GaAsP,
GaAs, InP, ZnO, ZnS, and ZnSe.
The arsenide-nitride materials of the present invention generally favor
growth of cubic zincblende crystal structures when arsenic is the majority
component in the group V lattice sites, up to about 4 or 5 percent
nitrogen and favor growth of hexagonal wurtzite crystal structures when
nitrogen is the majority component in the group V lattice sites, up to
about 20 percent arsenic. When growth techniques, such as metalorganic
chemical vapor deposition (MOCVD), are used to attempt growth of material
with between 5 and 90 percent nitrogen, mixed polycrystalline layers with
both zincblende and wurtzite crystals result. This immiscible region for
the arsenide-nitride material of the present invention necessitates the
use of growth techniques further away from thermodynamic equilibrium, such
as atomic layer epitaxy, to obtain defect-free monocrystalline material
with the required composition.
It has been found that lowering the growth rate of the AlGaAsN
semiconductor material allows a greater concentration of N to be
incorporated in the crystal. It is possible to increase the availability
of N for incorporation by epitaxial growth using low-pressure metal
organic vapor deposition (LP-MOCVD) utilizing N sources including
hydrazine, phenol-hydrazine and metal-organic N sources. Ammonia
(NH.sub.3) may be employed as an N source with catalysts such as arsine
and phosphine used to increase the cracking efficiency of NH.sub.3.
Alternatively, the N source can be pre-cracked. For example, in molecular
beam epitaxy (MBE) an electron cyclotron resonator can be used to crack
the NH.sub.3 molecule, resulting in a radical N. Plasma deposition systems
can also be used to increase the cracking efficiencies of NH.sub.3.
It is possible to obtain the correct concentration of N relative to As by
atomic layer epitaxy, a sequenced layering technique in which the growth
monolayers containing N is interspersed with the growth of other
monolayers containing As. The monolayers are thin enough to avoid defects
due to mismatched lattices, as differing lattice constants between the
monolayers tend to compensate for each other to produce a desired lattice
match. For instance, five monolayers of GaAs can be grown for every
monolayer of GaN, resulting in an overall Group V lattice site N
concentration of approximately 17% and As concentration of about 83%. It
is similarly possible to provide the relative concentrations of Al
compared to Ga with such layering. Thus, for example, having two
monolayers of Ga interspersed with four layers of Al results in a Ga
concentration of approximately 33% and an Al concentration of
approximately 67%. This process can be continued to form, for example,
alternating quarter wavelength (115 nm) optically reflective layers or
quantum wells formed of layers on the order of the electron wavelengths (5
nm) and other heterostructure lasing devices for which the exact depth of
the layers is important.
The previous discussion has centered upon the growth techniques for alloys
with N as the minority element on the Group V sublattice. Similar growth
technology may be employed to grow alloys where N is the majority element
on the Group V sublattice. However, the growth conditions themselves must
be changed. Experiments have shown that the solubility of As in GaN is
significantly greater than the solubility of N in GaAs; transmission
electron microscopy has shown greater than 10% incorporation of As in GaN
grown by conventional LP-MOCVD. Thus, for alloys with As content equal to
or less than 10% MOCVD, either atmospheric or low-pressure, growth can be
used with the ratio between N and As in the gas phase adjusted to provide
the required As content. For alloy compositions greater than 10% As,
multiple layer growth, as previously outlined, will be employed. For
example, the layer structure required for the case of 10% As would employ
nine layers of GaN, followed by a single layer of GaAs, etc.
Such materials can be used in a variety of applications. Semiconductor
lasers and diodes can be created that emit light of frequencies that more
than span the visible spectra, from infrared to ultraviolet. Conversely,
photodetectors and solar detectors can be fashioned that utilize the wide
range of available bandgaps to collect light of all colors. High
temperature transistors, diode rectifiers and other electronic devices can
also be produced using the high direct bandgap materials lattice matched
to Si, SiC or GaP.
Thus, active layers of the direct bandgap material GaAsN, when used with
cladding layers of GaN, AlGaN or AlGaAsN in a diode heterostructure, are
prime candidates for optical emission which can be placed under biaxial
compressive strain, a visible analog to the InGaAs strained layer
technology currently employed in the AlGaAs material system for diode
lasers or LED's. At a border between two of the semiconductor materials, a
p-n junction is formed by doping the respective materials with p-type and
n-type dopants. Introducing dopants, both acceptors, such as C, Mg, and
Zn, and donors, such as Si, Se, or Ge, into the GaAsN can be used to
fabricate p-type and n-type doping, respectively. A cubic crystal lattice
provides for easy formation of cleaved mirror and waveguide structures
that feedback the emitted light to cause laser amplification. The
resulting devices can efficiently emit coherent light spanning the visible
spectra, including the blue region.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a graph of the bandgap and lattice constant of AlGaInAs cubic
crystals superposed with the lattice constant of some common substrates.
FIG. 2 is a graph of the bandgap and lattice constant of AlGaInN wurtzite
crystals superposed with the lattice constant of some common substrates.
FIG. 3 is a table of tetrahedral covalent radii of group III and group V
elements used to estimate nearest neighbor bond lengths in the crystal
lattices of materials of the present invention.
FIG. 4 is a graph of the cubic and wurtzite lattice constants of GaAs.sub.x
N.sub.1-x for various concentrations (x) of N superposed with the lattice
constants of various substrates.
FIG. 5 is a graph of the cubic and wurtzite lattice constants of InAs.sub.x
N.sub.1-x for various concentrations (x) of N superposed with the lattice
constants of various substrates.
FIG. 6 is a graph of the lowest direct bandgap of GaAs.sub.1-x N.sub.x for
various concentrations (x) of nitrogen.
FIG. 7 is a graph of the lowest direct bandgap of InAs.sub.1-x N.sub.x for
various concentrations (x) of nitrogen.
FIG. 8 is a diagram of a LP-MOCVD apparatus used for manufacturing
materials and devices of the present invention.
FIG. 9 is a diagram of an MBE apparatus used for manufacturing materials
and devices of the present invention.
FIG. 10 is a perspective view of a light emitting device of the present
invention.
FIG. 11 is a cutaway perspective view of a light receiving device of the
present invention.
FIG. 12 is a side view of an AlGaAsN heterostructure of the present
invention matched to an SiC or Al.sub.2 O.sub.3 substrate.
FIG. 13 is a side view of an AlGaInAsN heterostructure of the present
invention matched to a ZnO substrate.
FIG. 14 is a side view of an LED with an active region made from material
of the present invention sandwiched between SiC layers.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
With reference to FIG. 1, the lattice constants and bandgap of AlGaInAs
semiconductor materials are plotted along with the lattice constants of
several common substrate materials. The points 10, 12 and 14 represent the
binary compounds AlAs, GaAs and InAs, respectively, while the lines
connecting these points represent the various ternary compounds AlGaAs,
AlInAs and GaInAs and the triangular region enclosed by those lines
represent the quaternary compounds AlGaInAs with various proportions of
Al, Ga and In. All of these materials have a native cubic zincblende
crystal structure. AlAs has a large bandgap of about 2.2 eV and a cubic
lattice constant of about 5.66 .ANG.. However, the binary compound AlAs,
represented by point 10, as well as the ternary and quaternary compounds
with a large proportion of aluminum, in the region near point 10, are
indirect bandgap materials. GaAs has a direct bandgap of about 1.4 eV and
a cubic lattice constant of about 5.65 .ANG.. Thus, AlAs and GaAs are
structural isomorphs, allowing free substitution of aluminum and gallium
in solid solution without any significant change in the lattice constant.
InAs is also a direct bandgap material with a bandgap energy of about 0.4
eV, but with a substantially larger lattice constant of about 6.06 .ANG..
The introduction of indium can be used to produce strain so as to alter
the bond structure in a layer and its neighboring layers of a
heterostructure, most notably affecting the heavy hole valence bond in a
way that can reduce laser threshold and improve operating efficiency of a
laser diode.
The lattice constants of several common substrates are also seen in FIG. 1.
The dashed line 16 represents both silicon and GaP substrates. Silicon has
a diamond-type (cubic) crystal structure with a lattice constant of about
5.43 .ANG., while GaP has a zincblende (also cubic) crystal structure with
a lattice constant of about 5.45 .ANG.. Both of these substrates have a
smaller crystal lattice than AlAs and GaAs, with a lattice mis-match of
about 4%. These substrates have been used to grow AlGaAsP layers (although
such compounds are indirect bandgap materials for large proportions of
either aluminum or phosphorus), but attempts to grow AlGaAs layers on
silicon or GaP have generally been unsuccessful, with the mismatch-strain
causing unacceptable defect concentrations in the resulting material. The
dashed line 17 represents both germanium and GaAs. As noted previously,
GaAs has a lattice constant of about 5.65 .ANG. making it an ideal
substrate for AlGaAs materials, as well as AlGaInP, AlGaAsP and InGaAs
material layers. Germanium, like silicon, has a diamond-type structure,
and also has a cubic lattice constant of about 5.66 .ANG., nearly
identical to GaAs. The dashed line 18 represents InP. It has a zincblende
structure with a lattice constant of about 5.87 .ANG. and is commonly used
as a substrate for long emission wavelength materials (with emissions
longer than 1100 nm) with a large proportion of indium, such as InGaAs and
InGaAsP. Due to its relatively large lattice, InP is considered to be a
potential substrate only for the mid-IR and far-IR emitting InAs1-xNx and
InSb.sub.y As1.sub.-x-y N.sub.x (x.ltoreq.0.05) materials of the present
invention, the other arsenide-nitride materials of the present invention
having too small a lattice for an InP substrate.
FIG. 2 shows another plot corresponding to FIG. 1, but for AlGaInN
semiconductor materials. These nitride materials have a native hexagonal
wurtzite crystal structure, but can also be induced to grow in a
metastable cubic zincblende crystal structure with a suitable choice of
substrate material and growth surface. That is, growth of a zincblende
structure is generally favored on the [001] surface of a cubic substrate,
while the active wurtzite structure generally results from growth on the
[0001] surface of a hexagonal substrate or on the [111] surface of a cubic
substrate. In the case of rhombohedral Al.sub.2 O.sub.3 (sapphire)
substrates, growth of a wurtzite structure is favored when the basal plane
surface is used, while growth of a zincblende structure is favored when
the r-plane surface is used. The plot in FIG. 2 assumes wurtzite growth
for these materials, and so the corresponding wurtzite lattice constants
are used.
The points 20, 22 and 24 represent the binary compounds AlN, GaN and InN,
respectively, while the lines connecting those points represent the
various ternary compounds AlGaN, AlInN and GaInN and the triangular region
enclosed by those lines represent the quaternary compounds AlGaInN with
various proportions of Al, Ga and In. It can be seen that AlN has a very
high bandgap of about 6.0 eV and a relatively small lattice constant of
about 3.11 .ANG.. GaN can be seen to have a lower bandgap of about 3.4 eV
and a slightly larger lattice constant of about 3.19 .ANG.. InN has a
still lower bandgap of about 2.0 eV and a considerably larger lattice
constant of about 3.53 .ANG..
Superposed on the plot of FIG. 2 are the lattice constants of some common
substrates. The dashed line 26 represents .alpha.-SiC, which has a lattice
constant of about 3.08 .ANG., less than the lattice constant of any
AlGaInN compound. The lattice mismatch with this substrate is about 1% for
AlN and about 3.5% for GaN, and so .alpha.-SiC is a useful substrate for
growing AlGaN layers, but less suitable for most of the arsenide-nitride
materials of the present invention, since the lattice mismatch becomes
worse as arsenic is incorporated into the lattice. However, the substrate
could be used in AlGaN heterostructures that include AlGaAsN or GaAsN
strained quantum wells. The dashed line 27 represents .alpha.-ZnO. (SiC
and ZnO have both wurtzite (.alpha.) and zincblende (.beta.) forms,
although the .alpha.-form is more common.) It can be seen that .alpha.-ZnO
has a lattice constant of about 3.25 .ANG., which nearly matches GaN (with
a mismatch of only 1.9%) and which corresponds to a wide range of AlGaInN
compounds, as seen by the dashed line 27 intersecting the triangular
region. .alpha.-ZnO is thus a promising substrate. Its use has been
limited so far by the relatively poor quality of available ZnO crystals,
but the quality is expected to improve. The dashed line 28 represents
basal plane sapphire (Al.sub.2 O.sub.3) substrates. The basal plane
growing surface of the rhombohedral Al.sub.2 O.sub.3 substrate has an
equivalent wurtzite lattice constant of about 3.62 .ANG., which can be
seen to be larger than any AlGaInN material. However, sapphire substrates
are considered promising for the arsenide-nitride semiconductor compounds
of the present invention, since the lattice size increases as arsenic is
added.
Because the arsenide-nitride materials of the present invention can have
either a zincblende structure or a wurtzite structure, with the zincblende
structure being the native crystal phase for those compounds with arsenic
as the majority element and nitrogen as a minority element at the group V
lattice sites, and with the wurtzite structure being the native crystal
phase for those compounds with nitrogen as the majority element and
arsenic as a minority element at the group V lattice sites of the crystal,
it is more convenient to use nearest neighbor bond length rather than
lattice constant in order to select suitable substrate materials and to
deter | | |
