1. Introduction
The technological
importance of GaN is due to its excellence in high-power, high-frequency, and
optoelectronic devices, which are significant for an assortment of
applications. The primary hindrance in the successful development of such a
variety of III-nitride-based devices is the lack of a native high-purity,
single-crystal GaN substrate with a low concentration of point and extended
defects. Commercial hydride vapor phase epitaxy (HVPE) GaN substrates contain
dislocation densities on the order of 105 cm−2, leading to critical limitations
and problems in devices, which can be illustrated, for example, by reduced
output and lifetime of III-nitride laser diodes as well as gate leakage current
in RF AlGaN/GaN high electron mobility transistors (HEMTs) . The development of a
true GaN boule growth technique with the potential for large-volume production
will alleviate these issues.
Under nitrogen ambient at
atmospheric pressure, gallium nitride is stable until only 1120 K , and it can only be
stabilized at higher temperatures by drastically increasing the nitrogen
pressure . The
standard crystal growth techniques used for the fabrication of semiconductor
wafers (Bridgman, Czochralski, etc.) cannot be developed for GaN, because at
pressures below 4.5 GPa GaN does not melt but decomposes.
There are two main
approaches to bulk GaN growth development—growth from gas phase and growth from
liquid phase. At this point, HVPE is the only technique that is able to
commercially produce GaN substrates. This technique has been used to grow
wafers up to 2 in diameter at relatively high growth rates (over
100 μm/h) .
Although these substrates have been used for device fabrication through
homoepitaxial growth, the quality of substrates (high dislocation density and
bowing) produced by this technique is a major concern for current device
development. Even areas with reduced dislocations within the best HVPE samples
have dislocation densities of at best 105 cm−2. In addition, the material grown by
HVPE has considerable residual stress level within the substrate.
An alternative technique
for single crystal growth involves deposition of GaN from the liquid phase,
i.e. from solution. Growth from the liquid phase has resulted in GaN single
crystals with dislocation densities less than 102 cm−2. High-pressure nitrogen solution
growth (HPS) method is carried out using high pressures (greater than
1.0 GPa) and high temperatures (greater than 1400 °C) to dissolve
nitrogen into Ga and has been successfully used to grow thin GaN crystal
platelets of up to 1.5 cm, laterally.
GaN also has been grown
at lower temperatures/pressures (∼5.0 MPa)
by the Na flux method . Addition of Na increases the reactivity and solubility of nitrogen
in the Ga and the gaseous nitrogen reacts with the flux/elemental gallium to
saturate the solution and deposit crystals. An ammonothermal method has also
been developed for GaN growth . This technique requires pressures in the range of 150–400 MPa
and growth rates are on the order of 1–2 μm/h.
This work is a
continuation of previous efforts and to
develop a new technique, which meets critical requirements for the production
of low defect density and low-cost GaN substrates. These requirements impose
limitations on the direction of the development. Perfection of the bulk
material requires crystal growth near equilibrium conditions (i.e. growth from
solution). Large crystals suitable for commercial production of high-quality
substrates require growth on a seed. In addition, size and cost appropriate for
industrial production require growth at moderate temperature and pressure. This
article describes a technical approach (Fig. 1a)
to dissolve a solid GaN source in the proper solvent to create a solution at
moderate pressure and temperature, and then to grow crystals on a seed from the
supersaturated solution by applying a temperature gradient.
Fig. 1. Schematics
of (a) the technical approach to GaN growth from solution, and (b) growth
reactor.
2.
Experimental procedure
Growth experiments were
carried out in a custom-designed growth reactor heated with a vertical tube
furnace from Mellen. The reactor allows the crucible with the charge to be
loaded from the bottom and is configured for top-seeded growth (Fig. 1b). A multi-component solvent was
developed to dissolve the GaN source and, subsequently, to grow GaN single crystals from the solution. Preparation of the multi-component solvent was
performed in a separate reactor, similar in design to the growth reactor. The
materials used for the solvent preparation have purity higher than 4 N.
Handling of the solvent components, which are sensitive to moisture and oxygen,
and charging the crucible, which has an inner diameter of 17 mm, were
carried out in a glove box under nitrogen atmosphere with O2 and H2O content below 1 ppm. Polycrystalline GaN
aggregate in the bottom of the crucible served as a source for the top-seeded
runs. A layer of the solvent was placed on the top of the source for these runs
(Fig. 2a). After charging in the glove box
the crucible and charge were transferred from the glove box to the growth
reactor. The crucible position in the reactor with respect to the heater (Fig. 1b) was chosen to induce an axial
thermal gradient inside of the solution. After loading, the reactor was pumped
and purged to create a pure nitrogen atmosphere. Purified house nitrogen
supplied the ambient during the experiments. The gallium nitride seed was
dipped and kept inside the solution at the growth temperature and pressure for
50–80 h, then it was removed and the crucible was cooled. Growth runs were
conducted at nitrogen pressure of 0.23–0.25 MPa and temperature of
800 °C. The crucible contents and seeds were washed in DI water and
hydrochloric acid to dissolve the remaining compounds. Optical microscopy,
micro Raman scattering (μRS), X-ray diffraction (XRD) and photoluminescence
(PL) spectroscopy were used to characterize the GaN crystals.
Fig. 2. Schematic
view and optical images of (a) the crucible with GaN source, solvent and HVPE
GaN polycrystalline aggregate as seed, (b) the seed—HVPE GaN polycrystalline
aggregate before growth run, (c) the seed—HVPE GaN polycrystalline aggregate
after growth run with grown GaN crystals, (d) and (e) zoomed images of the
crystals shown in the (c) arrow indicates m direction.
Micro Raman measurements were
performed at room temperature in the backscattering geometry, in order to
characterize the structural quality of the sample. The 1064.1 nm line of a
Nd:YAG laser was focused to a 2 μm diameter spot and used for excitation,
while the scattered light was analyzed by a triple spectrometer TRIVISTA 557
equipped with an OMA V 1024-1.7 liquid-nitrogen-cooled InGaAs linear array
detector. The FWHM spectral resolution of the system was 0.1 cm−1.
For XRD rocking curve
measurements of the freestanding crystals a Blake double-crystal diffractometer
equipped with a Si (1 0 0) beam conditioner and configured for the
004 diffraction with CuKα1 radiation was used. Compared to
typical diffractometers equipped with 4-bounce monochromators, the
double-crystal instrument provides nearly one order of magnitude higher beam
intensity at the sample and a much better resolution (5 versus 12 arcsec).
The beam size at sample was 80 μm wide and 200 μm high in cross-section.
Low-temperature PL
(LT-PL) measurements were performed to evaluate the optical and electronic
quality of the grown crystals. To maintain a temperature of ∼5 K during the LT-PL measurements, samples were placed
in a continuous helium flow cryostat. Samples were excited with the 325 nm
line from a He–Cd laser, and laser excitation intensity was maintained at
0.8 mW using neutral density filters. Light emitted by the sample was
dispersed using a double-grating spectrometer fitted with 1800 groves/mm.
The spectra were acquired using a UV-sensitive GaAs photomultiplier coupled to
a computer-controlled photon counter.
3. Results
and discussion
The critical part of the
solution growth approach is the choice of solvent. The solvent should
efficiently dissolve the GaN source and create a solution at moderate
temperatures and pressures at which GaN is stable. That solvent has been
developed to satisfy these criteria at near atmospheric pressure. As a result,
GaN has been grown on a GaN seed utilizing the proper solvent at a growth
temperature of 800 °C and pressures of 0.2–0.3 MPa. It should be
noted that GaN was also grown at lower pressures, but this was not explored
based on the developed technique, which is designed to maintain a small
overpressure of nitrogen in the growth ambient.
As previously mentioned,
the growth from the solution on a seed at moderate pressure and temperature is
one of the key features of the described technique. Demonstration of this
aspect was the main objective of this work. The first top-seeded growth runs
were conducted using HVPE GaN polycrystalline aggregates (Kyma Technologies,
Inc.) as seeds (Fig. 2). The shape of the polycrystalline
pieces facilitates easy mounting and makes it simple to observe dipping into
the solution. This type of seed also gives an opportunity to study the
nucleation and GaN growth on different crystal faces and in different
crystallographic directions during one growth run. Fig. 2b
shows such a polycrystalline seed before the growth run. The seed is mounted
with Ta wire and dipped partly into the solution as shown schematically in Fig. 2a.
Due to the crucible position inside of the furnace (Fig. 1b),
the temperature of the seed was slightly lower than the GaN source. After
68 h of exposure to the solution at 800 °C and 0.24 MPa, the
seed was removed. After cleaning the remaining solution from the seed, grown
crystals of different orientations were found on the immersed portion of the
seed. Most of the crystals formed as an epitaxial expansion of the crystallites
of the aggregate (Fig. 2c). Some crystals
were nucleated as twins on the very edge of the crystallites and developed as
freestanding crystals (Fig. 2d). Most of the
crystals grew epitaxially, with the most high growth rates in the non-polar
directions (Fig. 2e). All of the grown crystals were
transparent and colorless, with well-defined hexagonal morphology.
Examination of the grown
crystals with μRS spectroscopy in the z(xy,xy) 
id="" style='orphans: auto;widows: auto;-webkit-text-stroke-width: 0px;
word-spacing:-.15ex' alt="View the MathML source"
xmlns:xoe="http://www.elsevier.com/xml/xoe/dtd" class=imgLazyJSB border=0
title="View the MathML source" data-inlimgeid=1-s2.0-S0022024808004545-si1.gif
data-loaded=true v:shapes="_x0000_i1025"> geometry showed that the first
order allowed E21, E22 and
A1(LO) phonons with full-width at half-maximum (FWHM) of 0.26, 3.1 and
6.9 cm−1, respectively (Fig. 3).
The sharp line widths indicate high structural quality and low impurity
concentrations.
Fig. 3. Room
temperature micro Raman spectrum of the GaN crystal grown on the
polycrystalline GaN seed in the backscatter geometry.
High crystallinity
of the grown crystals was also confirmed by XRD. Note that the intensities
represented on the rocking curve are actual count rates. The sample tilt was
then adjusted for optimum rocking curve breadth and peak intensity. FWHM of
about 16 arcsec was obtained for the (0 0 0 4) rocking
curve, excluding the additional dispersion and convolution corrections that
would only enhance this number slightly (Fig.
4).
Fig. 4. XRD
rocking curve of the (0 0 0 4) reflection for the GaN crystal
grown on the polycrystalline GaN seed.
The position and intensity of the PL spectral peaks provide
information about the type and concentration of the impurities, respectively. A
dominant peak at 3.47 eV was observed in the spectra of crystals grown on
polycrystalline aggregate seed (Fig. 5a) and has been attributed to excitons bound to
neutral shallow donor impurities (XD0 or
D0X) . High crystalline quality of the grown crystals was verified by
improved XD0 line shape and linewidth, as compared with that
of the seed. The nature of the dominant peak at 3.42 eV, observed on the
spectrum of the seed (indicated with a question mark), is not yet determined
and is under investigation. Reduction of the yellow band (YB) intensity, shown
in Fig. 5b, is consistent with lower native defects
and/or residual impurity concentration in the growth crystal , as compared with that of the seed.
Fig. 5. (a)
LT–PL spectra of crystals grown on polycrystalline aggregate seed, (b)
reduction of yellow band in the grown crystals compared to the seed.
Growth runs were performed using HVPE GaN templates as a seed
to verify the prospect of the homoepitaxial growth on single-crystal GaN seeds.
The configuration of these runs was similar to those conducted with
polycrystalline GaN seeds. The seed was mounted with Ta wire and partly
immersed into the solution when the temperature in the solution reached
800 °C. A small positive temperature difference between the seed and GaN
source was induced by the chosen location of the crucible inside of the furnace
as shown in Fig. 1b. An epitaxially grown GaN layer was observed
on the part of the template immersed into the solution for 80 h (Fig. 6).
The part of the seed with the epitaxially grown layer showed enhanced
transparency due to N-face surface improvement as seen in Fig. 6a.
Fig. 6. Images
of the HVPE GaN template used as a seed with epitaxially grown GaN layer: (a)
nitrogen face, (b) gallium face.
LT-PL spectra covering the spectral range between 2.85 and
3.55 eV for four regions of the homoepitaxially grown layer are shown in Fig. 7a. The probed regions are the unpolished
nitrogen-face, polished gallium-face, and the homoepitaxial layer grown
simultaneously on both faces. Similar to the crystals grown on the
polycrystalline seed, LT-PL revealed a dominant peak at 3.47 eV for both
Ga- and N-faces. The intense luminescence band observed in the spectral range
at 3.0–3.3 eV is associated with the recombination process involving
donor–acceptor pair (DAP) with zero-phonon line (ZPL) at 3.27 eV, and
phonon replicas at 3.18 eV (1LO-DAP) and at 3.09 eV (2LO-DAP). The PL
spectra of the Ga-face show reduction of DAP band intensity, which suggests
lower incorporation of shallow acceptors, relative to the N-face of the grown
layer and the template. Fig. 7b
shows the PL spectra covering the region 1.75–2.75 eV. Both faces of the
layer showed a decrease in the YB intensities (2.2–2.4 eV) as compared to
that of the template, suggesting a decrease in the concentration of the native
defects and/or residual impurities.
Fig. 7. (a)
LT–PL spectra of epitaxially grown GaN layer, (b) reduction of yellow band in
the grown layer compared to the seed.
Fig. 8 displays an ω–2θ space
map of the symmetric (0 0 0 4) reflection for the Ga-face (a)
as-received GaN seed and (b) GaN grown crystal. The nearly 100 μm thick
homoepitaxially grown layer showed a two order-of-magnitude reduction in the
FWHM of the (0 0 0 4) XRD diffraction peak. FWHM of the X-ray
rocking curve measured on both the Ga- and N-face of the sample are 111 and
127 arcsec, respectively, compared to 2.15° and 2.45° for the Ga- and
N-face of the GaN seed, respectively.
Fig. 8. ω–2θ space map of the
symmetric (0 0 0 4) reflection for the (a) as-received GaN seed
and (b) Ga-face of the GaN grown crystal.
As a test for the new growth technique, runs with GaN seeds
at the bottom of the crucible and GaN source on the top of the solvent were
conducted (Fig. 9). In this case a temperature gradient opposite
to the top-seeded arrangement should be applied. In the present system, the
axial temperature gradient in the solution can be slightly varied only by
changing the crucible height along the vertical heater axis (Fig. 1a),
and for the bottom-seeded runs the crucible was located at the lower part of
the heater. Pressure and temperature for these runs were the same as for the
top-seeded runs. Only dissolution of the HVPE template at the bottom of the
crucible was observed in the first runs. Images of the seed before the run (Fig. 9a)
and after the run (Fig. 9c) visibly show the high
degree of the seed dissolution at the bottom of the crucible. The size of the
seed was found to be nearly three times smaller after this run, demonstrating
the extent to which the developed solution is able to dissolve GaN. Initial bottom-seeded
runs revealed that it is more difficult for this arrangement than for the
top-seeded arrangement to establish the correct temperature gradient. Moving
the crucible along the vertical heater axis does not allow sufficient control
of the temperature gradient within the highly thermoconductive solution located
inside the crucible, which is itself a similar high thermally conductive
material. Minor variations in the temperature gradient using only the vertical
heat distribution of the furnace cause dissolution of the HVPE template as
observed for a seed at the bottom of the crucible. When the correct temperature
difference between GaN source and seed was established, an epitaxial layer of
the GaN was grown on the seed located at the bottom of the crucible (Fig. 9d).
Runs in this configuration clearly demonstrate that the developed
multi-component solvent successfully dissolves GaN and forms a suitable
solution, and temperature gradients create the driving force for continuous
dissolution of the GaN source and the epitaxial growth of GaN on a seed. These
experiments also guided on-going modifications of the reactor heating system to
enable more flexibility to create and control a required axial temperature
gradient inside the solution.
Fig. 9. Optical
images and schematic view of the (a) HVPE GaN template as a seed before growth
run, (b) crucible with the charge for the bottom-seeded growth, (c) the same
HVPE GaN template partly dissolved during growth run when T2⩾T1, (d) GaN seed with grown GaN layer when T2<T1.
4.
Conclusion
High-quality GaN single
crystals have been grown for the first time from solution on a GaN seed at
nitrogen pressures less than 0.25 MPa and at 800 °C. The high
crystallinity and purity of the grown crystals were confirmed by μRS, PL and
XRD measurements. The essential feature of this new growth technique is the
ability to produce high-quality crystals even without tuning the growth
parameters. These results highlight the potential capability of this method.
Source: Journal of Crystal Growth
