Anda di halaman 1dari 14

Encyclopedia of

Nanoscience and
Nanotechnology

www.aspbs.com/enn

Phonons in GaN-AlN Nanostructures


J. Frandon, J. Gleize, M. A. Renucci
Universit Paul Sabatier, Toulouse, France

CONTENTS
1. Introduction
2. Raman Scattering in Hexagonal Crystals
and Nanostructures
3. Phonons in Quantum Wells and Superlattices
4. Phonons in Quantum Dots Structures
5. Summary
Glossary
References

1. INTRODUCTION
The scope of this chapter is the review of recent experimental studies of phonons in nanostructures made of nitride
semiconductors with an hexagonal structure. First let us
recall some basic denitions. A quantum well (QW) is
described as a thin layer exhibiting semiconducting properties, located between a couple of layers made of another
semiconductor and playing the role of barriers. Indeed, the
electronic bandgap energy in the latter is higher than its
counterpart in the former, thus favoring carrier connement
and electrical transport inside the well. Single or multiple
QW structures can be fabricated as well as superlattices (SL)
which are periodic arrays of QWs. A quantum dot (QD)
is an island made of a given semiconductor embedded in
another semiconductor acting as barrier. It is usually characterized by a pyramidal shape and a small height (typically
5 nm). The samples are periodic stackings of planes containing the QDs, which can be self-assembled by the effect of
vertical correlation.
The purpose of the extensive work recently devoted to
GaN-AlN or GaN-AlGaN nanostructures is the development of new devices, particularly laser diodes emitting in
the ultraviolet range. Their light emission is expected to be
much stronger than from heterostructures made of GaN and
AlN (or AlGaN) thick layers grown in the nineties, due to
the low-dimensional geometry of nanostructures. In addition, the crystallographic structure of nitride semiconductors
is responsible for their piezoelectric properties, generating
very strong electric elds in strained nanostructures; as a
ISBN: 1-58883-064-0/$35.00
Copyright 2004 by American Scientic Publishers
All rights of reproduction in any form reserved.

result, the emission is signicantly red-shifted with respect to


the absorption edge by the so-called conned quantum Stark
effect, and the luminescence can be tuned within a wide
interval from the visible range up to the near ultraviolet.
Such high-performance devices, grown using techniques
recently developed, are currently fabricated and are already
on the market. However, the knowledge of the structural
and optical properties involved in the light emission process must be improved. So numerous experimental studies
have been recently devoted to such nanostructures. Among
the various techniques used for this purpose, Raman spectroscopy is known to be a powerful probe of vibrational
modes (or phonons) in semiconductor materials. Measurements of phonon frequencies in nanostructures give the
opportunity of determining the strain state inside their constituent layers; these results are key data, directly related to
the light emission of the device. This kind of experiment is
rather simple, nondestructive, and usually needs no special
sample preparation. The size of the optical probe can be
very small (as low as 1 m). In addition, this technique is
accurate and very reproducible, allowing measurements of
the phonon energy within an uncertainty lower than 1 cm1
(about 0.1 meV).
This chapter is organized as follows: basic concepts concerning phonons in bulk nitride semiconductors and GaNbased nanostructures, as well as Raman scattering, are given
in Section 1. Section 2 is devoted to the Raman studies on
GaN-AlN (or GaN-GaAlN) QW structures and SLs, in nonresonant and resonant conditions. Section 3 of this chapter
deals mostly with phonons in QDs stackings.

2. RAMAN SCATTERING
IN HEXAGONAL CRYSTALS
AND NANOSTRUCTURES
2.1. Phonons in Bulk
GaN-Like Semiconductors
Only GaN, AlN, or AlGaN semiconductors with wurtzite
structure will be considered in the present article. Before
dealing with nanostructures, we must recall briey the
vibrational properties of bulk crystals, which were recently
Encyclopedia of Nanoscience and Nanotechnology
Edited by H. S. Nalwa
Volume 8: Pages (513526)

514

Phonons in GaN-AlN Nanostructures

reviewed by Frandon et al. [1]. In fact, almost all the samples


studied up until now are layers grown on a substrate and
on a buffer layer, but they behave usually as bulk materials due to their large thickness. Hexagonal nitride crys4
. The unit cell contains
tals belong to the space group C6v
two nitrogen atoms and two Ga or Al atoms [2]. The lattice constants a and c of GaN [3] and AlN [4] are given in
Table 1. In the following, the c axis of the hexagonal structure will be chosen as the z axis. Let us recall briey how
the symmetry of Raman active modes are derived in this
structure. Here we are only interested in the zone center
phonons, that is, characterized by a vanishing q wavevector.
The 12-dimensional representation of the atomic motions
in the unit cell can be reduced into irreducible representations of the C6v group. Besides the acoustic modes and the
silent (inactive) B1 optic phonons, one obtains four optical modes, two nonpolar (infrared inactive), and two polar
(both Raman and infrared active) phonons [5]. The nonpolar phonons exhibit the E2 symmetry, corresponding to
atomic motions perpendicular to the z axis. One of them
may be observed at low frequency. The other, denoted by E2
(high), shows up at a higher frequency with a high intensity
in Raman spectra recorded far from resonant conditions. In
addition, it cannot couple to plasmons and does not display
any angular dispersion, on account of its nonpolar character. Therefore, it is frequently used as a probe of structural
properties in the Raman characterization of nitride semiconductors.
Polar optic phonons are characterized by their A1 or E1
symmetries, related to atomic motions, parallel or normal
to the z axis, respectively. They may be either longitudinal
or transverse (LO or TO). In these ionic crystals, the longrange forces associated with the strong macroscopic electric
eld of longitudinal phonons are responsible for an important LO-TO splitting (202 cm1 for GaN). Moreover, due
to the uniaxial properties of these crystals, the LO and one
of the TO phonons are extraordinary modes, thus exhibiting an angular dispersion; their frequency varies with the
angle  between the phonon wavevector q and the z axis [6].
This variation range corresponds to the A1 -E1 splitting governed by short-range interatomic forces. However, it should
be noted that this variation (27 cm1 for TO phonons of
GaN) is much lower than the LO-TO splitting. The symmetry of the q = 0 extraordinary modes depends on the value
of ; for a vanishing angle, the LO and TO phonons exhibit
the A1 and E1 symmetry, respectively, and the reverse is
found for  = 90 . For intermediate q values, the extraordinary modes called quasi-LO and TO (QLO and QTO),
have a mixed symmetry. Finally, the ordinary TO phonon is
nondispersive and keeps the E1 symmetry when varying the
angle . Table 2 gives the frequencies of the q = 0 phonons
for relaxed bulk GaN and AlN [7, 8].
Table 1. Lattice constants a and c (nm) for wurtzite GaN and AlN.

GaNa
AlNb
a
b

From [3].
From [4].

0.31890
0.31106

0.51864
0.49795

Table 2. Long wavelength phonon frequencies (cm1 for wurtzite GaN


and AlN.
E2 (low) A1 (TO) E1 (TO) E2 (high) A1 (LO)
GaNa
AlNb

144
249

533
610

561
669

569
656

735
889

E1 (LO)
743
912

From [7].
From [8].
Note: The notation of phonons is that used in [5].

Note that q = 0 phonons, usually not involved in rstorder Raman scattering, must be invoked in some cases,
for example, when the translational symmetry is lost in
the sample under study. For hexagonal GaN, the phonon
dispersion
q , which is quite distinct from the angular
dispersion previously discussed, has been calculated in the
high-symmetry directions of the Brillouin zone [9, 10] and
have been recently determined by X-ray inelastic scattering
[11]. A scatter of results is observed, but the dispersion of
published data is always found much lower for the branch
starting from the E2 (high) phonon than for its LO counterpart, for example. We must keep in mind this difference
when effects of phonon connement in the nanostructures
will be considered later.
For bulk-disordered AlGaN solid solutions, the evolution of phonon frequencies versus their aluminum content is characterized either by a two-mode or a one-mode
behavior, depending on the symmetry of the vibrational
mode. In the rst case, corresponding to the E2 (high)
optic phonon for example, the GaN- and AlN-like oscillators
exhibit strengths of comparable orders of magnitude in most
parts of the composition range. In the second case, observed
for A1 (LO) and E1 (LO) phonons, the oscillator strength
is strongly transferred from one to the other type of oscillators, thus allowing the observation of one mode only in
the whole composition range of the alloy [1216]. Evidence
for two-mode behavior of the E1 (TO) mode has been given
from infrared measurements [17]. In contrast, the case of
the A1 (TO) phonon seems to be more complicated [16].
As previously mentioned, most samples are thick layers
grown on a substrate (sapphire, SiC, or Si) and on a buffer
layer (GaN or AlN). Due to the lattice mismatch and to the
different expansion coefcients of materials constituting the
nanostructure, the layers are usually submitted to a strong
biaxial stress acting in the plane normal to the z axis of the
hexagonal crystal. The induced strains zz and xx , respectively parallel and perpendicular to the z axis, modify the
phonon frequencies in the layers. For the mode , the corresponding frequency shift with respect to its value in the
relaxed material is given by the following linear equation

 = 2 a xx + b zz

(1)

where a and b are the deformation potentials of the


phonon . The knowledge of these parameters is of crucial
importance for evaluating the strains in the thick layers as
well as in the nanostructures. For GaN, the phonon deformation potentials were measured [1819] and calculated
[2021], but they are more controversial for AlN [2225].
Some of the calculated and experimental values of phonon

515

Phonons in GaN-AlN Nanostructures

deformation potentials are given for wurtzite GaN and AlN


in Table 3.
When the material is under biaxial stress  B ( B > 0 for
compressive stress), one can also use the Raman stress coefcient, dened by
d

d B

KB =

and AlAs [27]. Conned vibrational modes are oscillations


located in one type of layer and characterized by an effective wavevector qz . The latter is quantized, due to boundary
conditions at the interfaces with the adjacent layers
qz = n

(2)

If Hookes law is obeyed, K B can be expressed using the


deformation potentials and the elastic constants only. As
a rule, the observed frequency shift of phonons is positive
(resp., negative) if the material is under compressive (resp.,
tensile) biaxial stress.

2.2. Phonons in GaN-Based Nanostructures


Let us consider nanostructures made of alternate layers
exhibiting the wurtzite structure, grown along the z axis.
Their structural and optical properties have been extensively
studied during recent years. Under adequate growth conditions, the StranskiKrastanov mechanism of strain relaxation of GaN on AlN leads to the growth of strained GaN
islands on very thin (two monolayers thick) GaN wetting
layers, embedded in the AlN barriers, leading to QD structures [26]. Different experimental conditions allow a twodimensional growth, thus leading to strained QW structures
or to SLs; the internal strain of the layers depends on
numerous factors, including the nature of the underlying
buffer layer and the thicknesses of the layers in the structure. Obviously, as a result of internal strains, the phonons
from the wells and barriers will be shifted in the Raman
spectra according to Eq. (1). The same equation may also be
tentatively applied to QDs, assuming a nearly biaxial stress,
in consideration of their pyramidal shape and their lateral
size usually about 10 times larger than their height.
But new vibration modes must be considered in nanostructures. Some phonons can be conned inside one type
of layer. Their properties have been extensively studied in
the SLs made of IIIV cubic semiconductors, such as GaAs


d1

(3)

where n is an integer and d1 is the layer thickness. The frequencies of conned modes can be deduced from the LO
phonon dispersion
q of the bulk material along the z
direction; they correspond to the discrete qz values calculated using Eq. (3). If the top of the branch of the LO branch
is at the zone center  , as for the LO phonon of GaN, frequency shifts towards lower frequencies are expected and
can be measured, specially for very thin (a few nanometers
thick) layers. On the other hand, connement effects should
be negligible for the E2 (high) phonon, due to the weak dispersion of the corresponding branch in the Brillouin zone.
Other phonons specic to the SLs are the folded acoustic phonons. Indeed, if the dispersion of acoustic branches
is similar in both materials, that is, when their sound velocities are not too different, acoustic waves can propagate
through the whole nanostructure. Folding of the acoustic
phonon branches of the constituent layers into the reduced
Brillouin zone of the SL generate new vibrational modes,
which are characteristic of the periodicity d = d1 + d2 of the
SL (d1 and d2 stand for the thicknesses of wells and barriers, respectively). They may show up as doublets located in
the low-frequency range of the Raman spectra. The average
frequencies of these doublets are given by [27]

n = n

2 v
d

(4)

where n is an integer and v is an average sound velocity


in the nanostructure calculated from the sound velocities v1

Table 3. Phonon deformation potentials in Wurtzite GaN and AlN.


a (cm1 )

GaN, calculated
GaN, measured
AlN, calculateda
AlN, measured

E2 (low)

A1 (TO)

E1 (TO)

E2 (high)

A1 (LO)

E1 (LO)

+75
+115b
+149

640
630b
776
930d

717
820b
835
982d

742
850b , 818c
881
1092d , 1083e

664
685c
739
643d

775
867

b (cm1 )

GaN, calculateda
GaN, measured
AlN, calculateda
AlN, measured
a

From [21].
From [19].
From [18].
d
From [25].
e
From [23].
b
c

E2 (low)

A1 (TO)

E1 (TO)

E2 (high)

A1 (LO)

E1 (LO)

4
80b
223

695
1290b
394
904d

591
680b
744
901d

715
920b , 797c
906
965d , 1187e

881
997c
737
1157d

703
808

516

Phonons in GaN-AlN Nanostructures

and v2 in the wells and barriers of the SL according to


v=

v 1 v2
1  v2 + v1

Table 4. Selection rules for phonons in wurtzite crystals.

(5)

where  = d2 / d1 + d2 . The frequency difference between


the components of each doublet depends on the phonon
wavevector qz

= 2 qz v

(6)

The value of qz is determined by the experimental geometry


of Raman scattering, as it will shown in the following section.

2.3. Experiments
In the following, the incident (resp., scattered) light is
dened by its wavevector ki , its frequency
i , and its linear polarization ei (resp., kS ,
S and eS ). The experimental
conguration will be indicated by the usual Portos notation
ki ei eS kS . For bulk crystals, conservation law between the
initial and nal states is obeyed both by the wavevector and
the energy. In rst-order Raman scattering, the wavevector
q and the frequency
of the phonon involved in the scattering process are given by
q = ki kS

(7)

=
i
S

(8)

and

According to Eqs. (7) and (8), q is completely dened by the


experimental geometry and the laser energy. In the particular case of backscattering kS = ki , Eq. (7) can be written
as
q = 2ki  =

4 n
i

(9)

where n is the refractive index of the material at the frequency


i , and i is the wavelength of the incident light.
Most Raman spectra of GaN-AlN or GaN-AlGaN nanostructures reported in the literature have been recorded at
room temperature. The simplest experiment is performed
in a backscattering geometry along the growth axis z of the
nanostructure, with polarizations of incident and scattered
light either perpendicular or parallel, corresponding to the
congurations z xy z or z xx z, respectively. In the latter
case, both E2 and A1 (LO) phonons are allowed. Moreover, a micro-Raman spectrometer, using a small laser spot
whose diameter is lower than 1 m, allows the achievement
of other scattering conditions, particularly backscattering on
the edge of the sample under study. In this conguration,
the phonon wavevector q is perpendicular to the z axis and
additional phonons can be evidenced. The selection rules
for all the Raman active phonons in wurtzite materials are
recalled in Table 4.
The excitation of Raman scattering is usually made using
the monochromatic lines of an argon or krypton laser, either
in the visible or ultraviolet range (from 3.41 eV to 3.80 eV).
The experimental results strongly depend on the excitation
energy. If the latter corresponds to the visible range, that is,

Conguration
z xx z
z xy z
x zz x
x yy x
y zy x

Allowed phonons
E2 , A1 (LO)
E2
A1 (TO)
E2 , A1 (TO)
E1 (LO)

Note: The notation of phonons is that used in [5].

far from any electronic resonance, the dominant electronphonon interaction is the deformation potential process
and the nonpolar E2 (high) optic phonon exhibits a scattering cross section much stronger than the other phonons,
specially the A1 (LO) mode also allowed in the z xx z conguration. Note that the Raman spectra usually contain features from the GaN QDs or QWs and from the barriers of
the nanostructure, but also from the underlying thick buffer
layer (usually GaN or AlN), due to the negligible optical
absorption of the constituent layers in the visible range; thus
an unambiguous assignment of phonons may prove difcult.
However, the signature of the nanostructure itself can be
obtained using a confocal set-up with a low depth of focus,
as illustrated later.
On the other hand, a strong enhancement of Raman scattering by the polar LO phonon is observed under near resonant conditions, that is, if the energy of incident or/and
scattered photons is close to an electronic transition energy,
when the forbidden scattering process associated with
the intraband Frhlich electron-phonon interaction becomes
dominant [28]. With the present materials, resonant conditions can be achieved in the ultraviolet range only; indeed
the room temperature bandgap energy is about 3.4 eV and
6.1 eV for bulk GaN and AlN, respectively. The scattering
cross section of the A1 (LO) phonon can be enhanced by
several orders of magnitude [29], thus allowing the signature
of very small volumes inside the sample. However, it should
be noted that the penetration depth of the incident light in
the sample is strongly reduced under ultraviolet excitation,
due to high optical absorption. In addition, an intense photoluminescence (PL) band may show up in the spectra, making an observation of faint Raman features superimposed
on the PL signal very difcult.

3. PHONONS IN QUANTUM WELLS


AND SUPERLATTICES
3.1. Nonresonant Raman Scattering:
Signature of Wells and Barriers
in Superlattices
When Raman experiments are performed far from resonant
conditions, the signal coming from the nanostructure itself
is rather weak; it can be easily measured only if the sample
is thick enough (about 0.5 m), due to the low-scattering
cross section of GaN, and specially of AlN. Actually, this
kind of research study is still scarce; the main motivation is
to determine the nature of the vibrational modes in the SLs
and to derive the built-in strain of the constituent layers.

Phonons in GaN-AlN Nanostructures

The rst Raman study of an SL with a wurtzite structure


was published by Gleize et al. [30]. This structure was made
of a hundred periods of undoped GaN wells and AlN barriers, with nominal thicknesses of 6.3 nm and 5.1 nm respectively, deposited by MBE on a thick AlN buffer layer and a
sapphire substrate. The layer thicknesses were large enough
for neglecting the frequency shift of phonons induced by
connement, specially for the E2 mode. Micro-Raman spectra were recorded in backscattering geometry under various
polarization congurations, under a 2.54 eV excitation, far
from resonant conditions. Thanks to selection rules, most
observed features could be unambiguously assigned to all
optical phonons but the E1 (LO), either from the underlying AlN buffer layer or from each type of SLs layers. Due
to internal strains, phonons from wells and barriers were
found signicantly shifted with respect to their frequency in
relaxed materials. Using the phonon deformation potentials
of GaN, a biaxial stress of 6.3 GPa and an in-plane strain
xx = 1 3% were derived for the GaN layers of the SL from
the measured shift of the E2 (high) phonon. This strain is
close to the value expected for a free-standing state of the
present nanostructure. On the other hand, the measured frequency shifts of AlN phonons, actually larger than those of
GaN phonons, could not be used to determine the internal
strain in the barriers of the SL, because the corresponding
deformation potentials were not known at this time. Note
that a feature observed at 560 cm1 in x yy x Raman spectra could not be associated to an A1 (TO) phonon from
strained GaN layers, although it obeyed the regular selection
rules; it was thus tentatively assigned to an interface mode.
Finally, micro-Raman spectra were also recorded from place
to place on a bevel made by mechanical polishing of the
nanostructure; the angle between the beveled surface and
the (0001) plane was about 1 . The strain-induced shift of
the E2 (GaN) phonon was found higher in the deeper part
of the SL than near the surface, giving evidence for a partial
strain relaxation in the rst layers of the structure.
Schubert et al. [3132] performed other investigations
combining Raman scattering and infrared ellipsometry
experiments. The latter is an indirect technique for investigating the polar phonons, specially the E1 (TO) phonon
when a near normal incidence is used for the measurements.
A standard calculation allows the reproduction of the experimental infrared spectrum; this model needs several tting
parameters, including the TO and LO phonon frequencies
of the constituent materials, together with the concentration and mobilities (parallel and perpendicular to the z axis
of the SL) of free carriers. The latter data must be introduced when the sample under study is doped, intentionally or not. Indeed, the free-carrier plasmon and the LO
phonon, which are both longitudinal excitations, can couple together if their energies are close to each other, thus
shifting in frequency the high frequency component of the
coupled mode with respect to the uncoupled phonon [33].
In [32], eight GaN-AlGaN SLs, grown by MOVPE or by
MBE on thick GaN layer and on sapphire, exhibiting typical layer thicknesses of 25 nm and various aluminum contents in the barriers (0 08 < x < 1), were compared. As
expected, the strain-induced frequency shifts of the nonpolar
E2 phonon for each constituent layer, derived from Raman
measurements, proved that GaN (resp., AlGaN) layers were

517
submitted to an internal compressive (resp., tensile) stress.
A detailed discussion of the Raman data suggested that
these SLs were in a free-standing state: both kinds of layers adopted a common in-plane lattice parameter, different
from the one of the underlying thick GaN layer. The analysis
of infrared ellipsometry measurements led to the determination of an electron concentration higher than 1018 cm3
in the GaN layers of the SLs; these free carriers could originate from dislocation-activated donors or from the AlGaN
layers. Moreover, the in-plane mobility introduced for tting
the infrared spectra was found higher than its out-of-plane
counterpart: this result indicated a connement of free carriers inside the wells along the z axis, the AlGaN layers of
the SL acting as barriers for the free electrons as expected.
Another more recent article combining Raman and X-ray
diffraction measurements [34] deals with a GaN-AlN SL
grown on an AlN buffer layer and on a 6H-SiC substrate;
the GaN wells were specially thin (1.5 nm) in this sample,
compared to the AlN barriers (10 nm). The map of the
reciprocal lattice, which was deduced from X-ray diffraction
experiments, gave the average value of both lattice constants
a and c of the whole SL. Figure 1 shows three micro-Raman
spectra recorded in the z xx z conguration, when the laser
spot was focused slightly higher and higher upon the surface
of the sample. The variation of relative intensities of the
experimental features made their assignment easy either to
the SLs layers or to the underlying buffer layer, allowing the
measurement of the strain-induced frequency shift of the E2
(high) phonons from the SL. Combining all these data and
taking into account the measured in-plane strain of GaN
layers ( xx = 2 35%) derived from in-situ RHEED experiments, both components of the biaxial strain of AlN and
GaN layers could be determined. The out-of-plane strain
zz was found almost negligible for both types of layers,
giving for the ratio zz / xx a value quite different from
that predicted from the simple elastic theory (2C13 /C33 =
0 51). This difference is likely due to the large spontaneous
and piezo-electric polarization effects which can signicantly
decrease this strain ratio in the hexagonal SL, as demonstrated in a calculation by Gleize et al. [35].

3.2. Resonant Raman Scattering: Signature


of Single or Multiple QWs
The rst resonant Raman study of single GaN QWs was
published by Behr et al. [36]. In these samples grown by
MOCVD directly on a sapphire substrate without any buffer
layer, the wells lying between thick Ga0 85 Al0 15 N layers were
2 nm, 3 nm, or 4 nm thick. As expected, only the phonons
from the alloy were evidenced when the excitation was
achieved in the visible range, due to the negligible volume
of the QW. However, when the 3.54 eV laser line was used
for the excitation, that is, for an energy close to the estimated fundamental electronic transitions in the wider QWs,
the resonance conditions were nearly fullled and Frhlichinduced scattering by the GaN A1 (LO) mode was favored.
As can be shown in Figure 2, a Raman feature was observed
at 732 cm1 for the structures containing the 3nm- and 4nmwide QWs. The origin of this feature could be unambiguously assigned to the single well; its measured frequency
shift was negligible, because both connement and strain

518

Figure 1. Micro-Raman spectra recorded in backscattering along the


z axis, on a GaN (1.5 nm)-AlN (10 nm) SL grown on an AlN BL and
a SiC substrate. The excitation was made at 2.33 eV. Spectra 1 to 3
were obtained by focusing farther and farther away from the surface of
the sample. Modes from the SL and the BL are indicated by arrows.
Asterisks mark phonons from the substrate. Reprinted with permission
from [34], J. Frandon et al., Physica E (2003). 2003, Elsevier Science.

effects were probably weak in the well. In contrast, the LO


feature was broadened and shifted towards higher frequencies for the 2 nm thick QW. This observation was explained
by cation intermixing at the GaN-AlGaN interface, which
could not be neglected in that case. Smoothing of interfaces
likely due to poor growth conditions prevents the signature
of connement effects, which would induce an opposite frequency shift.

Figure 2. Resonant Raman spectra of GaN single QW with different


widths embedded in Al0 15 Ga0 85 N barriers, recorded under the 3.54 eV
excitation. Reprinted with permission from [36], D. Behr et al., Appl.
Phys. Lett. 70, 363 (1997). 1997, American Institute of Physics.

Phonons in GaN-AlN Nanostructures

Further resonant scattering experiments on single or multiple GaN-AlGaN QW structures will be reported in the following. Later Ten GaN QWs of 1.5 nm width, embedded
in 5 nm wide Ga0 89 Al0 11 N barriers and grown on a GaN
buffer layer, have been investigated under various ultraviolet excitations [37]. In the present structure, the QWs
were nearly relaxed, whereas the barriers were submitted
to an internal compressive stress, due to the presence of
the underlying buffer layer. As expected, rst-order scattering by the A1 (LO) phonons from GaN wells, favored by
the Frhlich interaction, was found strongly enhanced by
electronic resonance in the incoming channel when excitation is achieved using the 3.54 eV laser line. Indeed, the
energy of the incident photon was very close to the QWs
fundamental transition, as estimated using the simple model
of independent square potential wells of nite height. In
these experimental conditions, the second-order scattering
by the LO phonons from the thick buffer layer was dominant. Under a 3.70 eV excitation, LO phonons from GaN
wells were clearly evidenced in the second-order scattering
signal, due to electronic resonance in the outgoing channel;
weak contributions from the barriers were also evidenced.
Similar features were also observed in third-order scattering
under 3.80 eV excitation.
Another study performed in resonant conditions has been
devoted to a collection of four different single GaN QWs
separated by 10 nm thick Ga0 83 Al0 17 N barriers and grown
on a thick GaN buffer layer [38]; their thicknesses were
1 nm, 2 nm, 3 nm, and 4 nm (corresponding, respectively, to
4, 8, 12, 16 monolayers). The fundamental excitonic transitions in these wells were already determined from previous
PL measurements performed at a low temperature. In this
work, thick GaN and Ga0 83 Al0 17 N layers were also investigated in the same experimental conditions for comparison
with the structure under study. The Raman spectra recorded
at room temperature in backscattering geometry are shown
in Figure 3. The PL signature of the various wells was clearly
observed in the spectra, thus accurately giving the energy of
the fundamental transitions in the wells at room temperature. Under excitation at 3.53 eV, rst-order scattering by
GaN A1 (LO) phonons conned in one of the wider QWs
(3 nm) showed up in the spectra, enhanced by electronic
resonance in the outgoing channel; no similar observation
could be achieved on the thick GaN layer used as reference,
since the excitation was too far from resonance in the bulk
material. The measured frequency was close to 734 cm1 ,
corresponding to relaxed GaN, because the well involved
was almost unstrained in the nanostructure. In this case, the
Raman signature of the 3 nm-thick well, whose PL signal
was located at 3.43 eV, close to the observed Raman feature,
could be obtained. On the other hand, when the 3.70 eV
excitation was used, a rst-order Raman feature peaking
at higher frequency (775 cm1 and a weak contribution
at lower frequency was observed. They are clearly related
to phonons of the Ga0 83 Al0 17 N barriers and of the GaN
wells, respectively. Scattering by vibrational modes of both
layers requires an extended intermediate electronic state
of the QW with a signicant penetration into the barriers.
This delocalization implies high-lying states whose energy is
close to the bandgap of the alloy. Resonance most likely

519

Phonons in GaN-AlN Nanostructures

Figure 3. Resonant Raman spectra of four single GaN QW in


Al0 17 Ga0 83 N barriers, recorded using 3.53 eV and 3.70 eV excitations.
The thicknesses of these QWs were 4, 8, 12, and 16 monolayers. Arrows
mark the PL signal from the QWs. Reprinted with permission from [38],
F. Demangeot, Phys. Stat. Sol.(B) 216, 799 (1999). 1999, Wiley-VCH.

occurred in the incoming channel, as the incident photon


energy was almost tuned on the bandgap of the alloy. Moreover, the Raman signal was found just superimposed on the
PL signal from the thinner (1 nm) QW, centered at 3.60 eV;
no similar Raman features could be observed with thick
GaN or Ga0 83 Al0 17 N thick layers. The signal was thus likely
enhanced by a double resonance effect, as the second intermediate state involved in the Raman process was a bound
electronic state localized in the 1 nm-thick QW.

3.3. Calculations of Lattice-Dynamical


Properties of Wurtzite Nanostructures
Komirenko et al. [39] rst investigated the extraordinary
polar phonons of wurtzite single QWs in the framework of
a dielectric continuum model. The anisotropy of the materials was taken into account by introducing both components

and z (
) of the frequency-dependant dielectric tensor of the constituent materials; the latter involve the frequencies of A1 and E1 (LO and TO) phonons, polarized
along the z axis and in the (xOy) plane, respectively. The
electrostatic boundary conditions on the interfaces, together
with the assumption of a scalar potential vanishing far from
the wells, were used. For a long wavelength phonon characterized by the in-plane q wavevector, the angular dispersion
can be deduced from the following equation:

q2 + z
qz2 = 0

(10)

where qz is the z component of an effective wavevector.


Depending on the sign of the product z in each
medium, qz is found either real or imaginary, leading to linear superposition of oscillating or decaying solutions, respectively. Therefore, phonons could be classied according to
their localization. Conned modes are mainly located inside

the well but can penetrate signicantly into the surrounding


layers, interface modes decay exponentially from the interfaces in both types of layers, and propagating modes oscillate
in the whole structure. Their frequency variation versus the
product q d was given for a GaN well of width d embedded in innite barriers made of AlN or of Ga0 85 Al0 15 N.
Note that in another article, the same authors extended their
study to the calculation of scattering rates by the Frhlich
electron-phonon interaction in QW structures [40].
Very few investigations of lattice dynamics in hexagonal
superlattices by ab initio calculations have been published up
until now. Wagner et al. [41] computed the structural, dielectric, and lattice-dynamical properties of short-period hexagonal GaN-AlN SLs, which were compared to their cubic
counterparts. In this calculation, the widths of barriers as
well as of QWs were as low as two monolayers. Both types
of layers were assumed to be pseudomorphically strained
on the same in-plane lattice constant; the latter could be
either that of relaxed constituent materials (AlN or GaN)
or an average value corresponding to an elastically relaxed
SL. The goal of this work was to study the angular dispersion of phonons in the structures. The frequency
of
the zone center phonons was calculated as a function of ,
the tilt angle of the phonon wavevector with respect to the
z axis, within the framework of the density-functional perturbation theory. Most calculated modes were assigned either
to conned phonons or to interface phonons. The atoms
involved in each vibration mode, together with the strength
of the associated dynamical polarization, were also given.
The angular dispersion of these modes is shown in Figure 4
for a GaN-AlN wurtzite SL in several strain states. Actually, only a few differences were evidenced in the angular
dispersions of cubic and hexagonal SLs. In both cases, for
an increasing in-plane lattice constant, all modes decrease
in frequency, except the folded TA modes; the downward
shift was found more pronounced for LO phonons than for
the TO phonons. The only remarkable difference affects the
TO modes conned in the GaN layers of the hexagonal SL,
spreading in a narrower range than their counterparts in the
cubic nanostructure.
However, the latter results cannot be easily checked.
Indeed, experimental studies are usually performed on
nanostructures with much larger periods. An alternative
calculation based on the dielectric continuum model, rst
developed in [39] for the GaN QWs, was applied to wurtzite
SLs with more realistic periods [4244]. Within this framework, the vibrational modes were described by the dynamical polarization associated with the atomic motions in each
type of layer. The Maxwell equations, the electric boundary conditions at each interface, and the Blochs theorem,
taking into account the SLs periodicity, were used together
with the q = 0 phonon frequencies of the two types of layers in the strain state actually achieved in the SL. Note
that the phonon dispersion
q of the bulk constituents
was ignored in this model. For GaN-AlN SLs, this calculation predicted two types of polar phonons characterized by
an angular dispersion
 . The rst ones were the interface modes which have been already found for SLs of cubic
structure; the corresponding amplitude of atomic motions
decreases from the interface in both types of layers. Their

520

Phonons in GaN-AlN Nanostructures

Figure 4. Angular dispersion of zone-center phonons of an ultra-thin hexagonal GaN (0.5 nm)-AlN (0.5 nm) SL calculated in different strain
situations. From (l)(r), the results are given for SLs pseudomorphic on AlN, elastically relaxed on an average in-plane lattice constant, and
pseudomorphic on GaN. Reprinted with permission from [41], J. M. Wagner et al., IPAP Conference Series 1, 669 (2000). 2000, Institute of Pure
and Applied Physics.

dispersion is related to the anisotropy of the nanostructure. In contrast, another type of phonon, the quasi-conned
modes, exhibit an angular dispersion originating from the
anisotropy of the constituent materials. Indeed, the frequencies of the A1 and E1 phonons of GaN and AlN, in the
TO as well in the LO range, are the limits of their spectral
regions; the corresponding intervals are nite for wurtzite
crystals, but vanish for isotropic materials. Quasi-conned
modes correspond to oscillations in one type of SLs layers
but, in contrast with conned modes in cubic SLs, the associated electric eld penetrates into the adjacent layer with
a decay length much larger than the lattice constant: that is
why they are called quasi-conned. Note that modes delocalized throughout the whole nanostructure, characterized
by an oscillatory behavior in both types of layers, were found
only in GaN-GaAlN SLs with barriers made of Ga-rich
alloys within this dielectric approach.
Very recently, Romanov et al. [4546] calculated the polar
phonons of a single GaN QD, within the framework of a
macroscopic continuum dielectric model. They obtained formal analytical solutions for the surface vibrations of a GaN
QD exhibiting an oblate spheroidal form, embedded in AlN.
These modes are not discrete, in contrast with their counterparts in cubic GaAs-AlAs QDs, and they are found inside
a continuous, allowed frequency range, due to the crystal
anisotropy. In addition, two other types of phonons were
found: runaway modes that freely leave the QD surface and
quasi-stationary leaky modes.
The inuence of strong electric elds present in GaNAlN nanostructures has been discussed by Coffey and Bock
[47]. These authors calculated the wavefunctions associated
with electron and holes conned in strained GaN-AlN QWs.
The inuence of the electric elds (up to 1 MV/cm) induced
by internal strains on these states was shown in the particular case of a 2.6 nm-thick QW: holes and electrons are

spatially separated in the QW by the conned quantum


Stark effect. Strong effects on the Raman cross section by
A1 (LO) phonons conned in the GaN layer were proven
to proceed from the breaking of symmetry with respect to
the center plane of the well. In the calculated cross section, both contributions to the electron-phonon interaction
associated with deformation potentials and Frhlich processes were taken into account. For vanishing electric elds,
only conned phonons characterized by a quantum number
of even parity are allowed. In contrast, it was found that
conned phonons with an odd parity dominate the calculated Raman spectrum when strong elds are present in the
involved layer. Therefore, it was suggested that this breakdown of parity selection rules could be used for measuring
the electric eld in wurtzite nanostructures. Unfortunately,
the frequencies given for conned modes is questionable,
considering the dispersion
q calculated by the authors for
the LO phonon of bulk GaN, which is in agreement neither
with measurements nor with calculations already published
[911].

3.4. Experimental Investigations of Phonons


in Nitride-Based Superlattices
As previously shown for superlattices made of cubic IIIV
semiconductors, the Raman signatures of disordered solid
solutions and ordered nanostructures with the same mean
composition are quite different, except for SLs containing
ultra-thin (thinner than three monolayers) layers. Evidence
for SL ordering can thus be achieved by Raman spectroscopy, which is a nondestructive technique in contrast to
transmission electron microscopy (TEM) needing cross sections of the samples. However, Raman spectra of GaN-AlN
SLs and AlGaN alloys can at rst sight exhibit some similarities, specially concerning the frequencies of E2 (high)

521

Phonons in GaN-AlN Nanostructures

phonons, due to the two-mode behavior of the latter


modes in the alloy (see Section 2.2). Gleize et al. [30] have
observed that the ambiguity could be lifted by considering
the A1 (LO) phonon which is, on the contrary, characterized
in alloys by its one-mode behavior. Indeed, the phonon frequency expected for the A1 (LO) mode of the Ga1x Alx N
alloy with the mean Al content of the SL under study (x =
0 45) would be 821 cm1 [13]. Actually, the only feature obeying the appropriate selection rules, found at a much lower
frequency (738 cm1 in Raman spectra, was assigned to A1
(LO) phonons conned in the wells, slightly shifted from
the corresponding frequency in relaxed GaN by the opposite effects of connement and strain. It should be noted
that the A1 (LO) phonon conned in the barriers could not
be observed, likely due to the low-scattering cross section of
AlN. Finally, it was concluded that the SLs modes could be
unambiguously probed by nonresonant Raman scattering.
At this point of the discussion, one can wonder if Raman
experiments can determine the nature of the SL phonons,
delocalized or conned in one type of layer. A simple criterion has been suggested by Davydov et al. in several articles
[4850]; if two distinct phonons of the same symmetry can
be evidenced in Raman spectra from short period SLs, the
corresponding modes are claimed to be conned in one type
of layer. In the opposite case, the modes can be considered
as delocalized. However, it should be noted that the weakness of the Raman signal coming from one type of layer can
make a convenient use of this criterion difcult. The same
approach has been used by Gleize et al. [30], who implicitly
considered phonons conned in each type of layer, with the
exception of a phonon exhibiting the A1 (TO) symmetry,
assigned to an interface mode.
Chen et al. [51] claimed to obtain the rst evidence for
a connement effect of the A1 (LO) phonon in GaN layers of an hexagonal superlattice. In this study, three GaNGa0 8 Al0 2 N SLs were investigated; the thickness of the GaN
layers was 1.2 nm, 2.4 nm, and 3.6 nm, respectively. The
authors concentrated on the A1 (LO) mode observed in
z xx z micro-Raman spectra. The corresponding broad feature was slightly red shifted in frequency, compared to
the relaxed material; this shift was found increasing for
shrinking GaN wells. An approximate t of the measured
frequencies was obtained, using
qz sin2 qz c/4 for
the dispersion of the LO branch of GaN. However, it should
be noted that the SLs contained only 30 periods and that the
total thickness of GaN in the SL was as low as 36 nm for the
thinner layers; one can thus wonder if a clear signature of
the wells could be really obtained in nonresonant scattering
conditions, considering the presence of an underlying thick
GaN buffer layer.
Davydov et al. [4850] investigated by nonresonant
Raman scattering a set of GaN-Ga1x Alx N superlattices,
grown by MOCVD on buffer layers and sapphire substrates.
The SL period was varied between 2.5 nm and 320 nm. For
the sake of simplicity, the thicknesses of wells and barriers in the various nanostructures were the same in all the
nanostructures investigated. The aluminum content x in the
alloy of the barriers was lower than 50%. The total thickness
of the SLs under study was sufcient for achieving Raman
experiments in nonresonant scattering, with an excitation at
2.54 eV. In the recorded spectra, E2 (high) and E1 (TO)

phonons from GaN were observed and assigned to conned


modes in the wells, whereas their counterparts could not be
evidenced for the AlGaN layers of the SL. The lack of signal
from the barriers has been tentatively attributed to the lowscattering cross section by phonons of the AlGaN alloy. On
the other hand, the E2 (low) phonon conned in the barriers of the SL was also found in the low-frequency range of
the spectra, together with that from GaN layers [50]. Due to
its high sensitivity to the Al content (but not to the built-in
strain), the E2 (low) phonon from the alloy could be used
as a probe of the composition in the barriers of the SL. The
same authors also gave evidence for connement of the A1
(LO) phonon either in GaN or in GaAlN layers. In Figure 5,
two modes are clearly observed in z yy z spectra from various SLs where the alloy content of the barriers was kept
constant (28%). Note that the range of investigated periods
was very wide, between 5 nm and 3 m.
In contrast, the behavior of E1 (LO) and A1 (TO)
phonons was found quite different. Actually, each of them is
always observed as a single line; its location in Raman spectra is similar to that in a Ga1x Alx N disordered alloy, whose
composition corresponds to the mean Al content in the SL

x = x

d2
d1 + d 2

(11)

where d1 and d2 are the width of wells and barriers, respectively. Accordingly, these modes are considered as delocalized in the whole structure. The frequency variations of the

Figure 5. Raman spectra in the region of the A1 (LO) phonons,


recorded in backscattering along the z axis, on a set of GaN
Al0 28 Ga0 72 N SL with different periods: (1) 3 m, (2) 640 nm,
(3) 320 nm, (4) 160 nm, (5) 80 nm, (6) 40 nm, (7) 20 nm, (8) 10 nm,
(9) 5 nm. The thicknesses of wells and barriers in the SLs were the same
(unpublished results). Reprinted with permission from V. Yu. Davydov.

522
A1 (TO) phonons from the SLs under study and from the
corresponding ternary alloy were found rather similar. The
same observation was made on the E1 (LO) phonons. This
kind of evolution seems to be consistent with results of a
calculation performed in the framework of a dielectric continuum model, where the nanostructure was treated as an
homogeneous and anisotropic crystal, characterized by an
average dielectric constant z
.
Another possible evidence for the nature, localized or not,
of SLs phonons could be found by measuring their angular
dispersion
 experimentally, if the latter is signicantly
different from that of the bulk material. This has been
done by Gleize et al. [44] for polar phonons through an
experimental study of a GaN-AlN SL, performed for checking the validity of the continuum dielectric model previously developed [42]. Micro-Raman spectra were recorded
under 2.54 eV excitation, in backscattering geometry on the
top surface ( = 0), on the edge ( = 90 ), and also on a
bevel at 45 fabricated on the edge of the nanostructure
by ion etching. The variation of the angle  around the
above values was achieved by tilting the sample with respect
to the incident light beam; in this experiment, the uncertainty on  was lower than 5 . The measured frequencies of
various polar modes from the SL could be compared with
the predicted ones. As observed in Figure 6, the agreement
was found rather satisfactory, particularly for the TO modes
quasi-conned in GaN and AlN layers, which could not be
confused with the dispersive extraordinary TO modes from
the underlying thick AlN buffer layer.

3.5. Folded Acoustic Phonons in


Hexagonal Superlattices
The work published by Davydov et al. [49] on GaN-AlGaN
superlattices has evidenced for the rst time zone-centerfolded acoustic phonons in hexagonal SLs, characteristic of

Figure 6. Angular dispersion of polar phonons of a GaN (5 nm)- AlN


(5 nm) superlattice. Full circles and full lines correspond to the experimental values and to the calculated variation, respectively. The angular
dispersion of the relaxed thick buffer layer is indicated by dashed lines.
Reprinted with permission from [44], J. Gleize et al., Phys. Stat. Sol.
(A) 195 (2003). 2003, Wiley-VCH.

Phonons in GaN-AlN Nanostructures

their periodicity. As shown in Figure 7, remarkable sharp


Raman features were detected in the low-frequency range of
the spectra recorded in the z yy z scattering conguration,
from a few nanostructures made of GaN and Al0 28 Ga0 72 N
layers of same thicknesses, with SLs periods ranging between
6.1 nm and 23.8 nm. Actually, only the doublet expected
from the rst folding of acoustic branches was clearly
observed in each case. As expected, the mean frequency of
the doublet increases for decreasing SLs periods. The measured frequencies of both components of the doublet obeyed
fairly well Eqs. (4) and (6) appearing in Section 1.2. An average sound velocity v = 8140 m/s in the nanostructure was
derived from these measurements. Finally, different scattering congurations were achieved for changing the z component qz of the wavevector of the phonon involved in Raman
scattering. This experiment allows the probing of the frequency dispersion of the folded acoustic modes [27]. Indeed,
the authors found that the spectral spacing of both lines in
the doublet was decreasing with qz .

4. PHONONS IN QUANTUM
DOTS STRUCTURES
4.1. Nonresonant Raman Scattering
The rst Raman signature of QD structures has been published by Gleize et al. [52]. The samples were stackings of
GaN-AlN QDs grown along the (0001) direction on an AlN
buffer layer and a sapphire substrate. As demonstrated by
TEM studies [53], GaN islands exhibited a pyramidal shape
with a broad basis (about 30 nm) and a typical height of
4 nm, as shown in Figure 8. Micro-Raman spectra were
recorded under 2.54 eV excitation, in a backscattering geometry along the z axis. However, the whole sample was probed

Figure 7. Raman spectra of GaNAl0 28 Ga0 72 N superlattices with different periods (23.8 nm, 12.8 nm, and 6.1 nm), recorded under a 2.54 eV
excitation. Only the low frequency part of the spectra, exhibiting the
folded acoustic phonons, is shown. Reprinted with permission from [49],
Phys. Stat. Sol. (A), 188, 863 (2001). 2001, Wiley-VCH.

Phonons in GaN-AlN Nanostructures

Figure 8. Cross section of a GaN-AlN QD stacking, observed by transmission electron microscopy. The typical height of GaN QDs was 4
nm. Reprinted with permission from [53], C. Adelmann, Compte-Rendus
Academ. Sci. (Paris) 1, Serie IV, 61 (2000). 2000, Bruno Daudin.

under visible excitation and the distinction between signals


originating from the stacking and the buffer layer was not
straightforward. In order to overcome this difculty, the confocal conguration was used and the laser spot was focused
higher and higher above the surface, inducing a relative
intensity variation of the E2 phonons from the buffer layer
and from the QDs, as already illustrated for SLs. It allowed
the unambiguous signature of the GaN islands. A slight
(tensile) strain of the AlN spacers was deduced from the
measured (negative) frequency shift of the E2 phonon. The
effect of vertical correlation of QDs in these structures can
be evidenced from small changes of the strain measured for
the spacers.
Another study has been devoted to a set of GaN-AlN QD
stackings characterized by various heights and densities of
dots, deposited either on sapphire or on silicon [54]. Using
various backscattering geometries, most Raman-active optic
phonons from the structure could be observed. Except for
the A1 (LO) mode, the measured phonon frequencies were
rather close to those found in the similar disordered AlGaN
alloy. The observed frequency shifts were assigned to strain
effects, as in the case of SLs. An in-plane strain in GaN
dots of 2 4% or 2 6% was estimated from the frequency
shift (as high as +35 cm1 or +38 cm1 of the corresponding E2 phonon, using the corresponding deformation potentials. The QDs were found completely strained on AlN in
all the structures under study, as expected. Obviously, the
mean strain in the AlN spacers, tentatively derived from the
experimental data, was found much lower.

523
were used. In fact, large internal electric elds take place
in this kind of nanostructure, thus lowering the fundamental bandgap energies signicantly. In the present structure,
this quantum conned Stark effect was specially strong. The
room temperature PL originating from GaN QDs was centered around 2.35 eV, much lower than the excitation energy,
and thus did not merge the Raman signal. The more interesting result was obtained using the 3.80 eV laser line. A feature clearly showed up in the rst-order scattering range
at 744 cm1 in Raman spectra from the nanostructure, in
contrast to those recorded on the GaN layer used as a reference (see Fig. 9). The observed peak, which could not originate from the underlying GaN buffer layer of the sample,
was assigned to the polar A1 (LO) phonon from the QDs.
A weaker Raman feature located at 602 cm1 was associated with the nonpolar E2 phonon from the strained dots.
It should be noted that the latter mode could be observed
under the 2.33 eV excitation, that is, at an energy close
to the PL maximum, in contrast to the A1 (LO) phonon
from QDs. These results gave evidence for a strong resonant enhancement of the scattering by polar phonons in the
incoming channel at 3.80 eV, implying probably an excited
state of the dots.
Another original study was carried out by Kuball et al.
[56] on a single plane of self-assembled GaN QDs grown
on a Al0 15 Ga0 85 N layer, using silicon as anti-surfactant.
Two clearly distinct distributions of QD sizes were revealed
by means of atomic force microscopy. The Raman spectra
from the nanostructures are shown on the top of Figure 10.
Under a 3.53 eV excitation, rst-order scattering located at
736 cm1 , superimposed onto a broad PL band, was assigned
to the A1 (LO) phonon from the large (about 40 nm high)
dots, where connement effects on the LO phonon frequency are almost negligible. In the same experimental conditions, a phonon was observed at the same frequency but
with a lower intensity, on another sample grown in similar

4.2. Resonant Raman Scattering


Room temperature micro-Raman experiments have been
also performed on QD structures using ultraviolet laser
lines, in order to enhance the scattering intensity by means
of electronic resonance. In the paper by Gleize et al. [55]
the sample under study was a stacking of 39 periods of GaN
QDs embedded in 18 nm thick AlN spacers, grown on both
GaN and AlN buffer layers and on a Si (111) substrate.
A thick GaN unstrained layer was also available for comparison. Three laser lines at 3.41 eV, 3.70 eV, and 3.80 eV

Figure 9. Raman spectra of a thick GaN layer (a) and of a GaNAlN


quantum dot structure (b) grown on GaN and AlN buffer layers and on
a Si substrate, recorded in backscattering along the z axis under 3.41 eV
and 3.80 eV excitations. Reprinted with permission from [44], J. Gleize
et al., Phys. Stat. Sol. (A), 95 (2003). 2003, Wiley-VCH.

524

Phonons in GaN-AlN Nanostructures

diffraction. The shift towards lower frequencies observed in


the Raman spectra for the E2 (high), E1 (TO), and A1 (TO)
phonons of GaN was associated with nanosize effects.
Demangeot et al. [5859] studied structures made of
nanometer-size GaN pillars designed for photonic crystals.
The samples were obtained by reactive ion-etching using a
mask technique on 300 nm thick GaN layers grown on a
thick AlN buffer layer and on a sapphire substrate. Individual pillars of 300 nm in height, with diameters ranging between 5 m and 100 nm, were fabricated in this
way. The samples were investigated by micro-Raman spectroscopy under a 2.54 eV excitation. No one-dimensional
effects were observed as expected, considering the pillar size.
Spectra recorded in the z xx z conguration showed the
allowed phonons from each pillar. Due to the compressive
strain of the larger pillars, the E2 (high) mode was slightly
shifted from the frequency of relaxed material; in contrast,
strain relaxation was evidenced in smaller pillars. Moreover,
the upward shift of the LO phonon in the smaller pillars
could be explained by angular dispersion (see Section 2.1).
Indeed, the wavevector transferred to the phonon by the
incident and scattered light entering and coming out, respectively, through the facets of the pillars was tilted with respect
to the z axis, leading to the observation of a shifted quasiLO mode. The measured frequency shift was found in good
agreement with the value of the facet angle (25 ).

Figure 10. Raman spectra of self-assembled GaN QDs grown on


Al0 15 Ga0 85 N using Si as antisurfactant A , and of a continuous GaN
layer grown on Al0 15 Ga0 85 N B . The spectra have been recorded at
3.53 eV (top) and at 5.08 eV (bottom). In the latter case, a reference
spectrum of an Al mirror shows the system response function. Asterisks
and circles mark laser plasma lines and the Raman signal from the SiC
substrate, respectively. Reprinted with permission from [56], M. Kuball
et al., Appl. Phys. Lett. 78, 987 (2001). 2001, American Institute of
Physics.

conditions but without Si, made of a continuous 0.3 m


thick GaN layer on the alloy layer. The spectra recorded
at 3.70 eV and 3.81 eV did not give any signature of the
QDs. But under excitation in the far ultraviolet (5.08 eV),
a signicant shift (9 cm1 towards lower frequencies was
observed for the A1 (LO) phonon from the GaN dots, as
shown on the bottom of Figure 9. In the same experimental
conditions, no frequency shift could be evidenced with the
continuous GaN layer. The feature found at 727 cm1 was
thus assigned to the A1 (LO) phonon in small QDs; the
shift should be due to connement effects in dots exhibiting
smaller heights of about 23 nm. The electronic transition
favoring the resonant effect under the 5.08 eV excitation
was not yet identied.

4.3. GaN Nanowires and Pillars


Several investigations of samples containing GaN nanowires
have been recently published. For example, Jun-Zhang
and Lide-Zhang [57] studied by Raman spectroscopy such
as nanostructures, made of nanowires embedded in the
nanochannels of an anodic alumina membrane. The hexagonal structure of the nanowires was checked by X-ray

5. SUMMARY
Hexagonal GaN-AlN or GaN-GaAlN QW and QD structures have been extensively studied in the last few years, on
account of their important applications in opto-electronics.
The lattice-dynamical properties of two-dimensional systems
have been reviewed in this article. Only a few calculations of
phonons in such nanostructures have been performed yet.
Most articles published are actually devoted to experimental studies by Raman spectroscopy, or less often by infrared
measurements. The signature of QW and QD structures has
been obtained from nonpolar or polar phonons, by means
of nonresonant or resonant Raman scattering, respectively.
The results allowed the probing of internal strains and connement effects in the constituent materials. Up until now,
specic modes of GaN-AlN or GaN-GaAlN SLs have been
the subject of a few investigations. However, phonon connement in layers of the structures has been evidenced for
SLs in a wide range of periods. Moreover, the angular dispersion of quasi-conned and interface modes has been predicted and checked experimentally for the former. Finally,
GaN pillars recently investigated by Raman spectroscopy did
not exhibit lattice dynamical properties characteristic of onedimensional systems.

GLOSSARY
Quantum dot (QD) A small island made of a given semiconductor embedded in another semiconductor.
Quantum well (QW) A thin layer made of a given semiconductor located between two other semiconductors called
barriers.
Superlattice (SL) A periodic array of wells and barriers.

Phonons in GaN-AlN Nanostructures

REFERENCES
1. J. Frandon, F. Demangeot, and M. A. Renucci, Optoelectronic
Properties of Semiconductors and Superlattices Vol. 13, IIINitride Semiconductors Optical Properties I, pp. 333378, (M. O.
Manasreh and H. X. Jiang, Eds.). Philadelphia, Taylor & Francis,
2002.
2. R. W. G. Wyckoff, Crystal Structures, Vol. 1, Chap. III,
pp. 11112, 2nd ed. Interscience, 1963.
3. M. Leszcynski, H. Teisseyre, T. Suski, I. Grzegory, M. Bockowski,
J. Jun, S. Porowski, K. Pakula, J. M. Baranowski, C. T. Foxon, and
T. S. Cheng, Appl. Phys. Lett. 69, 73 (1996).
4. M. Tanaka, S. Nakahata, K. Sogabe, H. Nakata, and M. Tobioka,
Japanese J. Appl. Phys. Part 2, 36, L 1062 (1997).
5. W. Richter, Springer Tracts in Modern Physics 78, 121 (1976).
6. R. Loudon, Advances in Physics, 13, 423 (1964).
7. T. Deguchi, D. Ichiryu, K. Sekiguchi, T. Sota, R. Matsuo,
T. Azuhata, M. Yamaguchi, T. Yagi, S. Chichibu, and S. Nakamura,
J. Appl. Phys. 86, 1860 (1999).
8. J. M. Hayes, M. Kuball, Y. Shi, and J. H. Edgar, Japanese J. Appl.
Phys., Part 2, 39, 703 (1999).
9. H. Siegle, G. Kaczmarczyk, L. Filippidis, A. P. Litvinchuk,
A. Hoffmann, and C. Thomsen, Phys. Rev. B 55, 7000 (1997).
10. V. Yu. Davydov, Y. E. Kitaev, I. N. Goncharuk, A. N. Smirnov,
J. Graul, O. Semchinova, D. Uffmann, M. B. Smirnov, A. P.
Mirgorodsky, and R. A. Evarestov, Phys. Rev. B 58, 12899 (1998).
11. T. Ruf, J. Serrano, M. Cardona, P. Pavone, M. Pabst, M. Krisch,
M. DAstuto, T. Suski, I. Grzegory, and M. Lezczynski, Phys. Rev.
Lett. 86, 906 (2001).
12. F. Demangeot, J. Groenen, J. Frandon, M. A. Renucci, O. Briot,
S. Clur, and R. L. Aulombard, Proc. 2nd Eur. GaN Workshop
Internet J. Nitride Semicond. Res. 2, 40 (1997).
13. F. Demangeot, J. Groenen, J. Frandon, M. A. Renucci, O. Briot,
S. Clur, and R. L. Aulombard, Appl. Phys. Lett. 72, 2674 (1998).
14. N. Wieser, O. Ambacher, H. Angerer, R. Dimitrov, M. Stutzmann,
B. Stritzker, and J. K. N. Lindler, Proc. 3rd Int. Conf. Nitride
Semicond. Montpellier, France, Phys. Stat. Sol. (B) 216, 807 (1999).
15. A. Cros, H. Angerer, R. Handschuh, O. Ambacher, and
M. Stutzmann, Solid State Commun. 104, 35 (1997).
16. A. A. Klochikhin, V. Yu. Davydov, I. N. Goncharuk, A. N.
Smirnov, A. E. Nikolaev, M. V. Baidakova, J. Aderhold, J. Graul,
J. Stemmer, and O. Semchinov, Phys. Rev. B 62, 2522 (2000).
17. P. Wisniewski, W. Knap, J. P. Malzac, J. Camassel, M. D. Bremser,
R. F. Davis, and T. Suski, Appl. Phys. Lett. 73, 1760 (1998).
18. F. Demangeot, J. Frandon, M. A. Renucci, O. Briot, B. Gil, and
R. L. Aulombard, Solid State Commun. 100, 207 (1996).
19. V. Y. Davydov, N. S. Averkiev, I. N. Goncharuk, D. K. Nelson,
I. P. Nikitina, A. S. Polovnikov, A. N. Smirnov, and M. A.
Jacobson, J. Appl. Phys. 82, 5097 (1997).
20. J. M. Wagner and F. Bechstedt, Appl. Phys. Lett. 77, 346 (2000).
21. J. M. Wagner and F. Bechstedt, Phys. Rev. B 66, 115202 (2002).
22. T. Prokofyeva, M. Seon, J. Vanbuskirk, M. Holtz, S. A. Nikishin,
N. N. Faleev, H. Temkin, and S. Zollner, Phys. Rev. B 63, 125313
(2001).
23. A. Sarua, M. Kuball, and J. E. Nostrand, Appl. Phys. Lett. 81, 1426
(2002).
24. V. Darakchieva, P. P. Paskov, T. Paskova, J. Birch, S. Tungasmita,
and B. Monemar, Appl. Phys. Lett. 80, 2302 (2002).
25. J. Gleize, M. A. Renucci, J. Frandon, E. Bellet-Amalric, and
B. Daudin, J. Appl. Phys. 93, 2065 (2003).
26. B. Daudin, F. Widmann, G. Feuillet, Y. Samson, M. Arlery, and
J. L. Rouvire, Phys. Rev. B 56, R 7069 (1997).
27. B. Jusserand and M. Cardona, Light Scattering in Solids V, pp.
6073 (M. Cardona and G. Gntherodt, Eds.). Springer-Verlag,
Berlin, 1989.
28. M. Cardona, Light Scattering in Solids II, pp. 128135
(M. Cardona and G. Gntherodt, Eds.). Springer-Verlag, Berlin,
1982.

525
29. R. M. Martin and L. M. Falicov, Light Scattering I, pp. 79145,
(M. Cardona, Ed.). Springer-Verlag, Berlin, 1983.
30. J. Gleize, F. Demangeot, J. Frandon, M. A. Renucci, F. Widmann,
and B. Daudin, Appl. Phys. Lett. 74, 703 (1999).
31. M. Schubert, A. Kasic, J. Sik, S. Einfeldt, D. Hommel, V. Hrle,
J. Off, and F. Scholz, Materials Sci. Engin. B 82, 178 (2001).
32. M. Schubert, A. Kasic, J. Sik, S. Einfeldt, D. Hommel, V. Hrle,
J. Off, and F. Scholz, Internet J. Nitride Semicond. Res. 5, Suppl. 1
(2000).
33. G. Abstreiter, M. Cardona, and A. Pinczuk, Light Scattering
IV, pp. 1820 (M. Cardona and G. Gntherodt, Eds.). SpringerVerlag, Berlin, 1984.
34. J. Frandon, M. A. Renucci, E. Bellet-Amalric, C. Adelmann,
B. Daudin, in Proc. Int. Conf. on Superlattices, Nanostructures
and Nanodevices, Toulouse, France, Physica E (2003), to appear.
35. J. Gleize, J. Frandon, M. A. Renucci, and F. Bechstedt, in Proc.
Mater. Res. Soc. Spring Meeting, San Francisco, CA, USA, 680 E
(2001).
36. D. Behr, R. Niebuhr, J. Wagner, K. H. Bachem, and U. Kaufmann,
Appl. Phys. Lett. 70, 363 (1997).
37. J. Gleize, F. Demangeot, J. Frandon, M. A. Renucci, M. Kuball, N.
Grandjean, and J. Massies, in Proc. Eur. Mat. Res. Soc., Strasbourg, France, Thin Solid Films 364, 156 (2000).
38. F. Demangeot, J. Gleize, J. Frandon, M. A. Renucci, M. Kuball,
N. Grandjean, and J. Massies, Phys. Stat. Sol. (B) 216, 799 (1999).
39. S. M. Komirenko, K. W. Kim, M. A. Stroscio, and M. Dutta, Phys.
Rev. B 59, 5013 (1999).
40. S. M. Komirenko, K. W. Kim, M. A. Stroscio, and M. Dutta, Phys.
Rev. B 61, 2034 (2000).
41. J. M. Wagner, J. Gleize, and F. Bechstedt, in Proc. Int. Workshop on Nitride Semicond. IWN 2000, Nagoya, Japan, IPAP Conf.
Series 1, 669 (2000).
42. J. Gleize, M. A. Renucci, J. Frandon, and F. Demangeot, Phys.
Rev. B 60, 15985 (1999).
43. J. Gleize, J. Frandon, F. Demangeot, M. A. Renucci, M. Kuball,
J. M. Hayes, F. Widmann, and B. Daudin, Mater. Sci. and Engin. B
82, 27 (2001).
44. J. Gleize, J. Frandon, and M. A. Renucci, in 2nd Int. Workshop
Physics of Light-Matter Coupling in Nitrides, Rethimnon, Crete
(2002), Phys. Stat. Sol. (A) 195, 605 (2003).
45. D. A. Romanov, V. V. Mitin, and M. A. Stroscio, Phys. Rev. B 66,
115321 (2002).
46. D. A. Romanov, V. V. Mitin, and M. A. Stroscio, Physica B
316317, 359 (2002).
47. D. Coffey and N. Bock, Phys. Rev. B 59, 5799 (1999).
48. V. Yu. Davydov, A. A. Klochikhin, I. N. Goncharuk, A. N.
Smirnov, A S. Usikov, W. V. Lundin, E. E. Zavarin, A. V. Sakharov,
M. V. Baidakova, J. Stemmer, H. Klausing, and D. Mistele, in
Proc. Int. Workshop on Nitrides, Nagoya, Japan (2000), IPAP
Conf. Series 1, 665 (2000).
49. V. Yu. Davydov, A. A. Klochikhin, I. E. Kozin, V. V. Emtsev,
I. N. Goncharuk, A. N. Smirnov, R. N. Kyutt, M. P. Schglov,
A. V. Sakharov, W. V. Lundin, E. E. Zavarin, and A. S. Usikov,
Phys. Stat. Sol. (A) 188, 863 (2001).
50. V. Yu. Davydov, A. N. Smirnov, I. N. Goncharuk, R. N. Kyutt,
M. P. Scheglov, M. V. Baidakova, W. V. Lundin, E. E. Zavarin,
M. B. Smirnov, S. V. Karpov, and H. Harima, in Proc. Int. Workshop on Nitrides 2002, Aachen, Germany, Phys. Stat. Sol. B 234,
975 (2003).
51. C. H. Chen, Y. F. Chen, An Shih, S. C. Lee, and H. X. Jiang, Appl.
Phys. Lett. 78, 3035 (2001).
52. J. Gleize, F. Demangeot, J. Frandon, M. A. Renucci, M. Kuball,
F. Widmann, and B. Daudin, Phys. Stat. Sol. B 216, 457 (1999)
53. C. Adelmann, M. Arlery, B. Daudin, G. Feuillet, G. Fishman, and
Le Si Dang, Compte-Rendus Academ. Sci. (Paris), 1, Serie IV, 61
(2000).

526
54. J. Gleize, J. Frandon, M. A. Renucci, C. Adelmann, B. Daudin,
G. Feuillet, B. Damilano, N. Grandjean, and J. Massies, Appl.
Phys. Lett. 77, 2174 (2000).
55. J. Gleize, F. Demangeot, J. Frandon, M. A. Renucci, M. Kuball,
B. Damilano, N. Grandjean, and J. Massies, Appl. Phys. Lett. 79,
686 (2001).
56. M. Kuball, J. Gleize, S. Tanaka, and Y. Aoyagi, Appl. Phys. Lett.
78, 987 (2001).

Phonons in GaN-AlN Nanostructures


57. Jun-Zhang and Lide-Zhang, J. Appl. D. Applied Phys. 35, 1481
(2002).
58. F. Demangeot, J. Gleize, J. Frandon, M. A. Renucci, M. Kuball,
D. Peyrade, L. Manin-Ferlazzo, Y. Chen, and N. Grandjean,
J. Appl. Phys. 91, 2866 (2002).
59. F. Demangeot, J. Gleize, J. Frandon, M. A. Renucci, M. Kuball,
D. Peyrade, L. Manin-Ferlazzo, Y. Chen, and N. Grandjean,
J. Appl. Phys. 91, 6520 (2002).

Anda mungkin juga menyukai