Intersubband Transition-Based Processes and Devices in AlN/GaN-Based Heterostructures

Abstract

We report on the physics, epitaxial growth, fabrication, and characterization of optoelectronic devices based on intersubband transitions in the AlN/GaN material system. While in 1999, only results of optical absorption experiments could be shown, photodetectors and modulators with operation frequencies beyond 10 GHz as well as optically pumped light emitters have been demonstrated recently. This is the reason for a comprehensive report on the most important properties of such devices. Beside some basic theoretical considerations, we will concentrate on the fabrication and characterization of modulators, switches, photodetectors, and light emitters. At the end of this paper, an outlook to future trends and developments in this emerging field will be given.

Full text

Page 1

IntersubbandTransition-Based
ProcessesandDevicesin
AlN/GaN-Based Heterostructures
This review covers the physics, epitaxial growth, fabrication, and characterization
of optoelectronic devices for use in video players and other consumer electronics
as well as in commercial systems.
By Daniel Hofstetter, Esther Baumann, Fabrizio Raphael Giorgetta,
Ricardo The ´ron, Hong Wu, William J. Schaff, Member IEEE, Jahan Dawlaty,
Paul A. George, Lester F. Eastman, Fellow IEEE, Farhan Rana, Prem K. Kandaswamy,
Fabien Guillot, and Eva Monroy
ABSTRACT | We report on the physics, epitaxial growth,
fabrication, and characterization of optoelectronic devices
based on intersubband transitions in the AlN/GaN material
system. While in 1999, only results of optical absorption
experiments could be shown, photodetectors and modulators
with operation frequencies beyond 10 GHz as well as optically
pumpedlight emittershavebeen demonstratedrecently.Thisis
the reason for a comprehensive report on the most important
properties of such devices. Beside some basic theoretical
considerations, we will concentrate on the fabrication and
characterization of modulators, switches, photodetectors, and
light emitters. At the end of this paper, an outlook to future
trends and developments in this emerging field will be given.
KEYWORDS | III-nitrides; intersubband transitions; optical
rectification; quantum cascade detectors; quantum wells
I. INTRODUCTION
Intersubband transitions have already quite a long
history; their first description dates back to 1982 by
Ando et al. [1]. These three authors observed a novel kind
of optical transition between bound electronic levels
which were present in the two-dimensional carrier gas at
a semiconductor heterojunction. Because of the small
energetic separation between adjacent levels, the ob-
served absorption effects were visible only under
excitation with far-infrared radiation. Therefore, the
practical use of this effect was not immediately evident.
After first proposals of using such intersubband transi-
tions in infrared photodetectors by Coon et al. in 1984
[2], West and Eglash demonstrated in 1985 intersubband
absorption in a very thin semiconductor layer sandwiched
between two higher bandgap semiconductors. This
Bquantum well[ (QW) configuration exhibited basically
the same behavior as the heterojunction three years
before: it absorbed infrared light [3], but this time at
shorter midinfrared wavelengths. The big advantage of
this new result was its possibility of being tailored: by
simply growing thicker QWs, one could reduce the
D. Hofstetter, E. Baumann, F. R. Giorgetta, and R. The ´ron are with the University of
Neuchatel, 2009 Neuchatel, Switzerland (e-mail: Daniel.Hofstetter@unine.ch;
Esther.Baumann@unine.ch; Fabrizio.Giorgetta@unine.ch; Ricardo.Theron@unine.ch).
H. Wu, W. J. Schaff, J. Dawlaty, P. A. George, L. F. Eastman, and F. Rana are with
Cornell University, Ithaca, NY 14853 USA (e-mail: hwu@smics.ca; wjs2@cornell.edu;
jd234@cornell.edu; pag25@cornell.edu; lfe2@cornell.edu; Farhan.Rana@cornell.edu).
P. K. Kandaswamy, F. Guillot, and E. Monroy are with Equipe mixte CEA-CNRS
Nanophysique et Semiconducteurs, INAC/SP2M/PSC, CEA-Grenoble, 38054 Grenoble
Cedex 9, France (e-mail: premkumaraint@gmail.com; Fabien.Guillot@cea.fr;
Eva.Monroy@cea.fr).
Published in
Proceedings of the IEEE 98, issue 7, 1234-1248, 2010
which should be used for any reference to this work
1

Page 2

energy of the absorbed photons. Two years after this
important milestone, the first quantum-well infrared
photodetector (QWIP) was realized by Levine et al. [4].
This was a very crucial step towards practical devices
such as focal plane arrays for military and surveillance
applications [5], [6]. Another seven years later, in 1994,
the first laser emission, from the quantum cascade (QC)
laser, was demonstrated by a research group headed by
Capasso at Bell Laboratories [7]. This was the beginning
of an impressive development that culminated in a room-
temperature midinfrared continuously driven semicon-
ductor laser in 2002 [8], [9]. Nowadays, QWIPs and QC
lasers are commercialized and are present in many
important applications. An especially interesting aspect
of intersubband physics has been addressed only in
1999: it was the demonstration of intersubband transi-
tions approaching the technologically important wave-
length range around 1.55 ?m. Such transitions are
possible in carefully selected material systems offering a
sufficiently high conduction band discontinuity. One
possibility is the use of AlN/GaN, which exhibits a band
offset of nearly 2 eV; this is clearly sufficient for the
exploitation of near-infrared intersubband transitions
[10]–[17]. However, one has to take into account certain
particularities of this material system such as its
hexagonal crystal structure and its pyro- and piezoelectric
behavior. These properties lead to a relatively complicated
band structure in which the simulation of electronic
levels is substantially more difficult than for instance in
GaAs/AlGaAs [18]–[20]. Nevertheless, recent experi-
ments with AlN/GaN-based superlattice structures have
revealed very interesting physical properties, which have
eventually resulted in high-speed optoelectronic devices for
the 1.3/1.55 ?m wavelength region. That this statement is
actually more than just a hypothesis has been confirmed
already by several experiments which demonstrated inter-
subband scattering times on the order of hundreds of
femtoseconds [21], [22].
This paper is organized as follows. In Section II, we
will present some theoretical aspects of intersubband
physics and introduce Fermi’s golden rule. In Section III,
some important facts of crystal growth will be described.
Section IV deals with devices such as photodetectors, light
emitters, and saturable absorbers. In Section V, we will
discuss future trends and possible new developments in
this field.
II. PHYSICS/THEORY
A. Physics of Intersubband Transitions
The physics of intersubband transitions can be
understood on the base of Fermi’s golden rule. In our
case of a two-level system (with energies Ef and Ei)
interacting with electromagnetic radiation, it describes the
probability Wif of an electron passing from the lower
(energy Ei) to the upper level (energy Ef). In very general
terms, Fermi’s golden rule writes as
Wif¼2?
? h
ijH0j f
??
????2? ?ðEf? Ei? ? h!Þ
(1)
with ? h! being the photon energy, H0the interaction
Hamiltonian,and iand fthewavefunctionsoftheinitial
and the final electronic states, respectively. If we use the
dipole approximation, then the interaction Hamiltonian
can be expressed as H0¼ ðq2E2
the elementary charge, E0 the amplitude of the electric
field, m?the effective mass, ! the angular frequency of the
incoming radiation, ~ p the momentum operator, and ~ e the
unit vector of the light’s oscillation direction. If this term is
inserted into (1), then we get
0=4m?2!2Þ~ e ?~ p with q being
Wif¼2?
? h
q2E2
4m?2!2
0
ij~ e ?~ pj f
??
????2? ?ðEf? Ei? ? h!Þ: (2)
In the next step, the absorption probability given by (2) is
normalized towards the number of incoming photons and
multiplied with a geometrical factor to take into account
the angle between the incoming radiation and the surface
normal. Integration over all allowed directions has to be
done to get rid of the scalar product in the matrix element,
while the energy-conserving delta function is replaced by a
Lorentzian lineshape [23]. All these operations result in a
compact formula for the peak absorbance ?tQW
? ? tQW¼2?2q2
"0n?? ns? nQW? hzifi
jj2?sin2?
cos??
1
?Eif
(3)
where ? is the incidence angle measured towards the
surface normal, zifthe dipole matrix element, nsthe sheet
carrier density, nQWthe number of QWs, n the refractive
index of the material under investigation, ? the wave-
length corresponding to the energy difference between
states i and f, "0the vacuum permittivity, and ?Eifthe full
width at half-maximum of the optical transition. Careful
analysis of the formula shown above reveals that interac-
tion between light and an intersubband transition is
possible only for transverse magnetic (TM) polarized light.
This interesting feature is known under the term
polarization selection rule. Unfortunately, it prevents us
from using the most convenient geometry for all of our
experiments, namely, vertical incidence. The latter would
be possible only if one would add a surface grating for a
redirection of the light within the sample. For any
particular sample geometry, for instance, a multipass
zigzag waveguide, (3) contains the number of passes and
2

Page 3

the number of active region QWs. According to these
considerations, a typical 1.5-nm-thick GaN QW sand-
wiched between AlN barriers and with a silicon doping
level on the order of 1? 1020cm?3will produce roughly
0.5% absorption per pass. This is valid for an incidence
angle of 45?into the sapphire substrate. Using, for
instance, 40 active region periods, it is therefore possible
to fabricate structures which absorb nearly 50% of the light
per double-pass. An example of such absorbing structures
is given in Fig. 1; it compares the absorption curves for five
double-passes through structures with different QW
thickness. Due to the decreasing size of the dipole matrix
element in thinner QWs, the absorption goes down if the
well thickness is reduced. From the broadening of the
absorption curves at higher transition energy, an inter-
facial roughness of about one monolayer can be estimated.
This number will be confirmed below by investigations
using a transmission electron microscopy (TEM).
B. Saturation of Absorption
Another interesting aspect of intersubband transitions
is the effect of absorption saturation. In optical pump and
probe measurements on AlN/GaN superlattice structures,
several authors have reported very short carrier relaxation
times. For nitride structures absorbing at 1.55 ?m, the
most typical absorption recovery times are on the order
of 170–370 fs [21], [24], [25]. Based on this value, one
can compute at which intensity saturable absorption
should occur. The basic relation describing this behavior
is given by
?=?0¼
1
1 þ ðIin=IsatÞ
(4)
where ? and ?0 are the actually measured absorption
under saturation and the maximally possible absorption
without saturation, and Iinand Isatthe incoming and the
saturation intensities, respectively. Since pump probe
experiments using different input intensities reveal the
saturation intensity of this two-level system, these mea-
surements allow us an indirect determination of the
lifetime of the system [26]. This lifetime is in fact linked to
the saturation intensity via
?life¼?E12
2Isat
?cos?
sin2??nQW? ns
?0
?
1
tQW:
(5)
In (5), ?E12 is the photon energy, while the other
quantities are defined as in the previous formulae. It turns
out that the saturation intensity for nitride-based QW
structures is quite high, on the order of 1 GW/cm2.
Although certain other applications might be able to
produce such a high intensity, the main application field
will most likely lie in ultrafast solid-state lasers with
saturable absorbers. First experiments have indeed
revealed a great potential for such switching devices. In
the nitride material system, it seems to be crucial to adapt
the device geometry to achieve saturation intensities. By
making waveguides with a very small cross-section, it is
possible to fabricate saturable absorbers which can be
operated with moderate power levels. This technique has
been successfully implemented by the Iizuka group [27],
and it has been demonstrated by the Moustakas group at
Boston University that one can still refine it in order to
further reduce the required saturation power [28].
C. Photodetectors
Intersubband photodetectors based on AlN/GaN super-
lattices show an interesting mode of functioning that could
be clarified only several years after their first experimental
demonstration. For a correct understanding of their
mechanism, it is crucial to know the exact size and
orientation of the internal polarization fields. According to
the theory presented in Ambacher et al. [29], these internal
pyro- and piezoelectric polarizations can be treated like
PAlGaN
sp
ðxÞ ¼ ?0:090x ? 0:034ð1 ? xÞ
þ 0:021xð1 ? xÞ
ðxÞ ¼0:026ð1 ? xÞ þ 0:0248xð1 ? xÞ
ðxÞ ¼ ?0:0525x þ 0:0282xð1 ? xÞ
(6)
PAlAlN=GaN
piezo
PAlGaN=GaN
piezo
(7)
(8)
where x is the Al concentration of the AlGaN compound and
PAlGaN
sp
ðxÞ is the pyroelectric polarization for AlGaN.
Equation (6) describes the spontaneous polarization of
AlGaN compounds, (7) stands for the piezoelectric polari-
zation PAlAlN=GaN
piezo
ðxÞ of an AlGaN layer grown fully strained
on AlN, and (8) describes the piezoelectric polarization
Fig.1.Transmissionspectraofaseriesoffivesampleshavingdifferent
QW thicknesses (15, 20, 25, 30, and 35 A ˚) and showing the quantum
confinement shift [36, Fig. 7].
3

Page 4

PAlGaN=GaN
piezo
to the work of Bernardini and Fiorentini [30], the internal
fields Fint in the superlattice structure are related to the
respective polarizations and layer thicknesses by
ðxÞ for an AlGaN layer on top of GaN. According
FGaN
int¼
tAlN
PGaN
sp
??????? PGaN
???þ PAlN
piezo
???
???
???? PAlN
???? PGaN
sp
???
???
???? PAlN
???þ PGaN
piezo
??? ???
??
"AlNtGaN þ "GaNtAlN
(9)
FAlN
int¼
tGaN
PAlN
sp
???
piezo
sp
piezo
??? ???
??
"AlNtGaN þ "GaNtAlN
(10)
where "AlNand "GaNare the dielectric constants for AlN
and GaN, respectively, while tAlN and tGaN are the
corresponding layer thicknesses. For AlN/GaN hetero-
interfaces with all layers fully strained on the underlying
AlN substrate, one obtains thus polarization fields on the
order of 5–8 MV/cm. The resulting asymmetry of the QW
has important consequences for the functioning of
intersubband photodetectors. As shown in the schematic
picture of Fig. 2, the ground state wavefunction is slightly
displaced towards the left (i.e., in the [0001] growth
direction) while the first excited state is displaced towards
the right (i.e., in the [000-1] direction). Therefore,
excitation of an electron into the excited state is always
accompanied by a displacement of the electron along the
growth axis. However, since the positively charged doping
atom does not change its place, a small dipole moment
occurs for each electron/dopant atom pair. If, due to strong
absorption, a certain fraction of electrons will undergo
such a transition, the resulting total light-induced polar-
ization can become appreciable. It will eventually be large
enough to produce an electrical voltage between a dark and
an illuminated area of the device [31]. The photovoltage
generated by this effect can be described by
Vrect¼
nQWtQW
"0"stat
GaNnGaN? ?12? ?12? n3D? hn12i
jj2?q3E2
2?? h
sin2?
cos?
(11)
where tQW¼ 1:5 nm is the thickness of the QW, nQW¼ 80
(according to one double-pass through the active region) is
the number of QWs, ?12¼ 370 fs is the lifetime of the
upper quantized state, ?12¼ 3:0 A˚is the lateral displace-
ment of the electrons, n3D¼ 1020cm?3is the volume
carrier density, n12¼ 3:5 A˚is the dipole matrix element of
the optical transition, ? ¼ 37?is the angle of incidence
within the GaN material (diffracted by the sapphire
substrate), E2¼ ð2 IÞ=ðc "0Þ ¼ 7:5 ? 1010V2=m2is the
square of the incoming electric field amplitude (10 mW
focused to an area of 10 ? 10 ?m2), 2? ¼ 80 meV is the
full width at half-maximum of the transition, nGaN¼ 2:2 is
the refractive index of the QW material, and "stat
the static dielectric constant of GaN. Using the parameters
stated above results in a value of 80 ?V/W, which needs to
be compared to the experimental value of 130 ?V/W at
150 K. It has to be stressed, however, that in particular, the
lifetimes and the lateral displacements suffer from large
uncertainties. This is so because of an eventual transfer of
carriers into additional impurity levels being resonant with
the upper detector state. The disadvantage of such an optical
rectification type of operation is obviously that the lateral
displacements are relatively small, and the lifetimes of the
excited electrons are short; both effects limit the size of the
photovoltaicsignaltorelativelysmallvalues.However,since
again the lifetime is short, the system recovers very quickly
so that high-frequency operation is easily possible. In the
discussion about the experimental results, we will see that
such detectors have already been demonstrated for signal
frequenciesinexcessof10GHz.Recentexperiments usinga
quantum dot-based active region have revealed a marked
performance improvement, most likely due to lateral
coupling of the excited quantum dot state with the ground
state of the wetting layer.
GaN¼ 10:2
D. QC Lasers and Light Emitters
As a final point, quantum cascade lasers remain to be
demonstrated in this material system. Although there has
been quite a considerable effort in this direction, results
are somewhat discouraging so far. Since, in addition, no
truly convincing evidence of resonant tunneling has been
reported at this point, an electrically injected QC light
emitter remains an extremely challenging target. Never-
theless, there has been substantial progress in terms of
optically pumped light emission. The most advanced
experiments have been conducted at the University of
Paris-Sud in the group of Julien [32].
Fig. 2. Schematic representation of a GaN-based QW with a slightly
‘‘diagonal’’ transition due to the potential asymmetry induced
by the piezo- and pyroelectric field. The [0001] growth
direction is marked by an arrow [36, Fig. 16].
4

Page 5

One of the most serious problems for a QC laser in
this material system is clearly the short lifetime of an
electron in the upper quantum state. Even if this value
could be made as Blong[ as, say, approximately 350 fs
using tricks like diagonal transitions, the lower state
lifetime should still be considerably shorter in order to
establish a sufficiently large population inversion. Con-
sidering the good laser performance achieved with the
so-called bound-to-continuum design in InGaAs/InAlAs-
based QC lasers, such a design is also proposed for
nitride-based light emitters. It is shown in Fig. 3. The
injector is a regular superlattice whose electronic states
split into a miniband under application of a small bias
voltage. Owing to the equal layer thickness of GaN wells
and AlN barriers, strain will be fully compensated in the
injector. The gain region itself is designed as two strongly
coupled QWs. The lower lasing state is aligned with the
uppermost downstream injector states, while the upper
lasing state is a hybrid state between the two QWs of the
gain region. It is energetically aligned with the lowest
upstream injector states and has a good overlap with the
injector wavefunctions. Thanks to the continuum-like
lower lasing state, the electron lifetime could in theory be
reduced to 150 fs, which is short enough to achieve an
inversion. The proposed laser structure is designed to emit
at a photon energy of roughly 500 meV (2.5 ?m). Since
each period would show a voltage drop of at least 600 mV,
a maximum of 15–20 active region periods can be grown.
Otherwise, the total voltage drop on the structure would
exceed 12 V, which can be considered as maximum value
to be tolerated. The waveguide of such a device can be
formed using the sapphire substrate as a lower cladding
layer. The upper cladding is then fabricated with an
evaporated SiO2layer; altogether this would result in an
appreciable overlap factor for the active region. Based on
the different lifetimes and the expected population
inversion, an estimation of the maximal current density
Jmaxis possible
?n ¼ n3? n2¼Jmax
q
?3 1 ??2
?32
??
(12)
where n2 and n3 are the sheet carrier densities in the
lower and upper laser levels 2 and 3, respectively, ?2and
?32 are the lifetimes of the lower and the upper laser
level (considering only the 3 ! 2 transition), and
??1
3
¼ ??1
ering carrier loss into level 1 as well. Using the
lifetimes cited earlier and a population inversion of
?n ¼ 3 ? 1010cm?2, we get Jmax¼ 24 kA/cm2. In addi-
tion, it is important to know the approximate threshold
current for a given structure. This quantity can be
estimated if parameters like spectral linewidth
2? ¼ 0:1 eV, oscillator strength of the transition f ¼ 0:6,
effective mass m?¼ 0:2 me, and combined waveguide/
mirror losses ?mþ ?WG¼ 1 cm?1are known. We then get
32þ ??1
31is the total lifetime of level 3 consid-
Jth¼"0
4?q?
1
?3 1 ??2
?32
???2?nrefrLp?ð?mþ ?WGÞ
?optz2
ij
:
(13)
When using realistic values of ? ¼ 2:5 ?m, ?3¼ 150 fs,
?opt¼ 40%, ?opt¼ 40%,nrefr¼ 2:5,andLp¼ 15 nm,this
computation yields minimal values on the order of
Jth¼ 10 kA/cm2. In contrast to passive devices such as
photodetectors or optical switches, which rely on intersub-
band absorption effects only, such lasers would unfortu-
nately not be ultrafast. This might be somewhat
counterintuitive at first glance, but as described in a paper
by Bouadma et al. [33], a short laser cavity with a very short
photon lifetime offers a very efficient way to push the
relaxation oscillation frequency of a semiconductor laser to
its highest values. Since GaN QC lasers will obviously be
workinginalow-loss/low-gainregime,theywillmostlikely
haverelativelyhighthresholdsandlongcavities.Therefore,
they are not very good candidates for high-speed modula-
tion. For this reason, only moderate maximal modulation
frequencies in the low gigahertz range can be expected.
III. FABRICATION
AlN/GaN QW structures can be grown either by molecular
beam epitaxy (MBE) or by metal-organic vapor phase
epitaxy (MOVPE). Most of the results to be presented here
are based on material grown by plasma-assisted MBE
(PAMBE). The low growth temperature of this technique
results in sharp AlN/GaN interfaces, which is a critical
Fig. 3. Design of a nitride-based QCL structure. The layer thicknesses
in A ˚are as follows: 8=8==8=8==8=8==8=8==8=8==8=8==8=8==17=5==5=7:
Boldface means GaN wells, italics stands for AlN barriers, and
underlining marks the doped layers.
5

Page 6

requirementtoachievedecentdeviceoperationat1.55?m.
Here we give only some general information concerning
the growth procedure; details can be found in the work of
Monroy [34], [35]. All PAMBE samples are grown on AlN-
on-sapphire templates fabricated by MOVPE. The AlN
layer has a thickness of roughly 1 ?m and the orientation
of the substrate is C-face. During the growth of the
PAMBE layers, different methods have been investigated
in order to guarantee metal-rich growth conditions, which
in turn result in smooth surfaces. More particularly, In
used as a surfactant, Ga excess, and growth interruptions
along with Ga and Al excess for GaN and AlN layers,
respectively, have been explored [35]. From a morpho-
logical point of view, the three different methods gave
comparable results; however, the crystalline quality of the
Ga excess samples was highest, so that most layers of this
paper were grown using this method. A typical charac-
terization of sample morphology and interface quality
using atomic force microscopy (AFM) and TEM is
presented in Fig. 4(a) and (b), respectively. The AFM
analysis showed a root-mean-square surface roughness of
about 0.6 nm in an area of 2.5? 2.5 ?m2. From the
analysis of the TEM picture, we concluded that the AlN/
GaN interfaces are abrupt at the monolayer scale [36].
If, on the other hand, MOVPE is used for growth of
such structures, one has to deal with somewhat larger
interface roughness [37], [38]. However, for certain device
applications involving thick layer stacks, it is obviously an
advantage to have a higher growth rate like is typically
achieved using MOVPE. Different authors have also
reported structures that contain layers from both growth
techniques; in these samples, all thick cladding layers are
typically grown using MOVPE, while the thin QWs and
barriers in the active region are grown by MBE. This is an
ideal combination of the high growth rate and good crystal
quality of MOVPE and the high interface control of MBE.
A good example of such a device is the lattice-matched
InAlN/GaN waveguide reported by Lupu et al. [39].
IV. DEVICES
In terms of devices, we will first briefly talk about saturable
absorbers, optical switches, and modulators. Since consider-
able work has been devoted to photodetectors, it will consti-
tute the main part of this paragraph. Lastly, some results of
luminescence devices will be presented and discussed.
A. Saturable Absorbers/Modulators
As we have seen in the theory paragraph, the
intensities required to achieve saturable absorption in
GaN are quite considerable. Nevertheless, some successful
demonstrations have been published. The most advanced
structures are those reported by Iizuka et al. from Toshiba
Corporation in Japan [40], [41]. As shown in the
schematic cross-section of Fig. 5, the device was grown
on a 70-nm-thick high-temperature AlN (HT-AlN) buffer
followed by a 500 nm AlN/GaN multiple intermediate
layer (MIL). The active zone consisted of a 480-nm-thick
GaN lower waveguide layer, 10 AlN/GaN multi-QWs (Si,
5? 1019cm?3), and a 960-nm-thick GaN upper wave-
guide layer. The QW thickness was 1.5 nm while the
barriers measured 2 nm. Most recent devices were
mounted in a butterfly package and fiber-pigtailed for
further use, as presented in the photograph of Fig. 6.
Fabrication was based on lateral waveguide etching with
an electron cyclotron resonance reactive ion beam etching
Fig. 4. (a) Atomic force microscopy surface scan and (b) transmission electron microscopy image of a superlattice sample grown by
PAMBE [36, Fig. 4].
6

Page 7

(ECR-RIBE) system. The switch consisted of a 400 ?m
long and 1.5 ?m wide waveguide with tapered ends on
both sides. Antireflection coatings helped to minimize
the in- and out-coupling losses to very small values. The
total insertion loss of the component was on the order of
6.6–8.6 dB for TE mode.
Fig. 7 shows a series of experiments proving the correct
functioning of the device. Using the signal and idler beams
of an optical parametric oscillator (OPO) as signal and
control pulse generators, respectively, switching times of
below 1 ps could be achieved. Thanks to the waveguide
scheme, the switching energy was reduced to a value on
the order of 100 pJ. Extinction ratios of the saturable ab-
sorption in excess of 11.5 dB were seen. Since the recovery
times were on the order of G 1 ps, a series of pulses
separated by only 1 ps in time could be demultiplexed
using these structures.
As far as modulators are concerned, several prototype
devices were demonstrated first by the authors at the
University of Neuchatel [42], but later mainly from the
Julien team at the University of Paris-Sud. While our
device was based on an electron transfer between the two-
dimensional carrier gas underneath a 5 QW active region
(see Fig. 8), their first tests were performed using coupled
double QWs with a very thin tunneling barrier in between.
With our prototype modulator, we faced the problem of an
insufficient signal on–off ratio. There was a strong
absorption tail from the two-dimensional electron gas’
reaching into the wavelength region of interest. Since this
absorption could not be suppressed, the on–off ratio did
not become higher than a couple of decibels. This is shown
in Fig. 9, which presents the absorption spectra for three
different voltages on the top contact. On the other hand,
the main problem of the double QW structures was to
adjust the doping levels and the layer thicknesses in a
suitable way, which would allow moving the carriers from
one well into its neighbor [43]. The published work
showed a modulator with a quite convincing performance,
however. The device consisted of 20 periods of a 3-nm-
thick reservoir GaN QW coupled via an ultrathin AlN
barrier of 1 nm to an active GaN (Si, 5? 1019cm?3) QW
of 1 nm thickness. This active region was sandwiched
between two Al0:6Ga0:4N (Si, 5? 1019cm?3) contact
layers. The samples were processed as square mesas with
700 ?m side length having 500? 500 ?m2large windows
in the center. This process allowed testing under Brewster
geometry. As shown in Fig. 10, electromodulated absorp-
tion in two different wavelength windows around 1.2–1.67
Fig. 5. Schematic representation of a GaN-based switch as used by
Iizuka. The structure was grown on an HT-AlN buffer layer,
followed by a MIL; while the mesa was etched with an
ECR-RIBE system [41, Fig. 1].
Fig. 6. Photograph of a GaN-based switch as used byIizuka. Fibers for
optical input and output are coming in from the left/right, while the
electrical connections are shown on the bottom.
Fig. 7. Experimental results of a GaN-based all-optical switch as used
by Iizuka. Whenever signal and control pulse coincided in time,
a bleached absorption (increased transmittance) was observed.
The shortest pulse interval demonstrated was 0.67 ps [41, Fig. 6].
7

Page 8

and 2.1–2.4 ?m was achieved. The maximum modulation
depth was 44% and the RC-limited cutoff frequency
reached a value of 11.5 MHz.
B. Photodetectors
In some sense, AlN/GaN-based infrared photodetectors
play a privileged role because they constituted the first
working optoelectronic device using intersubband transi-
tions in this material system [44]. It took, however, quite a
long time until the functioning of these components was
well understood. Here, we review the current status and
give an outlook of future developments.
The generic structure for such photodetectors is based
on an MBE grown, 40-period AlN/GaN superlattice with
layer thicknesses of 1.5 nm for the Si-doped (0.5–
1:0 ? 1020cm?3) GaN QWs, and between 1.5 and 15 nm
for the AlN barriers (undoped). This active region is
typically grown on a 500-nm-thick AlN buffer layer and
covered with a 100 nm AlN cap. Growth is performed on
top of MOVPE-grown templates consisting of a 1-?m-thick
AlN layer. Characterization of such structures is done in
two stages requiring almost identical sample preparation:
Ti/Au (10/400 nm) based contacts are being evaporated on
the sample surface; the patterning is stripe-shaped for low-
frequency electrical tests and square-shaped for high-
frequency measurements. Absorption experiments do
usually not require any electrical contacts. The stripes are
800 ?m wide and 3 mm long, while the squares measure
typically 100 ?m across. In order to allow an efficient
coupling of the incoming radiation into the sample, two
Fig. 8. Electrooptical modulator prototype with an applied bias of ?2 V (a) at the top contact and (b) without bias. Under the action of the
applied bias voltage, the carriers will be transferred from or to the two-dimensional electron gas [42, Fig. 3].
Fig.9. Experimentalresultsofthemodulatorpresentedschematically
in Fig. 8. An on–off ratio on the order of 3 dB was seen [42, Fig. 4].
Fig.10. ExperimentalresultsofthemodulatorpresentedbytheJulien
group. An on–off ratio on the order of 11.5 dB was seen [43, Fig. 3].
8

Page 9

parallel 45?inclined facet mirrors are being polished on
both sides. To form a zigzag waveguide, the back of the
sample needs to be polished as well. A schematic picture of
suchasamplealongwiththedirectionoftheincominglight
is shown in Fig. 11. For simple absorption measurements, it
is sufficient to mount this waveguide sample on a holder. A
comparison between normalized absorption data for both
TE and TM polarization allows the correct extraction of
absorption curves. The absorbance is then given by
?ð?ÞL ¼ ln
ITEð?Þ
ITMð?Þ
??
(14)
where ITMand ITEare transmitted intensities for both TM
and TE polarized light and L the effective interaction
length. Experiments involving photodetection were done
by illumination of the device at only one of the two
electrical contacts, while leaving the other one in the dark.
The exact reason for using this technique can be explained
by the nonlinear optical rectification mechanism used in
these devices: the lateral conductivity between the two
contacts is quite low. According to (11), illumination
underneath one contact and at the correct wavelength
induces an optical polarization, whereas the dark contact
does not produce any effect. A measurement between dark
and illuminated contacts therefore results in a measurable
photovoltage. Typical spectral characteristics of the first
such detector are shown in Fig. 12 [44]. It worked at a peak
wavelength of 1.85 ?m, but due to its relatively broad
sensitivity curve, it could nevertheless detect down to the
technologically important wavelength of 1.55 ?m. This
feature was tested with a superluminescent diode. The
maximum operating temperature was 170 K. The high-
frequency behavior of this first device was determined by
several inherently slow processes and the large size of the
device; they limited its operation to values below 100 kHz.
In the following years, considerable optimization work
was done, mainly in terms of crystal growth but also in
terms of process and characterization [45], [46]. Later
generations of such photodetectors delivered responsivity
curves as shown in Fig. 13. In this experiment, three
different samples with 3 QW thicknesses were grown; this
resulted in a series of photodetectors being sensitive
between 2.33 and 1.49 ?m [47].
As mentioned in the Introduction, it was only in 2007
when the functioning of these devices could be explained
in a satisfactory way. This was also the moment when more
serious high-frequency tests could be run [48]. For these
experiments, a more sophisticated mounting directly on a
Fig. 11. Schematic cross-sections through a typical intersubband
absorption sample using nitride semiconductors [36, Fig. 5].
Fig. 12. Energetic peak positions of absorption (red line 77 K,
green line 300 K) and photovoltage (blue dashed line, 10 K)
for the first GaN-based intersubband photodetector
demonstrated in 2003 [44, Fig. 2].
Fig.13. Photovoltageresponsivityspectraofaseriesofthreesamples
having different QW thicknesses (17, 22, and 38 A ˚) showing the
quantum confinement shift. The sample with the thickest QW shows
two higher order transitions [47, Fig. 2 (bottom)].
9

Page 10

subminiature Version A connector socket with short bond
wires was used. Otherwise, parasitic effects due to the
highly inductive bonds severely hampered the high-
frequency characteristics of the detector. For character-
ization of a high-speed device, we carefully separated
optical signal generation and detection. On the generation
side, we used a commercially available continuous wave
operated telecommunication laser at 1.55 ?m. Its output
was directly delivered to an optical modulator with a
maximum modulation frequency going up to 10 GHz. The
signal then entered a 30-m-long optical single mode fiber
and was directed towards the AlN/GaN-based detector.
The latter was held at room temperature and mounted on
an x-y-z-stage for alignment purposes. The result of these
measurements is shown in Fig. 14, which presents
electrical detector signal as a function of optical input at
frequencies between 100 kHz and 10 GHz. The rollover
was seen at 43 MHz, while a signal was detected up to
modulation frequencies of nearly 3 GHz. It is obvious that
parasitic effects limited the performance of this device; as
a matter of fact, Julien et al. recently reported a nitride-
based quantum cascade detector with appropriate high-
frequency stripe-line mounting (see below) [49]. This
device worked well beyond 10 GHz, which is very
encouraging for future developments.
In a different line of research, we tested such a detector
also for the 1 ! 3 intersubband transition, which can be
exploited via exactly the same detection mechanism [50].
The detection wavelength is then half of 1.55 ?m, namely,
775 nm. For this wavelength, there exist very powerful
pulsed solid-state light sources such as Ti:sapphire or
Nd:YAG lasers. By using such a laser device, but still the
same detectorasforthe3GHzexperiment, wewereable to
go to considerably higher frequencies as well. In the signal
versus frequency characteristic, a pronounced change in
slope from ?20 to ?40 dB/dec can be seen. It happens at
exactly the same frequency as in the 1.55 ?m experiment,
suggesting an extrinsic mechanism as the limiting factor.
Similar to a high-frequency experiment involving a
midinfrared quantum cascade detector (QCD), where a
second-order low-pass filter characteristic was observed,
we suspect here that the parasitic inductance of the bond
wire and the parasitic capacitance of the detector contacts
produce this frequency behavior [51]. This is shown in
Fig.15,wherethemeasuredfrequencyresponseisreported
for different experimental configurations. In the first test,
where the 1 ! 2 transition was excited using an Yb-doped
mode-locked fiber laser, the signal was amplified by an
SHF806P broadband amplifier. This resulted in a high-
frequency cutoff of 2.2 GHz. Thanks to the better signal-to-
noise ratio achieved with an additional NexTec NBL 00558
amplifier, the maximum measurement frequency could be
pushed further, up to 5 GHz. When finally using a 780 nm
pulsed high-power Nd:YAG laser by exploiting the 1 ! 3
optical detection transition and the same amplification
scheme as in the previous experiment, the highest detec-
tion frequency was 13.3 GHz.
C. Quantum Cascade Detectors
Recent experiments using a QCD based on a periodic
GaN/AlGaN structure revealed even higher detection fre-
quencies [49]. The device structure is shown in Fig. 16,
whilethefrequencyresponseofthedetectorispresentedin
Fig. 17. This detector contains 40 active region periods,
each comprising a doped (Si, 1? 1019cm?3) GaN-based
main well and an AlN/Al0:25Ga0:75N-based (6 ? 1 nm AlN/
1 nm Al0:25Ga0:75N) carrier extraction region. The latter is
coupled to the precedent and the following stage’s main
well. Besides the possibility of actively designing a real
transport mechanism for the carriers, another important
Fig. 14. Typical detector signal versus measurement frequency plot
for an AlN/GaN-based intersubband detector working at 1.55 ?m.
A maximum frequency of 3 GHz was achieved in this experiment
[48, Fig. 3].
Fig. 15. Testing of the detector’s maximal frequency range using an
ultrashort pulse solid state laser source (blue and red curves).
In this experiment, the 1 ! 3 transition was exploited as well
(black curve) [50, Fig. 4].
10

Page 11

advantage of this device is its small size of 17? 17 ?m2and
its correct high-frequency mounting using coplanar stripe
line technology. As Fig. 17 shows, a ?3 dB frequency of
11.4 GHz was seen, along with a cutoff beyond 30 GHz. A
responsivity on the order of 2–2.5 mA/W was reported for
this device. Altogether, these are impressive results that
clearly reveal the future potential of GaN-based photo-
detectors for telecommunication wavelengths. Whether
the demonstration of this device already paves the way
towardsamoreadvancedintersubbandsystemsuchasaQC
laser remains an open question. For a QC detector, the
designed potential asymmetry next to the active QW along
with some sort of impurity-assisted electron scattering will
suffice to allow a more or less correct function. For a QC
laser, however, an extremely well-defined resonant tunnel-
ing process would be required. Although resonant tunneling
cannot be excluded in the reported QCDs, the presently
existing level broadening will render injection of electrons
into an upper laser state very difficult and inefficient.
D. Light Emitters
A considerable amount of theoretical work has been
published on designs for QC light emitters and lasers. The
majority of these designs use the classical approach of the
electron-longitudinal optical (LO-) phonon scattering to
depopulate the lower laser level [52], [53]. Although this
was also the first experimentally explored approach of the
University of Paris-Sud group, no successful experimental
demonstration of this concept has been published so far.
An alternative to the LO-phonon resonance design is the
bound-to-continuum design presented in Section II-D. Yet
another approach was used for the first functional
intersubband light emitter based on a 250-period regular
AlN/GaN-superlattice (2.1 nm GaN/3 nm AlN), which was
tested at the University of Paris-Sud as well by Nevou et al.
[32]. No doping was intentionally added to the structure;
the electron concentration is therefore intrinsic by
residual doping [54]. The estimated doping level through
this intrinsic doping was 4? 1012cm?2. In the device, the
1 ! 3 transition was exploited for optical pumping with
a 980 nm high-power semiconductor pump laser and the
3 ! 2 transition for the emission of radiation at 2.1 ?m.
In order to minimize the mirror losses, two parallel, 90?
oriented facet mirrors were polished onto the sample to
couple out the intersubband light; pumping took place
via two parallel, polished 45?facets (placed orthogonally
to the 90?facets) and a multiple zigzag path through
the sample. The power density of the pump laser was on
the order of 104Wm?2.
In order to measure the output radiation, lock-in
amplification at 4 kHz was necessary. As Fig. 18 shows, the
output spectrum is centered around 2.1 ?m and fully TM
polarized. A comparison with the corresponding intersub-
band absorption shows very good agreement. The mea-
sured luminescence efficiency is on the order of 10 pW/W.
A slightly better performance could be achieved in a more
recent sample having an active region with a 200-period
stack of quantum dots [55]. While pumping with an YVO4
solid-state laser took place at a wavelength of 1.34 ?m,
the emission from the quantum dots was exactly one
LO-phononenergyaway,namely,at1.48?m.Thisisshown
in Fig. 19. By fitting a Lorentzian lineshape to the sharply
dropping high energy shoulder of the emission spectrum, it
was possible to achieve an excellent matching with the
measurement.Thefullwidthathalf-maximumwasassmall
as 4 meV, roughly a factor of ten smaller than in standard
QW samples. Very recently, the group of Moustakas et al.
Fig. 16. Schematic band structure of a near-infrared QCD.
The transport direction is given by the slope of the potential
ramp [49, Fig. 1].
Fig. 17. High-frequency behavior of a QC detector with a size of
17 ? 17 ?m2(red curve) and of 25 ? 25 ?m2(blue curve).
The dotted lines are simulations taking into account the wire
inductance and the device capacitance. The insets show, on top,
the electrical equivalent circuit and, at the bottom, a microscope
picture of the finished device [49, Fig. 2].
11

Page 12

demonstrated such an optically pumped GaN-based QW
light emitter under pulsed excitation with an OPO. Thanks
to the much higher peak pump power, an increased output
power on the order of hundreds of nanowatts could be
observed [56].
V. OUTLOOK
After these encouraging results in different crucial fields
of optoelectronics, we believe that GaN is the material of
choice for several emerging applications. Ultra-high-speed
devices such as photodetectors or optical switches might
soon play an important role in modern telecommunication
systems. Since GaN and AlN are chemically inert and
mechanically very robust materials, applications under
harsh environmental conditions might become especially
appealing for optoelectronic components fabricated from
AlN/GaN. It is clear, too, that the development toward
higher frequencies will continue until the intrinsic
physical limits of the material system are reached. At
this moment, there is however no sign that this will
happen in the near future. Based on the short intersub-
band lifetime and the absence of an internal parasitic
capacitance, cutoff frequencies on the order of several
hundred gigahertz are expected. In this same context,
detectors with larger asymmetriesVbe it with asymmetric
potentials and physical transport of electrons or using
optical rectification employing additional electronic
levelsVwill be promising candidates for improved
responsivity and higher speed. Along the same lines,
highly sensitive photodetectors with quantum dot active
regions will be further developed. Preliminary experi-
ments have shown already a responsivity improvement of
a factor of 60.
An application with considerable potential for future
development is the use of GaN saturable absorbers as
mode-locking elements in solid-state laser cavities. Since
GaN is chemically robust, it will be able to withstand the
required high pump powers. The search for the ideal
device geometry and the suitable place within the laser
cavity are some of the important points to be resolved here.
Another interesting potential direction for novel devices
might be the use of nonpolar III-nitrides. The absence of
internal polarization fields would lift one of the big
uncertainties and would clearly constitute an advantage for
intersubband device design. As far as intersubband light
emitters are concerned, the general prognosis might be
somewhat less enthusiastic than for detectors and
switches. At this point, neither electroluminescence nor
optically pumped lasing has been demonstrated. Never-
theless, optically pumped luminescence has been ob-
served. Further progress will certainly be made using
device configurations involving distributed feedback, more
sophisticated pump geometries, and quantum-mechanical
optimizations such as the Raman laser. The most
important point for future QC laser development, how-
ever, is the use of high-quality dislocation-free GaN bulk
crystals. The unambiguous demonstration of a resonant
tunneling diode on such material will be a highly important
milestone that would give us confidence that the necessary
material quality for laser development is now reached.
Without this demonstration, transport measurements in
Fig. 19. Output spectra of the latest luminescence sample from the
Paris-Sud group. All possible configurations of pump/output
polarization have been tried out. The strongest signal occurs when
both pump and detection are in TM polarization [55, Fig. 3].
Fig. 18. Output spectra of the first luminescence sample from the
Paris-Sud group. The green dotted line corresponds to the absorption
of the e1–e3 transition, while the red and blue curves are emission in
TM and TE polarization, respectively [54, Fig. 2].
12

Page 13

any III-nitride device always suffer from a lack of clarity
and meaningfulness. Like 17 years ago with the violet-
blue laser diodes based on GaN, it is expected that a
particularly annoying bottleneck present in the Breal
world[ will be able to trigger a dedicated effort in the
development of nitride intersubband devices for practical
applications. In 1992, this bottleneck was the short
lifetime of blue lasers based on ZnSe. GaN was the only
material capable of outperforming ZnSe-based devices,
and it soon did. h
REFERENCES
[1] T. Ando, A. B. Fowler, and F. Stern,
BElectronic properties of two-dimensional
electronic systems,[ Rev. Mod. Phys., vol. 54,
no. 2, pp. 437–672, 1982.
[2] D. D. Coon and R. G. P. Karunasiri, BNew
mode of infrared detection using quantum
wells,[ Appl. Phys. Lett., vol. 45, no. 6,
pp. 469–471, 1984.
[3] L. C. West and S. J. Eglash, BFirst observation
of an extremely large-dipole infrared
transition within the conduction band of a
GaAs quantum well,[ Appl. Phys. Lett., vol. 46,
no. 12, pp. 1156–1158, 1985.
[4] B. F. Levine, K. K. Choi, C. G. Bethea,
J. Walker, and R. J. Malik, BNew 10 ?m
infrared detector using intersubband
absorption in resonant tunneling GaAlAs
superlattices,[ Appl. Phys. Lett., vol. 50,no. 16,
pp. 1092–1094, 1987.
[5] K.-K. Choi, B. F. Levine, C. G. Bethea,
J. Walker, and R. J. Malik, BMultiple quantum
well 10 ?m GaAs=AlxGa1?xAs infrared
detector with improved responsivity,[ Appl.
Phys. Lett., vol. 50, no. 25, pp. 1814–1816,
1987.
[6] H. C. Liu, J. Li, E. R. Brown, K. A. McIntosh,
K. B. Nichols, and M. J. Manfra, BQuantum
well intersubband heterodyne infrared
detection up to 82 GHz,[ Appl. Phys. Lett.,
vol. 67, no. 11, pp. 1594–1596, 1995.
[7] J. Faist, F. Capasso, D. L. Sivco, C. Sirtori,
A. L. Hutchinson, and A. Y. Cho, BQuantum
cascade laser,[ Science, vol. 264, no. 5158,
pp. 553–556, 1994.
[8] M. Beck, D. Hofstetter, T. Aellen, J. Faist,
U. Oesterle, M. Ilegems, E. Gini, and
H. Melchior, BContinuous wave operation of a
mid-infrared semiconductor laser at room
temperature,[ Science, vol. 295, no. 5553,
pp. 301–305, 2002.
[9] Y. Bai, S. R. Darvish, S. Slivken, W. Zhang,
A. Evans, J. Nguyen, and M. Razeghi, BRoom
temperature continuous wave operation of
quantum cascade lasers with watt-level optical
power,[ Appl. Phys. Lett., vol. 92, no. 10,
p. 101105, 2008.
[10] C. Gmachl, H. M. Ng, S. N. G. Chu, and
A. Y. Cho, BIntersubband absorption at
? ? 1:55 ?m in well- and modulation-doped
GaN/AlGaN multiple quantum wells with
superlattice barriers,[ Appl. Phys. Lett., vol. 77,
no. 23, pp. 3722–3724, 2000.
[11] C. Gmachl, H. M. Ng, and A. Y. Cho,
BIntersubband absorption in degenerately
doped GaN=AlxGa1?xN coupled double
quantum wells,[ Appl. Phys. Lett., vol. 79,
no. 11, pp. 1590–1592, 2001.
[12] K. Kishino, A. Kikuchi, H. Kanazava, and
T. Tachibana, BIntersubband transition in
(GaN)(m)/(AlN)(n) superlattices in the
wavelength range from 1.08 to 1.61 ?m,[
Appl. Phys. Lett., vol. 81, no. 7, pp. 1234–1236,
2002.
[13] A. Helman, M. Tchernycheva, A. Lusson,
E. Warde, F. H. Julien, K. Moumanis,
G. Fishman, E. Monroy, B. Daudin,
L. S. Dang, E. Bellet-Amalric, and D. Jalabert,
BIntersubband spectroscopy of doped and
undoped AlN/GaN quantum wells grown by
molecular-beam epitaxy,[ Appl. Phys. Lett.,
vol. 83, no. 25, pp. 5196–5198, 2003.
[14] N. Suzuki and N. Iizuka, BEffect of
polarization field on intersubband transition
in AlGaN/GaN quantum wells,[ Jpn. J. Appl.
Phys., vol. 38, no. 4A, pp. L363–L365, 1999.
[15] S. Nicolay, J. F. Carlin, E. Feltin, R. Butte,
M. Mosca, N. Grandjean, M. Ilegems,
M. Tchernycheva, L. Nevou, and F. H. Julien,
BMidinfrared intersubband absorption in
lattice-matched AlInN/GaN multiple
quantum wells,[ Appl. Phys. Lett., vol. 87,
no. 11, p. 111106, 2005.
[16] M. Tchernycheva, L. Nevou, L. Doyennette,
F. H. Julien, E. Warde, F. Guillot, E. Monroy,
E. Bellet-Amalric, T. Remmele, and
M. Albrecht, BSystematic experimental and
theoretical investigation of intersubband
absorption in AlN/GaN quantum wells,[ Phys.
Rev. B, vol. 73, no. 12, p. 125347, 2006.
[17] G. Cywicski, C. Skierbiszewski,
A. Fedunieiwcz-Z˙muda, M. Siekacz,
L. Nevou, L. Doyennette, M. Tchernycheva,
F. H. Julien, P. Prystawko, M. Krys ´ko,
S. Grzanka, I. Grzegory, A. Presz,
J. Z. Domagaaa, J. Smalc, M. Albrecht,
T. Remmele, and S. Porowski, BGrowth of
thin AllnN/GalnN quantum wells for
applications to high-speed intersubband
devices at telecommunication wavelengths,[
J. Vac. Sci. Technol. B, vol. 24, no. 3,
pp. 1505–1509, 2006.
[18] G. Martin, S. Strite, A. Botchkarev,
A. Agarwal, A. Rockett, W. R. L. Lambrecht,
B. Segall, and H. Morkoc ¸, BValence band
discontinuity between GaN and AlN
measured by X-ray photoemission
spectroscopy,[ J. Electron. Mater., vol. 24,
no. 4, pp. 225–227, 1995.
[19] S. Wei and A. Zunger, BValence band
splittings and band offsets of AlN, GaN,
and InN,[ Appl. Phys. Lett., vol. 69, no. 18,
pp. 2719–2721, 1996.
[20] G. Martin, A. Botchkarev, A. Rockett, and
H. Morkoc ¸, BValence-band discontinuities of
wurtzite GaN, AlN, and InN heterojunctions
measured by x-ray photoemission
spectroscopy,[ Appl. Phys. Lett., vol. 68,
no. 18, pp. 2541–2543, 1996.
[21] N. Iizuka, K. Kaneko, N. Suzuki, T. Asano,
S. Noda, and O. Wada, BUltrafast
intersubband relaxation (? 150 fs) in
AlGaN/GaN multiple quantum wells,[ Appl.
Phys. Lett., vol. 77, no. 5, pp. 648–650, 2000.
[22] Z. Wang, K. Reimann, M. Woerner,
T. Elsaesser, D. Hofstetter, J. Hwang,
W. J. Schaff, and L. F. Eastman,
BFemtosecond intersubband dynamics of
electrons in AlGaN/GaN-based
high-electron-mobility transistors,[ Semicond.
Sci. Technol., vol. 19, no. 4, pp. S463–S464,
2004.
[23] M. Helm, BThe basic physics of intersubband
transitions,[ in Semiconductors and
Semimetals, vol. 62.
Academic, 2000, pp. 1–99.
London, U.K.:
[24] J. D. Heber, C. Gmachl, H. M. Ng, and
A. Y. Cho, BComparative study of ultrafast
intersubband electron scattering times at
?1:55 ?m wavelength in GaN/AlGaN
heterostructures,[ Appl. Phys. Lett., vol. 81,
no. 7, pp. 1237–1239, 2002.
[25] Z. Wang, K. Reimann, M. Woerner,
T. Elsaesser, D. Hofstetter, E. Baumann,
F. R. Giorgetta, H. Wu, W. J. Schaff, and
L. F. Eastman, BUltra-fast hole-burning in
intersubband absorption lines of AlN/GaN
superlattices,[Appl. Phys. Lett., vol.89, no. 15,
p. 151103, 2006.
[26] F. H. Julien, J.-M. Lourtioz, N. Herschkorn,
D. Delacourt, J. P. Pocholle, M. Papuchon,
R. Planel, and G. Le Roux, BOptical saturation
of intersubband absorption in
GaAs?AlxGa1?xAs quantum wells,[ Appl.
Phys. Lett., vol. 53, no. 2, pp. 116–118, 1988.
[27] N. Iizuka, K. Kaneko, and N. Suzuki,
BSub-picosecond all-optical gate utilizing GaN
intersubband transition,[ Opt. Express, vol. 13,
no. 10, pp. 3835–3840, 2005.
[28] Y. Li, A. Bhattacharyya, C. Thomidis,
T. D. Moustakas, and R. Paiella, BUltrafast
all-optical switching with low saturation
energy via intersubband transitions in
AlN/GaN quantum-well waveguides,[
Opt. Express, vol. 15, no. 26,
pp. 17 922–17 924, 2007.
[29] O. Ambacher, J. Majewski, C. Miskys, A. Link,
M. Hermann, M. Eickhoff, M. Stutzmann,
F. Bernardini, V. Fiorentini, V. Tilak,
W. J. Schaff, and L. F. Eastman, BPyroelectric
properties of Al(In)GaN/GaN hetero- and
quantum well structures,[ J. Phys. Cond.
Matter, vol. 14, no. 13, pp. 3399–3434, 2002.
[30] F. Bernardini, V. Fiorentini, and
D. Vanderbilt, BSpontaneous polarization and
piezoelectric constants of III-V nitrides,[
Phys. Rev. B, vol. 56, no. 16,
pp. 10 024–10 027, 1997.
[31] E. Rosencher, P. Bois, J. Nagle, E. Costard,
and S. Delaitre, BObservation of nonlinear
optical rectification at 10.6 ?m in
compositionally asymmetric AlGaAs
multiquantum wells,[ Appl. Phys.
Lett., vol. 55, no. 16, pp. 1597–1599, 1989.
[32] L. Nevou, M. Tchernycheva, F. H. Julien,
F. Guillot, and E. Monroy, BShort wavelength
ð? ¼ 2:13 ?mÞ intersubband
luminescence from AlN/GaN quantum
wells at room temperature,[ Appl. Phys. Lett.,
vol. 90, no. 12, p. 121106, 2007.
[33] N. Bouadma, B. Sermage, and P. Delvodere,
BLow threshold current and high relaxation
oscillation frequency of short-cavity
integrable InP/InGaAsP BRS lasers,[ Jpn. J.
Appl. Phys., vol. 29, no. 12, pp. L2223–L2225,
1990.
[34] P. K. Kandaswamy, F. Guillot,
E. Bellet-Amalric, E. Monroy, L. Nevou,
M. Tchernycheva, A. Michon, F. H. Julien,
E. Baumann, F. R. Giorgetta, D. Hofstetter,
T. Remmele, M. Albrecht, S. Birner, and
L. S. Dang, BAlN/GaN short-period
superlattices for intersubband
optoelectronics: A systematic study of their
13

Page 14

epitaxial growth, design, and performance,[
J. Appl. Phys., vol. 104, no. 9, p. 093501, 2008.
[35] E. Monroy, B. Daudin, E. Bellet-Amalric,
N. Gogneau,D.Jalabert,F.Enjalbert,J.Brault,
J. Barjon, and L. S. Dang, BSurfactant effect of
In for AlGaN growth by plasma-assisted
molecular beam epitaxy,[ J. Appl.
Phys., vol. 93, no. 3, pp. 1550–1556, 2003.
[36] D. Hofstetter, E. Baumann, F. R. Giorgetta,
R. The ´ron, H. Wu, W. J. Schaff, J. Dawlaty,
P. A. George, L. F. Eastman, F. Rana,
P.K.Kandaswamy,S.Leconte,andE.Monroy,
BPhotodetectors based on intersubband
transitions using III-nitride superlattice
structures,[ J. Phys., Condens. Matter, vol. 21,
p. 174208, 2009.
[37] S. Nicolay, E. Feltin, J.-F. Carlin, M. Mosca,
L. Nevou, M. Tchernycheva, F. H. Julien,
N. Grandjean, and M. Ilegems, BIndium
surfactant effect on AlN/GaN
heterostructures grown by metal-organic
vapor-phase epitaxy: Applications to
intersubband transitions,[ Appl. Phys. Lett.,
vol. 88, no. 15, p. 151902, 2006.
[38] E. Baumann, F. R. Giorgetta, D. Hofstetter,
S. Golka, W. Schrenk, G. Strasser, L. Kirste,
S. Nicolay, E. Feltin, and N. Grandjean,
B1.5 ?m absorption and room temperature
photovoltaic response in AlN/GaN
superlattices grown by metalorganic vapor
phase epitaxy,[ Appl. Phys. Lett., vol. 89, no. 4,
p. 041106, 2006.
[39] A. Lupu, F. Julien, S. Golka, G. Pozzovivo,
G. Strasser, E. Baumann, F. R. Giorgetta,
D. Hofstetter, S. Nicolay, M. Mosca, and
N. Grandjean, BLattice-matched GaN/InAlN
waveguides at 1.55 ?m grown by metalorganic
vapor phase epitaxy,[ IEEE Photon. Technol.
Lett., vol. 20, no. 2, pp. 102–104, 2008.
[40] C. Kumtornkittikul, T. Shimizu, N. Iizuka,
N. Suzuki, M. Sugiyama, and Y. Nakano, BAlN
waveguide with AlN/GaN quantum wells for
all-optical switch utilizing intersubband
transition,[ Jpn. J. Appl. Phys., vol. 46,
no. 12–16, pp. L352–L355, 2007.
[41] N. Iizuka, K. Kaneko, and N. Suzuki,
BAll-optical switch utilizing intersubband
transition in GaN quantum wells,[
IEEE J. Quantum Electron., vol. 42, no. 7–8,
pp. 765–771, 2006.
[42] E. Baumann, F. R. Giorgetta, D. Hofstetter,
F. Guillot, E. Bellet-Amalric, and E. Monroy,
BElectrically adjustable intersubband
absorption of a AlN/GaN superlattice grown
on a transistorlike structure,[ Appl. Phys. Lett.,
vol. 89, no. 10, p. 101121, 2006.
[43] L. Nevou, H. Kheirodin, M. Tchernycheva,
L. Meignien, P. Crozat, A. Lupu, E. Warde,
F. H. Julien, G. Pozzovivo, S. Golka,
G. Strasser, F. Guillot, E. Monroy,
T. Remmele, and M. Albrecht,
BShort-wavelength intersubband
electroabsorption modulation based on
electron tunneling between AlN/GaN coupled
quantum wells,[ Appl. Phys. Lett., vol. 90,
no. 22, p. 223511, 2007.
[44] D.Hofstetter,S.-S.Schad,H.Wu,W.J.Schaff,
and L. F. Eastman, BAlN/GaN-based
quantum-well infrared photodetector
for 1.55 ?m,[ Appl. Phys. Lett., vol. 83, no. 3,
pp. 572–574, 2003.
[45] D. Hofstetter, E. Baumann, F. R. Giorgetta,
M. Graf, M. Maier, F. Guillot,
E. Bellet-Amalric, and E. Monroy,
BHigh-quality AlN/GaN-superlattice
structures for the fabrication of narrow-band
1.4 ?m photovoltaic intersubband detectors,[
Appl. Phys. Lett., vol. 88, no. 12,
p. 121112, 2006.
[46] F. R. Giorgetta, E. Baumann, F. Guillot,
E. Monroy, and D. Hofstetter, BHigh
frequency ðf ¼ 2:37 GHzÞ room temperature
operation of 1.55 ?m AlN/GaN-based
intersubband detector,[ Electron.
Lett., vol. 43, no. 2, pp. 185–187, 2007.
[47] E. Baumann, F. R. Giorgetta, D. Hofstetter,
H. Wu, W. J. Schaff, L. F. Eastman, and
L. Kirste, BTunneling effects and
intersubband absorption in AlN/GaN
superlattices,[ Appl. Phys. Lett., vol. 86, no. 3,
p. 032110, 2005.
[48] D. Hofstetter, E. Baumann, F. R. Giorgetta,
F. Guillot, S. Leconte, and E. Monroy,
BOptically nonlinear effects in intersubband
transitions of AlN/GaN-based superlattice
structures,[ Appl. Phys. Lett., vol. 91, no. 13,
p. 131115, 2007.
[49] A. Vardi, N. Kheirodin, L. Nevou,
H. Machhadani, L. Vivien, P. Crozat,
M. Tchernycheva, R. Colombelli, F. H. Julien,
F. Guillot, C. Bougerol, E. Monroy,
S. Schacham, and G. Bahir, BHigh speed
operation of AlN/GaN quantum cascade
detectors at ? ? 1:55 ?m,[ Appl. Phys. Lett.,
vol. 93, no. 19, p. 193509, 2008.
[50] D. Hofstetter, E. Baumann, F. R. Giorgetta,
J. Dawlaty, P. A. George, R. Fana, F. Guillot,
and E. Monroy, BHigh frequency
measurements on an AlN/GaN-based
intersubband detector at 1550 and 780 nm,[
Appl. Phys. Lett., vol. 92, no. 23, p. 231104,
2008.
[51] D. Hofstetter, M. Graf, T. Aellen, J. Faist,
S. Blaser, and L. Hvozdara, B23 GHz
operation of a quantum cascade detector
at 5.35 ?m,[ Appl. Phys. Lett., vol. 89, no. 6,
p. 061119, 2006.
[52] V. D. Jovanovic, Z. Ikonic, D. Indjin,
P. Harrison, V. Milanovic, and R. A. Soref,
BDesigning strain-balanced GaN/AlGaN
quantum well structures: Application to
intersubband devices at 1.3 and
1.55 ?m wavelengths,[ J. Appl. Phys., vol. 93,
no. 6, pp. 3194–3197, 2003.
[53] G. Sun, R. A. Soref, and J. B. Khurgin, BActive
region design of a terahertz
GaN=Al0:15Ga0:85N quantum cascade
laser,[Superlattices Microstruct., vol. 37, no. 2,
pp. 107–113, 2005.
[54] L. Nevou, F. H. Julien, R. Colombelli,
F. Guillot, and E. Monroy, BRoom
temperature intersubband emission of
AlN/GaN quantum wells at ? ¼ 2:3 ?m,[
Electron. Lett., vol. 42, no. 22, pp. 1308–1309,
2006.
[55] L. Nevou, F. H. Julien, M. Tchernycheva,
F. Guillot, E. Monroy, and E. Sarigiannidou,
BIntersubband emission at ? ? 1:48 ?m from
AlN/GaN quantum dots at room
temperature,[ Appl. Phys. Lett., vol. 92, no. 16,
p. 161105, 2008.
[56] K.Driscoll,Y.Liao,A.Bhattacharyya,L.Zhou,
D. J. Smith, T. D. Moustakas, and R. Paiella,
BOptically pumped intersubband emission of
short-wave infrared radiation with
AlN/GaN quantum wells,[ Appl. Phys. Lett.,
vol. 94, no. 8, p. 081120, 2009.
ABOUT THE AUTHORS
Daniel Hofstetter was born in Zug, Switzerland, in 1966. From 1988 to
1993, he studied physics at the Swiss Federal Institute of Technology,
Zurich. He received the Ph.D. degree from the Paul Scherrer Institute,
Zurich, in 1996.
His doctoral work included the design, fabrication, and testing of a
semiconductor-based monolithically integrated Michelson interferome-
ter for optical displacement measurement. For his diploma thesis, he
carried out photoacoustic spectroscopy on fatty acids using gas lasers.
After an apprenticeship as an Electrical Mechanic with Landis & Gyr, Zug,
from 1982 to 1986, he became a Physics Technician until 1988. Later, he
was with XEROX Palo Alto Research Center, Palo Alto, CA, developing
single-mode InGaN-based violet semiconductor lasers (1996–1998).
From 1998 to 2001, he was with the Mesoscopic Physics Group,
University of Neuchatel, Switzerland, where his work concentrated on
the fabrication and testing of single-mode distributed-feedback
quantum-cascade (QC) lasers and high-performance QC lasers. Since
2002, he has been an Assistant Professor at the University of Neuchatel.
His main activities included the development of novel types of
semiconductor-based devices, such as QC detectors, for the midinfrared
wavelength region.
Esther Baumann received the M.Sc. degree in electrical engineering from
the Swiss Federal Institute of Technology, Zurich, in 2003 and the Ph.D.
degree in physics from the University of Neuchatel, Switzerland, in 2007.
Her doctoral work was on photovoltaic light detection in III-nitrides
intersubband systems. Since 2008, she has worked on fiber laser
frequency combs as a Guest Researcher with the National Institute of
Standards and Technology, Boulder, CO.
Fabrizio Raphael Giorgetta received the M.Sc. degree in electrical
engineering from the Swiss Federal Institute of Technology, Zurich, in
2003 and the Ph.D. degree in physics from the University of Neuchatel,
Switzerland, in 2007.
His doctoral work was on quantum cascade detectors. Since 2008, he
has worked on fiber laser frequency combs as a Guest Researcher at the
National Institute of Standards and Technology, Boulder, CO.
Ricardo The ´ron, photograph and biography not available at the time of
publication.
14

Page 15

Hong Wu, photograph and biography not available at the time of
publication.
William J. Schaff, (Member, IEEE) photograph and biography not
available at the time of publication.
Jahan Dawlaty, photograph and biography not available at the time of
publication.
Paul A. George, photograph and biography not available at the time of
publication.
Lester F. Eastman, (Fellow, IEEE) photograph and biography not
available at the time of publication.
Farhan Rana, photograph and biography not available at the time of
publication.
Prem K. Kandaswamy, photograph and biography not available at the
time of publication.
Fabien Guillot, photograph and biography not available at the time of
publication.
Eva Monroy, photograph and biography not available at the time of
publication.
15