Cavity mode gain alignment in GaAsSb-based near-infrared vertical cavity lasers
studied by spectroscopy and device measurements
G. Blume, K. Hild, I. P. Marko, T. J. C. Hosea, S.-Q. Yu, S. A. Chaparro, N. Samal, S. R. Johnson, Y.-H. Zhang,
and S. J. Sweeney
Citation: Journal of Applied Physics 112, 033108 (2012); doi: 10.1063/1.4744985
View online: http://dx.doi.org/10.1063/1.4744985
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/112/3?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Detecting and tuning anisotropic mode splitting induced by birefringence in an InGaAs/GaAs/AlGaAs verticalcavity surface-emitting laser
J. Appl. Phys. 111, 043109 (2012); 10.1063/1.3682648
Optimal photonic-crystal parameters assuring single-mode operation of 1300 nm AlInGaAs vertical-cavity
surface-emitting laser
J. Appl. Phys. 105, 093102 (2009); 10.1063/1.3115449
Self-sustained pulsation in the oxide-confined vertical-cavity surface-emitting lasers based on submonolayer
InGaAs quantum dots
Appl. Phys. Lett. 91, 121106 (2007); 10.1063/1.2784937
Reverse-bias emission sheds light on the failure mechanism of degraded vertical-cavity surface-emitting lasers
J. Appl. Phys. 99, 123113 (2006); 10.1063/1.2206852
Blue vertical-cavity surface-emitting lasers based on second-harmonic generation grown on (311)B and (411)A
GaAs substrates
J. Appl. Phys. 87, 1597 (2000); 10.1063/1.372065

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.22 On: Wed, 04 Feb 2015 17:30:15

JOURNAL OF APPLIED PHYSICS 112, 033108 (2012)

Cavity mode gain alignment in GaAsSb-based near-infrared vertical cavity
lasers studied by spectroscopy and device measurements
G. Blume,1,a) K. Hild,1 I. P. Marko,1 T. J. C. Hosea,1,2,b) S.-Q. Yu,3,c) S. A. Chaparro,3
N. Samal,3 S. R. Johnson,3 Y.-H. Zhang,3 and S. J. Sweeney1,d)

1
Advanced Technology Institute and Department of Physics, University of Surrey, Guildford, Surrey GU2 7XH,
United Kingdom
2
Ibnu Sina Institute, Universiti Teknologi Malaysia, 81310 UTM Skudai, Johor, Malaysia
3
School of Electrical, Computer and Energy Engineering and Center for Photonics Innovation, Arizona State
University, Tempe, Arizona 85287-5706, USA

(Received 21 March 2012; accepted 11 July 2012; published online 8 August 2012)
We present a combination of spectroscopy and device measurements on GaAsSb/GaAs
vertical-cavity surface-emitting laser (VCSEL) structures to determine the temperature at which
the wavelength of the VCSEL cavity mode (CM) aligns with that of the quantum well (QW)
ground-state transition (GST), and therefore the gain peak. We find that, despite the achievement
of room temperature (RT) continuous wave lasing in VCSEL devices, the QW transition and the
CM are actually slightly misaligned at this temperature; room temperature electroluminescence
measurements from a cleaved edge of the VCSEL wafer indicate that the 300 K QW GST energy is
at 0.975 6 0.005 eV, while the CM measured in the VCSEL surface reflectivity spectra is at
0.9805 6 0.0002 eV. When the wafer sample is cooled, the CM and QW GST can be brought into
alignment at 270 6 10 K, as confirmed by temperature-dependent electro-modulated reflectance
(ER) and edge-electroluminescence spectroscopic studies. This alignment temperature is further
confirmed by comparing the temperature dependence of the emission energy of a fabricated
VCSEL device with that of an edge-emitting laser structure with a nominally identical active
region. The study suggests that for further device improvement, the room temperature CM and QW
GST energies should be more closely matched and both designed to a smaller energy of about
0.95 eV, somewhat closer to the 1.31 lm target. The study amply demonstrates the usefulness of
non-destructive ER characterisation techniques in VCSEL manufacturing with GaAsSb-based
C 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.4744985]
QWs. V
I. INTRODUCTION

Vertical-cavity surface-emitting lasers (VCSELs) emitting close to 1.3 lm are of considerable importance for the
development of metropolitan area networks as they provide
the necessary properties, such as low temperature dependence, low threshold currents, and circular beam profile, for
future fibre-to-the-home applications for the 1.31 lm zero
dispersion wavelength of standard silica fibres.1 Traditionally, lasers operating at this wavelength have been based on
InP. However, it is difficult to grow the high quality distributed Bragg reflectors (DBRs), which are needed for
VCSELs, based on InP, and complex approaches, such as
wafer-bonding, are required.2 Since DBRs at this wavelength
are relatively easily grown on GaAs using GaAs/Al(GaAs)
layer pairs, considerable effort has also been devoted to the
development of GaAs-based active regions which emit near
1.3 lm. Examples include highly strained InGaAs,3 InAs
quantum dots,4 and GaInNAs-based quantum wells (QWs)
all of which have been the subject of extensive research.
a)

Present
address:
Ferdinand-Braun-Institut,
Leibnitz-Institut
f€
ur
H€
ochstfrequenztechnik, Gustav-Kirchhoff-Straße 4, 12489 Berlin, Germany.
b)
Corresponding author e-mail: J.Hosea@surrey.ac.uk.
c)
Present address: Department of Electrical Engineering, University of
Arkansas, 3217 Bell Engineering, Fayetteville, Arkansas 72701, USA.
d)
e-mail: s.sweeney@surrey.ac.uk.
0021-8979/2012/112(3)/033108/7/$30.00

Laser operation has been demonstrated at 1.3 lm using
GaInNAs-based QWs,5 but their performance is limited by
difficulties related to nitrogen incorporation, resulting in
degraded material quality and lifetimes. Laser operation has
also been demonstrated at 1.3 lm with quantum dots,4 but
such active layers have relatively low gain, requiring long
cavity lengths for edge-emitting lasers (EELs), which is expensive in terms of wafer “real-estate.” VCSELs based upon
quantum dots suffer from the lack of gain, requiring a high
number of layers, leading to carrier transport problems. An
alternative approach, which we examine here, does not suffer
from such disadvantages but has been less exhaustively studied. This is the use of GaAsSb QWs in VCSELs grown on
GaAs substrates,6 to try to find a 1.3 lm temperature insensitive emitter with good beam qualities using well-established
GaAs technologies. It is possible to achieve GaAsSb-based
QW emission near 1.3 lm, despite the considerable compressive strain due to mismatch with the GaAs substrate. It was
initially found that the high compressive strain (>  2%)
caused material quality problems when growing a second
QW. Therefore, additional GaAsP layers were introduced,
which have a smaller lattice constant than GaAs and partly
compensate the strain.7–12 This enables one to grow multiple
GaAsSb QWs thereby increasing the luminescence intensity.
As is well known, an important condition necessary for
the efficient operation of a VCSEL structure is the alignment

112, 033108-1

C 2012 American Institute of Physics
V

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.22 On: Wed, 04 Feb 2015 17:30:15

033108-2

Blume et al.

of the wavelengths of the cavity mode (CM) dip of the
VCSEL reflectivity spectrum and the gain peak of the active
QW region. It is generally accepted that the achievement of
this alignment at the desired operating temperature should
give close to ideal device operating conditions when other
effects such as self-heating and non-radiative recombination
have been taken into account. A small change in this alignment can have a considerable effect on the threshold current
and output power.13 Here, we undertake a detailed spectroscopic investigation of the temperature dependence of the
relative alignment of the CM dip with the QW ground-state
transition (GST) energy (which largely dictates the peak
energy in the gain spectrum due to the active QW region) of
a GaAsSb/GaAs VCSEL structure, and corroborate these
with spectroscopic measurements of the temperature dependence of the emission wavelengths of processed devices.
II. SAMPLE AND EXPERIMENTAL DETAILS AND
INITIAL DEVICE RESULTS

The VCSEL structure studied here was grown by solid
source molecular beam epitaxy on an n-doped GaAs substrate (doped to 2  1017 cm3). It has top and bottom
stacks of Al0.91Ga0.09As/GaAs DBR mirrors; an n-doped bottom DBR with 35 pairs and a p-doped top DBR with 25
pairs. Low doping levels of 5 to 10  1017 cm3 were intentionally used throughout the entire mirrors in order to minimise free-carrier absorption. The mirror stacks enclose an
Al0.25Ga0.75As cavity region with an optical thickness equal
to the targeted room temperature (RT) free space emission
wavelength (k) of 1.26 lm (0.984 eV). Three QW structures
are centred near the central anti-node of the cavity standing
wave. Such 1  k symmetric cavities with a single central
anti-node in the cavity are typically used in order to obtain
efficient symmetric electrical injection efficiency. Each of
the three QW structures consists of a central GaAs0.65Sb0.35
QW of thickness 7 nm, sandwiched by two GaAs barriers of
thickness 3 nm, and further sandwiched by two 8 nm-thick
GaAs0.8P0.2 strain-compensation layers. For further sample
details see Dowd et al.10 For the spectroscopic investigations, the structure was capped with a 25 nm layer of
In0.5Ga0.5 P (i.e., close to lattice-matched) on top of the
GaAs to protect the GaAs surface during processing, and a
5 nm semi-transparent Pt film evaporated on top of a piece of
the VCSEL wafer to serve as an electrical contact.
Devices were also fabricated from the same VCSEL wafer, with different oxide aperture and optical window diameters.7,14 It was found that continuous wave (CW) lasing
operation could be obtained at RT for a VCSEL device with
a 9 lm diameter aperture with threshold currents above
approximately 1 mA (current density of 1.5 kA/cm.) As
described later, pulsed operation was achieved at RT at a
current density of 1.2 kA/cm2 for both 9 lm and 27 lm diameter devices. For these device studies, an edge-emitting laser
device with a nominally identical active region to that of the
VCSEL was also grown and fabricated.
A closed-cycle helium cryostat was used to study the
above wafer samples and the fabricated devices, as a function of temperature. All spectra were measured using the

J. Appl. Phys. 112, 033108 (2012)

conventional combination of a single-grating monochromator, InGaAs photodiode detector, and lock-in amplifier.15
III. ROOM TEMPERATURE SURFACE REFLECTIVITY
AND EDGE LUMINESCENCE

The aforementioned fact that the VCSEL devices operated at RT tentatively suggests that the QW gain peak and
the CM dip are aligned near the desired operating temperature of 300 K. In fact, one would normally interpret this as
meaning that the energy of the QW GST (EQW) is at a
slightly higher energy than that of the CM (EQW > ECM) at
RT, such that they will come into resonance at the slightly
elevated temperatures that occur during CW operation due to
self-heating. However, we shall see later that these initial
conclusions are misleading.
Turning to the spectroscopic studies of the pre-fabrication
VCSEL wafer; as is well known, the top-surface luminescence
spectrum of a VCSEL wafer is not a useful characterisation
probe since it essentially yields only a peak at the wavelength
of the CM, because appreciable light can escape from the top
only in the vicinity of this dip in the reflectivity spectrum.
However, one can sometimes gain more useful information by
studying the emission from the side, through a cleaved edge
of the wafer, which yields a spectrum essentially unperturbed
by the DBR mirror structures.16 Unfortunately, measurements
of the edge-photoluminescence spectra (not shown here)
proved unhelpful as the emission was very weak and noisy. It
was also extremely broad (full-width at half-maximum,
FWHM > 150 meV) probably because of additional photoinduced luminescence from surface and interface states.
Therefore, we instead investigated the edge emission electroluminescence (EL) using electrical excitation, as described in
Ref. 16.
An example of the RT edge-EL is shown in Fig. 1. Here,
the sample was biased in the forward direction with a voltage
of 2.8 V, and constant current pulses of 20 mA, with a

FIG. 1. Room temperature normal-incidence reflectivity (R) spectrum of the
GaAsSb VCSEL wafer sample, which shows that the position of the CM is
near ECM ¼ 0.98 eV. High resolution scans indicate that the CM dip is very
sharp with a FWHM 0.4 meV. The centre of the edge-emission EL peak is
at a lower energy, implying that EQW ¼ 0.975 eV with a FWHM of
90 meV. This result that ECM > EQW at 300 K implies that the VCSEL
device would need to be cooled for optimal performance.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.22 On: Wed, 04 Feb 2015 17:30:15

033108-3

Blume et al.

frequency of 15 Hz and a duty cycle of 1:10, to reduce heating effects. The resulting EL peak (which is near the QW
GST energy EQW) is at 0.975 6 0.005 eV (1272 6 7 nm) and
has a broad FWHM of CQW  90 meV. There may be several
contributory factors resulting in this large observed EL
width. First, the EL is due to the spontaneous emission of
carriers in the QW and is naturally broad, particularly at high
temperature. Second, the three QWs in the active region may
have slight structural and compositional differences resulting
in a spread of the associated EL peaks.
Also shown in Fig. 1 is the normalised front-surface
reflectivity (R) spectrum, which exhibits a CM dip energy
(ECM) near 0.98 eV. Higher resolution R scans indicated that
the CM dip is at 0.9805 6 0.0002 eV (1264.5 6 0.2 nm) and
is actually very sharp with a FWHM of CCM  0.4 meV, corresponding to a finesse of more than 2400. Comparing this
with the above result for the width of EL peak, we see that
CQW CCM. We will mention this important fact again in
discussing the results of Sec. IV.
The comparison of EL and R spectra in Fig. 1 actually
shows that EQW < ECM at the desired operating temperature
of 300 K. This suggests that the VCSEL growth is actually
slightly non-ideal. It would appear that the aforementioned
RT device operation is only made possible due to the large
broadening of the QW transition (CQW  90 meV).
Note that edge-EL is necessarily a destructive technique,
because of the need to cleave the VCSEL wafer, in contrast
to the results we show next using modulated reflectance
(MR). It is worth mentioning that these MR experiments
were actually performed before the destructive EL experiments, but, for the sake of clarity we do not present the
results in the order they were performed.
IV. TEMPERATURE DEPENDENT ER AND EL

As the EL showed that EQW < ECM at RT, the next logical step was to cool the wafer sample to bring both into
alignment; both energies increase with cooling but EQW
moves approximately five times faster than ECM (see later)
so that they will cross. It has been shown that MR is the
most sensitive technique to study the QW-CM alignment,
giving well-resolved features due to the CM and, under certain circumstances, additional MR features due to the QW
transitions, of much sharper resolution than is possible with
photoluminescence.17,18 Photo-modulated reflectance is the
preferred non-contact, non-destructive version of MR. However, due to the already deposited Pt electrodes, we chose to
use the electro-modulated reflectance (ER) version of the
MR technique in these temperature-dependent studies. We
investigated the ER as a function of applied square-wave
voltage amplitude and DC bias.
Figure 2 shows the temperature dependence of the normal incidence ER spectra in the region of the CM feature for
the case of an applied square-wave voltage of amplitude
62 V, zero DC bias, and frequency 333 Hz. The first thing to
note is that these ER spectra show only a single feature due
to the CM dip; no extra features due to QW transitions are
evident. The observation of separate QW and CM features in
MR relies on a range of circumstances including the require-

J. Appl. Phys. 112, 033108 (2012)

FIG. 2. Electro-reflectance (ER) signal in the vicinity of the VCSEL structure cavity-mode, as a function of temperature from 10 to 320 K in steps of
10 K, as indicated to the left of the spectra (main plot). The inset shows the
phase of the CM ER feature, obtained by fitting with an empirical Aspnes
lineshape. This indicates that the feature is nearly perfectly anti-symmetrical
(i.e., phase of 90 ) close to 260 K (spectrum highlighted in the main plot),
which indicates a crossing between the energies of the CM and QW ground
state transition.

ment that the upper DBR must not have too high a reflectivity and that the QW and CM MR features must be of
comparable broadening.17,18 These requirements are clearly
not satisfied here, especially the latter since we concluded
earlier that CQW  CCM. However, this does not mean one
cannot deduce information about the energy-position of the
QW transitions as it is also known that one can infer when
there is a CM-QW alignment by examining either the symmetry or the amplitude of the MR feature associated with the
CM dip.17,18 Furthermore, it has been shown that in situations where CQW  CCM the amplitude of the CM feature is
a quite misleading parameter, and rather it is the symmetry
which one must examine.17,18 When a resonance occurs
between a sharp CM dip and a much broader QW GST feature, a virtually perfect anti-symmetric MR line shape is
predicted.17,18
Bearing this in mind, it is clear from a careful examination of Fig. 2 that the CM ER feature at RT (300 K) starts
with a nearly symmetrical oscillatory line shape, and on
cooling passes through an intermediate state at 260 K
where the oscillator becomes almost perfectly antisymmetric, which by the above arguments indicates the
occurrence of the resonance between the CM and QW GST
features at this temperature. In order to quantify the degree
of anti-symmetry of the ER features, they were fitted with
the Aspnes third derivative functional form,19 which provides an empirical description of the symmetry of the line
shape via its phase parameter.18 The inset shows the results
for the fitted phase parameter, with a phase of 90 indicating
perfect anti-symmetry, showing a peak near 90 occurring at
260 6 5 K, as expected.
Upon further cooling, we observe that the CM ER signal
becomes more symmetrical again, but, unlike the high temperature spectra, this time the profile is positive-definite (i.e.,
lacking any significant negative-going oscillations). Meanwhile, its amplitude increases rapidly with cooling, reaching

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.22 On: Wed, 04 Feb 2015 17:30:15

033108-4

Blume et al.

J. Appl. Phys. 112, 033108 (2012)

a maximum at 120 K, after which it drops rapidly (by about
two orders of magnitude) while maintaining a roughly symmetrical positive-definite profile. The implications of these
lineshape changes are discussed in detail next.
As explained above, the peak in amplitude is of no significance in interpreting any CM-QW resonances but it is
helpful here to discuss further the behaviour of the CM ER
line shapes observed in Fig. 2. Here, we give only qualitative
arguments and refer the reader to the seminal papers for a
more detailed mathematical understanding.17,18 A modulated
reflectivity spectrum can be written in general as19
DR=R ¼ aDe1 þ bDe2 ;

(1)

where a and b are the Seraphin coefficients and De1 and De2
the real and imaginary parts of the modulated dielectric function, respectively.
It has been shown that, in the case of a VCSEL in the vicinity of ECM, a and b can be essentially assumed to result
from the modulation of the refractive index and absorption,
respectively, of the active cavity region, while, in the vicinity
of the QW GST, De1 and De2 result from the modulated complex dielectric function of the QW.17 Both pairs of parameters form Kramers-Kronig couples so that, while a has a
dispersive (i.e., anti-symmetric) shape, b has an absorptive
(symmetric) form—see Fig. 3(a) (left and right, respectively).
In contrast, De1 has an essentially absorptive (symmetric)
form while De2 is dispersive (anti-symmetric)—see Fig. 3(b).
However, in the present situation where CQW  CCM, then,
when the CM and QW energies are in close proximity and
the MR features overlap, both products aDe1 and bDe2 are

FIG. 3. Schematic lineshapes of: (a) the Seraphin coefficients a and b in the
vicinity of the CM energy (ECM) and; (b) of the modulated dielectric function (De1 and De2) in the vicinity of the QW ground-state transition energy
(EQW), in the situation where the broadening of the QW feature is much
larger than that of the CM. (c) shows their respective products aDe1, bDe2
near resonance, i.e., ECM  EQW. (d) shows the products when away from
resonance, e.g., when ECM < EQW.

essentially anti-symmetric in nature and have spectral profiles
with symmetries determined primarily by those of the Seraphin coefficients. This is illustrated qualitatively in Fig. 3(c),
left and right, respectively.
Note also from the example situation in Fig. 3(c), that in
the situation where ECM ¼ EQW the bDe2 term actually
becomes very small compared to aDe1, because De2 is weak
and broad and slowly crossing zero at EQW, while, though
De1 is also weak and broad, it is maximal at EQW. Thus, the
aDe1 term is actually dominant, resulting in a predicted line
shape that is virtually perfectly anti-symmetric when
ECM ¼ EQW.
As one moves slightly away from the resonance condition, for instance due to changing the temperature (see Figs.
3(d)), the amplitude of the bDe2 product may temporarily
increase significantly, and, because EQW now lies off-centre
from ECM, the bDe2 product can yield a roughly symmetric
line shape of quite large amplitude. On the other hand, the
aDe1 product is still approximately anti-symmetrical, but its
amplitude is strongly reduced in this “de-tuned” non-resonance condition. The net amplitude of the DR/R signal in this
case is given by the sum according to Eq. (1), but, since the
broad De1 and De2 profiles are approximately constant over
the region of the sharp CM feature, the DR/R signal is given
essentially by a linear combination of the Seraphin coefficients. Since the respective magnitudes and widths of a and
b depend on structural details of the particular sample, the
amplitude of the DR/R signal cannot be used as a reliable indicator of any resonance condition in these conditions. In our
case, it is clear from Fig. 2 that, as we cool below the resonance condition temperature, the product bDe2 is actually
stronger than the aDe1 component, so the observed ER amplitude increases and reaches a symmetrically shaped maximum at 120 K (Fig. 2). However, as explained earlier, this
is an artefact and cannot be used to infer any information of
use for our present purposes. Furthermore, the observed sign
of this bDe2 signal is of no significance and depends on the
phase settings of the lock-in amplifier.
Further cooling causes ECM and EQW to shift even further apart and one can see from Fig. 3 that it is inevitable
that the overall DR/R magnitude must eventually decrease to
very small values as the respective CM and QW features
move increasingly further out of each other’s influence. The
final drop in ER amplitude seen in Fig. 2 towards 10 K may
also be partly attributed to the freeze-out of the carriers in
the doped DBRs and thus an increase of the width of the
depletion region. Hence, the constant external modulation
voltage would then result in a smaller field modulation of the
actual QW region, so that the ER signal drops away.
To provide further confirmation of the CM/QW GST ER
resonance found at 260 K, we performed edge EL measurements similar to that shown in Fig. 1, but as a function of
temperature. The resulting peak positions of the edge EL are
plotted in Fig. 4 (filled squares) together with the CM positions obtained from normalised normal-incidence R measurements (open triangles), which were obtained simultaneously
with the ER measurements. The large error bars on the EL
peaks are due to the very broad EL spectra. The corresponding uncertainties in the CM energies are within the symbol

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.22 On: Wed, 04 Feb 2015 17:30:15

033108-5

Blume et al.

FIG. 4. Edge electro-luminescence (EL) peak positions (filled squares) and
CM positions from normalised normal-incidence reflectivity measurements
(open triangles) as a function of temperature. The curves, which are only
guides to the eye, cross at 280 K 6 20 K, thus confirming the ER result of
Fig. 2.

size of Fig. 4. We observe a crossing of the two data sets at
about 280 6 20 K in Fig. 4, which, though rather approximate, is equal to within experimental uncertainty to the resonance temperature 260 6 5 K found in the ER of Fig. 2.
Apart from the rather large uncertainty due to the broad EL
peaks, the difference in the two deduced tuning temperatures
may possibly also be due to the different applied electric
field amplitude in the two experiments; while the ER was
performed in an unbiased configuration, thus resulting in
only small applied internal fields, the EL experiments were
performed with a forward bias, which, due to the injection of
carriers, screens any built-in internal fields. Thus, in the EL
experiments the small red-shift in the transition energy
induced via the quantum confined Stark effect20 is slightly
reduced and would result in a slightly higher EQW transition
energy. This would then give a slightly higher crossover temperature in the EL results of Fig. 4 as compared to the ER
results in Fig. 2.
From the two sets of spectroscopic results in Figs. 2 and
4, we may conclude reasonably confidently that the CM and
QW GST are in resonance at a temperature of 270 6 10 K in
this VCSEL wafer.
V. DEVICE MEASUREMENTS—SHIFT OF LASING
WAVELENGTH WITH TEMPERATURE

We now return to the aforementioned studies on devices
fabricated from the VCSEL wafer. An EEL was also grown
and fabricated with nominally the same active region as the
VCSEL. The cladding in the EEL is Al0.65Ga0.35As. The
length of the fabricated EEL is 1000 lm with a 50 lm-width
ridge stripe. Fig. 5 shows the temperature dependence of the
lasing emission energies of the two devices, both under
pulsed operating conditions of 500 ns-long pulses at a duty
cycle of 10 kHz (for the VCSEL) and 100 Hz (for the EEL).
The VCSEL device diameter in these studies was 9 lm. The
different duty cycles were used in order to maximise the
strength of the measured integrated signal whilst minimising
Joule heating. The weaker luminescence from the VCSEL

J. Appl. Phys. 112, 033108 (2012)

FIG. 5. Shift of the pulsed lasing photon energy of the VCSEL (open
squares) together with the corresponding shift of the edge emitter (filled
circles) with nominally identical active region structure. The cross-over is at
267 K, which confirms the earlier spectroscopic results about the CM/QW
alignment. The fitted straight lines show that both behaviours are approximately linear over the studied temperature range. The inset shows difference
in lasing peaks between the EEL and VCSEL at 220 K.

(due to the smaller emitting area) and lower absolute threshold current meant that a higher repetition rate was needed for
the VCSEL measurements. The pulse width was fixed to
avoid increasing the heating during the pulse. Thus, for
improved signal (i.e., signal/noise ratio) the VCSEL was
operated at a higher frequency.
The shift of the EEL emission energy reflects the temperature dependence of the gain peak and thus that of the
emission peak due to the QW GST, which in turn is given by
the thermally induced shift in bandgap energy. In this temperature range (between 100 K and 310 K), the shift is
approximately linear and can be fitted with
EEEL ðTÞ ¼ 1:0648eV  T  0:300meV=K:

(2)

This is comparable to the corresponding shift in the edgeEL peaks seen in Fig. 4 which can be fitted with the straight
line
EedgeEL ðTÞ ¼ 1:0461eV  T  0:231meV=K:

(3)

These two fitted lines are equal to within the experimental uncertainties in the edge-EL energies (see Fig. 4).
The shift of the VCSEL emission with temperature is
determined only by the shift of the CM dip which is much
slower than the shift of the QW GST, and is due only to thermal changes in the cavity thickness and refractive index.21
This is also approximately linear in this temperature range
and can be fitted with
EVCSEL ðTÞ ¼ 0:9984eV  T  0:056meV=K:

(4)

This matches quite well with the shift of the CM in Fig. 4
measured from the reflectivity measurements on the prefabrication wafer, which, over the same temperature range as
Fig. 5 can also be satisfactorily fitted with a straight line
ECM ðTÞ ¼ 0:9969eV  T  0:056meV=K:

(5)

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.22 On: Wed, 04 Feb 2015 17:30:15

033108-6

Blume et al.

The slight bodily blue-shift of 1.5 meV of the VCSEL
fitted line (Eq. (4)) with respect to the CM fitted line (Eq.
(5)) is most probably due to slight differences in cavity thickness between the fabrication and pre-fabrication regions of
the wafer. The cross-over of the two straight lines fitted to
the device results in Fig. 5 (Eqs. (2) and (4)), and thus the
CM-QW alignment deduced from that, is found to occur at
T ¼ 272 6 5 K which agrees very well with the earlier wafer
spectroscopy result of 270 6 10 K.
These observations have consequences regarding the device performance; at the operating current the gain peak has
to be well aligned with cavity mode dip in an operating
VCSEL. If the CM and QW are not tuned at the desired operating temperature, this will necessitate higher threshold currents to enable lasing. In a recent separate set of
temperature-dependent studies of the spontaneous and lasing
emission characteristics of the EEL and VCSEL devices of
Fig. 5, we showed that the radiative component of the
threshold current is indeed minimal at the CM-QW tuning
temperature of 270 K deduced here from the spectroscopic
studies.22
In the absence of any other temperature dependencies,
the total threshold current will also be a minimum at this
temperature. However, it is important to note that there are
also highly temperature-dependent non-radiative contributions to the threshold current and that these compete with the
CM-QW tuning effect; with, say heating from low temperature, while the CM and QW are moving into better alignment, so producing a steady drop in the radiative component
of the threshold current, the non-radiative losses are increasing and divert ever-increasing proportions of the total driving
current into deleterious non-radiative recombination as the
device is warmed. The end result is a “trade-off” between
these two effects and so that the threshold current in the
VCSEL device is actually at a minimum at a somewhat
lower temperature (225 K) than the CM-QW tuning temperature deduced here from the purely spectroscopic studies.
These effects are described in full detail in Ref. 22 and must
be considered when optimising a VCSEL design.
VI. SUMMARY AND CONCLUSIONS

We have shown that a GaAs-based VCSEL with
GaAsSb QWs operates near the silica-fibre 1.31 lm zero dispersion wavelength in CW mode at 300 K, important for
metropolitan-area fibre-to-the-home networks. Nondestructive ER and destructive edge-EL spectroscopic measurements on pre-fabrication wafer samples show that the
growth has resulted in a slight misalignment between the
CM dip and the QW ground-state transition energies
(EQW < ECM) at 300 K and only come into the “tuned” state
(EQW ¼ ECM) at the lower temperature of 270 6 10 K. The
results also indicate that the aforementioned CW VCSEL
lasing observed at 300 K occurs because of the large broadening of the QW emission, which is approximately 200 times
wider than the CM dip in the VCSEL. We have explained
how this great difference in broadening means that the ER
spectral line shape becomes anti-symmetrical at the tuning
temperature. This effect has in fact been used to determine

J. Appl. Phys. 112, 033108 (2012)

the tuning temperature, by quantifying the degree of antisymmetry of the ER line shape as a function of temperature.
The spectral emission peaks of fabricated VCSEL devices
have also been studied as a function of temperature and compared with those of EELs with nominally identical active
regions to those of the VCSEL, showing that the EEL and
VCSEL emission wavelengths cross at the same tuning temperature as determined by the ER and EL studies on the prefabrication wafer samples. We have discussed how these
results indicate that the radiative component of the threshold
current should be minimal near 270 K. Finally, the results
reported here are discussed in light of our separate study of
the temperature dependence of the actual threshold current
in operating VCSEL devices, where the contribution of competing non-radiative recombination processes is examined in
detail. The techniques used here provide a useful insight into
the factors influencing the operating characteristics of the
VCSELs and provide a means of further device optimisation.
The utility of the non-destructive ER characterisation technique has been clearly demonstrated for these VCSEL
structures.
ACKNOWLEDGMENTS

We acknowledge support from the EPSRC(UK) under
grants GR/T21516/01 and EP/H005587/1. TJCH thanks the
“Malaysian University Grant Program Tier 1,” Universiti
Teknologi Malaysia (Q.J130000.7126.01H55) and M.O.H.E
for funding.
1

A. J. Mayhew, S. J. Page, A. M. Walker, and S. I. Fisher, BT Technol. J.
20, 91 (2002).
2
Y.-F. Lao, C.-F. Cao, H.-Z. Wu, M. Cao, and Q. Gong, Electron. Lett.
45(2), 105 (2009).
3
S. J. Sweeney, D. McConville, N. F. Masse, R.-X. Bouyssou, A. R. Adams,
C. N. Ahmad, and C. Hanke, Phys. Status Solidi B 241(14), 3391–3398
(2004).
4
O. B. Shchekin and D. G. Deppe, IEEE Photon. Technol. Lett. 14, 1231
(2002).
5
I. P. Marko, A. R. Adams, S. J. Sweeney, D. J. Mowbray, M. S. Skolnik,
H. Y. Liu, and K. M. Groom, IEEE J. Sel. Top. Quantum Electron. 11(5),
1041 (2005).
6
M. Yamada, T. Anan, K. Kurihara, K. Nishi, K. Tokutome, A. Kamei, and
S. Sugou, Electron. Lett. 36, 637 (2000).
7
P. Dowd, W. Braun, C. M. Ryu, C.-Z. Guo, S. L. Chen, U. Koelle, S. R.
Johnson, and Y.-H. Zhang, CLEO 1999 Tech. Dig. (IEEE Cat.
No.99CH37013), 1999, 489.
8
N. Samal, Y. Cao, D. Ding, J.-B. Wang, S.-Q. Yu, X. Jin, S. A. Chaparro,
Yu. G. Sadofyev, S. R. Johnson, and Y.-H. Zhang, in SPIE Photonics
West, San Jose, CA, January 2003.
9
S.-Q. Yu, D. Ding, J.-B. Wang, N. Samal, X. Jin, Y. Cao, S. R. Johnson,
and Y.-H. Zhang, J. Vac. Sci. Technol. B 25(5), 1658 (2007).
10
P. Dowd, S. R. Johnson, S. A. Feld, M. Adamcyk, S. A. Chaparro, J. Joseph, K. Hilgers, M. P. Horning, K. Shiralagi, and Y.-H. Zhang, Electron.
Lett. 39, 987 (2003).
11
X. Jin, S.-Q. Yu, Y. Cao, D. Ding, J.-B. Wang, N. Samal, Yu. G. Sadofyev, S. R. Johnson, and Y.-H. Zhang, in Proceedings of the IEEE Lasers
and Electro-Optics Society 16th Annual Meeting (IEEE Cat.
No.03CH37460) (2003), Vol. 1, p. 69.
12
Yu. G. Sadofyev, S. R. Johnson, S. A. Chaparro, Y. Cao, D. Ding, J.-B.
Wang, K. Franzreb, and Y.-H. Zhang, Appl. Phys. Lett. 84, 3546 (2004).
13
M. Zujewski and W. Nakwaski, Opto-Electron. Rev. 18(2), 181 (2010).
14
S. R. Johnson, C. Z. Guo, S. Chaparro, Yu. G. Sadofyev, J. Wang, Y. Cao,
N. Samal, J. Xu, S. Q. Yu, D. Ding, and Y.-H. Zhang, J. Cryst. Growth
251, 521 (2003).
15
T. K. Sharma and T. J. C. Hosea, Jpn. J. Appl. Phys. 48, 082301 (2009).

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.22 On: Wed, 04 Feb 2015 17:30:15

033108-7
16

Blume et al.

S. Ghosh, S. Constant, T. J. C. Hosea, and T. E. Sale, J. Appl. Phys. 88,
1432 (2000).
17
T. J. C. Hosea, Thin Solid Films 450, 3 (2004).
18
S. Ghosh, S. B. Constant, and T. J. C. Hosea, Appl. Phys. Lett. 78,
3250–3252 (2001).
19
D. E. Aspnes, Surf. Sci. 37, 418 (1973).

J. Appl. Phys. 112, 033108 (2012)
20

S. Schmitt-Rink, D. S. Chemla, and D. A. B. Miller, Adv. Phys. 38, 89
(1989).
21
G. Knowles, S. J. Sweeney, T. E. Sale, and A. R. Adams, Phys. Status
Solidi B 223, 581 (2001).
22
K. Hild, I. P. Marko, S. R. Johnson, S.-Q. Yu, Y.-H. Zhang, and S. J.
Sweeney, Appl. Phys. Lett. 99, 071110 (2011).

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.22 On: Wed, 04 Feb 2015 17:30:15