Spatially and spectrally resolved measurement of

Journal of Crystal Growth 230 (2001) 517–521
Spatially and spectrally resolved measurement of optical loss
in InGaN laser structures
H.D. Summersa,*, P.M. Smowtona, P. Blooda, M. Dineenb, R.M. Perksb,
D.P. Bourc, M. Kneisseld
Department of Physics and Astronomy, Cardiff University, P.O. Box 913, Cardiff, Wales, CF24 3YB, UK
Cardiff Semiconductor and Microelectronics Centre, Cardiff University, P.O. Box 913, Cardiff, Wales, CF24 3YB, UK
Agilent Laboratories, 3500 Deer Creek Road, Palo Alto, California, USA
Xerox Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, California, USA
The optical loss co-efficient in InGaN laser diodes, emitting at $410 nm, has been measured. The measurement
technique is based on the transmission of internally generated spontaneous emission through varying lengths of the
laser waveguide. It is unique in that it provides spectral and spatial information on the optical loss. The lasers studied
are typical of InGaN structures showing a high degree of waveguide loss, ai ¼ 40 cmÀ1 . The measurements also show
clear evidence of higher order transverse modes in the direction perpendicular to the growth plane with resonant
leakage of the optical field into the outer layers of the structure. This produces a modulation in the loss of these modes.
# 2001 Elsevier Science B.V. All rights reserved.
PACS: 42.55.P; 78.40.F; 42.25.B
Keywords: A3. Quantum wells; B1. Nitrides; B3. Laser diodes
1. Introduction
The accurate determination of optical loss
within laser diode structures is essential for the
optimisation of semiconductor laser performance.
The need to minimise the waveguide loss to ensure
a low threshold gain and hence a low threshold
current is self-evident. Within quantum well
systems this requirement has added importance
because there is strong saturation of the optical
gain at high carrier densities due to the constant
*Corresponding author. Tel.: +44-0-2920-874458; fax: +440-2920-874056.
E-mail address: [email protected] (H.D. Summers).
density of states within a single sub-band. Thus
operation at high threshold gain can produce
severe increases in current. Other aspects of the
laser performance are also compromised; the
differential gain is reduced thus affecting the
dynamic response and thermally activated carrier
leakage from the quantum well is increased due to
the high values of the quasi-Fermi level energies.
The optical loss within semiconductor lasers is
traditionally measured on devices of different
length. This provides a systematic variation in
the distributed mirror loss, am whilst the intrinsic
loss, ai remains fixed. Measurements of the slope
efficiency of the laser can be used to directly
compare am and ai [1] or the threshold current with
0022-0248/01/$ - see front matter # 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 0 2 4 8 ( 0 1 ) 0 1 2 8 4 - 2
H.D. Summers et al. / Journal of Crystal Growth 230 (2001) 517–521
length can be used as this reflects the changing
threshold gain due to the variation in the mirror
loss [2]. Both techniques are difficult to implement
accurately. Comparison of slope efficiencies requires measurements on large numbers of devices
to reduce uncertainties introduced by variations in
the facet quality and is only valid in the ‘ideal’
limit where complete pinning of the quasi-Fermi
levels occurs at the onset of lasing. In many laser
diode systems this is not the case due to the
presence of current spreading or thermally activated leakage currents [3]. Analysis of the threshold current is reliant upon a knowledge of the
gain–current relationship, this is not always
accurately known especially in new materials such
as InGaN. A further aspect of these techniques is
that they are based on an analysis of the lasing
mode and thus provide only limited information
on waveguide structures in which there are multiple modes. The loss is also determined solely at the
lasing wavelength.
Within InGaN/GaN/AlGaN laser structures
the measurement of the optical loss is particularly
important as it is difficult to produce well
controlled waveguiding of the optical emission
within this material system [4–7]. Growth
limitations on the thickness of the AlxGa1ÀxN
cladding layers and the presence of high refractive
index contact layers lead to structures which
are multi-moded in the direction perpendicular
to the epitaxial planes. They can also suffer
from high optical loss due to the presence of
large optical field intensities within the substrate
and outer buffer or capping layers of the laser.
In this paper we present details of a loss
measurement technique which allows us to determine the modal loss within InGaN laser diode
structures as a function of wavelength. The
technique also allows spatially distinct propagation modes to be distinguished.
2. Loss measurement technique
The measurement of the optical loss is based
upon a multi-section device configuration as
depicted in Fig. 1. A 4 mm wide trench is etched
through the top metallic contact and upper
Fig. 1. Schematic of the multi-section device configuration used
for the loss measurements.
capping layer in order to electrically isolate
300 mm length sections [8]. The electro-lumninescence generated within each section is collected
from the front of the sample. The emission from
the first section is taken as the source light, I0 . The
intensity collected when pumping subsequent
sections, distance La away from the facet, is then
related to the source by the expression
IðLa Þ ¼ I0 expðÀaLa Þ;
where a is the optical loss co-efficient of the
waveguide. At the long wavelength limit reabsorption of the luminescence within the unpumped sections is avoided because of bandgap
renormalisation within the pumped section which
red shifts the emission spectrum relative to the
passive material. Thus, the loss co-efficient correlates directly to the waveguide loss, ai . As the
source light is generated by spontaneous emission
within the semiconductor it is broadband and so
the technique provides spectral information on the
loss co-efficient. To obtain spatial discrimination
the collected light is passed through a monochromator and then focused onto a two-dimensional
H.D. Summers et al. / Journal of Crystal Growth 230 (2001) 517–521
CCD array. This gives the intensity profile at the
device facet in the vertical direction (perpendicular
to the growth plane). In the case of multi-mode
devices the emission from the different modes can
be isolated by choosing the relevant detector
3. Experimental results
The laser wafers studied were grown by metalorganic chemical vapour deposition (MOCVD)
on a (0 0 0 1)-oriented sapphire substrate. The
device consists of a 4 mm, n-doped, GaN buffer
layer, a 0.6 mm, n-doped Al0.08Ga0.92N cladding
layer, a 0.2 mm GaN waveguide region with five
In0.15Ga0.85N quantum wells, a 0.5 mm p-doped
Al0.08Ga0.92N cladding layer and a 0.1 mm,
p-doped GaN capping layer. The wafer was
processed into 50 mm wide, oxide isolated, stripe
devices with Ti/Al/Ti/Au contacts on the n-side
and Ti/Au metallisation on the p-side. As described above the p-type metallisation was separated into 300 mm length sections. The
spontaneous emission spectrum, collected from
the device facet, is shown in Fig. 2a. The luminescence is centred around 400 nm and has a 15 nm
full-width at half-maximum. The spectrum from
section 2 shows a reduction in intensity due to
optical loss within the foremost section. This is
particularly apparent at shorter wavelengths
(l5420 nm) due to absorption within the InGaN
quantum wells. The spectra were collected in the
transverse electric (TE) polarisation as this is the
lasing direction. The optical loss, derived from
these spectra, is shown in Fig. 2b. At longer
wavelengths, beyond the absorption edge of the
wells, the loss tends to a constant value of
$40 cmÀ1 which is typical of waveguide loss
values within InGaN lasers and in agreement with
previous measurements on these samples [9]. The
waveguide structure of the devices produces higher
order modes in the vertical direction in which the
optical field leaks out into n-doped buffer layer
and substrate. The spatial extent of this light is
depicted in Fig. 3 which shows an image from the
CCD array. The uppermost emission maximum
corresponds to a mode which is well confined
Fig. 2. (a) Spontaneous emission spectrum collected from the
device facet for electrical injection into sections 1 and 2,
respectively; (b) loss co-efficient calculated from the emission
Fig. 3. Detector image of the device facet showing wavelength
dependence in the x-direction and spatial variation on the yaxis. The dashed rectangle is a guide to the position of the facet.
H.D. Summers et al. / Journal of Crystal Growth 230 (2001) 517–521
within the waveguide region. It is this mode
(labelled A) which is shown in the earlier figures.
A second mode can be seen in the vertical direction
which has a high intensity within the GaN buffer
layer and the substrate (labelled B). In the case of
mode B there are a number of distinct maxima in
the horizontal direction indicating a spectral
modulation in the optical intensity. The optical
loss in modes A and B is shown in Fig. 4. At
shorter wavelength both modes show an increasing
loss due to absorption within the InGaN quantum
wells. This indicates that the higher order mode B
has some overlap with the waveguide region.
There is a clear modulation in the absorption loss
within mode B with a periodicity of 3.8 nm. Such
resonance effects have been observed within the
longitudinal mode spectra of InGaN lasers [10]
where they were attributed to scattering within the
laser producing coupled cavity effects. Similar
effects have also been observed in InGaAs
quantum dot lasers [11] where they have been
shown to be caused by resonant coupling of the
modes within the different layers of the structure
[12]. Theoretical studies on waveguiding in InGaN
lasers have also concentrated on the effects of such
mode coupling in the vertical direction [4]. This
produces energy transfer into the highly doped
GaN contact layers which increases the optical
loss. These experimental results support this
hypothesis as there is clearly a spectral modulation
Fig. 4. Optical loss spectra for the modes A and B indicated in
Fig. 3.
in the loss co-efficient. The modulation period of
3.8 nm corresponds to an interaction length, in the
vertical direction, of $8 nm which agrees with the
observation from Fig. 3 that the higher order
mode is guided within the GaN buffer layer and
sapphire substrate. At long wavelength the waveguide loss, ai is the same for both modes despite
their differing spatial profiles. We believe that this
loss is due to optical scattering at dislocation sites
as the dislocation density is constant through the
structure and hence such losses will be independent
of the mode profile.
4. Summary
In summary, a multi-section device structure has
been used to measure the optical loss co-efficient in
InGaN laser structures. By implementing the
technique using a CCD detector array the different
waveguide modes perpendicular to the epi-layers
have been isolated. The results show a waveguide
loss of 40 cmÀ1. There is also evidence of resonant
coupling of the higher order modes with the GaN
buffer layer leading to a spectral modulation of the
[1] J.R. Biard, W.N. Carr, B.S. Reed, Trans. Metall. Soc.
AIME 230 (1964) 286.
[2] G.P. Agrawal, N.K. Dutta, Semiconductor Lasers, Van
Nostrand Reinhold, New York, 1993.
[3] P.M. Smowton, P. Blood, IEEE J. Select. Topics Quantum
Electron. 3 (1997) 491.
[4] P.G. Eliseev, G.A. Smolyakov, M. Osinski, IEEE J. Select.
Topics Quantum Electron. 5 (1999) 771.
[5] D.K. Young, M.P. Mack, A.C. Abare, M. Hansen,
L.A. Coldren, S.P. Denbaars, E.L. Hu, D.D. Awschalom,
Appl. Phys. Lett. 74 (1999) 2349.
[6] S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa,
T. Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto,
T. Kozaki, H. Umemoto, M. Sano, K. Chocho, Appl.
Phys. Lett. 73 (1998) 832.
[7] M. Onomura, S. Saito, K. Sasanuma, G. Hatakoshi,
M. Nakasuji, J. Rennie, L. Sugiura, S. Nunoue, J. Nishio,
K. Itaya, IEEE J. Select. Topics Quantum Electron. 5
(1999) 765.
[8] J.D. Thomson, H.D. Summers, P.J. Hulyer, P.M. Smowton, P. Blood, Appl. Phys. Lett. 75 (1999) 2527.
H.D. Summers et al. / Journal of Crystal Growth 230 (2001) 517–521
[9] D.P. Bour, M. Kneissl, L.T. Romano, R.M. Donaldson,
C.J. Dunnrowicz, N.M. Johnson, Appl. Phys. Lett. 74
(1999) 404.
[10] S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa,
T. Yamada, T. Matsushita, Y. Sugimoto, H. Kiyoku,
Appl. Phys. Lett. 70 (1997) 2753.
[11] P.M. Smowton, E.J. Johnston, S.V. Dewar, P.J. Hulyer,
H.D. Summers, A. Patane, A. Polimeni, M. Henini, Appl.
Phys. Lett. 75 (1999) 2169.
[12] E.P. O’Reilly, A.I. Onischenko, E.A. Avrutin, D. Battacharyya, J.H. Marsh, Electron. Lett. 34 (1998) 2035.