Thin Solid Films 367 (2000) 290±294
www.elsevier.com/locate/tsf
MBE growth of planar microcavities with distributed Bragg re¯ectors
K. ReginÂski*, J. Muszalski, M. Bugajski, T. Ochalski, J.M. Kubica, M. Zbroszczyk,
J. KaÎtcki, J. Ratajczak
Institute of Electron Technology, Warsaw, Poland
Abstract
We discuss some problems of molecular beam technology (MBE) technology of planar microcavities based on GaAs, AlGaAs and InGaAs
compounds. This technology needs speci®c methods of in situ control and postgrowth characterization. One of the crucial problems is that of
controlling the substrate temperature and the growth rate with very high accuracy. We demonstrate the usefulness of substrate temperature
oscillations observed by infrared pyrometry for both the temperature and the growth rate control. For studying the perfection of layers and
interfaces the cross-sectional transmission electron microscopy has been used. To optimize the optical characteristics of the microcavities,
several experimental methods have been applied. The Bragg re¯ectors were investigated by optical re¯ectivity. For selective excitation of a
quantum well (QW) in a cavity active layer, the Ti-sapphire tuneable laser have been used. The ®ne tuning between the QW emission and the
cavity Fabry±Perot resonance has been investigated by photoluminescence for varying temperature of the sample. q 2000 Elsevier Science
S.A. All rights reserved.
Keywords: Molecular beam technology growth of Bragg re¯ectors; Planar microcavities; Vertical cavity surface emitting lasers
1. Introduction
Vertical cavity surface emitting lasers (VCSELs) have
been attracting much attention in recent years and a lot of
different devices have been demonstrated by using various
growth techniques [1±3]. The laser cavity of VCSEL is
usually constructed normal to the substrate plane by stacking multilayer ®lms including an active region and two
dielectric mirrors. Such a structure forms a Fabry±Perot
cavity resonator. A dielectric mirror can be formed with a
periodic stack of quarter wavelength thick layers of alternating high and low refractive index material. Such a mirror
is referred to as a distributed Bragg re¯ector (DBR). The
dielectric layers can be semiconductor layers deposited via
molecular beam epitaxy (MBE) growth. The active region
consists usually of a layer of GaAs of the thickness of one
wavelength or half a wavelength and of one or several quantum wells (QWs) of InGaAs (in our case In0.20Ga0.80As has
been applied). The quantum wells are typically situated in
the antinodes of a standing electromagnetic wave [4]. An
example of such a structure is presented in Fig. 1. This is the
whole VCSEL structure designed for l 1000 nm emission. A l-cavity is made of GaAs and contains in the center
one QW made of In0.20Ga0.80As.
* Corresponding author. 32/46. Lotnikow Avenue, PL-02-668 Warsaw,
Poland. Tel.: 1 48-22-548-7920; fax: 1 48-22-847-0631.
E-mail address: [email protected] (K. ReginÂski)
The procedure of designing the MBE process for growing
the VCSEL structure comprises two classes of problems.
The ®rst one is connected with growing the materials of
high quality for all parts of the structure. This optimization
concerns GaAs, Al(Ga)As, and InGaAs compounds. The
problem is typical for MBE technology and can be solved
by standard methods: We optimize the MBE technology by
testing different substrate temperatures, different growth
rates, and different arsenic to metal ratios. Similarly, the
problem of smoothness of interfaces is solved typically by
comparing the interfaces made with and without the growth
stopping.
The second group of problems is characteristic of the
VCSEL structure. Optimization of such a structure requires
the precise tuning of its main parts ± Bragg re¯ectors, GaAs
cavity and QWs. From theoretical modeling as well as from
test processes it follows that the structure perfection is very
sensitive to variation in thickness of the separate layers and
their composition [4]. The required accuracy can hardly be
achieved without additional non-standard internal control in
the MBE system. Thus, besides the careful calibration of the
growth rate and composition, some non-standard methods
of internal control should be applied. Moreover, to test the
grown structures, we should combine several methods of
postgrowth characterization. The experimental works
should also be combined with theoretical calculations of
optical properties of separate DBRs and the whole micro-
0040-6090/00/$ - see front matter q 2000 Elsevier Science S.A. All rights reserved.
PII: S00 40-6090(00)0069 0-8
K. ReginÂski et al. / Thin Solid Films 367 (2000) 290±294
cavities (theoretical re¯ectivity spectra and emission characteristics).
2. Precision requirements
The optimization of the microcavity requires proper
tuning of the wavelength of radiation emitted from the
active region, the peak re¯ectivity of the DBRs, and the
cavity resonance. It is the reason why the structure performance is very sensitive to variations in thickness of layers
and their compositions. The wavelength of radiation from
the QW depends on both the composition and thickness. The
re¯ectivity of DBRs in case of GaAs/AlAs re¯ectors
depends on the layer thickness in the mirrors. Similarly,
the position of the cavity resonance depends on the thickness of the spacer layers between the mirrors and the QWs
region and the phase of the re¯ection from the mirrors.
Thus, the optimal growth of the structure requires simultaneous alignment of all three features. However, some variation of those parameters can be tolerated depending on how
much variation in threshold and ef®ciency is acceptable
from device to device.
From theoretical considerations it follows that even a
10% error in layer thickness does not signi®cantly degrade
the peak re¯ectivity, but in this case the stop band is shifted
in wavelength [5]. Also of interest is the effect of having one
layer in the DBR thinner than a quarter wavelength and one
thicker. The result of the offsetting thickness variations is to
decrease slightly the peak re¯ectivity of the mirror. Thus,
the goal in growing the mirrors for a VCSEL is to get the
layers each approximately a quarter wavelength thick and to
get the re¯ection band centered at the right wavelength. If
the layers slightly differ from a quarter wavelength, it is not
so important as long as the re¯ection band is situated in the
right place. The position of the cavity resonance will ultimately determine the lasing wavelength. From theoretical
considerations it follows that in a properly fabricated laser
Fig. 1. Layer structure of a vertical cavity surface emitting laser designed
for l 1000 nm emission.
291
structure the wavelength of the re¯ectivity peak may be
shifted by ^2%, or each layer thickness may vary by
^2%. It means that the accuracy of the control of gallium
and aluminium ¯uxes should be of the order of 2%.
The ®nal issue to consider is the effect that a change in
layer thickness and composition has on the wavelength
emitted from the active region. For an InGaAs quantum
well there are two effects to account for: a change in band
gap with composition and a change in electron and hole
con®nement energies with well width. From simple calculations it follows that even an error in Ga ¯ux of 5% causes
negligible change in the transition energy since the
increased con®nement energy is compensated by the
decreased band gap. The same calculations for a 5% error
in the In ¯ux show a wavelength shift one order of magnitude greater. Whereas a small change in the Ga ¯ux will
cause no noticeable shift in the transition energy, the In ¯ux
should be extremely accurate. In reality, it is very dif®cult to
obtain stabilization of the ¯ux better than 2% without a realtime control of the growth rate in the MBE system.
3. Experimental set-up and growth of the structures
The growth processes reported in this study have been
performed by Elemental Source MBE technique on RIBER
32P machine equipped with ABN 135L evaporation cells.
Certain details of the control system installed in the growth
chamber of this machine are presented in Fig. 2.
The molecular ¯uxes were measured by a Bayard±Alpert
gauge mounted on the sample manipulator. The manipulator
could be rotated so as to place the Bayard±Alpert gauge at
Fig. 2. Details of the control system installed in the growth chamber of
RIBER 32P MBE machine.
292
K. ReginÂski et al. / Thin Solid Films 367 (2000) 290±294
the standard position of the substrate. Thus the gauge could
measure the beam equivalent pressure (BEP) of the As4, Ga,
Al, and In beams.
For monitoring the state of the crystal surface during the
growth the re¯ection high energy electron diffraction
(RHEED) system with a 10 keV electron gun was used.
The RHEED patterns were registered by CCD camera and
then processed in real-time and recorded by a computer
acquisition system. This system consists of a personal
computer equipped with a frame grabber, two monitors,
and a high resolution laser printer. The RHEED pictures
could be processed by a specially written computer program
and stored on a hard disc for further reference. The system
also enables us to register RHEED intensity oscillations
and, as a result, to determine the growth rate of a layer.
The substrate temperature was measured with an thermocouple and simultaneously with an IRCON Modline Plus
pyrometer. This particular model of pyrometer is specially
designed to measure the GaAs surface temperature by monitoring the radiation emitted in a narrow range of wavelengths (0.940 ^ 0.03 mm) which is shorter than the band
edge of GaAs (but longer then AlxGa12xAs, x . 0:25) at the
temperatures which are of interest for MBE (400±7508C).
The pyrometer is connected with a computer system
enabling registration the pyrometric temperature as a function of time. During the growth of the structure, even if the
temperature registered by a thermocouple is constant, the
pyrometric temperature oscillates in time. The source of
those apparent temperature oscillations is the interference
of the radiation emitted from the substrate and the radiation
re¯ected at the underlying layer/grown layer and grown
layer/vacuum interfaces. The phase information and the
period of oscillations provide information on the actual
growth rate, whereas the mean value of the pyrometer readout correlates with the true substrate temperature. Analysis
of the oscillation curve after the growth gives as direct
information about thickness of all the layers in the grown
structure. Thus, the interference pyrometry offers a convenient alternative to a standard laser re¯ectometry when it is
necessary to control growth rate of layers with high precision. More details about this control system one can ®nd in
[7].
The calibration of the growth rate of the layers was
performed in several steps. The preliminary calibration
has been performed by combining two methods. First, the
necessary values of temperatures of effusion cells were
found by using temperature±¯ux calibration curves and
well known formulae for converting BEP into absolute
values of ¯uxes [6]. Next, RHEED intensity oscillations
were measured during the growth processes of all the semiconductor compounds applied in the structure. The results
of the measurements gave some corrections into the values
of effusion cell temperatures. The ultimate precision of the
growth rate calibration was achieved during the test growth
of the structures by using the system of registering the
apparent substrate temperature oscillations. It enabled to
control the growth rate of the individual layers with accuracy about 1%. As an example, a pyrometer readout during
the MBE growth of the test VCSEL structure is shown in
Fig. 3.
In a series of processes the separate elements of the
VCSEL structure as well as the whole VCSELs were
grown. The VCSEL structures were designed for l
1000 nm emission. They consisted of two Bragg re¯ectors
composed of 14.5 and 23.5 pairs of AlAs/GaAs quarter
wavelength layers. A l-cavity was made of GaAs and
contained in the center a QW made of In0.2Ga0.8As. This
structure is presented in Fig. 1.
The growth has been performed on the GaAs p2substrate
of (001) orientation. The optimized growth temperatures
were as follows: the GaAs:Be buffer layer was grown at
5908C; the AlAs/GaAs Bragg re¯ectors, GaAs cavities
and In0.2Ga0.8As QW were grown at constant temperature
5208C. The growth rate for GaAs and AlAs was 1 mm/h and
that for AlGaAs was 2 mm/h. The background pressure
during the growth was 5 £ 10 8 Torr. The whole process
was performed without any stopping. During the growth
of the main structure, the substrate temperature shown by
a thermocouple was kept constant. The pyrometer readout
shown in Fig. 3 demonstrates that this method of regulation
leads to the really good thermal stabilization of the
substrate.
4. Postgrowth characterization
The process of optimizing the technology comprised
postgrowth characterization of the separate Bragg re¯ectors,
the microcavities, and the complete VCSEL structures.
Cross-sectional transmission electron microscopy
(XTEM) has been used to observe the perfection of layers
and interfaces. Cross-sectional specimens were prepared by
the method described in [8]. The specimens were studied in
a JEM 200CX (200 kV) microscope. A XTEM micrographs
of the structure consisting of Bragg re¯ectors, l-cavity and
In0.2Ga0.8As QW are presented in Figs. 4 and 5. Whole the
structure is visible in Fig. 4. The interfaces between separate
layers are sharp and the periodicity of the AlAs/GaAs layers
Fig. 3. Pyrometer readout during the MBE growth of the test structure of a
vertical cavity surface emitting laser.
K. ReginÂski et al. / Thin Solid Films 367 (2000) 290±294
293
Fig. 4. Transmission electron microscopy micrograph of the laser l-cavity
consisting of Bragg re¯ectors, GaAs region, and In0.2Ga0.8As quantum well.
is in a good accordance with the project. The region of the
In0.2Ga0.8As QW is presented in Fig. 5. The interfaces
between QW and cavity as well as those between the cavity
and the Bragg re¯ectors layers are straight and sharp. No
defects in the structure are visible.
Optical properties of grown structures were investigated
by optical re¯ectivity and photoluminescence (PL). The
re¯ectivity spectra were measured in a system by using a
halogen lamp as a source of incident white light. After
re¯ection the relevant wavelength is selected by monochromator and registered by Ge detector. Fig. 6 shows a typical
re¯ectance spectrum measured from the structure consisting
of Bragg re¯ectors, l-cavity and In0.2Ga0.8As QW. A re¯ectance spectrum for a similar structure, but with l/2-cavity, is
shown in Fig. 7. A re¯ectance spectrum for the whole
VCSEL structure is presented in Fig. 8. The measured spectra are similar to those calculated for ideal structures and
exhibit the cavity resonance centered in the high re¯ectivity
band of the Bragg re¯ectors. All the spectra exhibit broad
stop bands where the plot is almost ¯at. The re¯ectivity
within those frequency intervals attains nearly 100%. The
stop bands are bounded by deep minima which also testi®es
to the good quality of DBRs.
PL measurements give us important information about
the InGaAs quantum wells. To excite the quantum well, a
Fig. 5. Transmission electron microscopy micrograph of the central part of
the laser l-cavity. The region of In0.2Ga0.8As quantum well is shown in
detail.
Fig. 6. Re¯ectance spectrum and photoluminescence from the structure
consisting of Bragg re¯ectors, l-cavity and In0.2Ga0.8As quantum well.
Ti-sapphire tuneable laser has been used. Depending on the
size of the cavity we observed either an enhancement or
suppression of the coupling of 1e±1hh electronic dipole
transitions into the resonant Fabry±Perot mode of the planar
microcavity structure.
The structures were excited with a beam perpendicular to
its surface. The emission was registered either from the edge
of the structure (in plane PL) or in the direction perpendicular to its surface. The emission from the edge does not
show any modi®cations (comparing with a separate QW)
and can be treated as a reference. The emission from the
surface shows the resonance with a Fabry±Perot resonator.
The large amplitude of this signal can be regarded as an
evidence of good tuning between the wavelength emitted
from the quantum well and the cavity resonance. Some
examples of PL measurements are shown in Figs. 6 and 7.
Fig. 6 shows a typical PL spectrum measured from the
Fig. 7. Re¯ectance spectrum and photoluminescence from the structure
consisting of Bragg re¯ectors, l/2-cavity and In0.2Ga0.8As quantum well.
PL, photoluminescence perpendicular to the Bragg re¯ector; PL in plane,
photoluminescence from the edge of the structure.
294
K. ReginÂski et al. / Thin Solid Films 367 (2000) 290±294
This type of measurements enabled us to accelerate the
procedure of designing and testing the laser structures.
5. Conclusions
Fig. 8. Re¯ectance spectrum from a vertical cavity surface emitting laser
structure of l-type with an In0.2Ga0.8As quantum well.
structure consisting of Bragg re¯ectors, l-cavity and
In0.2Ga0.8As QW. The narrow PL peak corresponds to the
case of good tuning. The PL spectra for a similar structure,
but with l/2-cavity, are shown in Fig. 7. Both the types of
registration of photoluminescence signal, from the edge
(denoted as PL in plane) and from the surface (denoted as
PL), are presented.
The ®ne tuning between the QW emission and the cavity
Fabry±Perot resonance has also been investigated by PL for
varying temperature of the sample. Even when the whole
system is not well tuned we can test the structure by registering luminescence as a function of temperature. By changing the temperature we can shift the frequency of the wave
emitted from the QW and achieve the resonance for a
certain temperature. An example of such measurements is
presented in Fig. 9. The tested structure consisted of Bragg
re¯ectors, l-cavity and In0.2Ga0.8As quantum well.
Applying several methods of in situ control and combining selected methods of postgrowth characterization we
have developed the MBE technology of growing the
VCSEL structures. We have demonstrated that for reproducible growth of microcavities and VCSELs by MBE, the
growth rate of individual layers has to be controlled with
accuracy better than 2%. To achieve this accuracy we have
applied a precise procedure of preliminary calibration
combined with registering the pyrometric temperature oscillations. We have shown that the interference pyrometry
offers a convenient alternative to a standard laser refractometry when it is necessary to control the growth rate of
layers with high precision.
By using the above described methods we have grown a
number of microcavities and VCSEL structures of good
quality con®rmed by transmission electron microscopy,
photoluminescence, and optical re¯ectivity. The structures
have proper characteristics of the Bragg re¯ectors and the
quantum well emission wavelength precisely tuned to the
cavity resonance.
Acknowledgements
This work has been supported by the State Committee for
Scienti®c Research (Poland) under Contract Nos. PBZ 28.11/P7 and 8.T11B.064.14.
References
Fig. 9. Photoluminescence spectra for varying temperature of the sample.
The structure consists of Bragg re¯ectors, l-cavity and In0.2Ga0.8As quantum well.
[1] T. Baba, T. Hamano, F. Koyama, K. Iga, IEEE, J. Quantum Electron.
27 (1991) 1347.
[2] C.J. Chang-Hasnain, in: G.P. Agrawal (Ed.), Semiconductor Lasers;
Past, Present, and Future, AIP Press, Woodbury, 1995, p. 145.
[3] L.A. Coldren, E.R. Hegblom, Y.A. Akulova, J. Ko, E.M. Strzelecka,
S.Y. Hu, VCSELs in 98: What we have and what we can expect, in:
K.D. Choquette, R.A. Morgan (Eds.), Vertical Cavity Surface Emitting
Lasers II, Proceedings of SPIE, Vol. 3286, 1998, pp. 2±16.
[4] T.E. Sale, Vertical Cavity Surface Emitting Lasers, Research Studies
Press, Taunton (1995) 55.
[5] R.S. Geels. PhD thesis, University of California, Santa Barbara, 1991.
[6] G.J. Davies, D. Williams, in: E.H.C. Parker (Ed.), The Technology and
Physics of Molecular Beam Epitaxy, Plenum, New York, 1985, p. 38.
[7] J. Muszalski, Thin Solid Films (this issue) p. 299.
[8] J. Katcki, J. Ratajczak, A. Malag, M. Piskorski, in: A.G. Cullis, A.
Staton-Bevan (Eds.), Microscopy of Semiconducting Materials, IOP,
Bristol, 1995, p. 273.