APPLIED PHYSICS LETTERS
VOLUME 80, NUMBER 13
1 APRIL 2002
Frequency response of current-driven polarization modulation
in vertical-cavity surface-emitting lasers
Guy Verschaffelt,a) Jan Albert, Irina Veretennicoff, and Jan Danckaert
Department of Applied Physics and Photonics (TW–TONA), Vrije Universiteit Brussel, Pleinlaan 2,
1050 Brussels, Belgium
Sylvain Barbayb) and Giovanni Giacomelli
Istituto Nazionale di Ottica Applicata and INFM, Largo E. Fermi 6, 50125 Florence, Italy
Francesco Marin
Department of Physics, University of Firenze and INFM and LENS, Largo E. Fermi 2,
50125 Florence, Italy
共Received 3 July 2001; accepted for publication 30 January 2002兲
We present an experimental study of the current-driven polarization modulation properties of
vertical-cavity surface-emitting Lasers 共VCSELs兲. Some VCSELs exhibit a high-contrast
polarization flip for a particular value of pump current. By modulating the current around this value
we measure the critical modulation amplitude necessary to force current-driven polarization
switching as a function of the modulation frequency. We thus obtain the polarization modulation
frequency response, which shows the same cut-off frequency as the thermal frequency response.
This indicates the necessity to incorporate a temperature-dependent variable in realistic models that
describe the polarization behavior of these VCSELs. © 2002 American Institute of Physics.
关DOI: 10.1063/1.1465113兴
The polarization state of the light emitted by verticalcavity surface-emitting lasers 共VCSELs兲 is not defined a priori by their structure. Nevertheless, VCSELs do actually
emit linearly polarized light in one of two orthogonal linear
polarization modes 共PMs兲, which are usually oriented according to specific crystallographic directions. The two PMs
have a difference in frequency, which can vary between 1
and 40 GHz.1 In some VCSELs, polarization switching 共PS兲
between PMs is observed as the current is changed. During
the past few years experimental2–7 and theoretical work8 –11
has been aimed at understanding this effect. Generally, the
different proposals for explaining the observed PS can be
divided into two categories: those invoking slow 共lattice兲
thermal mechanisms2,4 – 6 and those relying on other, faster
mechanisms.3,9–11 The dynamic mechanism and the dominating time scales of PS are not only interesting from a physical
point of view, but also of great practical importance if one
wishes to exploit the VCSEL’s polarization properties for
applications in which fast controlled PS is required.
In this letter we present the results of a systematic experimental investigation of the current-driven polarization
modulation properties of single mode VCSELs12,13 as a function of both the amplitude and frequency of modulation. We
have studied different types of VCSELs 共proton implanted,
air post and oxide confined兲, operating at different wavelengths 共850 and 980 nm兲 that present different types of
switching 共from the lower to the higher optical frequency
mode with an increase in current, called type II PS, and vice
versa, called type I PS兲. We focus in this letter on the results
a兲
Electronic mail: guy.verschaffelt@vub.ac.be
Present address: Laboratoire Photonique et Nanostructures, 196 Avenue H.
Ravera, BP 107, 92225 Bagneux Cedex, France.
b兲
obtained on proton-implanted VCSELs from VIXEL Corporation.
While the steady state characteristics and some of the
dynamic features of these devices have already been
reported,4 – 6,13,14 the experimental study presented here is
original and represents a critical bench test for models describing the polarization features of VCSELs. Our gainguided proton-implanted GaAs/AlGaAs quantum-well VCSEL operates around 850 nm and exhibits a threshold current
close to 7mA. The VCSEL shows type II PS in its singletransverse mode regime when operated under continuous
wave 共cw兲 conditions on the high frequency side of the gain
maximum.5 The switching current can be tuned over the entire current region of single-transverse mode operation by
applying a controlled amount of stress with our selffabricated laser holder.6
In our setup, the light emitted by the VCSEL is collimated with a lens and sent through a linear polarizer to select
one of the two PMs. The polarization resolved intensity impinges on a 2 GHz bandwidth detector whose output signal is
visible on an oscilloscope. All the optical components are
slightly misaligned to prevent optical feedback, which could
cause extra instability in the VCSEL. The bias current generated by a stabilized current source and the modulation signal from a function generator are combined in a bias T and
subsequently sent through the VCSEL.13 The temperature of
the VCSEL package is actively stabilized up to a few mK.
The frequency responses of all the individual electronic components were calibrated beforehand. The overall amplitude
response 共from the signal generator to the laser current兲 was
checked as being linear up to at least several hundred MHz
in the small amplitude modulation regime.
When the VCSEL is biased close to the PS current, random mode hopping between the two orthogonally polarized
0003-6951/2002/80(13)/2248/3/$19.00
2248
© 2002 American Institute of Physics
Downloaded 18 Oct 2002 to 192.84.145.98. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
Appl. Phys. Lett., Vol. 80, No. 13, 1 April 2002
FIG. 1. Time trace of the polarization-resolved laser intensity when modulating the VCSEL around a PS current of 10.3 mA with a 200 kHz sinusoidal
signal and amplitude of 0.62 mA.
fundamental modes occurs,7,13–15 indicating that a small
bistable region is present. The average residence time 共or
dwell time兲 in the mode-hopping regime strongly scales with
the switching current: from ns when the PS current is close
to the laser threshold to nearly seconds when the PS current
is about two times the threshold. This stochastic hopping due
to spontaneous emission also shows up in modulation
experiments.13 While modulating around the switching current, PS occurs with a random delay 共i.e., jitter兲. If this delay
is longer than half a modulation period a polarization switch
will be missed. An example of a polarization-resolved time
trace with three switches missed is shown in Fig. 1. We
define a ‘‘successful’’ modulation period as one where the
laser undergoes the complete PS cycle.
Throughout the measurements we track the average time
spent in each of the PMs in real-time by means of an oscilloscope with built-in statistical functions.14 This allows us to
bias the VCSEL at the PS current where the randomly jumping laser spends on average an equal time in the two PMs.
After setting the bias current, we add a sinusoidal modulation signal with fixed frequency and varying amplitude. In
Fig. 2 we plot the fraction of successful modulation periods
measured as a function of the amplitude of a 200 kHz sinu-
Verschaffelt et al.
2249
FIG. 3. PS frequency response for switching currents of I sw⫽9.85 and
10.3 mA.
soidal modulation signal. Low amplitudes have only a small
fraction of successful cycles. This fraction grows rapidly as
the modulation amplitude is increased: through a narrow
transition region (⬃0.3 mA) one attains a regime of nearly
100% successful cycles. One should keep in mind that we
are actually modulating the current around the center of a
bistable region. In order to have PS, the modulation amplitude should exceed the width of the bistable region. Due to
stochastic effects, however, the transition is smoothened 共see
Fig. 2兲.
Based on Fig. 2, we define the minimum amplitude for
current-driven polarization modulation as the amplitude that
yields 80% successful modulation cycles. We have measured
this minimum amplitude as a function of the modulation frequency. In Fig. 3 we plot the inverse of this minimum amplitude as a function of the modulation frequency, henceforth
called the PS frequency response, that can be compared with
the response of a standard first order system 共see below兲.
These measurements were performed for PS currents of 9.85
and 10.3 mA, corresponding to dwell times of 10 ms and 5 s,
respectively. We have chosen these high switching currents
because the associated long dwell times reduce the occurrence of PS due to spontaneous mode hopping with respect
to PS forced by modulation. Still, as we will discuss below,
the effects of spontaneous mode hopping cannot be discarded
completely.
In Fig. 3 two regimes can clearly be distinguished and
they are separated by a cut-off frequency of about 90 kHz. In
the low frequency range there is a clear 共albeit weak兲 monotonic increase of the PS frequency response with a decrease
in frequency. Above 90 kHz the PS frequency response decreases drastically with an increase in frequency. For higher
frequencies 共above 3 MHz兲 the measurements are disturbed
by the fact that the minimum modulation amplitude is so
large that it drives the VCSEL from the threshold up into the
higher transverse mode regime. When we compare the experiments for the two different switching currents, we notice
a vertical shift of the measured PS response curve. This is
because the bistable region is wider for higher switching currents. However, the overall shape of the PS frequency response is not changed.
The PS cut-off frequency in VCSELs is much lower than
would be expected from the measured switching time of 10
ns,13,16 and is also much lower than the one encountered in
FIG. 2. Fraction of successful cycles as a function of the modulation amplitude. The VCSEL is biased at PS current of 9.85 mA and modulated with
a 200 kHz sinusoidal signal.
Downloaded 18 Oct 2002 to 192.84.145.98. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
2250
Verschaffelt et al.
Appl. Phys. Lett., Vol. 80, No. 13, 1 April 2002
FIG. 4. Thermal frequency response measured at bias current of 11.1 mA
and modulation amplitude of 0.1 mA 共closed squares兲 and fit to a first-order
response function with a cut-off frequency of 90 kHz 共solid line兲.
edge-emitting lasers 共where 500 MHz PS has been
reported17兲. The low cut-off frequency of 90 kHz observed in
our VCSELs cannot be explained on the basis of existing
models8 –10 that only take much faster time scales into account 共such as photon and carrier lifetimes兲. Thermal effects
are an obvious possibility as an explanation. To test this conjecture, we have evaluated the thermal frequency response of
our VCSEL with a scanning Fabry-Pérot spectrum analyzer
by measuring the width of the modulation-induced shift 共or
chirp ⌬F兲 of the optical frequency emitted as a function of
the modulation frequency. The optical frequency emitted is
indeed determined by the cavity resonance, which in turn
depends on the lattice temperature of the laser cavity. For the
frequency range measured, the chirp ⌬F is mainly due to
Joule heating. The result of this measurement is shown in
Fig. 4. This thermal response corresponds very well to the
transfer function of a linear first-order system, given by
A
⌬F
,
⫽
⌬I 冑1⫹ 共 f / f 0 兲 2
where f is the modulation frequency, f 0 is the cut-off frequency, A is a proportionality constant and ⌬I is the current
modulation amplitude, which is kept constant throughout the
measurement.
When we compare the PS and thermal frequency responses 共see Figs. 3 and 4兲, we notice two main differences.
First, for low frequencies the PS response is not as flat as the
thermal response. Second, the PS frequency response curve
does not decrease as quickly as the thermal one in the high
frequency region. The origin of these differences requires
further investigation. Numerical simulations and a preliminary theoretical analysis indicate that the slope in the PS
response at low frequencies can be explained by switching
events triggered by spontaneous mode hopping. In any case,
both response curves show a cut-off frequency of about 90
kHz, suggesting that thermal effects play an essential role in
PS in these VCSELs. Different thermal mechanisms have
been proposed to explain both type I 共Refs. 2 and 4兲 and type
II 共Ref. 5兲 PS. Type I PS can be understood to be the result of
a thermally induced frequency shift of the cavity modes with
respect to the laser’s gain curve.2 The type II PS studied here
can be explained on a thermal basis by two mechanisms 共or
a combination of them兲. The first takes the temperature dependence of the loss curve into account5 in addition to the
thermal shift of the gain curve. The second relies on lifting of
the gain degeneracy of the two PMs due to strain.6 As a
function of temperature the two resulting gain curves move
with respect to one another. Both mechanisms can explain
thermally induced type II PS and are therefore compatible
with our experimental results.
In conclusion, we have studied the current-driven polarization modulation properties of an 850 nm VCSEL as a
function of both the amplitude and frequency of modulation.
The device exhibits type II polarization switching. We measured the PS frequency response for different values of the
switching current. These frequency responses display a cutoff of about 90 kHz that coincides with the thermal cut-off
frequency of the VCSEL measured, indicating that it is necessary to incorporate a temperature-dependent variable in
models that describe the dynamic polarization behavior of
these VCSELs.
This research was supported in Belgium by FWO, IUAP,
GOA and the OZR at Vrije Universiteit Brussel. Two of the
authors 共G.V. and J.D.兲 acknowledge the Fund for Scientific
Research, Flanders 共FWO兲 for fellowships. The collaboration
between the Brussels and the Florence groups was supported
by EU Programs COST268 and by the RTN network VISTA
共Contract No. HPRN-CT-2000-00034兲. The authors would
like to thank D. and T. Westmalle for stimulating the discussions.
1
M. P van Exter, A. K. Jansen van Doorn, and J. P. Woerdman, Phys. Rev.
A 56, 845 共1997兲, and references therein.
2
K. D. Choquette, D. A. Richie, and R. E. Leibenguth, Appl. Phys. Lett. 64,
2062 共1994兲.
3
J. Martı́n-Regalado, J. L. A. Chilla, J. J. Rocca, and P. Brusenbach, Appl.
Phys. Lett. 70, 3350 共1997兲.
4
K. Panajotov, B. Ryvkin, J. Danckaert, M. Peeters, H. Thienpont, and I.
Veretennicoff, IEEE Photonics Technol. Lett. 10, 6 共1998兲.
5
B. Ryvkin, K. Panajotov, A. Georgievskii, J. Danckaert, M. Peeters, G.
Verschaffelt, H. Thienpont, and I. Veretennicoff, J. Opt. Soc. Am. B 16,
2106 共1999兲.
6
K. Panajotov, B. Nagler, G. Verschaffelt, A. Georgievskii, H. Thienpont, J.
Danckaert, and I. Veretennicoff, Appl. Phys. Lett. 77, 1569 共2000兲.
7
M. B. Willemsem, M. U. F. Khalid, M. P. van Exter, and J. P. Woerdman,
Phys. Rev. Lett. 82, 4815 共1999兲.
8
M. San Miguel, O. Feng, and J. V. Moloney, Phys. Rev. A 52, 1728
共1996兲.
9
S. Balle, E. Tolkachova, M. San Miguel, J. R. Tredicce, J. Martı́nRegalado, and A. Gahl, Opt. Lett. 24, 1121 共1999兲.
10
A. Valle, L. Pesquera, and K. S. Shore, IEEE Photonics Technol. Lett. 9,
557 共1997兲.
11
B. Ryvkin and A. Georgievskii, Semiconductors 33, 813 共1999兲.
12
K. D. Choquette, K. L. Lear, R. E. Leibenguth, and T. M. Asom, Appl.
Phys. Lett. 64, 2767 共1994兲.
13
G. Verschaffelt, J. Albert, M. Peeters, K. Panajotov, J. Danckaert, I.
Veretennicoff, H. Thienpont, F. Monti di Sopra, S. Eitel, R. Hoevel, M.
Moser, H.P. Zappe, and K. Gulden, Proc. SPIE 3946, 246 共2000兲.
14
J. Albert, K. Panajotov, G. Verschaffelt, B. Nagler, H. Thienpont, I.
Veretennicoff, J. Danckaert, S. Barbay, G. Giacomelli, F. Marin, F. Monti
di Sopra, and S. Eitel, Proc. SPIE 4286, 34 共2001兲.
15
G. Giacomelli, F. Marin, M. Gabrysch, K. H. Gulden, and M. Moser, Opt.
Commun. 146, 136 共1998兲.
16
M. B. Willemsem, M. P. van Exter, and J. P. Woerdman, Phys. Rev. Lett.
84, 4337 共2000兲.
17
A. Klehr, R. Müller, M. Voss, and A. Bänwolff, Appl. Phys. Lett. 64, 830
共1994兲.
Downloaded 18 Oct 2002 to 192.84.145.98. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp