Content uploaded by Daan Lenstra
Author content
All content in this area was uploaded by Daan Lenstra
Content may be subject to copyright.
Ultrafast all-optical wavelength conversion using nonlinear
polarization rotation in a semiconductor optical amplifier
A.K. Mishraa, X. Yanga, D. Lenstraa, b, G.D. Khoea and H.J.S. Dorrena
aCOBRA Research Institute, Eindhoven University of Technology,
P.O. Box 513, 5600 MB, Eindhoven, the Netherlands
bVrije Universiteit, FEW, Department of Physics and Astronomy,
de Boelelaan 1081, 1081 HV, Amsterdam, the Netherlands
Tel: (31) 40 2475479, Fax: (31) 40 245 5197, e-mail: a.k.mishra@tue.nl
ABSTRACT
We demonstrate wavelength conversion based on nonlinear polarization rotation driven by ultrafast carrier relaxation in
an InGaAsP-InGaAs multi-quantum-well (MQW) semiconductor optical amplifier (SOA). A continuous wave probe
beam at a center wavelength of 1555 nm, and a control pulse of duration of 120 fs (FWHM) at a center wavelength of
1520 nm were used. We have presented wavelength conversion for different injection currents and for different control
pulse energies. The conversion efficiency of 12 dB was obtained for the control pulse energies of 10 pJ.
I. INTRODUCTION
Nonlinear phenomena in semiconductor optical amplifiers (SOAs) such as cross-gain modulation, cross-phase
modulation and four-wave mixing have been widely utilized for wavelength conversion and optical switching.
Wavelength conversion based on nonlinear polarization rotation in SOAs is presented in [1-5]. This concept has been
described in [1] using optical pulses with duration of 47 ps. In brief, this kind of wavelength conversion is caused by
polarization-dependent gain saturation, and a polarization-dependent nonlinear index change is introduced by pump
light in the SOA [5]. Thus, the pump beam creates additional birefringence in the SOA, which makes the polarization
angle of the probe light rotated while propagating through the SOA. It has been shown in [5] that this concept can lead
to error free inverted and non-inverted wavelength conversion at a bit rate of 10 Gbit/sec.
In this paper, we investigate wavelength conversion driven by optical pulses with duration of 120 femtosconds in a
nonlinear polarization switch configuration. Using sub-picosecond optical pulses with a high peak power, nonlinear
gain and index dynamics driven by two-photon absorption (TPA) and free-carrier absorption (FCA) plays an important
role in the amplifier recovery. We investigate experimentally the wavelength conversion efficiency as a function of the
injection current and the pulse intensity. We examine numerically, the amplifier recovery using the SOA model
presented in [6]. We discus differences between the ultrafast recovery of a wavelength converter based on cross-phase
modulation in a semiconductor optical amplifier placed in an asymmetric Mach-Zehnder interferometer and a
wavelength converter based on nonlinear polarization rotation. It is shown numerically, that wavelength converters
based on nonlinear polarization rotation, allow full recovery of the switch at sub-picosecond timescales.
II. EXPERIMENTS AND RESULTS
The scheme of our wavelength converter is shown in Figure 1. The SOA used in this experiment has active layers of
MQW InGaAsP-InGaAs with a central length of 750 µm and at both sides a taper zone of 400 µm. A beam of optical
pulses with duration of 120 fs (FWHM) at a central wavelength of 1520 nm were generated by an optical parametric
oscillator (OPO) at a repetition rate of 75.82 MHz. The OPO output was attenuated using a half-wave plate (HW-1) and
a polarizer. A second half-wave plate (HW-2) was used to set the polarization of the laser beam to the TE mode. A
tunable laser emits a continuous wave (CW) probe beam at wavelength 1555 nm. A variable attenuator was used to
control power of probe beam and a polarization controller (PC-1) controls the polarization state. The pump and probe
beam were combined by a beam splitter (BS-1) and fed into the SOA by using microscope objectives. After passing
through PC-2, the probe beam was filtered by a band pass filter (BPF) at the SOA output. The BPF with a bandwidth of
1 nm was used to remove the pump light and also suppressed the amplified spontaneous emission generated by the
SOA. The transmitted light through the polarizing beam splitter (PBS) was measured.
In the first experiment, the polarization-dependent gain of the SOA was measured as a function of pump pulse energy.
The SOA injection current was 200 mA. The results are shown in Figure 2, in which the amplification for TE and TM
modes are plotted as a function of the injected pulse energy. The curve with the maximum amplification is attributed to
the TE mode and the curve with the minimum amplification is attributed to the TM mode. The solid lines (dashed lines)
in Figure 2 represent the computed amplifications for the TE (TM) mode, while the diamond-shaped (star-shaped)
points represent the measured data [8]. We corrected the coupling and component losses that were estimated to be 12.0
dB, which includes two times 3.0 dB facet losses and 6.0 dB for the components used in the experimental setup. If we
increase the pulse energy to 8.6 pJ, the gain of the corresponding modes drops down to -3.1dB for the TE mode and
–4.1 dB for the TM mode. It follows from Figure 2 that the experimental and numerical results are in good agreement.
OPO
PBS PC-2 BS-2
BS-1
M
M
M
HW-2
A-2
A-1
PHW-1
PC-1
Figure 1: Experimental implementation of the
nonlinear polarization switch. OPO: optical
parametric oscillator, HW: half-wave plate, P:
polarizer, M: mirror, PBS: polarizing beam-splitter,
BS: beam-splitter, L: lens, A: attenuator, BPF: band-
pass filter, PC: polarization controller, CW:
Continuous wave tunable laser.
0 1 2 3 4 5 6 7 8 9
-5
0
5
10
15
20
Pulse Energy (pJ)
Amplification (dB)
TE
TM
Figure 2: Measured and computed polarizations
dependent gain for the TE and TM modes as a function
of the pump pulse energy.
Wavelength conversion was realized in this set-up by setting the linear polarization of the probe beam by PC-1
approximately 450 with respect to the SOA active layers. PC-2 was adjusted so that initially no light can pass through
the PBS. The whole set-up was placed in a box to shield the polarization switch from thermal and mechanical
disturbances. When saturating pump pulses were injected in the SOA, the polarization-dependent gain saturation will
lead to pump-induced birefringence. The pump-induced birefringence makes the TE component of the probe beam
experiences a different refractive index compared to the TM component of the probe beam, and thus a phase difference
between the TE and TM modes of the probe signal which causes a rotation of its polarization state [1,2]. As a
consequence, the powermeter could detect some probe light passed through the PBS. This means that at the PBS
output, the pump pulse was converted to the wavelength of the probe light and the rotation of the polarization state was
obtained by measuring the transmission through the PBS. The discrete points in Figure 3 show the observed PBS output
for various pump pulse energies while the SOA injection current is 200 mA and the average power of the CW probe
beam is 3 dBm. The solid and dashed lines represent computed results, based on the model, a conversion efficiency
larger than 12 dB was obtained for pulses with energy of 10 pJ. It is clearly visible that our SOA model leads to results
that are in good agreement with the experimental data. This experiment was repeated in the case probe power of 0 dBm
and we found similar results.
0 2 4 6 8 10 12
-38
-36
-34
-32
-30
-28
-26
-24
-22
Inj ected Pump E nergy (pJ)
Output Power in (dBm)
Figure 3: Measured and computed output power of the
nonlinear polarization switch as a function of pump
pulse energy.
20 40 60 80 100 120 140 160 180 200
-50
-45
-40
-35
-30
-25
-20
Injected Current in SOA (mA)
Output Power in (dBm)
Figure 4: Measured and computed output power of the
nonlinear polarization switch as a function of the SOA
injection current.
We have also investigated wavelength conversion as a function of the injection current for different pump pulse
energies. The power of the CW probe beam was 3 dBm. The results are shown in Figure 4. The diamond-shaped points
represent the results for pump energies of 10 pJ and the star-shaped points represent the results for pump pulse energies
of 6.3 pJ. The solid and dashed lines represent computed results for pump pulse energies of 10 pJ and 6.3 pJ
respectively. Based on the model, it is observed that the averaged converted power of the light that passes through the
PBS increases as a function of current. If we account for current dependency on population imbalance factor (0.6~0.8)
for the currents (0~200 mA), our experimental results are in good agreement with the computational results for the
current above transparency current (50mA).
The expression for the )(t
NL
φ
∆is the pump-induced nonlinear phase difference between the TE and TM modes per
unit length which can be expressed as
)]()([
)( tgtg
z
tTM
CW
TE
CW
NL −=
∂
∆∂
α
φ
. (1)
Here, α is the linewidth enhancement factor and )(tgTE
CW and )(tgTM
CW represent the gain that accounts for TPA and FCA.
Note that Eq. (1) differs from its counterpart in [9] since in Eq. (1) both modes propagate through the same SOA, the
contribution to the nonlinear phase shift due to TPA is canceled out. As a result, the operation of a nonlinear
polarization switch operated by femtosecond optical pulses differs fundamentally from a similar functionality based on
nonlinear gain and index dynamics of a SOA placed in a Mach-Zehnder interferometer [6]. Figure 5a shows a
simulation of the nonlinear phase shift as a function of the time. It follows from Figure 5 that the nonlinear phase shift
)(t
NL
φ
∆ has a long-lived tail that is much smaller than 0.1 radians. However, since the PBS output power is
proportional to the cosine of the nonlinear phase shift, the effect of the long-lived tail has vanished in the PBS output
power. This is visible in Figure 5b, which shows a simulation of the pulse that outputs the nonlinear polarization
switch. Figure 5 also shows that the nonlinear phase shift recovers in 500 fs so that the duration of the pulse that
outputs the nonlinear polarization switch is also approximately 500 fs (FWHM).
00.5 11.5
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
Time(ps)
Phase difference(radian)
(a)
00.5 11.5
0
1000
2000
3000
4000
5000
6000
7000
Time(ps)
Transmitted Out Put (photon/um3)
(b)
Figure 5: a): Computed nonlinear phase shift )(t
NL
φ
∆
as a function of time using Eq.(1). b): Computed pulse
transmission through the PBS as a function of time. In both cases, the pump pulse energy was 10 pJ and the power of
the CW probe light was 3 dBm. The SOA injection current was 200 mA.
III. CONCLUSIONS
We have discussed wavelength conversion using a nonlinear polarization switch that is driven with optical pulses with
duration of 120 fs and demonstrated the static conversion efficiency greater than 12 dB. We have also shown that the
operation of a nonlinear polarization switch differs from the operation of a nonlinear optical switch based on a SOA
placed in a Mach-Zehnder interferometer. This is due to both TE and TM modes propagate through the SOA, in
contrast to a Mach-Zehnder interferometer where only TE mode of the probe beam propagates through the SOA [7]. It
was argued in [10] that the nonlinear phase shift contains two contributions, one due to the phase shift introduced by
the carrier depletion and the other due to the direct nonlinear phase shift introduced by TPA. Our model reveals that
there is a direct effect of the modulated pump light on the CW probe beam due to cross TPA modulation. Since in a
nonlinear polarization switch both the TE and TM modes propagate through the same SOA, the direct contribution due
to TPA is cancelled out. Due to a precise cancellation, this effect plays no role in our present switching configuration.
This implies that the width of the pulse that outputs the nonlinear polarization switch only depends on the nonlinear
carrier dynamics in the SOA. On the other hand, there is an indirect and much slower effect due to ultrafast nonlinear
index and carriers dynamics that is driven by TPA and FCA, which does play an important role to realize the ultrafast
wavelength conversion here demonstrated.
ACKNOWLEDGEMENTS
This work was supported by the Netherlands Organization for Scientific Research (NWO), the Technology Foundation
STW and the Ministry of Economic Affairs through respectively the NRC Photonics grant and the Innovational
Research Incentives Scheme programme.
REFERENCES
[1] M. F. C. Stephens, M. Asghari, R. V. Petty, and I. H. White, IEEE Photon. Technol. Lett., 9 (1997) 449.
[2] H. Soto, D. Erasme, and G. Guekos, IEEE Photon. Technol. Lett., 11 (1999) 970.
[3] H. J. S. Dorren, et al., J. Quantum Electron., 39 (2003) 141.
[4] R. J. Manning, A. Antonopoulos, R. Le Roux, and A. E. Kelly, Electron. Lett., 37 (2001) 229.
[5] Y. Liu, et al., IEEE Photon. Technol. Lett., 15 (2003) 90.
[6] X. Yang, D. Lenstra, G. D. Khoe and H.J.S. Dorren, Optics Communications, 223 (2003) 169.
[7] X. Yang, et al., Accepted for publication in Optic Comm., 2004.
[8] J. Mark and J. M∅rk, Appl. Phys. Lett., 61 (1992) 2281.
[9] H. Kuwatsuka, T. Simoyama, and H. Ishikawa, IEEE J. Quantum Electron., 35 (1999) 1799.
[10] H. J. S. Dorren, et al., IEEE Photon. Technol. Lett., 15 (2003) 792.