Magia
Magia
Magia
Abstract—We report on the wavelength-dependent non- The step-like I/O characteristic (also known as differential
linear-gain properties of a vertical-cavity semiconductor optical gain) may be used to perform all-optical logic regeneration,
amplifier. A step-like output versus input transfer curve and and optical logic, with input optical switching power on the
bistability are found on the long wavelength side of the resonant
peak. The characteristics are caused by a dispersive nonlinearity order of 1 W [10]. It is also an important aspect to consider
arising from gain saturation in the device. The nonlinear transfer in modeling amplifier linearity.
of the amplifier could be used for logic regeneration and all-optical These I/O curves are recognizable as similar to those of
logic operations, while the optical injection power required to FPSOAs with nonlinear-gain media [11], which have been
switch the device is an order of magnitude lower than previously analyzed by Adams [12]. We compare the predictions of that
reported with edge-emitting devices.
analysis to experimental results in Section II. To clarify later
Index Terms—Nonlinearities, optical amplifiers, optical logic de- discussion, the approach is summarized here. A nonlinear index
vices, optical resonators, optical switches. of refraction is assumed, coupled to the strength of the optical
field in the cavity. The optical field is in turn coupled to
I. INTRODUCTION the carrier density through the carrier rate equation. The gain
and single-pass phase change (or detuning) are calculated by
Manuscript received August 8, 2002; revised November 18, 2002. II. RESULTS AND DISCUSSION
The authors are with the ECE Department, University of California at San
Diego, La Jolla, CA 92093-0407 USA (e-mail: [email protected]). To provide experimental verification of these effects in VC-
Digital Object Identifier 10.1109/LPT.2003.809310 SOAs, the optical output versus input optical power is measured
1041-1135/03$17.00 © 2003 IEEE
508 IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 15, NO. 4, APRIL 2003
Fig. 1. Calculated output versus input of VCSOA using parameters in Table I. Fig. 2. Output versus input of 20-m VCSOA showing transition to bistable
Symbols are measured data, solid lines are simulation results. Measured regime. Solid lines connect data points. Arrows show the direction of the
wavelength and detuning parameter used in simulation: (a) = 851:560 nm, hysteresis loops. (a) 842.045 nm, (b) 842.048 nm, (c) 842.052 nm, and
2
= 9:0 10 , (b) = 851:580 nm, = 0:0, (c) = 851:600 nm, (d) 842.055 nm.
0 2
= 7:5 10 , and (d) 851.620 nm, = 13:0 10 . 0 2
onset of bistability with detunings to the long wavelength side
at several wavelengths across the gain window. In this experi- of resonance.
ment, two electrically pumped proton-implanted VCSELs (Em- Several device parameters must be determined in order
core model 8085-1100) are biased below the lasing threshold to accurately compare the predictions of (1) and (2) to the
to be used as amplifiers. The lasing wavelengths are about 850 experimental data. The mirror reflectivities and cavity length are
and 842 nm, and the VCSELs have quantum well gain regions. estimated using a scanning electron microscope cross section
The apertures are 8 and 20 m in diameter, but otherwise the of the laser die. The number of mirror pairs is counted and
devices are identical in structure. The devices are operated in the reflectivities calculated using well-known equations, as in
reflection mode, so the light enters and exits the device through [13]. The confinement factor is typically nearly unity in large
the top mirror. Despite their circular symmetry, the VCSOAs aperture VCSELs [13], so a value of 1.0 is assumed. Then,
exhibit some variation in gain with polarization direction. One the standard equation for Fabry–Pérot gain is used to fix the
polarizer is used to align the polarization of the input light with value of by comparing the calculated gain with measured
the direction corresponding to the maximum gain of the ampli- values. For this calibration step, the gain versus wavelength is
fier. A second polarizer is placed closer to the source laser to measured with very small input powers (100 nW). The effect
control the input light intensity. A tunable laser is adjusted to of the nonlinearity at this injection level is greatly reduced, so
vary the detuning while the VCSOA is stabilized in a temper- the measurement gives an estimate of the zero-order linewidth
ature-controlled mount. A wavemeter (HP 86 120B) is used to and resonant wavelength of the cavity [14]. The bias
monitor the input wavelength. A high quality microscope ob- condition is determined from the measured bias current.
jective lens is used to couple the input light into the amplifier There remain two free variables: and . The latter term
and collect the output. A beamsplitter directs a fraction of the is determined in part by the material gain coefficient and
input and output light into power meters (Newport 818-SL w/ recombination lifetime [12]. These are unknown parameters,
1830-C) for detection. and so is treated as a fitting variable. The value of
The continuous-wave input and output power is measured at scales the results for intensity to the measured values of optical
several wavelengths around the resonant gain peak. The output power, and this scaling is asymmetric because of the unknown
versus input curves for the 8- m device are shown in Fig. 1 with coupling loss to the amplifier. Thus, two values are used, one
a bias 95% of threshold . Note that at short wavelengths the for the input and one for the output . The linewidth
transfer curve is approximately linear because the gain falls off enhancement factor is adjusted to achieve the best fit to the
very quickly with increasing input power. This occurs because shape of the I/O curves measured. The final parameters used
the increasing photon density pushes the gain peak farther away are given in Table I.
from the input wavelength. As the input wavelength is increased, The calculation results are plotted along with the experi-
the nonlinearity becomes more noticeable, and the peak gain mental data points in Fig. 1. Reasonable agreement is found
occurs at nonzero input intensity when the detuning is to the across the wavelength range of interest, and the detunings used
long wavelength side. This shift in peak gain to nonzero input in the simulation match the measured wavelengths within the
power results in the differential gain characteristic. However, accuracy of the wavemeter. The gain is somewhat overestimated
as the wavelength is further increased, the gain saturates before by the theory, particularly for positive detuning (short ). The
the large optical intensity needed to pull the cavity mode to the theoretical curve also is smoother than the measured data. That
input wavelength is reached. The gain saturates too quickly in is, the calculated result cuts the corners of the step in Fig. 1(c).
the 8- m device to allow for a bistable transition, but that is not The comparison suggests that some of the simplifications made
the case with the 20- m VCSOA. The 20- m device was biased should be reconsidered for accurate modeling of VCSOAs. In
closer to threshold (99% ), and also should have a larger sat- particular, the differences may be due to the absence of higher
uration power, due to the larger active volume. Fig. 2 shows the order terms in the carrier rate equation, such as those describing
SÁNCHEZ et al.: NONLINEAR GAIN IN VERTICAL-CAVITY SEMICONDUCTOR OPTICAL AMPLIFIERS 509