Gain dynamics in quantum experimental comparison well lasers and optical amplifiers: An N. Tessler, J. Mark,a) G. Eisenstein, J. Mdrk,a) U. Koren,b) and C. A. Burrusb) Electrical Engineering Department Technion, Haifa 32000, Israel (Received 6 December 1993; accepted for publication 7 February 1994) We describe an experimental comparison of gain dynamics in quantum well lasers and optical amplifiers. We demonstrate an approximately 30% increase in the time constant describing the gain recovery on the -1 ps time scale in a laser above threshold. The increase is due to the high rate of stimulated emission which modifies the relative significance of the various mechanisms contributing to the gain recovery. We suggest an explanation based on the coupling of two processes: Carrier capture and carrier cooling. We conclude that laser gain dynamics contain details that cannot be revealed in experiments on optical amplifiers due to the vast differences in operating conditions between a laser and an optical amplifier. Carrier and gain dynamics are fundamental properties that govern operation of diode lasers. In addition to the obvious interest in the nature of these basic properties, they are important due to their relationship to the laser gain nonlinearity.‘*’ It is well known that the gain nonlinearity dictates3 limitations to the modulation capability as well as .~ many static properties. Gain dynamics are usually studied experimentally in the small signal regime. Namely, the dynamics are measured in experiments which for a particular operating condition, introduce a small gain perturbation to a static operating point. Common experiments include modulation response measurements (by electrical4 or optical5 gain modulation), short pulse pump-probe,6V7and nondegenerate wave mixing.8 Such experiments are interpreted using a variety of models56,7710 that are based on the assumption of a linear behavior. These experiments4-7 can be divided into two groups. Modulation response measurements are performed on lasers and their aim is to extract parameters of the gain nonlinearity.’ Pumpprobe and wave mixing experiments, on the other hand, serve for the identification and characterization of physical mechanisms that contribute to the gain nonlinearity.27’*0 Reported pump-probe and wave mixing experiments have used optical amplifiers’+8 and not lasers. The fact that the two groups of experiments have been performed under fundamentally different operating conditions (stemming from, the differences between lasers and optical amplifiers) causes difficulties in quantifying the implications to gain nonlinearity of the details of the relevant physical processes. The differences between optical amplifiers and lasers operating above threshold result primarily from stimulated emission. Stimulated emission causes: (a) Shortening of the carrier lifetime, (b) a deviation from equilibrium at the lattice temperature of the carrier density distribution, and (c) near pinning of the carrier density and Fermi level. In structures based on a quantum well (QW) gain region, the differences are significantly sharpened since the shortening of the carrier lifetime (in the QWj causes an accumulation of carriers in the barrier and confinement regions.‘r’*’ This effect, known as the carrier injection bottleneck effect, has recently been demonstrated under static injection conditions.r3 This letter reports measurements comparing gain dynamics under different operating conditions, namely: (a) a QW laser above threshold, (b) a QW laser below threshold, and (c) a QW optical amplifier. In order for the comparison to be reliable, all measurements were performed using a single, specially designed, device.14 This ensures that the observed differences are exclusively due to the different operating conditions. The specially designed laser diode is depicted schematically in Fig. 1. It is a two-section distributed Bragg reflector (DBR) laser width, a QW gain region, and an antireflection (AR) coating on the back of the Bragg mirror. The laser has a short gain section and a narrow-band Bragg mirror. The length of the gain and Bragg sections as well as the Bragg coupling coefficient are such that the cavity mode spacing is larger than the Bragg mirror bandwidth. Current injection ‘)Tele Danmark Research, DK-2970 Horsholm, Denmark. “AT&T Bell Laboratories, Holmdel, New Jersey 07733. FIG. 1. Schematic description of the special DBR laser, the cavity modes, and the Bragg filter reflection function. 2050 Appl. Phys. Lett. 64 (16), 18 April 1994 1 --q,y,,,/j I, I ~~~__I__..._ Bragg Reflection Cavity / i,=5mA 0003-6951/94/64(16)/2050/3/$6.00 \ I,=0 Mode h Q 1994 American Institute of Physics Downloaded 27 Jan 2004 to 132.68.1.29. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp 1 1,=4mA (b) I,=4mA t t -77-77 -2 3 b 3 2 Laser 0 0 IO. Current 20 (mA) - I,=0 - II 1545 1555 Wavelength 1565 (nm) FIG. 2. Measured L-Z curves (a) and emission spectra (b) for the laser and the amplifier. The emission spectra were measured for a bias level of 20 mA. into the Bragg section causes the reflection to tune away from the cavity mode (without reaching the next mode). Consequently, the two conditions needed for oscillations (proper round-trip phase accumulation and gain loss balance) are not satisfied at the same wavelength and the laser operates as an optical amplifier.14 The AR coating on the Bragg side prevents multiple reflections at wavelengths outside the Bragg mirror bandwidth. Measured L-1 curves and spectra under the two different operating conditions are shown in Fig. 2. The spectra are shown near 1550 nm where we note that the power contained in the laser oscillating line is reduced by four orders of magnitude when the laser is turned into an optical amplifier. The spectra near 1490 nm are not shown but they contain no ripples since the output reflectivity is close to zero outside the Bragg reflection region. Gain dynamics in the special device were measured using a cross polarized pump-probe technique. Pump-probe measurements in a laser require special care in the choice of pulse energy which must be kept extremely low in order for the device to remain in the linear regime. The reason for the strong limitation on pulse energy lies in the inherent nonlinearity of the laser oscillation which takes place simultaneously with the pulse propagation in the cavity. The pump and probe energies in the present experiment were 40 and 4 fJ, respectively. The pulse wavelength is chosen to be 1490 and 60 nm shorter than the DBR laser wavelength. This ensures that in the laser case, the dynamics are examined sufficiently far from the oscillating line so as to not be affected by the spectral hole caused by the DBR laser (since spectral hole burning has a limited spectral extent2). It also ensures that the output (Bragg section) reflectivity is approximately zero. The experimental setup contains a standard cross polarized pump-probe apparatus. The pulses used were generated by an additive pulse mode-locked (APM) color center lasersV7 and are of 120 fs duration. In order to reduce the background light generated by the DBR laser, we used, in addition to the usual polarizer, a 20-nm-wide optical filter centered around the probe pulses, at the laser output (before the detection Appl. Phys. Lett., Vol. 64, No. 16, 18 April 1994 Amplifier - I,=5 i 0 2 Pump 4 6 Probe 8 0 2 Delay 4 6 8 ips) FIG. 3. Measured gain evolution for the laser (a) and the optical amplifier (b). system). The output probe signal was measured using a standard lock-in technique. ?tyo sets of pump-probe measurements are described in Fig. 3. In the first, described in Fig. 3(a), the Bragg contact was grounded so the device operating condition was that of a laser. In the second, described in Fig. 3(b), a 5 mA bias was injected into the Bragg section changing to an optical amplifier operating condition. For each case, the gain dynamics were measured for several drive currents to the gain section ranging from 4 to 25 mA. The laser threshold current was 8 mA. Although the APM laser was tuned 60 nm shorter than the laser oscillating wavelength, all but the curve for the 4 mA case, exhibit a decrease in the long lived gain, typical of operation in the gain regime.6T7The time constant describing the gain recovery during the first few ps was extracted for each of the measured traces, some of which are shown in Fig. 3. In order to obtain accurate time constants, it was necessary to extend the measurements to a time delay of at least 30 ps. The time constant was extracted iu the following manner: First the final value was subtracted. Next, the subtracted curves were displayed on a logarithmic scale where an exponential recovery appears as a straight line whose slope represents the recovery time constant. The data were sufficiently clear to identify a single time constant in the 1 ps time scale. The dependence of this time constant on operating condition is obtained from its bias dependence in both the optical amplifier and laser modes of operation. The results are depicted in Fig. 4. The laser L-f curve is superimposed on Fig. 4 for reference. At low currents, below threshold, the laser and amplitude exhibit essentially the same time constant -0.7 ps, consistent with Refs. 1, 6, and 7. The fact that the time constants are similar assures that there are no artifacts produced by the current injection into the Bragg section. At higher currents, the time constant of the amplifier remains essentially unchanged while in the laser case, it increases sharply above threshold to -0.9 ps. This result shows that the high rate of stimulated emission in the laser above threshold slows down the gain recovery during the first one picosecond following the gain perturbation. This is Tessler et a/. 2051 Downloaded 27 Jan 2004 to 132.68.1.29. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp Laser i I I \ /’ Amplifier I’ I,=5mA /! 0 10 Current 20 (mA) FIG. 4. Gain recovery time constant for a laser and an amplifier. The laser L-I curve is superimposed on the plot for reference. most important since it suggests that the nonlinear gain coefficient is power dependent. The limited accuracy of the measurement (seen by the displayed error bars) prevented us from observing this bias dependence above threshold. Only the large differences between below and above threshold conditions are clearly observable as seen in Fig. 4. The results shown in Fig. 4 can be explained by considering the local carrier capture phenomenon associated with the gain recovery. There are two possible carrier capture mechanisms; phonon mediated capturei and carrier-carrier scattering mediated capture.‘5.16 In the case of phonon mediated capture, no extra heat is producedr’ and the capture process simply serves as an additional channel for carrier cooling. This increases the carrier density (and hence the gain) at the probe wavelength and the measured time constant has some effective value that couples the two processes. Since these two processes have very similar time constants, they are inseparable in single wavelength pump-probe measurements. Above threshold, there is an increase of carriers in the barrier and confinement (3D) states due to the injection bottleneck effect.‘“‘3 This increases the relative contribution of the capture process. If the gain recovery time that is associated with capture is larger than the cooling time, the increase in density of three-dimensional carriers results in an 2052 Appl. Phys. Lett., Vol. 64, No. 16, 18 April 1994 increase of the effective (measured) recovery time. In the case of carrier-carrier scattering mediated capture, the relevant time constant may be as short as a few tens offs (on the scale of the spectral hole burning time constant). In this case however, the capture process produces extra heat” and slows down the cooling process.17 Since the carriercarrier scattering mediated capture time is density dependent, the relative importance of the two capture mechanisms depends on the carrier density at threshold. The effective recovery time constant is expected to increase therefore for either capture mechanism, as demonstrated in Fig. 4. To conclude, we have shown that laser gain dynamics are unique and contain details that cannot be revealed by studying optical amplifiers. The fast gain recovery time (-1 ps) was found to increase by -30% above threshold. This result is in good agreement with static gain spectrum measurements under similar conditions.13 It suggests that the change in the time constant is indeed related to the carrier capture phenomenon and the injection bottleneck associated with it. However, further study is needed to clearly verify and quantify the exact role of the carrier capture process. ‘M. P. Kcsler and E. P. Ippen, Appl. Phys. Len. 51, 1765 (1987). ‘M. Wileatzen, A. Uskov, J. M&k, H. Olesen, B. Tromborg, and A. P. Jauho, IEEE Photon. Tech. Lett. 3, 606 (1991). 3 I. P. Kaminow and R. S. Tucker, Guided-Wave Optoelectronics, edited by T. Tamir (Springer, Berlin, 1988). p. 211. 4J. E. Bowers and M. A. Pollack, Optical Fiber Telecommunications IZ? edited by S. E. Miller and I. P. Kaminow (Academic, San Diego, 1988), p. 509. 5C. H. Lange and C. B. Su, Appl. Phys. Lett. 55, 1704 (1989). “K. L. Hall, G. Lenz, E. P. Ippen, U. Karen, and G. Raybon, Appl. Phys. Lett. 61, 2512 (1992). 7J. Mark and J. M&k, Appl. Phys. Lett. 61, 2281 (1992). ‘K. K&u&i, M. Kakui, C. E. Zah, and T. P. Lee, IEEE J. Quantum Elec-~ tron. 28, 151 (1992). ‘S. Murata, K. Kaniwae, J. Shimizu, M. Nido, A. Tomita, and A. Suzuki, Electron. Lett. 28, 1456 (1992). “B. N. Gomatam and A. P. DeFonzo, IEEE J. Quantum Electron. 26, 1689 (1990). ” N. Tessler and G. Eisenstein, IEEE J. Quantum Electron. 29, 1586 (1993). ‘aW. F. Shartin, J. Schlafer, W. Rideout, B. Elman, R. B. Lauer, J. Lacourse, and F. D. Crawford, IEEE Photon. Tech. Lett. 3, 193 (1991). 13N. Tessler, R. Nagar, G. Eisenstein, S. Chandrasekhar, C. H. Joyner, A. G. Dentai, U. Koren, and G. Raybon, Appl. Phys. Lett. 61, 2383 (1992). 14M Margalit, R. Nagar, N. Tessler, G. Eisenstein, and M. Grenstein, Opt. Lit. 18, 610 (1993). “P. W. M. Blom, J. E. M. Haverkort, P. J. Van Hall, and J. H. Walter, Appl. Phys. Lett. 62, 1490 (1993). 16M. Svendsen, J. M&k, and H. Haug, Phys. Rev. B 49 (1994). r7L. F. Lester and 8. K. Ridley, J. Appl. Phys. 72, 2579 (1992). Tessler et a/. Downloaded 27 Jan 2004 to 132.68.1.29. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
© Copyright 2026 Paperzz