Abstract
We investigate the influence of thermal effects on the highspeed performance of 1.3μm InAs/GaAs quantumdot lasers in a wide temperature range (5–50°C). Ridge waveguide devices with 1.1 mm cavity length exhibit small signal modulation bandwidths of 7.51 GHz at 5°C and 3.98 GHz at 50°C. Temperaturedependent Kfactor, differential gain, and gain compression factor are studied. While the intrinsic dampinglimited modulation bandwidth is as high as 23 GHz, the actual modulation bandwidth is limited by carrier thermalization under continuous wave operation. Saturation of the resonance frequency was found to be the result of thermal reduction in the differential gain, which may originate from carrier thermalization.
Keywords:
Molecular beam epitaxy; Temperature; Modulation; Quantumdots; Semiconductor lasersIntroduction
Hightemperature stability in laser operation is an essential characteristic required for the longwavelength semiconductor lasers in optical communication systems. Realization of uncooled highspeed operation of 1.3μm quantumdot (QD) lasers has attracted intensive research interests due to its application in optical communication. Over the past decade, promising dynamic properties of QDs such as large differential gain, high cutoff frequency, and small chirp were reported in devices with emission wavelength less than 1.2 μm [1]. Improved temperature characteristics of QD lasers, such as temperatureinvariant threshold current [1], high characteristic temperature (T_{o}) [2], and linewidth enhancement factor [3], have been realized through pdoping technique. However, quantumdots emitting at 1.3 μm and above have not fulfilled the initial expectation of improved temperatureinsensitive modulation bandwidths, which have largely remain below 12 GHz [4]. With the increase in QD size and the strain effect of the cap layer, the selfassembled (SA) InAs/GaAs QDs can emit at 1.3 μm. The energy levels are still discrete. However, the number of energy level increases and the level separation, especially for holes, becomes much narrower (8–11 meV for hole) than that in the shortwavelength QDs. This results in significant hole thermalization [5]. Other problems reported in the 1.3μm SA QDs include the finite GaAs barrier and thin wetting layer [6]. These disadvantages consequently lead to the temperaturesensitive performance observed in 1.3μm QD lasers, such as the low characteristic temperature at or above room temperature [7] and strong temperaturedependent maximum gain. Fiore et al. [8] have studied the effects of intradot relaxation on the Kfactor and differential gain of quantumdot lasers. Deppe et al. [9] have reported the role of density of states, especially thermalization of holes due to their closely spaced discrete energy levels. This limits the modulation speed of QDs with deep confinement potentials such as the 1.3μm InAs/GaAs QDs. Many theoretical [8,9] and experimental [3,10,11] investigations have been performed to study the bandwidth limitations in longwavelength QD lasers. According to these investigations, Kfactor [8,11] has been recognized to be one of the limiting factors for the modulation bandwidth of QD lasers, which accounts for the effect of photon lifetime, differential gain, and nonlinear gain compression factor. Despite the theoretical and experimental investigations on the effect of differential gain on the DC performance of 1.3μm QD laser and directly modulated uncooled 1.3μm QD laser [12], the effect of carrier thermalization on the highspeed performance of 1.3μm QD laser has not been analyzed systematically. Obviously, the modulation speed (or bandwidth) of the 1.3μm QD lasers should be temperaturedependent due to the temperaturesensitive gain profile of QDs. As there are few investigations on the effect of temperature on the bandwidth of 1.3μm QD lasers, a study on this aspect will provide greater understanding on the differential gain and carrier dynamics in longwavelength QD lasers.
In this paper, we investigate the influence of thermal effects on the highspeed modulation characteristics of 1.3μm InAs/GaAs QDs by studying the temperaturedependent small signal modulation behavior. The effects of temperature on the Kfactor, differential gain, and nonlinear gain compression will be presented here.
Experimental Details
The tenlayer selfassembled InAs/GaAs QD laser structure, as shown in Figure 1, was grown on GaAs (100) substrate by molecular beam epitaxy (MBE). The structure consists of QD active region sandwiched between two 1.5μm C and Sidoped Al_{0.35}Ga_{0.65}As cladding layers. The active layer comprises 2.3 monolayer (ML) of InAs QDs capped by a 5nm In_{0.15}Ga_{0.85}As layer. A 33nm GaAs layer is used to separate the two QD layers [13]. The wafer was processed into 4μmwide ridge waveguide (RWG) lasers by standard photolithography process and wet chemical etching at room temperature (RT) [14]. Ridge height of approximately 1.3 μm was obtained before the pulsed anodic oxidation (PAO) process. A 200 ± 5 nmthick oxide layer was formed by PAO method, whose experimental setup can be found in [15]. Subsequently, pcontact layers (Ti/Au, 50/300 nm) were deposited by electron beam evaporation, while ncontact layers (Ni/Ge/Au/Ni/Au, 5/20/100/25/300 nm) were deposited on the backside of the substrate following lapping down to ~100 μm. Finally, the wafer was cleaved into laser bars and the cleaved facets were left uncoated. The devices were mounted pside down on a heat sink for measuring the small signal modulation characteristics. The small signal modulation response of the QD lasers was measured under continuous wave (CW) biasing condition using a vector network analyzer (VNA), a highspeed photoreceiver and laser diode current source. A thermoelectric temperature controller was used to regulate and monitor the device temperature during measurements.
Figure 1. The schematic diagram of the InAs/GaAs tenlayer QD laser structure.
Results and Discussions
The measured CW PowerCurrent performance of a device with cavity length of 1.1 mm shows that the threshold current (I_{th}) and slope efficiency are 55 mA and 0.27 W/A at room temperature, respectively. Maximum output power of 96 mW occurred at injection current of 395 mA. Figure 2 shows the lasing spectrum of the laser device under injection current of 100 mA at RT for verification. The lasing wavelength is centered at 1,306.5 nm. Furthermore, no lasing at excited state was observed. Characteristic temperature T_{o} is around 41 K from 5 to 50°C. The small signal modulation response under different injection current levels is shown in Figure 3. At room temperature, the highest bandwidth of 6.1 GHz was obtained at injection current level of 390 mA. For injection current more than 395 mA, the resonance frequency f_{r} decreases with increasing injection current. This is because, there are two competing factors affecting the resonance frequency: (1) increase in resonance frequency with injection current and (2) decrease in resonance frequency due to internal heating. Therefore, when injection current increases higher than 395 mA, the internal heating resulted from the increased current becomes dominant and leads to the decrease in resonance frequency. The small signal modulation response was further fitted into a transfer function that accounts for the intrinsic response of the laser as well as the extrinsic effects [16]:
Figure 2. The lasing spectrum from the InAs/InGaAs QD laser (4 × 1,100 μm^{2}) with injection current of 100 mA at RT.
Figure 3. Small signal modulation response measured at RT under different injection current levels.
where f_{r} is the resonance frequency, γ is the damping rate, and f_{p} is the parasitic cutoff frequency. From the fitting, we obtained values of the damping rate γ and resonance frequency f_{r} at different bias currents. The parasitic cutoff frequency is almost temperatureindependent and only restricts the bandwidth minimally. According to the plot of f_{r} vs. the square root of the normalized bias current (I – I_{th})^{1/2}, the slope (known as Dfactor or modulation efficiency) is obtained to be 0.28 GHz/mA^{1/2} at RT. The relationship between resonance frequency and damping rate defines the Kfactor, which is 0.83 ns at RT. Furthermore, the Kfactor is directly related to the dampinglimited bandwidth (f_{3dB, damping}) by [16]:
The internal quantum efficiency (η_{i}) and internal optical loss (α_{i}) of the devices were estimated to be 51% and 4 cm^{1} by measuring lasers with different cavity length (1–3 mm) [17,18]. The internal quantum efficiency and internal optical loss exhibit weak dependency on temperature. With the values of internal quantum efficiency and internal optical loss, the differential gain (dg/dn) and nonlinear gain compression factor (ε) are extracted. The gain derivatives with respect to the carrier population defines differential gain, while the nonlinear gain compression factor is used to describe the gain dependence on the photon density. From the value of the Dfactor, the differential gain is obtained to be 11.1 × 10^{15} cm^{2} at RT, which is almost ten times higher than that reported in literature [19] (differential gain of 1 × 10^{15} cm^{2} at 300 K for a device emitting at 1,263 nm). The nonlinear gain compression factor is determined to be 12 × 10^{16} cm^{3} at RT. Note that the results are different from that reported recently [20]. We believe that the differences are due to the different device dimensions considered since the performance depends strongly on the device dimensions [21,22].
Measurements of direct small signal modulation of the QD laser were carried out from 5 to 50°C. Figure 4 shows the maximum measured (triangles) bandwidth (f_{3dB, measured}) as function of temperature. The maximum measured bandwidth decreases almost linearly with temperature as temperature increases from 5 to 50°C. The highest f_{3dB, measured} of 7.51 GHz occurred at 5°C. The Dfactor is 0.36 GHz/mA^{1/2} at 5°C and 0.15 GHz/mA^{1/2} at 50°C as shown in Figure 5 (solid circles). The differential gain from 5 to 50°C decreases following increase in temperature as shown in Figure 6. Figure 7 shows the calculated Kfactor of the QD laser as function of temperature. There is a significant increase in the Kfactor as temperature increases. The calculated Kfactor increases approximately by a factor of three over the temperature range of 5–50°C.
Figure 4. Calculated thermal (squares) and dampinglimited (circles) bandwidth (BW) and plot of the measured (triangles) bandwidth at different temperatures.
Figure 5. The dependence of Dfactor (solid circle) and nonlinear gain compression (hollow circle) on temperature.
Figure 6. The differential gain at different temperatures.
Figure 7. Plot of temperaturedependence of Kfactor.
From (2), the f_{3dB,damping} of the QD laser is 23 GHz at 5°C and 8.9 GHz at 50°C, which is limited by the carriercapture time and modal gain via the Kfactor [11,23]. This shows that the 1.3μm InAs/GaAs QD lasers can potentially operate at very high frequencies. However, our experimental data shows much lower bandwidth. This can be attributed to the thermal effects, i.e., the thermal saturation of photon number (S_{o}) at rollover current injection due to the selfheating. It could also be caused by serious hole thermalization due to the closely spaced hole levels in 1.3μm InAs QDs. Considering the dependence of bandwidth on resonance frequency, this suggests that the saturation of the bandwidth is caused by saturation of the photon density. Saturation of the photon density could possibly be caused by the strong gain compression. Meanwhile, the nonlinear gain compression factor is in the order of 10^{16} cm^{3} and shows a relatively weak dependence on temperature [refer to Figure 5 (hollow circle)]. The ε · S_{o} product is less than 0.1, which suggests that the effect of the gain compression on the resonance frequency is relatively small. At relatively small damping effects, the thermallimited bandwidth (f_{3dB,thermal}) is related to f_{r} by [16]:
where f_{r, max} is the maximum resonance frequency at a constant temperature. The f_{r, max} of 6.6 GHz at 5°C and 2.5 GHz at 50°C would give a thermallimited bandwidth of 10 GHz and 3.9 GHz (squares in Figure 4), respectively. This suggests that the main limitation on the bandwidth might be due to the decrease in differential gain, which may result from the thermal effects related to carrier thermalization in the multistack quantumdots. The origin of the temperaturedependent differential gain is currently under investigation. The incorporation of ptype modulation doping and tunnel injection might be useful to improve the QD laser performance by reducing the thermal effects.
Finally, the calculated intrinsic dampinglimited bandwidth (squares) and thermallimited bandwidth (circles) are shown in Figure 4 in comparison with the experimental results f_{3dB, measured} (triangles). The thermallimited f_{3dB,thermal} is in close agreement with the experimental results, indicating that the bandwidth measured in this study was limited by thermal effects.
Conclusion
In conclusion, we have studied the influence of thermal effects on the small signal modulation characteristics of undoped InAs/GaAs QD lasers. The role of temperaturedependent differential gain and nonlinear gain compression factor in determining the frequency bandwidth was investigated. Calculation of the temperaturedependent bandwidth of the undoped QD laser shows close agreement between the thermallimited bandwidth and the measurement results. The bandwidth of the undoped InAs/GaAs QD lasers is mainly limited by thermal effects, which may result from carrier thermalization in the undoped QD laser structure.
Acknowledgements
The authors would also like to acknowledge the assistance of Dr. Ngo Chun Yong and Dr. Loke Wan Khai for their useful inputs to this research.
References

Fathpour S, Mi Z, Bhattacharya P:
IEEE Photonics Technol Lett. 2005, 17:2250. Publisher Full Text

IEEE Photonics Technol Lett. 2002, 14:1231. Publisher Full Text

Cong DY, Martinez A, Merghem K, Ramdane A, Provost JG, Fischer M, Krestnikov I, Kovsh A:
Appl Phys Lett. 2008, 92:191109. Publisher Full Text

Kim SM, Wang Y, Keever M, Harris JS:
IEEE Photonics Technol Lett. 2004, 16:377. Publisher Full Text

Appl Phys Lett. 2002, 80:2758. Publisher Full Text

IEEE J Quantum Electron. 2008, 44:879. Publisher Full Text

Mukai K, Nakata Y, Otsubo K, Sugawara M, Yokoyama N, Ishikawa H:
IEEE J Quantum Electron. 2000, 36:472. Publisher Full Text

IEEE J Quantum Electron. 2007, 43:287. Publisher Full Text

Deppe DG, Huang H, Shchekin OB:
IEEE J Quantum Electron. 2002, 38:1587. Publisher Full Text

Gyoungwon P, Shchekin OB, Deppe DG:
IEEE J Quantum Electron. 2000, 36:1065. Publisher Full Text

Ishida M, Hatori N, Akiyama T, Otsubo K, Nakata Y, Ebe H, Sugawara M, Arakawa Y:
Appl Phys Lett. 2004, 85:4145. Publisher Full Text

Todaro MT, Salhi A, Fortunato L, Cingolani R, Passaseo A, De Vittorio M, Della Casa P, Ghiglieno F, Bianco L:
IEEE Photonics Technol Lett. 2007, 19:191. Publisher Full Text

Cao Q, Yoon SF, Liu CY, Ngo CY:
Nanoscale Res Lett. 2007, 2:303. Publisher Full Text

Wang R, Tong CZ, Yoon SF, Liu CY, Zhao HX, Cao Q:
IEEE Electron Device Lett. 2009, 30:1311. Publisher Full Text

Yuan S, Jagadish C, Kim Y, Chang Y, Tan HH, Cohen RM, Petravic M, Dao LV, Gal M, Chan MCY, Li EH, O JS, Zory PS Jr:
IEEE J Sel Top Quantum Electron. 1998, 4:629. Publisher Full Text

Kirstaedter N, Schmidt OG, Ledentsov NN, Bimberg D, Ustinov VM, Egorov AY, Zhukov AE, Maximov MV, Kopev PS, Alferov ZI:
Appl Phys Lett. 1996, 69:1226. Publisher Full Text

IEEE J Quantum Electron. 2006, 42:942. Publisher Full Text

Kuntz M, Ledentsov NN, Bimberg D, Kovsh AR, Ustinov VM, Zhukov AE, Shernyakov YM:
Appl Phys Lett. 2002, 81:3846. Publisher Full Text

Zhao HX, Yoon SF, Ngo CY, Wang R, Tong CZ, Liu CY, Cao Q:
IEEE Photonics J. 2010, 2:630. Publisher Full Text

Liu CY, Qu Y, Yuan S, Yoon SF:
Appl Phys Lett. 2004, 85:4594. Publisher Full Text

Liu CY, Yoon SF, Fan WJ, Uddin A, Yuan S:
IEEE Photonics Technol Lett. 2006, 18:791. Publisher Full Text

IEEE J Lightwave Technol. 1999, 17:1634. Publisher Full Text