Opt Quant Electron (2008) 40:41–46 DOI 10.1007/s11082-008-9230-9

Characterization and optimization of high-power InGaAs/InP photodiodes Huapu Pan · Andreas Beling · Hao Chen · Joe C. Campbell

Received: 4 November 2007 / Accepted: 8 June 2008 / Published online: 2 July 2008 © Springer Science+Business Media, LLC. 2008

Abstract The performance of 34-µm-diameter InGaAs/InP charge compensated modified uni-traveling carrier photodiodes (CC MUTCs) is studied using a commercial device simulator based on a drift-diffusion model. Excellent agreement has been achieved between simulations and experiments. The origin of the degradation of responsivity at high optical injection level is investigated. Parameters of the photodiode are further optimized to maximize the saturation current without sacrificing bandwidth and responsivity. Keywords

Photodiode · High-power · Saturation · Modeling

1 Introduction Analog optical links are attractive for microwave applications due to their inherent advantages such as low loss and high bandwidth. Photodiodes as optical-to-electrical converter must be able to provide high photocurrent so as to satisfy the requirement of high dynamic range of the optical link (Seeds 2002). Several photodiode structures have been developed to increase the output current, including the uni-traveling-carrier (UTC) photodiode (Li et al. 2004) and the partially-depleted-absorber (PDA) photodiode (Li et al. 2003). Here we demonstrate a normal-incidence back-illuminated 34-µm-diameter photodiodes with modified UTC (MUTC) structure (Jun et al. 2006), with both high saturation current of 100 mA and high responsivity of 0.75 A/W at 1,550 nm wavelength. The origin of current saturation is studied using a commercial device simulator APSYS (Crosslight 2007) based on the drift-diffusion model. Good agreement has been achieved between the simulation and the experiment. The role of the cliff layer is investigated, and the thickness of the inserted undoped InGaAs absorption layer is optimized to maximize the saturation current.

H. Pan (B) · A. Beling · H. Chen · J. C. Campbell Department of Electrical and Computer Engineering, University of Virginia, 351 McCormick Rd., Charlottesville, VA 22904, USA e-mail: [email protected]

123

42

H. Pan et al.

Fig. 1 InGaAs/InP MUTC photodiode structure

2 Device structure As shown in Fig. 1, the depletion region consists of a 605 nm InP depletion layer, two 15 nm undoped InGaAsP “bandgap smooth” layers, and a 200 nm depleted InGaAs absorbing layer. The undepleted absorbing region consists of four InGaAs layers with graded p-doping. More details can be found in our previous work (Wang et al. 2007).

3 Experiments and simulations A key figure of merit for high power photodiodes is saturation current, which is defined as the photocurrent at which the RF output power at the 3 dB cut-off frequency is compressed by 1 dB. The MUTC photodiodes reported here achieved high saturation current of 100 mA at 5 V reverse bias (Wang et al. 2007). Since photodiodes are more easily saturated at lower bias, the RF power and responsivity were measured at zero bias from low-level to high-level optical injection. The measured RF power peaks at about 8 mW of incident optical power and then decreases with input optical power. The measured responsivity remains constant at 0.75 A/W when the optical power is small and begins to decrease near 18 mW (Fig. 2, bottom). This phenomenon is replicated by our commercial device simulator APSYS based on drift-diffusion model. The 50
load resistor and a series resistance of 5.6
are considered in simulation. Thermal effects are also considered following the experiment data (Duan et al. 2006). In general, the simulated results agree well with the measured values, as is shown in Fig. 2.

123

Characterization and optimization of high-power InGaAs/InP photodiodes

43

Fig. 2 Measured and simulated RF output power (top) and responsivity (bottom) versus optical power at zero bias

In order to study the origin of responsivity degradation, the electrical fields in our devices are simulated at zero bias for different optical injection levels. Figure 3 shows that as the optical power increases, the electric field in the depleted InGaAs absorption layer is reduced due to the screening caused by the photo-generated electrons and holes, which is referred to as space charge effect. As a consequence, the band discontinuity in the conduction band at the step graded interface between the InGaAs layer and the InP layer become more pronounced and behaves as a barrier to impede the electrons from flowing out the absorption layers, which degrades the responsivity. This is similar to the case in UTC photodiodes (Srivastava et al. 2004). This explains the falloff in the responsivity of our devices at about 18 mW and the falloff in the RF power at 8 mW. Because both RF power degradation and responsivity degradation are caused by space charge effect, the saturation current is closely related to the photocurrent at which the responsivity begins to degrade. Due to the self-induced field, the responsivity always increases slightly with photocurrent at low optical injection level (Srivastava et al. 2004), and then decrease when space charge effect becomes pronounced. So in the following we refer the photocurrent at which the responsivity reaches its peak as the “degradation current”. Since saturation current is more difficult to simulate, in the next section we will evaluate the “degradation current” to estimate the ability of a photodiode to sustain high optical injection.

4 Device optimization According to our previous work (Wang et al. 2007), the bandwidth of the 34 µm MUTC photodiode is RC limited. In order to maintain the bandwidth, the total thickness of the depletion

123

44

H. Pan et al.

Fig. 3 Electric field at zero bias for the optical power of 0, 8 and 16 mW. Undepleted InP layer is from x = 0.1 µm to x = 0.705 µm, and undepleted InGaAs absorption layer is from x = 0.735 µm to x = 0.935 µm

Fig. 4 Distribution of electric field in the photodiode with different cliff layer thickness. Doping level of all the cliff layer is 5 × 1017 cm−3

region (835 nm), and thus the capacitance, was kept constant. A cliff layer, which is a thin moderately doped n-type InP layer, is added between the semi-intrinsic InGaAs layer and the semi-intrinsic InP layer to reduce space charge effect (Srivastava et al. 2004). The role of the cliff layer is shown in Fig. 4. The electric field in the depleted InGaAs absorption layer is increased by the cliff layer, which makes the photodiodes less susceptible to the space charge effect, while the electric field in the depleted InP layer is reduced. As shown in Fig. 4, when the doping level of this cliff layer is fixed to 5 × 1017 cm−3 , the semi-intrinsic InP layer is still fully depleted when the thickness of the cliff layer is 5 nm, but full depletion cannot be maintained as the thickness increases to 10 nm. With the purpose of maintaining the capacitance,

123

Characterization and optimization of high-power InGaAs/InP photodiodes

45

Fig. 5 Simulated responsivity versus optical power at 5 V reverse bias when the depleted absorber constitutes 1/4, 1/2 and 7/8 of the total depletion region

we choose the thickness of the cliff layer to be 5 nm and the doping to be 5 × 1017 cm−3 for our optimized device structure to ensure the full depletion of the semi-intrinsic InP layer. With the n-type cliff layer inserted, the percentage of the depleted InGaAs absorber in the total depletion region was varied to maximize the degradation current. The responsivity at 5 V reverse bias is simulated versus optical power under isothermal condition when the fractions of the depleted InGaAs layer were varied from 1/4 to 7/8. As shown in Fig. 5, responsivity increases monotonically with the percentage of the depleted absorber. However, the corresponding “degradation currents” are 200, 206 and 179 mA, respectively. The 1/4 device does not produce high degradation current because the relatively thin depleted absorber makes the device more susceptible to the conduction band barrier at the heterojunction interface. Besides, its relatively low responsivity also requires higher optical power to reach a certain current level and thus thermal failure may happen earlier. On the other hand the 7/8 device is more susceptible to the space charge effect because of its thick depleted absorber. So the optimum percentage of the depleted absorber should be approximately 50%. Although the degradation current is calculated under isothermal condition and the “degradation current” is not identical to the saturation current, these results provide useful insight to device design and optimization.

5 Conclusion In conclusion, responsivity degradation at high optical injection levels in InGaAs/InP MUTC photodiodes has been studied and good agreement has been achieved between measurements and simulations. The origin of the reponsivity degradation is investigated. Further simulation results show that for high-power photodiodes whose bandwidth is RC-limited, higher saturation current can be achieved without sacrificing bandwidth and responsivity by properly designing the cliff layer and the depleted InGaAs absorber. The depleted InGaAs absorber should constitute approximately 50% of the whole depletion region to maximize the saturation current.

123

46

H. Pan et al.

References Crosslight: www.crosslight.com, SimuApsys version 2007.3 (2007) Duan, N., et al.: Thermal analysis of high-power InGaAs/InP photodiodes. IEEE J. Quantum Electron. 42, 1255–1258 (2006) Jun, D.H., et al.: Improved efficiency-bandwidth product of modified uni-traveling carrier photodiode structures using an undoped photo-absorption layer. Jpn. J. Appl. Phys. 45(4B), 3475–3478 (2006) Li, X., et al.: High-saturation-current InP-InGaAs photodiode with partially depleted absorber. Photonics Technol. Lett. 15, 1276–1278 (2003) Li, N., et al.: High-power charge-compensated unitraveling-carrier balanced photodetector. Photonics Technol. Lett. 16, 2329–2331 (2004) Seeds, J.: Microwave photonics. IEEE Trans. Microw. Theory Tech. 50, 877–887 (2002) Srivastava, S., Roenker, K.P.: Numerical modeling study of the InP/InGaAs uni-travelling carrier photodiode. Solid-State Electron. 48, 461–470 (2004) Wang, X., et al.: InGaAs/InP photodiodes with high responsivity and high saturation power. IEEE Photonics Technol. Lett. 19, 1272–1274 (2007)

123