IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 17, NO. 4, APRIL 2005

881

Characteristics of Waveguide Photodetectors at High-Power Optical Input Joong-Seon Choe, Yong-Hwan Kwon, and Kisoo Kim

Abstract—High-power characteristics of waveguide photodetectors (PDs) with various absorbing layer thicknesses were analyzed. When large photocurrent was generated by high-power optical input, the saturation of radio frequency response was closely related with the optical absorption efficiency. The device with a single quantum well absorbing layer showed no saturation behavior until the device failure, while PDs with a thicker absorbing layer were saturated at photocurrent of a few milliamperes. It was also found that as optical confinement decreases, the maximum photocurrent the device can endure increases by more than three times. Index Terms—Optical confinement factor, photodetector (PD), quantum well, reliability, saturated output.

Fig. 1. Device structure used in this analysis. Polyimide-passivated ridge structure was adopted. Except for InGaAs absorbing layer, all the epitaxial layers were grown identically for four kinds of wafers.

I. INTRODUCTION

P

HOTODETECTORS (PDs) are essential in many applications where optical-to-electrical conversion is needed. Among those, some require the PD to receive high-power optical input to generate large photocurrent, as in CATV and radio-over-fiber, etc. [1], [2]. A large number of photogenerated carriers screen the electric field by space-charge effect, leading to a reduction in carrier drift velocities and a corresponding build-up of carriers [3], and consequently deteriorating high-speed performance. In addition, large photocurrent may cause a reliability problem due to the increase in device temperature [4] caused by Joule heating. There are many reports on high output current PD such as unitraveling-carrier PD [5], [6], PD with partially depleted absorber [7], evanescently coupled double taper PD [8], and velocitymatched distributed PD [9], etc. Their high output performance was improved by adopting sophisticated structures compared with conventional waveguide PD (WGPD). In this study, we used conventional WGPDs with four different absorbing layer thicknesses and investigated the feasibility of simple WGPD as a high-power PD in view of radio frequency (RF) response and device reliability. II. EXPERIMENT Fig. 1 shows the schematic device structure used in this experiment. Waveguide core layers are composed of InGaAsP m and InGaAs. Both of the upper and lower InGaAsP layers are 1.3 m thick, and InGaAs absorbing layer has variation in its thickness: 100, 500, 1000, and 2000 . InGaAsP layers are partly doped (n-type: Si, cm ; p-type: Zn, Manuscript received August 12, 2004; revised December 22, 2004. The authors are with Electronics and Telecommunications Research Institute, Daejeon 305-350, Korea (e-mail: [email protected]; [email protected]; [email protected]). Digital Object Identifier 10.1109/LPT.2005.844003

Fig. 2. Schematic setup for measuring the saturation behavior of the PDs. The thin lines indicate optical fibers and the thicker ones RF or bias cables.

cm ) so that the total undoped region thickness is 1 m. The asymmetric doping scheme is adopted for higher transit-time-limited bandwidth [10], [11]. Upper and lower InP cm and cm , respecclad layers are doped to tively. The waveguide structure, 3 m wide, was formed by conventional photolithography and dry-etching processes. The exposed surface was passivated by polyimide (PI-2723) which was cured on a hot plate, and ground-signal-ground (GSG) Au-electrode was deposited. The input facet was formed through mechanical cleaving and no antireflection coating was deposited. The length of devices was 300 m. In the following, we will denote devices from wafer with InGaAs of 2000, 1000, 500, and 100 as PD1, PD2, PD3, and PD4, respectively. For all the devices, dark current was as small as 20 nA at 3 V. Shown in Fig. 2 is the schematic diagram for the measurement setup. The modulated beam of 1.55- m wavelength was generated by a tunable laser source (HP 8168F) and an electrooptic

1041-1135/$20.00 © 2005 IEEE

882

IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 17, NO. 4, APRIL 2005

TABLE I 1-dB COMPRESSION CURRENT OF EACH DEVICES AT VARIOUS FREQUENCIES

input. Under only 1550-nm input of 1 mW, the devices showed bandwidth of about 15 GHz and photocurrent of 400 A. In spite of the difference in the absorbing layer thickness, the four devices show similar responsivities owing to the sufficiently long cavity length, that is explained by the following: (1)

Fig. 3. Relative RF responses of four devices (a) PD1, (b) PD2, (c) PD3, and (d) PD4. All the data are normalized to the response at the lowest optical input level. The data of (c) and (d) are vertically translated for the sake of clarity.

modulator. Between the laser and modulator, a polarization controller was used for transverse-magnetic-polarized input to modulator. In order to control the optical input level, 1480-nm highpower continuous-wave laser diode was used. The two kinds of laser outputs, combined by 9 : 1 optical coupler, are incident on the facet of PD through a tapered fiber. Using the 9 : 1 coupler is to supply PD with sufficient optical input (90%) from high-power laser diode. Before the coupler input, Er-doped fiber amplifier boosted the modulator output compensating for the loss the coupler would bring about. The generated photocurrent was collected by GSG probe which was connected with Anritsu 37 000C vector network analyzer (VNA) through RF cable. A reverse bias of 3 V was applied to the PD by Keithley 236 source-measure unit (SMU) through a bias tee between VNA and PD. SMU and VNA measured dc photocurrent and RF signal response of the PD, respectively. III. RESULT Fig. 3 shows the relative RF response data of (a) PD1, (b) PD2, (c) PD3, and (d) PD4 at various photocurrent levels. All the data are normalized to the responses when there is no 1480-nm

, , , and are responsivity, maximum rewhere , sponsivity, optical confinement factor, absorption coefficient, and cavity length, respectively. Here the facet reflection was ignored. For small , the responsivity depends strongly upon , but the influence decreases with . From another experiment, we have ascertained that absorbing layer thickness does not affect the responsivity when is larger than 200 m. As can be seen clearly, each device shows quite different relative RF responses. The response of PD1, whose absorbing layer is 2000- -thick bulk InGaAs, shows saturated behavior with optical input power. As the photocurrent increases with optical input, the relative frequency falls abruptly at low frequency. At 10 GHz, the response reduces by 8 dB when the photocurrent increases to 3 mA, compared with the response when photocurrent was 400 A. When carriers are generated by absorbed photons, the electric field is altered as the carriers are redistributed [12]. Because the carriers drift along the field under reverse bias, the modified electric field distribution changes the carrier drift velocity, which affects the RF response as in [10]. PD2 shows similar behavior with PD1 [Fig. 3(b)]. The response data at 5 mA is, however, less degraded than PD1. At 5 mA, PD2 shows a response similar to that of PD1 at 3 mA. In both of the cases, abrupt decrease of response occurs below 5 GHz and above it further severe degradation does not occur. Compared to above two cases, PD3 shows quite sustained response [Fig. 3(c)]. At a photocurrent level of 4 mA, it did not show 1-dB response compression before 10 GHz, and at a higher photocurrent, the diminution of response becomes severe below 5 GHz, too. Fig. 3(d) is the response of PD4, whose absorbing layer is 100- -thick InGaAs–InGaAsP single quantum well. Up to 17 mA, the response shows no apparent difference within the frequency range measured. In our experimental setup, we could not observe the saturation behavior of PD4 within the frequency range measured before the device failure. From Fig. 3, 1-dB compression photocurrent is readily at 2.4, 5.8, obtained at a specific frequency. Table I shows 10, and 20 GHz. PD1 and PD2 show similar results, and PD3 exceeds them by about two times. For PD4, was found to be above 17 mA, however, the exact value could not be obtained due to the breakdown of the device before the response saturation. As the degradation with photocurrent results from the field screening by generated carriers [3], the response can be considerably restored by increasing applied reverse bias [13]. Fig. 4 shows calculated through the beam propagation method (BPM) and maximum photocurrent that the

CHOE et al.: CHARACTERISTICS OF WGPDs AT HIGH-POWER OPTICAL INPUT

883

related with optical input level, and it was found that just by reducing the optical confinement factor , conventional WGPD can show good high output current characteristics. The WGPD with 100- -thick InGaAs–InGaAsP quantum well kept its RF response up to the photocurrent level of 17 mA where the device failure occurred. It was also found that the maximum photocurrent the WGPD can stand is strongly influenced by , and increases by about three times as decreases from 16.3% to 0.5%. REFERENCES

Fig. 4. Maximum photocurrent the devices can endure and the optical confinement factor calculated as functions of the absorbing layer thickness.

devices can endure as functions of the absorbing layer thick. While increases from 0.5% to 16.3% with , ness decreases from 17 to 5 mA. This is explained by the fact that the failure mechanism is strongly related with the local Joule heating [4]. The carrier distribution conforms to the photon density as follows: (2) and are a normalization constant and position where along the propagation direction, respectively. From (2), the carrier density at a given is proportional to under the same total current. It means that near the input facet, which is crucial for device reliability, PD1 has approximately 30 times larger current density than PD4 at the same photocurrent level. The excessive current density caused the facet to be melted down, which was observed through scanning electron microscope. Although some researchers adopted small optical confinement to improve the power behavior of WGPD [14], to the best of our knowledge, there has been no report relating the optical confinement factor and the maximum photocurrent before of PD4 is not as large as expected when device failure. only the difference in is considered since PD4 has the largest thermal impedance due to the large thermal conductivity of InGaAs [15]. In addition, near the facet, has a different value from that obtained by BPM because of the transformation from the fiber mode to waveguide mode just after the incidence. A rough estimation gives that of PD1 at the input facet is smaller than 20 times that of PD4 considering the overlap of InGaAs layer and Gaussian incident beam. As noted previously, the raised operation voltage may compensate the degraded RF response. However, it causes the larger heat dissipation that reduces . Thus, WGPD can work as a high-power PD only when is small enough to guarantee both the RF response and reliability. From Fig. 4, should be less than 1% in order to make a WGPD operate at photocurrent level over 10 mA. IV. CONCLUSION In this letter, we have presented the characteristics of WGPDs under high optical input condition. The RF response was closely

[1] K. J. Williams and R. D. Esman, “Photodiode DC and microwave nonlinearity at high current due to carrier recombination nonlinearities,” IEEE Photon. Technol. Lett., vol. 10, no. 7, pp. 1015–1017, Jul. 1998. [2] R. P. Braun, G. Grosskopf, H. Heidrich, C. von Helmolt, R. Kaiser, K. Kruger, U. Kruger, D. Rohde, F. Schmidt, R. Stenzel, and D. Trommer, “Optical microwave generation and transmission experiments in the 12and 60-GHz region for wireless communications,” IEEE Trans. Microw. Theory Tech., vol. 46, no. 4, pp. 320–330, Apr. 1998. [3] P.-L. Liu, K. J. Williams, M. Y. Frankel, and R. D. Esman, “Saturation characteristics of fast photodetectors,” IEEE Trans. Microw. Theory Tech., vol. 47, no. 7, pp. 1297–1303, Jul. 1999. [4] K. J. Williams and R. D. Esman, “Design considerations for high-current photodetectors,” J. Lightw. Technol., vol. 17, no. 8, pp. 1443–1454, Aug. 1999. [5] Y. Muramoto, H. Fukano, T. Furuta, and Y. Matsuoka, “A polarization-independent high-efficiency refracting-facet unit-traveling-carrier photodiode with a bandwidth over 50 GHz,” in Proc. Eur. Conf. Optical Communications 2000, Munich, Germany, Sep. 2000, pp. 109–110. [6] N. Li, X. Li, S. Demiguel, X. Zheng, J. C. Campbell, D. A. Tulchinsky, K. J. Williams, T. D. Isshiki, G. S. Kinsey, and R. Sudharsansan, “High-saturation-current charge-compensated InGaAs-InP uni-traveling-carrier photodiode,” IEEE Photon. Technol. Lett., vol. 16, no. 3, pp. 864–866, Mar. 2004. [7] X. Li, N. Li, S. Demiguel, X. Zheng, J. C. Campbell, H. H. Tan, and C. Jagadish, “A partially depleted absorber photodiode with graded doping injection regions,” IEEE Photon. Technol. Lett., vol. 16, no. 10, pp. 2326–2328, Oct. 2004. [8] S. Demiguel, L. Giraudet, L. Joulaud, J. Decobert, F. Blache, V. Coupé, F. Jorge, P. Pagnod-Rossiaux, E. Boucherez, M. Achouche, and F. Devaux, “Evanescently coupled photodiodes integrating a double-stage taper for 40-Gb/s applications—compared performance with side-illuminated photodiodes,” J. Lightw. Technol., vol. 20, no. 12, pp. 2004–2014, Dec. 2002. [9] M. S. Islam, S. Murthy, T. Itho, M. C. Wu, D. Novak, R. B. Waterhouse, D. L. Sivco, and A. Y. Cho, “Velocity-matched distributed photodetectors and balanced photodetectors with p-i-n photodiodes,” IEEE Trans. Microw. Theory Tech., vol. 49, no. 10, pp. 1914–1920, Oct. 2001. [10] J.-S. Choe, Y.-H. Kwon, K. Kim, J. Kim, S.-C. Kong, and Y.-W. Choi, “Traveling-wave photodetector with asymmetrically heterostructured intrinsic region,” Jpn. J. Appl. Phys., vol. 8A, pp. 5105–5109, Aug. 2004. [11] S.-C. Kong, S.-H. Ok, Y.-W. Choi, J.-S. Choe, Y.-H. Kwon, and J.-H. Kim, “Numerical analysis and design of a new traveling-wave photodetector with an asymmetric i-layer cross section,” IEEE J. Sel. Topics Quantum Electron., vol. 9, no. 3, pp. 770–774, May/Jun. 2003. [12] I. S. Ashour, J. Harari, J.-P. Vilcot, and D. Decoster, “High optical power nonlinear dynamic response of AlInAs/GaInAs msm photodiode,” IEEE Trans. Electron Devices, vol. 42, no. 5, pp. 828–833, May 1995. [13] C.-K. Sun, I.-H. Tan, and J. E. Bowers, “Ultrafast transport dynamics of p-i-n photodetectors under high-power illumination,” IEEE Photon. Technol. Lett., vol. 10, no. 1, pp. 135–137, Jan. 1998. [14] S. Jasmin, N. Vodjdani, J. Renaud, and A. Enard, “Diluted- and distributed-absorption microwave waveguide photodiodes for high efficiency and high power,” IEEE Trans. Microw. Theory Tech., vol. 45, no. 8, pp. 1337–1341, Aug. 1997. [15] M. Gruden and J. Piprek, “Material parameters of quarternary III-V semiconductors for multilayer mirrors at 1.55 m wavelength,” Modeling Simul. Materi. Sci. Eng., vol. 4, pp. 349–357, Jul. 1996.