Uni-traveling-carrier waveguide photodiodes (PDs) with a variable optical confinement mode size transformer are demonstrated. The optical mode is large at the input for minimal front-end saturation and the mode transforms as the light propagates so that the absorption profile is optimized for both high-power and high-speed performance. Two differently designed PDs are presented. PD A demonstrates a 3-dB bandwidth of 12.6 GHz, and saturation currents of 40 mA at 1 GHz and 34 mA at 10 GHz. PD B demonstrates a 3-dB bandwidth of 2.5 GHz, a saturation current greater than 100 mA at 1 GHz, a peak RF output power of + 19 dBm, and a third-order output intercept point of 29.1 dBm at a photocurrent of 60 mA.
©2011 Optical Society of America
High-power and high-speed photodiodes (PDs) are critical components for microwave photonic systems such as fiber-optic links. For such systems, it is desirable that the PDs demonstrate high saturation current, high linearity, high responsivity, and sufficient bandwidth for the system of interest.
State of the art surface-illuminated PDs have demonstrated a saturation current of 144 mA, a third-order output intercept point (OIP3) > 39 dBm, a responsivity of 0.69 A/W, and a bandwidth of 24 GHz . Although impressive performance has been demonstrated, it is challenging to simultaneously achieve high responsivity and high bandwidth with a surface-illuminated PD. Also, the illumination is generally not uniform, so the PD can saturate in the center, which affects the linearity performance. To overcome this limitation and more uniformly illuminate the PD, the working distance of the optical fiber can be increased or the PD can incorporate a graded-index lens [2,3]. While these approaches can improve the saturation characteristics and linearity, typically a penalty in responsivity is incurred.
For waveguide PDs (WGPDs), responsivity and bandwidth are somewhat decoupled because the responsivity is dependent on input coupling efficiency and absorption length whereas the bandwidth is dependent on the intrinsic layer thickness. WGPDs can therefore have a higher bandwidth-efficiency product. Additionally, WGPDs can be monolithically integrated with other optical components, and can incorporate traveling wave designs to overcome RC bandwidth limitations. State-of-the-art WGPDs have demonstrated a 1-dB compression current of 80.5 mA at a frequency of 1 GHz, and OIP3 of 46.1 dBm [4,5]. These WGPDs utilized a lowering of the optical confinement to increase the absorption length and in turn improve the saturation characteristics. For higher speed applications, evanescent coupling has been utilized to optimize the tradeoff between bandwidth and efficiency, and to increase the fiber-coupling efficiency [6,7]. Using evanescent coupling, the authors in  achieved a responsivity of 1.02 A/W, a bandwidth of 48 GHz, and a 1-dB compression current of 11 mA at 40 GHz. The power handling capabilities of WGPDs, however, are ultimately limited by front-end saturation associated with an exponential absorption profile, and by the optical mode size. One attempt to significantly increase the optical mode size to improve power-handling performance utilized a distributed-absorption waveguide where the distribution of generated photocarriers is designed to be uniform . This device demonstrated a responsivity of 1 A/W, a bandwidth of 29 GHz, and a linear photocurrent of 8 mA at 20 GHz. The limited photocurrent was attributed to poor contact electrodes by the authors.
In this work, a uni-traveling-carrier photodiode (UTC-PD) is incorporated into a variable confinement slab-coupled optical waveguide structure. In addition to the variable confinement waveguide architecture, the optical mode at the input is made large to reduce the optical power density using slab-coupled optical waveguide technology . The performance of two differently designed lumped-element PDs are presented. PD A demonstrates a bandwidth of 12.6 GHz, saturation currents of 40 mA at 1 GHz and 34 mA at 10 GHz, and a responsivity of 0.7 A/W. PD B demonstrates a bandwidth of 2.5 GHz, saturation current greater than 100 mA at 1 GHz, and a responsivity of 0.8 A/W. Previous record performance was 80.5 mA at 1 GHz as reported in ; therefore, to the best of our knowledge, the data presented in this paper represents the now highest reported saturation current for an InP-based WGPD. Additionally, results on the peak RF output power generated and the OIP3 performance are presented for PD B.
2. Device design and fabrication
The uni-traveling-carrier variable confinement slab-coupled optical waveguide PD (UTC-VCSCOWPD) incorporates a lateral taper mode size transformer to vary the optical mode size and confinement along the length of the device. The mode at the input of the device is large and nearly circular, the latter of which lends itself to a low input fiber coupling loss. Because of the large mode size and low confinement, the optical power density of the mode and the absorption are low at the input. The front-end saturation is therefore low. Varying the mode size and confinement along the length allows for optimization of the absorption profile. Although the front-end saturation is low, the device is not excessively long; therefore it is possible to simultaneously achieve high-power and high-speed performance with this novel PD.
The UTC-PD operates such that light is absorbed in a p-type absorber layer [10,11]. Photogenerated electrons subsequently diffuse to a wide bandgap collector layer, and then are swept out of the active region by the applied field. For certain absorber and collector layer thicknesses, the carrier transport is governed primarily by the electron drift velocity, which is higher than that of holes. As such, the mobile space charge density is lower than that for typical PIN-PDs for the same current density. The UTC-PD can therefore achieve higher speed and higher power performance.
A plan-view schematic of a UTC-VCSCOWPD following the ridge formation fabrication steps is shown in Fig. 1(a) along with a top-view of a completed device inset. The indices of refraction used for simulations are shown. The top ridge width (W) varies along the length, while the lower mesa width remains fixed at 6 µm. The optical confinement factor, defined as the overlap of the optical mode with the absorber layer, varies as the light propagates in the waveguide. A comparison of the simulated absorption profiles of a conventional WGPD and a UTC-VCSCOWPD is shown in Fig. 1(b). These simulations were performed using BeamPROP, a commercial beam propagation method simulation tool by the RSoft Design Group. For a conventional WGPD, most of the light is absorbed at the front of the device. For a UTC-VCSCOWPD, on the other hand, the absorption is low at the front-end and gradually increases as the top ridge width increases. In this way, a UTC-VCSCOWPD can simultaneously achieve low front-end saturation and high-speed performance.
The device structures are grown on semi-insulating InP substrates by metalorganic chemical vapor deposition. The structures consist of a 1-μm thick InP buffer layer, followed by a 4.97-μm thick InGaAsP waveguide. Part of the waveguide is doped n-type for forming topside n-contacts. The upper 1 μm of the waveguide serves as the collector layer for the UTC-PD structure. The remaining epitaxial layers include band smoothing layers, a 50-nm thick InGaAs absorber layer, a p-cladding with a graded doping profile, and lastly an InGaAs p-contact layer.
UTC-VCSCOWPDs are fabricated by first etching the top ridge using inductively-coupled plasma reactive ion etching (ICP-RIE). Next, the lower mesa is etched, also using ICP-RIE. The PDs are then passivated using silicon nitride (SixNy). This is followed by patterning benzocyclobutene (BCB), which is used for reducing parasitic pad capacitance. Topside n-contacts are then deposited and annealed. Vias are etched through the BCB and SixNy to the p-contact layer on the top ridge, and p-metal contacts are deposited and annealed. The samples are then thinned, cleaved for bar singulation and facet formation, and antireflection coated. Devices are soldered to copper submounts to facilitate characterization.
Two photodiode designs are compared. PD A was designed for high speed and moderate saturation current, and PD B was designed for moderate speed and high saturation current. PD A is 3 µm wide at the input and flares linearly to 6 µm over a length of 200 µm. It is then followed by a 100-µm long straight section (no flare) that is 6 µm wide. PD B has a 2-µm-wide, 300-µm-long section at the input. This is followed by a section that flares linearly from 2 µm to 3 µm over a length of 600 µm, a section that flares linearly from 3 µm to 6 µm over a length of 600 µm, and a 250-µm-long straight section. With the ridge width at the input smaller for PD B than PD A, the mode size at the input of PD B is larger. For both PDs, increasing the width along the length of the device increases the optical confinement and reduces the mode size.
3. Results and discussion
The frequency responses of the PDs were measured using a lightwave component analyzer and the results are shown in Fig. 2 . These measurements were taken at a photocurrent level of 20 mA and a bias of −3.5 V. At these operating conditions, the 3-dB bandwidth of PD A is 12.6 GHz and that of PD B is 2.5 GHz. The responsivity varies somewhat with photocurrent level and bias. For a photocurrent level of 40 mA and a bias of −3.5 V, the measured responsivity is 0.7 A/W for PD A and 0.8 A/W for PD B.
To observe the saturation characteristics of the PDs, the frequency response was measured as a function of photocurrent level and bias. The modulation depth for these measurements was approximately 20%. Figure 3 shows the results for both PDs at frequencies of 1 GHz and 10 GHz, where the data is normalized to an ideal line with a slope of 20 dB/decade. Plotted this way, an ideal linear characteristic would have a normalized RF power of 0 dB as a function of increasing photocurrent before reaching the onset of saturation. As shown in Fig. 3, initially the power increases significantly and then decreases as the PD begins to saturate. This has previously been observed for UTC-PDs . The conventional definition of 1-dB compression current is the photocurrent level at which the measured RF power degrades by 1 dB from the expected linear response . According to this definition, this would be the photocurrent level at which the RF power decreases to −1 dB in the plots shown in Fig. 3. Using this traditional definition, the 1-dB compression current of PD A would be greater than 60 mA at 1 GHz. PD A, however, shows signs of saturation at a lower photocurrent level than that defined by this traditional definition. PD A begins to saturate at a photocurrent of 36 mA, which is extremely high for a waveguide PD with a bandwidth greater than 10 GHz. For the 1-dB compression current figure of merit, we instead adopt a different definition as suggested in , which is the photocurrent level at which the RF power degrades by 1 dB from its peak value. This definition yields a lower 1-dB compression current than that obtained using the traditional definition. Using this definition, the 1-dB compression current of PD A is 40 mA at 1 GHz and 34 mA at 10 GHz.
In comparison, PD B can operate at significantly higher photocurrent levels. As shown in Fig. 4 , PD B shows no sign of saturation for up to 100 mA of photocurrent, making it the highest saturation current InP-based WGPD ever reported. The response was not measured at the compression point because the device underwent thermal runaway and failed prior to reaching this point. The 1-dB compression current is, however, projected to be at least 150 mA. Improvements in thermal management should increase the maximum current level that can be measured at a particular bias voltage. The RF power was also measured under large signal modulation for PD B by overdriving a Mach-Zehnder modulator to nearly 100% modulation depth, and the results are shown in Fig. 5 for a frequency of 900 MHz. A peak power of + 19 dBm was achieved.
Intermodulation distortion measurements were performed to qualify the PDs for high spurious-free dynamic range microwave photonic link applications. The third-order intermodulation distortion (IMD3) was characterized using a three-tone setup. Compared to a traditional two-tone setup, the source modulators do not contribute to the nonlinear power measured at the IMD3 frequencies using a three-tone setup. This is particularly useful for characterizing very linear PDs. The three-tone OIP3 can also be translated readily to the more commonly used two-tone OIP3 as discussed in .
For the three-tone measurements performed, the signal frequencies used were f1 = 885 MHz, f2 = 900 MHz, and f3 = 920 MHz. Figure 6(a) shows a measurement taken for PD B at a photocurrent level of 60 mA and a bias of −3.5 V. The two-tone OIP3 was measured to be 29.1 dBm at these operating conditions. The OIP3 was also measured as a function of photocurrent and the results are shown in Fig. 6(b). A maximum OIP3 of 29.7 dBm was achieved at 40 mA. Although these IMD3 results are sufficient for some applications, we believe they will improve with device optimization. Currently, the OIP3 appears to be limited by responsivity nonlinearities. Further investigation is underway.
UTC-VCSCOWPDs have been demonstrated. These PDs have a large mode at the input that transforms in size as the light propagates in the waveguide by way of a lateral taper. The optical power density is low at the input and the absorption length is optimized, simultaneously allowing for minimal front-end saturation and high-speed performance. Two different PDs were characterized. PD A demonstrated a 3-dB bandwidth of 12.6 GHz, and 1-dB compression currents of 40 mA at 1 GHz and 34 mA at 10 GHz. PD B, designed for higher power-handling performance, demonstrated a 3-dB bandwidth of 2.5 GHz, and a 1-dB compression current > 100 mA at 1 GHz, which is a record for an InP-based WGPD. A peak RF output power of + 19 dBm at 900 MHz was achieved for PD B. The IMD3 characteristics of PD B were measured yielding a maximum OIP3 of 29.7 dBm at 40 mA and an OIP3 of 29.1 dBm at 60 mA. These PDs are suitable for microwave photonic link applications.
This work was supported by the Defense Advanced Research Projects Agency (DARPA) under United States Air Force contract No. FA8721-05-C-0002. The opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the United States Government.
References and links
1. Z. Li, H. Pan, H. Chen, A. Beling, and J. C. Campbell, “High-saturation-current modified uni-traveling-carrier photodiode with cliff layer,” IEEE J. Quantum Electron. 46(5), 626–632 (2010). [CrossRef]
2. A. Beling, H. Pan, H. Chen, and J. C. Campbell, “High-power modified uni-travling carrier photodiode with > 50 dBm third order intercept point,” in IEEE MTT-S Microwave Symposium Digest (2008), pp. 499–502.
3. A. Joshi, S. Datta, and D. Becker, “GRIN lens-coupled top-illuminated highly linear InGaAs photodiodes,” IEEE Photon. Technol. Lett. 20(17), 1500–1502 (2008). [CrossRef]
4. J. Klamkin, A. Ramaswamy, N. Nunoya, L. A. Johansson, J. E. Bowers, S. P. DenBaars, and L. A. Coldren, “Uni-traveling-carrier waveguide photodiodes with >40 dBm OIP3 for up to 80 mA of photocurrent,” in Device Research Conference (2009).
5. A. Ramaswamy, J. Klamkin, N. Nunoya, L. A. Johansson, L. A. Coldren, and J. E. Bowers, “Three-tone characterization of high-linearity waveguide uni-traveling-carrier photodiodes,” in IEEE Lasers and Electrooptics Society Conference (IEEE, 2008), pp. 286–287.
6. L. Giraudet, F. Banfi, S. Demiguel, and G. Herve-Gruyer, “Optical design of evanescently coupled waveguide-fed photodiodes for ultrawide-band applications,” IEEE Photon. Technol. Lett. 11(1), 111–113 (1999). [CrossRef]
7. S. Demiguel, N. Li, X. Li, X. Zheng, J. Kim, J. C. Campbell, H. Lu, and A. Anselm, “Very high-responsivity evanescently coupled photodiodes integrating a short planar multimode waveguide for high-speed applications,” IEEE Photon. Technol. Lett. 15(12), 1761–1763 (2003). [CrossRef]
8. 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. 45(8), 1337–1341 (1997). [CrossRef]
9. S. M. Madison, J. J. Plant, D. C. Oakley, A. Napoleone, and P. W. Juodawlkis, “Slab-coupled optical waveguide photodiode,” in Conference on Lasers and Electro-Optics/Quantum Electronics and Laser Science Conference and Photonic Applications Systems Technologies, OSA Technical Digest (CD) (Optical Society of America, 2008), paper CWF4.
10. J. Klamkin, Y.-C. Chang, A. Ramaswamy, L. A. Johansson, J. E. Bowers, S. P. DenBaars, and L. A. Coldren, “Output saturation and linearity of waveguide uni-traveling-carrier photodiodes,” IEEE J. Quantum Electron. 44(4), 354–359 (2008). [CrossRef]
11. T. Ishibashi, T. Furuta, H. Fushimi, S. Kodama, H. Ito, T. Nagatsuma, N. Shimizu, and Y. Miyamoto, “InP/InGaAs uni-traveling-carrier photodiodes,” IEICE Trans. Electron. E83-C, 938–949 (2000).
12. K. J. Williams and R. D. Esman, “Large-signal compression-current measurements in high-power microwave pin photodiodes,” Electron. Lett. 35(1), 82–84 (1999). [CrossRef]