Placing a quantum well modulator in an asymmetric Fabry-Perot cavity enables significantly higher contrast ratios than are possible in a conventional surface-normal quantum well modulator. However, fixed-cavity asymmetric Fabry-Perot quantum well modulators require extremely precise and uniform crystal growth and are sensitive to small fluctuations in temperature or angle of incidence. Here, we experimentally demonstrate an InP-based microelectromechanically tunable asymmetric Fabry-Perot quantum well modulator that operates in the optical C-band. By actuating a suspended InGaAlAs reflector, the cavity mode can be perfectly matched to the appropriate quantum well absorption wavelength. The devices exhibit contrast ratios over 30 (15 dB) at 8 volts quantum well bias and modulation speeds of 1 MHz.
© 2008 Optical Society of America
Surface-normal quantum well electroabsorption modulators [1, 2, 3] are important components of optical systems that require polarization insensitivity or large modulation areas , and can be used for dense modulator arrays for optical interconnects  or pulse shaping . Multiple quantum well (MQW) electroabsorption modulators in the optical C-band have proven extremely effective at operating with large modulation areas, high data rates, and low voltage, but suffer from low contrast ratios: typically about 2:1 in a double-pass configuration[7, 2]. MQWmodulators with higher contrast ratios could improve the modulator efficiency and eliminate CW backgrounds, thereby enabling their use in high-sensitivity applications that require excellent off-state extinction.
Asymmetric Fabry-Perot cavities have been investigated in recent years to as a means to increase the contrast ratio in these modulators . This is achieved by placing an absorber (the multiple quantum well) with an electrically variable loss coefficient aQW (varied using the quantum-confined Stark effect) into a Fabry-Perot cavity with mismatched mirror reflectivities, Rhigh and Rlow. For low-loss or high-loss absorber states, the effective mirror reflectivities are mismatched such that even at cavity mode wavelengths a large amount of optical power is reflected from the cavity. However, when the product of the quantum well double-pass transmission and the high mirror reflectivity, Rhighe -2aQW, is equal to Rlow, and the wavelength is tuned to the cavity mode, the cavity is in the “matched” condition. Matching the effective mirror reflectivities allows for near-zero reflected optical power. Though fixed-cavity asymmetric Fabry-Perot quantum well modulators (AFPQWM’s) have previously been demonstrated, including those that operate in the optical C-band [9, 10], they are intolerant to small cavity length or excitonic resonance inaccuracies caused by growth imperfections or nonuniformities. In addition, fixed-cavity AFPQWM’s cannot compensate for temperature fluctuations (which typically shift the quantum well excitonic resonance more than the cavity mode) or variations in the angle of incidence (which change the effective cavity mode wavelength) and have a limited optical bandwidth. In applications of AFPQWM’s in free-space modulating retroreflector links , which operate at data rates of tens of Mbs, achieving high contrast modulation over a range of temperatures and angles is critical.
Tunable mirrors utilizing microelectromechanical systems (MEMS) in the optical C-band have previously been used with electro-optically active III-V semiconductor devices such as resonant photodetectors , tunable semiconductor optical amplifiers  and tunable VCSELS’s . In this work, we report the integration of an InGaAlAs microelectromechanical mirror into an AFPQWM. A bias voltage between the InGaAlAs mirror and the etch stop layer electrostatically actuates the mirror height. This movable mirror serves as the front reflector of the asymmetric Fabry-Perot cavity, allowing the cavity mode wavelength to be dynamically tuned and matched to a specific wavelength near the excitonic resonance. Our earlier demonstration of a microelectromechanically tunable AFPQWM  operated at 980 nm in a GaAs/AlGaAs system with a gold micromirror. Here, by using an all-semiconductor heterostructure grown on InP, we are able to operate at wavelengths in the optical C-band and avoid thermal and fabrication complexities associated with metallic microelectromechanical systems on semiconductor substrates. The demonstration of an AFPQWM in the optical C-band is critical for such emerging applications as free-space laser communication and optical interconnects that operate in this wavelength range, and requires the growth and fabrication of a state-of-the-art InP-based distributed Bragg reflector (DBR) and novel InGaAlAs MEMS.
2. Growth and fabrication
Our AFPQWM is designed for operation at wavelengths near 1.55 µm using a compound semiconductor heterostructure grown using molecular beamepitaxy (MBE) on a semi-insulating InP substrate. As shown in Fig. 1, it is a reflection modulator intended for interrogation from the top side. The three main components of the device are (i) an InP/InGaAlAs DBR grown on the InP substrate (the back mirror); (ii) an InGaAs/InAlAs multiple quantum well grown on the DBR (the electroabsorption layers); and (iii) a 170 nm thick InGaAlAs microbridge layer (the top mirror) on an InAlAs sacrificial layer. All layers are grown to be lattice-matched except for the InGaAlAs mirror, which is grown slightly tensile to insure flatness upon release.
The 20-period DBR consists of alternating layers of 114 nm thick In.53Ga.39Al.08As and 125 nm thick InP. The InGaAlAs alloy fractions are chosen to have as high an index of refraction as possible without adding excessive absorption in the C-band . Separate measurements performed on calibration samples show this mirror to have a 3dB bandwidth of approximately 140 nm, a maximum reflectance (Rhigh) of 95%, and an absorption edge at approximately 1500 nm. The top 4 layers of the DBR are n-doped (n-doping=5×1018 cm-3) to allow reverse biasing of theMQW. A 31 nm undoped InAlAs layer is grown between the DBR and theMQW. TheMQWis designed to have a heavy-hole excitonic resonance at 1550 nm and consists of 32-periods of alternating layers of 8 nm thick InGaAs and 6 nm thick InAlAs. Above this layer is a 37 nm thick undoped InAlAs layer, followed by 100 nmof p-InAlAs (p-doping=3×1018 cm-3), and then by a 30 nm p-InGaAs layer (p-doping=1×1019 cm-3). This p-InGaAs layer serves as an etch stop during sacrificial wet etching of the microbridge as well as the contact layer for the ground probe during electrical testing. The final layers of the device are a 1520 nm undoped InAlAs sacrificial layer, followed by a 170 nm thick n-In.52Ga.28Al.20As (n-doping=5×1018 cm-3) bridge layer that has a reflectance (Rlow) of approximately 50% when released. The thickness of this bridge layer is chosen to give a reflectivity near 50%in the optical C-band.
The devices are fabricated using photolithography and wet etching to define the “bottom” level, on which the quantum well contacts are deposited; the “middle” level, a mesa on which the ground contacts are deposited; and the “top” level, a second mesa on which the bridge contacts are deposited (see Fig. 1). Then, selective etching of the InAlAs sacrificial layer releases the InGaAlAs microbridge. The microbridges range in length from 30 µm to 120 µm, in width from 6 µm to 18 µm, and show a downward bowing in the range of 100–400 nm. InGaAlAs is used for the microbridge because it is transparent at 1550 nm, it is highly selective against the HCl etch used to remove the InAlAs, and it can be grown with precise tensile strain by controlling the indium alloy fraction. Unlike a conventional asymmetric Fabry-Perot modulator, this is a three-terminal n-p-n device, with separate front mirror (Vbridge) and quantum well (VQW) reverse bias contacts. Dark currents at 10 V reverse bias are typically less than 100 nA and 1 nA for V QW and V bridge, respectively. Fig. 1(b) shows a scanning electron microscope (SEM) image of a fabricated device.
Figure 2 shows electrostatic actuation of a 120 µm long released microbridge measured using a white light interferometer. V bridge is the reverse bias applied between the middle p-contact and the top n-contact. The data show that actuation voltages below 10 V displace the microbridges by a few hundred nanometers, a range sufficient to achieve cavity matching in these structures. The alloy fraction of the bridge layer is chosen to give a uniform tensile strain of about 0.0006, an amount that previous investigations have shown to result in extremely flat bridges (tens of nm of vertical bowing over bridge lengths of hundreds of microns) without dislocations or cracks. However, these electrostatic actuation measurements combined with measurements of mechanical resonant frequencies and flatness in this sample suggest that the bridges have a vertical gradient of tensile strain. This strain gradient is likely a result of partial strain relaxation.
3. Electro-optic characterization
Figure 3(a) shows a one-dimensional transfer matrix calculation of the reflectivity of our device for a range of microbridge heights. The exact layer thicknesses for the calculation are determined by cross-sectional SEM measurements. The calculation shows that wavelengths near the heavy-hole excitonic resonance of the quantum well (here 1554 nm) satisfy the “matching” condition. Thus, as the microbridge mirror height in our device is reduced, not only does the Fabry-Perot cavity mode blueshift, but the minimum reflectivity also decreases to near zero, as expected. Optical reflectance measurements are obtained by focusing a tunable laser to the center of the microbridge and collecting the reflected light using the same focusing objective into a single mode optical fiber. This reflectance data is shown in Fig. 3(b) for a range of microbridge heights for a typical microbridge that is 120 µm long and 10 µm wide. The agreement is excellent, showing that at V bridge ≈5 V and a wavelength of 1554 nm, the microbridge tunes the cavity into a near-perfect match with maximal extinction at the cavity mode. The primary source of discrepancy between the calculated and the measured spectra arises from the uncertainty in the specific spectral shape of the quantum well absorption profile.
Figure 4(a) shows a detailed reflectivity spectrum for the same device as shown in Fig. 3 at a fixed microbridge height and for VQW=0 V and VQW=8 V. At a wavelength of 1554 nm, VQW=0 V, and Vbridge=4.8 V, a minimum reflectivity of 0.0029 is measured. The effect of the applied bias across the MQW is to redshift the heavy-hole excitonic resonance and reduce the oscillator strength, thereby decreasing the absorption at 1554 nm and destroying the “matched” cavity condition. The ratio of these reflectivity curves (the contrast ratio) is plotted in Fig. 4(b) in logarithmic units for a range of microbridge heights. The data show that excellent contrast ratios can be achieved duringMQW modulation by actuating the top mirror to an optimal height. Such contrast ratio optimization is not possible for a fixed Fabry-Perot cavity. Investigations of a number of different devices show maximum measured contrast ratios between 30 (15 dB) and 40 (16 dB) for VQW between 0 V and 8 V. Only small improvements in the contrast ratio are observed for VQW>8 V since the heavy-hole excitonic resonance has shifted well to the red of 1554 nm at these field strengths.
Because our device allows for dynamic tuning of the cavity length, the contrast ratio can be maximized for a given wavelength by tuning VQW and Vbridge for optimum extinction. The resulting maximum contrast ratio (achieved by modulating VQW to 8 V) and insertion loss (the reflectivity at VQW=8 V) as a function of wavelength is shown in Fig. 5. The wavelength range for which our AFPQWM achieves a contrast ratio over 10 dB is 9.6 nm, compared to a range of 6.6 nm if our cavity was fixed at the optimal position. The maximum contrast ratio drops for wavelengths outside of this range due to the decreased absorption of the heavy-hole excitonic transition away from its peak wavelength. Also note that the insertion loss for this device is 8.5 dB at wavelengths corresponding to optimal contrast.
Themodulation speed of the AFPQWMcan be investigated by driving VQW with a sinusoidal signal between 0 V and 8 V while monitoring the time-dependent optical reflectance signal. This data is shown in Fig. 6 for a 1 MHz signal for a device characterized by a matching condition of Vbridge=5.8 V at a wavelength of 1558.6 nm. The contrast is excellent, showing that even at modulation speeds of 1 MHz, our AFPQWM provides extremely high extinction at the optimal wavelength and microbridge height. Also shown in Fig. 6 is the reflected optical power measured for a non-optimizedmicrobridge height (Vbridge=2.0 V), emphasizing the importance of cavity tunability to achieve high extinction. The frequency doubling for the non-matched data is due to the proximity of the quantum well’s absorption inflection point to the operation wavelength. Small signal frequency domain measurements (not shown) indicate that the 3dB frequency bandwidth of our devices to be approximately 1.5 MHz. This value agrees with the calculated RC-limited 3 dB bandwidth of our p-contact layers, 1.6±0.8 MHz, which is dominated by the calculated p-layer resistance of approximately 2 kΩ.
4. Discussion and summary
Due to careful MBE growth and microfabrication, the measured characteristics of our AFPQWM are in remarkable agreement with our calculations. The measured insertion loss for our devices at the matched wavelength and microbridge height is typically 10 dB. Next generation devices can be designed to have a lower microbridge reflectivity and more quantum wells, enabling significantly lower insertion losses in the matched condition. Such modifications would also increase the optical bandwidth over which high extinction modulation can be obtained. The contrast ratios of the current devices are most likely limited by a small amount of microbridge curvature (a few hundred nanometers over microbridge lengths of about 100 µm), diffraction losses caused by narrow microbridges, and the lack of an anti-reflection coating on the backside of the substrate. Although the measured contrast ratios are sufficient for many applications and represent an enormous improvement over conventional quantum well modulators, we anticipate that future devices with flatter and wider microbridges and backside anti-reflection coatings will improve the contrast ratios even further.
Our current modulators were designed for low-speed operation, though even modulation rates of a few MHz are sufficient for applications in free-space laser communication. Future generation modulators will operate with a decreased resistivity in the middle ground contact layer, achieved with a combination of different mesa sizes, better placement of metal contacts, and alternate doping schemes (such as p-n-p designs). These changes, combined with the use of coupled quantum wells  for lower electroabsorption voltages, should enable high-contrast, low-voltage modulation at speeds in the range of tens of MHz.
In summary, we have demonstrated a microelectromechanically tunable asymmetric Fabry-Perot quantum well modulator that operates in the optical C-band. By electrostatically actuating a suspended InGaAlAs microbridge above a quantum well modulator grown onto a InP/InGaAlAs DBR, we can achieve near-perfect cavity matching. This enables a surface-normal MQW modulator to be used in applications that demand high-extinction-ratio modulation, such as high-sensitivity photonic networks that utilize photon-counting photoreceivers. In addition, by using an approach that exploits the electronic, optical, and mechanical properties of epitaxially grown III-V semiconductors, our device avoids the use of more complicated fabrication steps such as wafer bonding, dielectric mirror growth, regrowth, or metallic micromachining.
References and links
1. H. Mohseni, W. K. Chan, H. An, A. Ulmer, and D. Capewell, “Tunable surface-normal modulators operating near 1550 nm with a high-extinction ratio at high temperatures,” IEEE Photon. Technol. Lett. (18), 214–216 (2006). [CrossRef]
2. R. N. Pathak, K. W. Goossen, J. E. Cunningham, and W. Y. Jan, “InGaAs-InP (MQW)-N Surface-Normal Electroabsorption Modulators Exhibiting Better Than 8:1 Contrast Ratio for 1.55 µm Applications Grown by Gas-Source MBE,” IEEE Photon. Technol. Lett. (6), 1439–1441 (1994). [CrossRef]
3. T. H. Stievater, W. S. Rabinovich, P. G. Goetz, R. Mahon, and S. C. Binari, “A Surface-Normal Coupled-Quantum-Well Modulator at 1.55 Microns,” IEEE Photon. Technol. Lett. (16), 2036–2038 (2004). [CrossRef]
4. W. S. Rabinovich, P. G. Goetz, R. Mahon, L. Swingen, J. Murphy, M. Ferraro, J. H. Ray Burris, C. I. Moore, M. Suite, G. C. Gilbreath, S. Binari, and D. Klotzkin, “45-Mbit/s cat’s-eye modulating retroreflectors,” Opt. Eng. (46), 104001 (pages 8) (2007). [CrossRef]
5. H. Liu, C. C. Lin, and J. S. Harris, “High-speed, dual-function vertical cavity multiple quantum well modulators and photodetectors for optical interconnects,” Opt. Eng. (40), 1186–1191 (2001). [CrossRef]
6. Y. Ding, R. M. Brubaker, D. D. Nolte, M. R. Melloch, and A. M. Weiner, “Femtosecond pulse shaping by dynamic holograms in photorefractive multiple quantum wells,” Opt. Lett. (22), 718–720 (1997). [CrossRef] [PubMed]
7. P. G. Goetz, R. Mahon, T. H. Stievater, W. S. Rabinovich, and S. C. Binari, “High-Speed Large Area Surface-Normal Multiple Quantum Well Modulators,” in Free-Space Laser Comm. & Active Laser Illumination III, D. G. Voelz and J. C. Ricklin, eds., pp. 346–354 (2004).
8. M. Whitehead, A. Rivers, G. Parry, J. S. Roberts, and C. Button, “Low-voltage multiple quantum well reflection modulator with on-off ratio greater than 100:1,” Electron. Lett. (25), 984–985 (1989). [CrossRef]
9. S. J. B. Yoo, M. A. Koza, R. Bhat, and C. Caneau, “1.5 µm asymmetric Fabry-Perot modulators with two distinct modulation and chirp characteristics,” Appl. Phys. Lett. (72), 3246–3248 (1998). [CrossRef]
10. R. I. Killey, C. P. Liu, M. Whitehead, P. Stavrinou, J. B. Song, J. S. Chadha, D. Wake, C. C. Button, G. Parry, and A. J. Seeds, “Multiple-Quantum-Well Asymmetric Fabry-Perot Modulators for Microwave Photonic Applications,” IEEE Trans. Microwave Theory Tech. (49), 1888–1892 (2001). [CrossRef]
11. G. L. Christenson, A. T. T. D. Tran, Z. H. Zhu, Y. H. Lo, M. Hong, J. P. Mannaerts, and R. Bhat, “Long-Wavelength Resonant Vertical-Cavity LED/Photodetector with a 75-nm Tuning Range,” IEEE Photon. Technol. Lett. (9), 725–727 (1997). [CrossRef]
12. Q. Chen, G. D. Cole, E. S. Bjorlin, T. Kimura, S. Wu, C. S. Wang, N. C. MacDonald, and J. Bowers, “First demonstration of a MEMS tunable vertical-cavity SOA,” IEEE Photon. Technol. Lett. (16), 1438–1440 (2004). [CrossRef]
13. D. Vakhshoori, P. Tayebati, C.-C. Lu, M. Azimi, P. Wang, J.-H. Zhou, and E. Canoglu, “2 mW CW single-mode operation of a tunable 1550 nm vertical cavity surface emitting laser with 50 nm tuning range,” Electron. Lett. (35), 900–901 (1999). [CrossRef]
14. W. S. Rabinovich, T. H. Stievater, N. A. Papanicolaou, D. S. Katzer, and P. G. Goetz, “Demonstration of a micro-electromechanical tunable asymmetric Fabry-Pérot quantum well modulator,” Appl. Phys. Lett. (83), 1923–1925 (2003). [CrossRef]
15. M. H. M. Reddy, T. Asano, R. Koda, D. A. Buell, and L. A. Coldren, “Molecular beam epitaxy-grown Al-GaInAs/InP distributed Bragg reflectors for 1.55 µm VCSELs,” Electron. Lett. (38), 1181–1182 (2002). [CrossRef]