We demonstrate a ring-resonator modulator based on a silicon-polymer hybrid slot waveguide with a tunability of 12.7 pm/V at RF speeds and a bandwidth of 1 GHz, for optical wavelengths near 1550 nm. Our slot waveguides were fabricated with 193 nm optical lithography, as opposed to the electron beam lithography used for previous results. The tunability is comparable to some of the best ring-based modulators making use of the plasma dispersion effect. The speed is likely limited only by resistance in the strip-loading section, and it should be possible to realize significant improvement with improved processing.
©2011 Optical Society of America
Ring resonator modulators are promising candidates for a number of applications in silicon photonics. These modulators have until now been based almost exclusively on the free-carrier plasma dispersion effect . In recent work such devices have been used, for example, as switches , tunable filters , and modulators [4,5]. The advantages of ring resonator modulators include their compact footprint, low drive-voltages, and the ability to drive them as lumped RF elements, eliminating the need for traveling-wave design . However, free-carrier plasma dispersion based ring-resonator modulators have several key limitations. Perhaps most importantly, the tunability is still fairly low. Most free-carrier based modulators are based on a carrier depletion mechanism, and the best of these devices typically achieve tunabilities on the order of 10-20 pm/V [6–8]. Using carrier injection in forward-biased P-I-N junctions, much larger DC tunabilities of up to 138 pm/V have been shown . However, the latter approach suffers from a severe bandwidth limitation; a 6.9 Vpk-pk drive voltage was required to drive the device at 1.5 Gb/s suggesting a tunability of only 4 pm/V at this speed. A more recent result attained a speed of 18 Gb/s in a carrier-injection device, but this was done with a similarly large modulating signal of 4 Vpk-pk combined with an additional +/− 2V of pre-emphasis . While more recent work has shown dramatically lowered drive voltages at 1 Gb/s speeds, increasing the speed to 10 Gb/s is projected to still require significantly larger drive voltages . This limitation is likely intrinsic for the forward-biased mechanism, and large improvements in speed while maintaining high tunability may not be possible. Carrier depletion based modulators have bandwidth limitations as well, though at higher speeds; to our knowledge, the highest bandwidth ring resonator modulator in silicon based on the free-carrier plasma dispersion effect is 35 GHz .
Silicon-polymer hybrid slot waveguide modulators have the potential to address both of these limitations. First, the tunability can potentially be substantially higher , once narrower slots and higher activity polymers are used. In fact, tunabilities of 42 pm/V have already been demonstrated, though at speeds around 6 MHz . Additionally, silicon-polymer hybrid slot waveguides also show a path to significantly higher bandwidth operation; phase modulation at 100 Gb/s has been demonstrated in such waveguides , and polymer-based rings have been shown to have bandwidths of over 165 GHz . The bandwidths of these polymer-based modulators are well beyond anything that has been achieved with free-carrier plasma dispersion devices. The reason for the difference in performance is that free-carrier based devices rely on changing a concentration of electrons or holes in the optical path. As a result, they will ultimately be limited by the carrier drift velocities. However, slot waveguide modulators only utilize the silicon as a transparent conductor, and are not limited by drift velocities as a result.
The performance achieved thus far in slot ring modulators has fallen far short of the theoretical limits. In fact, to our knowledge the highest bandwidth slot ring modulator demonstrated has been the aforementioned result of 6 MHz . This speed is sufficiently slow that it even potentially raises questions about the mechanism of the result; thermal effects, which typically cause large shifts in resonance peaks, have been observed in silicon waveguides at speeds of 100 kHz . Here we demonstrate a silicon-polymer hybrid slot waveguide ring-resonator modulator with a 6 dB bandwidth of 1 GHz, a high-speed device tunability of 12.7 pm/V, and a Q value of 5000. This is an increase in speed by well over an order of magnitude over the previous best result. The factor of 3.5 decrease in tunability compared to the previous result  is readily explained by the use of a different non-linear material for the cladding as well as geometric differences between the two devices [12,13].
2. Design and fabrication
We fabricated our slot ring modulator with 193 nm optical lithography. Two self-aligned patterning steps and two dry etching steps were performed on a silicon-on-insulator wafer with a 220 nm-thick top silicon layer and a 2 µm-thick buried oxide layer. This enabled us to define regions of un-etched silicon, partially etched silicon regions having a thickness of 77 nm (for strip-loading), and finally completely etched silicon. A 10 nm-thick thermal oxide was then grown on the silicon for the purpose of surface passivation, consuming approximately 5 nm of silicon.
As a post-processing step, the whole chip was given a blanket implant of 5x1017 cm−3 P. A subsequent masked implant of 1x1020 cm−3 P and metal deposition process (100 nm Au on top of 1.5 µm Al) were used to define contact pads. The masked post-processing steps were performed with relatively inaccurate alignment (due to the limitations of our contact photolithography systems) and required a large clearance of approximately 8 µm between the metal pads and the waveguides. These large clearances contribute to the bandwidth limitation of our device; a much smaller clearance should be possible in future devices. Using the results of FDTD simulations, we estimate that a pad clearance of 3 µm would result in only 0.1 dB/cm of excess loss.
The final geometry of the strip-loaded slot waveguide utilized for the modulator consisted of 230 nm wide arms, and a slot that was 200 nm wide. Losses in this waveguide geometry of 8 dB/cm have been demonstrated , though we anticipate that the losses are higher here due to post-processing. Figure 1 shows an SEM micrograph of the cross-section of a typical strip-loaded slot waveguide, as well as a contour plot of the solved mode pattern. A ring resonator was constructed out of this waveguide with a bend radius of 60 µm. A dark-field optical micrograph is shown in Fig. 2 , as well as a passive wavelength sweep through the device. Coupling on and off chip was achieved with grating couplers .
AJSP-series electro-optic polymers are a recently developed series of materials and have been shown to exhibit high electro-optic activity at telecommunication wavelengths, with r33 values of 50-200 pm/V . It is important to distinguish this parameter, which measures the polymer nonlinearity, from the resonator tunability, a quantity also reported in pm/V. Our device was spin-coated with PMMA doped with 15 wt% ALJZ53 chromophore to form the cladding, after an oxide etching step. This material has a measured refractive index of 1.54 at a wavelength of 1550 nm and has been demonstrated to have an r33 of 60 pm/V as a thin film. In order to make electrical contact with the device, the polymer was cleared from contact pads using a laser ablation system, leaving bare metal. The material was poled by applying 20 V across the slot at a temperature of 100° C for 30 seconds, followed by a rapid cooling to room temperature. This was done by contacting the pads with a 150 µm-pitch GSG configured probe with the poling voltage applied to the center pad and the outer two pads held at ground.
3. Experimental results
A tunable fiber-coupled laser provided continuous-wave optical input to our device, and TE polarization was ensured before coupling to the chip. In order to make electrical contact with the device, a 150-µm-pitch RF microprobe was brought into contact with the metal pads. The two outer pads were held at ground while the signal voltage was applied to the pad inside the ring structure.
3.1. DC measurement
The DC tunability of the device was determined by measuring transmission as a function of wavelength with the device biased at several DC voltages, using a low-speed optical power detector. This was done in a non-sequential fashion to exclude thermal drift as the cause of the peak shifting behavior. The behavior is illustrated in Fig. 3 . A linear regression  of the peak shifts showed a device tunability of 16.5 ± 0.6 pm/V (Fig. 3 inset). This tunability is comparable to the best reported value for depletion-based ring modulators . The implied polymer r33 is calculated to be 19 pm/V , as compared to the peak r33 expected for this polymer of 60 pm/V. Further work will be required to determine the source of the poor poling efficiency achieved. Had the peak material r33 value been obtained, device tunability near 52 pm/V would have been demonstrated. This can also be favorably compared with the tunability reported in polymer-based rings that do not utilize silicon slot waveguides – to our knowledge, the best result is 7 pm/V, achieved with a polymer having a larger material r33 of 30 pm/V .
As determined at the peak used for RF measurements, the free-spectral range of the ring was 2 nm and the Q-factor was approximately 5000. This Q value is lower than the ultimate Q of 18k  that could be achieved with a waveguide loss of 8 dB/cm. Loss measurements of non-metalized strip-loaded slot waveguides on the same chip as the device show a significantly higher loss of around 35 dB/cm, in close agreement with the measured Q value of the ring. However, loss measurements taken for the same waveguides on an un-doped chip coated with the same polymer cladding yielded loss values of approximately 10 dB/cm. We can thus discount both the metal contacts and the polymer cladding as major sources of loss. One possibility is that the excess loss is due to improperly annealed dopants acting as absorption centers; the measured resistivity of the un-etched silicon is 0.15 Ω⋅cm, compared with an expected value of 0.03 Ω⋅cm for the given doping. It is also possible that the surface of the silicon endured a large amount of damage during the oxide-clearing etch, which would explain both the higher than expected resistance and waveguide loss.
3.2. RF measurement
High-frequency measurements were taken with a New Focus 1647 1.1 GHz bandwidth avalanche photodiode (APD). The wavelength of the source was set to 1552.15 nm, and varied such that the maximum response was obtained for each measurement, biasing the device at the location of the steepest slope in the resonator wavelength spectrum. The drive signals used were sufficiently small to ensure a predominantly linear response of the device. A variation in the wavelength of the source of only ± 20 pm was required over the course of all measurements to keep the laser at the point of maximum RF response. For the frequency range from 30 kHz to 20 MHz, the modulator was driven by a signal generator while the output from the APD (AC coupled) was fed into a lock-in amplifier. For the frequency range from 20 MHz to 5 GHz, a virtual network analyzer (VNA) was used to provide the input and read the output. The two configurations are illustrated schematically in Fig. 4 .
The S21 parameter of a system is the ratio of the system output RF power to the input RF power. The experimental setup described above allowed us to measure this parameter for the electro-optic system contained between the source and receiver of the electrical signal. The resulting S21 value can be described by the equation:Figure 5 displays the S21 parameter of the device, normalized against the measured value over the flat response region located between 10 MHz and 100 MHz. The 6 dB roll-off in S21, corresponding to the half-power point in the electro-optic response of the device, was determined to be around 1 GHz. This is likely a result of the parasitic RC circuit created by the slot in series with the strip load. The bandwidth was confirmed by measurements taken with a 25 GHz photodetector to rule out the influence of the APD cutoff. A typical signal-to-noise value observed for both the RF lock-in measurements and the S-parameter measurements was 50 dB. A significantly lower value of 30 dB was seen in the frequency range from 20 MHz to 40 MHz. A solid black line in Fig. 5 indicates the expected S21 parameter based on the DC tunability, the photodetector responsivity, and the measured optical losses in the system.
It is important to note that the agreement between the RF measurements and the predicted value from the DC measurements do not exactly agree at 10 MHz and beyond, even though the response is relatively flat; there is approximately a 2.3 dB discrepancy. There is a slight increase in response at lower frequencies, with nearly perfect matching obtained around 100 kHz. We attribute this discrepancy to a portion of the ring not being charged at higher frequencies, due to extra parasitic resistances. In Fig. 2, one can see that the top and bottom portions of the ring do not have metal in close proximity to the slot waveguide. In the case of the bottom portion, this was unavoidable due to the presence of the add waveguide. The result of these non-contacted areas is that the equivalent circuit model of the device is a network of resistive and capacitive elements, with each capacitor representing a different portion of the ring. As a result, different parts of the ring have different cut-off frequencies.
Figure 6 (a) is an illustration of this model. The cut-off frequency of the bulk of the ring is highest due to its greater proximity to the contacts. As seen in the figure, there are two regions of the ring that do not have grounding electrodes in close proximity to the waveguide; there is a region at the top of the ring where the metal has been discontinued, though a relatively wide strip-loaded portion remains, and a region at the bottom of the ring, where the add waveguide passes the ring. In the bottom region, the metal is excluded due to the presence of an additional waveguide. Additionally, the close proximity of a second slot waveguide limits the lateral size of the silicon region where current can flow. For the ring tested, the edge to edge spacing of the slot waveguides is only 0.55 µm. The length of this region is approximately 80 μm, as compared to the entire circumference of the ring, which is 400 μm. This amounts to 20% of the ring length. As a result, if this region were not fully charged at higher frequencies, one would expect the tunability of the ring to decrease by around 20%; this is because the tunability should be linearly proportional to the fraction of the ring that is fully charged at a given frequency. In fact, this is close to the observed behavior; for frequencies above 20 MHz, the 2.3 dB decrease in S21 could be explained, via Eq. (1), as a 23% reduction in tunability. This is the source of our high-speed tunability figure of 12.7 pm/V, a reduction of 23% of the DC value of 16.5 pm/V. We conclude that it is likely that in the frequency regime from 20 MHz to the final roll-off, the bottom portion of the ring is not being fully charged. The top portion likely begins to be not fully charged out at a frequency closer to the overall bandwidth of the structure, and so does not appear to be distinct from the final roll-off in the S21 trace.
In order to more fully understand this effect, we present an equivalent circuit model for the bottom portion of the ring that is not contacted, which is presented in Fig. 6 (b). Here, the circuit that forms the bottom portion of the ring consists of a series of differential elements, with a perfect voltage source driving one half of the slot through the a resistor, while the outer portion of the ring consists of a series of resistors, due to the outer rim of silicon. Using the silicon resistivity measured in section 3.1, we calculate a value for R0 of around 30 kΩ/μm. Using a dielectric constant of 4 for the polymer, finite element simulations suggest a value for C of around 0.07 fF/μm . Based on the device bandwidth, we can then determine R1 to be = 1.2 MΩ⋅µm. As it turns out, the value of R1 has very little effect on the results of the analysis described below. Now, take z as the direction of travel around the edge of the ring, take L to be the length between the two outer ground electrodes, and take z = 0 to be the halfway point between the two ground electrodes in the bottom coupling portion of the ring. Analysis of this differential element yields the following system of differential equations under the assumption that V0(t) = V0exp(iωt):
In order to solve for the frequency response of these sections, the voltage across the differential capacitance must be integrated from the edge of one contacted region to the edge of the other along the z-direction (for frequencies lower than the cutoff of the main part of the ring):
The quantity [Eq. (6)] is directly proportion to the amount of tunability that the bottom region of the ring contributes to the net tunability of the ring. At low frequencies, the outer portion of the ring remains at ground, and so Eq. (6) simply reduces to V0L. As the frequency, and therefore the magnitude of α, increases, the quantity in Eq. (6) will fall. At 20 MHz, with the values for R0, R1, L and C reported above, the quantity in Eq. (6) takes on a value that is still more than 97% of its DC value in magnitude; this is not in agreement with our experimental data, in which this portion of the ring appears to be nearly completely uncharged at 20 MHz and above. We believe the discrepancy is due to a change in the sheet resistance of the unetched silicon layer in the region of our device, as compared to the control structures. The sheet resistance in our control structures was measured from ridges that were 0.5 μm, while the ridges in the bottom portion of the ring are 0.23 μm in width. If we postulate that R0 is larger by a factor of 10, then the decrease in tunability for this portion of the ring would be 60% at 20 MHz, placing the predicted roll-off nearly within a dB of the observed value. It is also possible that electrostatic coupling to the adjacent waveguide plays a role. In any case, further study will be needed to determine the precise RF bandwidth limitation for this portion of the ring.
For some applications, a relatively flat response from 20 MHz and higher may be acceptable, but in some situations the lower speed discrepancy may prove problematic. Unlike carrier depletion modulators, there is no easy way to end the region of modulation in a slot waveguide ring, since the slot must continue around the entire device. It will be difficult to extend the metal around the entire length of the ring, and so there will always be at least a small portion of the slot with poor electrical contact. Further design work will be needed to address this fundamental challenge of slot waveguide ring modulators, and to make the frequency response completely flat. One possible approach will to use a masked implant, in order to raise the resistivity of this portion of the ring, pushing the low-frequency transition region below 1 MHz. It might also be possible to shorten the coupling region with more careful design work.
We have demonstrated a slot waveguide ring resonator modulator with a high-speed device tunability 12.7 pm/V and a 6 dB bandwidth of 1 GHz. This is an improvement of more than two orders of magnitude with respect to the previous highest bandwidth device of this kind and represents a significant step toward making slot waveguide based ring modulators competitive with state-of-the-art plasma dispersion-based devices. This result not only provides conclusive evidence of the electro-optic modulation mechanism, but also lays down a clear path to further improving the metrics of similar devices.
Decreasing the resistance between the contacts and the slot should improve the speed of the device. We expect multiple implant steps and improved annealing to improve this metric significantly, while also reducing waveguide loss and increasing the quality factor of the ring. Reducing the metal-to-waveguide spacing and redesigning the layout so that more of the ring is in closer proximity to the contacts should help as well. There are also straightforward ways of improving device tunability. Because the electric field in the slot increases almost exactly as the reciprocal of the width , we can expect to improve the tunability greatly by simply reducing the slot width in our waveguide geometry. We can expect to improve performance based on improved materials as well; material r33 values of 250 pm/V have already been demonstrated in thin films at telecommunication wavelengths . Taken together, these facts suggest that an order of magnitude improvement should be possible in tunability, which would make future devices comparable in performance to the current best forward biased P-I-N based resonators [4,9] but without the bandwidth limitation. We believe that silicon-polymer hybrid ring modulators will likely become important components in integrated photonics systems.
The authors would like to thank Gernot Pomrenke, of the Air Force Office of Scientific Research, for his support through a Presidential Early Career Award (FA9550-10-1-0053), and the STTR grant (FA9550-10-1-0053), and Warren Herman of the UMD Laboratory for Physical Sciences. The authors would like to acknowledge support from the National Science Foundation STC MDITR Center, DMR0120967 and from the Washington Research Foundation. M. Gould would like to thank NSERC for support through the CGS fellowship.
References and links
1. G. T. Reed, G. Mashanovich, F. T. Gardes, and D. J. Thomson, “Silicon optical modulators,” Nat. Photonics 4(8), 518–526 (2010). [CrossRef]
2. H. L. R. Lira, S. Manipatruni, and M. Lipson, “Broadband hitless silicon electro-optic switch for on-chip optical networks,” Opt. Express 17(25), 22271–22280 (2009). [CrossRef]
3. P. Rabiei and W. H. Steier, “Tunable polymer double micro-ring filters,” IEEE Photon. Technol. Lett. 15(9), 1255–1257 (2003). [CrossRef]
5. F. Y. Gardes, A. Brimont, P. Sanchis, G. Rasigade, D. Marris-Morini, L. O’Faolain, F. Dong, J. M. Fedeli, P. Dumon, L. Vivien, T. F. Krauss, G. T. Reed, and J. Martí, “High-speed modulation of a compact silicon ring resonator based on a reverse-biased pn diode,” Opt. Express 17(24), 21986–21991 (2009). [CrossRef] [PubMed]
6. P. Dong, S. Liao, D. Feng, H. Liang, D. Zheng, R. Shafiiha, C. C. Kung, W. Qian, G. Li, X. Zheng, A. V. Krishnamoorthy, and M. Asghari, “Low Vpp, ultralow-energy, compact, high-speed silicon electro-optic modulator,” Opt. Express 17(25), 22484–22490 (2009). [CrossRef]
7. J. B. You, M. Park, J. W. Park, and G. Kim, “12.5 Gbps optical modulation of silicon racetrack resonator based on carrier-depletion in asymmetric p-n diode,” Opt. Express 16(22), 18340–18344 (2008). [CrossRef] [PubMed]
8. P. Dong, R. Shafiiha, S. Liao, H. Liang, N. N. Feng, D. Feng, G. Li, X. Zheng, A. V. Krishnamoorthy, and M. Asghari, “Wavelength-tunable silicon microring modulator,” Opt. Express 18(11), 10941–10946 (2010). [CrossRef] [PubMed]
9. S. Manipatruni, Q. Xu, B. Schmidt, J. Shakya, and M. Lipson, “High speed carrier injection 18 Gb/s silicon micro-ring electro-optic modulator,” in Lasers and Electro-Optics Society (IEEE, 2007), pp. 537–538.
11. D. Gill, M. Rasras, K. Tu, Y. Chen, A. E. White, S. S. Patel, D. Carothers, A. Pomerene, R. Kamocsai, C. Hill, and J. Beattie, “Internal bandwidth equalization in a CMOS-compatible Si-ring modulator,” IEEE Photon. Technol. Lett. 21(4), 200–202 (2009). [CrossRef]
12. M. Hochberg, T. Baehr-Jones, G. Wang, J. Huang, P. Sullivan, L. Dalton, and A. Scherer, “Towards a millivolt optical modulator with nano-slot waveguides,” Opt. Express 15(13), 8401–8410 (2007). [CrossRef] [PubMed]
13. T. Baehr-Jones, M. Hochberg, G. Wang, R. Lawson, Y. Liao, P. A. Sullivan, L. Dalton, A. K.-Y. Jen, and A. Scherer, “Optical modulation and detection in slotted Silicon waveguides,” Opt. Express 13(14), 5216–5226 (2005). [CrossRef] [PubMed]
14. W. Freude, J. Leuthold, L. Alloatti, T. Vallaitis, D. Korn, R. Palmer, C. Koos, J. Brosi, P. Dumon, R. Baets, M. Scimeca, I. Biaggio, B. Breiten, F. Diederich, A. Barklund, R. Dinu, and J. Wieland, “100 Gbit/s electro-optic modulator and 56 Gbit/s wavelength converter for DQPSK data in silicon-organic hybrid (SOH) technology,” in Photonics Society Summer Topical Meeting Series (IEEE, 2010), pp. 96–97.
15. B. Bortnik, Y.-C. Hung, H. Tazawa, B.-J. Seo, J. Luo, A. K.-Y. Jen, W. H. Steier, and H. R. Fetterman, “Electrooptic polymer ring resonator modulation up to 165 GHz,” IEEE J. Sel. Top. Quantum Electron. 13(1), 104–110 (2007). [CrossRef]
16. R. L. Espinola, M.-C. Tsai, J. T. Yardley, and R. M. Osgood, “Fast and low-power thermooptic switch on thin silicon-on-insulator,” IEEE Photon. Technol. Lett. 15(10), 1366–1368 (2003). [CrossRef]
17. R. Ding, T. Baehr-Jones, W. J. Kim, X. Xiong, R. Bojko, J. M. Fedeli, M. Fournier, and M. Hochberg, “Low-loss strip-loaded slot waveguides in silicon-on-insulator,” Opt. Express 18(24), 25061–25067 (2010). [CrossRef] [PubMed]
18. L. L. Hope, “Theory of optical grating couplers,” Opt. Commun. 5(3), 175–182 (2010).
19. J. Luo, X.-H. Zhou, and A. K.-Y. Jen, “Rational molecular design and supramolecular assembly of highly efficient organic electro-optic materials,” J. Mater. Chem. 19(40), 7410–7424 (2009). [CrossRef]
20. W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in C: The Art of Scientific Computing, 2nd ed. (Cambridge University Press, 1992).
21. B. A. Block, T. R. Younkin, P. S. Davids, M. R. Reshotko, P. Chang, B. M. Polishak, S. Huang, J. Luo, and A. K.-Y. Jen, “Electro-optic polymer cladding ring resonator modulators,” Opt. Express 16(22), 18326–18333 (2008). [CrossRef] [PubMed]
22. A. Yariv, “Universal relations for coupling of optical power between microresonators and dielectric waveguides,” Electron. Lett. 36(4), 321–322 (2000). [CrossRef]
23. J. Witzens, T. Baehr-Jones, and M. Hochberg, “Design of transmission line driven slot waveguide Mach-Zehnder interferometers and application to analog optical links,” Opt. Express 18(16), 16902–16928 (2010). [CrossRef] [PubMed]
24. T. Kim, J. Luo, J. Ka, S. Hau, Y. Tian, Z. Shi, N. M. Tucker, S. Jang, J. Kang, and A. K.-Y. Jen, “Ultralarge and thermally stable electro‐optic activities from Diels–Alder crosslinkable polymers containing binary chromophore systems,” Adv. Mater. (Deerfield Beach Fla.) 18(22), 3038–3042 (2006). [CrossRef]