Graphene electro-optic modulator with 30 GHz bandwidth Christopher T. Phare 1, Yoon-Ho Daniel Lee 1, Jaime Cardenas 1, and Michal Lipson 1,2,* 1School of Electrical and Computer Engineering, Cornell University, Ithaca, NY 14850, USA 2Kavli Institute at Cornell for Nanoscale Science, Cornell University, Ithaca, NY 14850, USA *Corresponding Author: ml292@cornell.edu Transmission through a bus waveguide coupled to a ring resonator is described by: PP oooooo = PP iiii aa 2 aaaa cos φφ + tt 2 1 2aaaa cccccc φφ + aa 2 tt 2 φφ = 2ππ λλ nn ggll Where P in is the incident optical power, a is the round-trip ring transmission, t is the ringwaveguide self-coupling coefficient (controlled by the gap between ring and waveguide; t=1 indicates no ring-waveguide coupling), λ is the incident wavelength, n g is the ring group index, and L is the ring circumference. To model experimental data in Figure 3, we simultaneously fit all six datasets, allowing each curve to have an independent a and n gl, while forcing t and P in to be shared among all curves. Modulating the carrier concentration on the graphene modulates the round-trip loss a. NATURE PHOTONICS www.nature.com/naturephotonics 1
Supplementary Figure 1: Transmission Contours Transmission on-resonance (ϕ=0), neglecting graphene phase modulation effects, for t = {0.7, 0.8, 0.9, 0.975}. Black squares indicate the operating points of our experimental device, which has t = 0.70. Note that lower losses in the ring resonator would permit devices with lower insertion loss and higher transmission slopes. Graphene s loss under bias, added to any loss from the waveguide, sets the parameter a. Residual loss (from, e.g., not completely gating the graphene or optical absorption in metal contacts) sets a maximum possible round-trip transmission. We then can choose t arbitrarily for a tradeoff between insertion loss (1-T) and voltage sensitivity ( T/ a a/ V). Operating near critical coupling gives the best sensitivity to changes in round-trip loss (Supplementary Figure 2). 2 NATURE PHOTONICS www.nature.com/naturephotonics
SUPPLEMENTARY INFORMATION Supplementary Figure 2: Coupling Optimization Transmission sensitivity to changes in roundtrip loss (db scale) as a function of ring-waveguide coupling t, for a=0.676 (the maximum roundtrip transmission achieved in experiment). The vertical red line indicates our coupling coefficient; the dashed black line is the theoretical optimum at critical coupling (a=t). NATURE PHOTONICS www.nature.com/naturephotonics 3
Supplementary Figure 3: Graphene Raman Spectra Raman spectra for bottom (black) and top (red) layer of graphene, with 514 nm excitation. Baseline photoluminescence from other materials has been subtracted. G peaks are at 1601 and 1597 cm -1 and 2D peaks at 2717 and 2708 cm -1, respectively. These peak positions indicate a p-type doping of approximately 5 x 10 12 cm -2. 4 NATURE PHOTONICS www.nature.com/naturephotonics
SUPPLEMENTARY INFORMATION Supplementary Figure 4: Broadband Resonance Modulation Transmission spectrum of the device from 1520 nm to 1620 nm, for 0 V (black) and -50 V (pink) DC bias. Note that modulation is significant for every resonance, as expected from graphene s broadband absorption. NATURE PHOTONICS www.nature.com/naturephotonics 5
We perform electrical measurements on ALD Al 2O 3 layers to verify their robustness under high bias voltages. On a silicon wafer, we deposit 50 nm of platinum via DC magnetron sputtering and then coat this layer with varying thicknesses of Al 2O 3. We then evaporate 1 cm squares of 5 nm Ti / 45 nm Au, forming metal-insulator-metal capacitors. We access the platinum bottom electrode via selective buffered oxide etch of the dielectric. We characterize the dielectric breakdown voltage by I-V curve analysis and extract a breakdown field of 1.0 GV/m from the linear fit (Supplementary Figure 5). In comparison, typical field strength in our graphene devices is 460 MV/m at 30 V DC bias. Supplementary Figure 5: Al 2O 3 Breakdown Voltage We extract breakdown voltage from I-V analysis of Pt/Al 2O 3/Ti/Au capacitors with varying thicknesses of ALD Al 2O 3. The linear fit (red) gives a breakdown field of 1.0 GV/m. Capacitance of the dielectric film is reasonably constant with applied voltage, as shown in C-V measurements for the 45 nm-thick capacitor (Supplementary Figure 6). Capacitance changes less than 5% over a ±25 V bias range. Frequency response of our modulator should thus be relatively independent of bias voltage. The dielectric is also robust over time; we detect a capacitance change of less than 0.3% over 30 minutes of continuous 25 V DC bias. 6 NATURE PHOTONICS www.nature.com/naturephotonics
SUPPLEMENTARY INFORMATION Supplementary Figure 6: Al 2O 3 Capacitance-Voltage Curve Capacitance of the 45 nm-thick MIM capacitor vs. bias voltage, measured with an LCR meter at 1 khz. NATURE PHOTONICS www.nature.com/naturephotonics 7
Supplementary Figure 7: Eye Diagram Test Setup The testing setup consists of an optical (red) and electrical (blue) arm. The external-cavity diode laser is amplified twice to compensate for chip facet and fiber component insertion losses, then sent through a bandpass grating filter to remove amplified spontaneous emission noise. RF PRBS signals (Centellax TG1P4A) are amplified in a 40 Gbps modulator driver (Centellax OA4MVM3) and biased before contacting the device with a GGB 40A picoprobe. A second probe blocks DC bias and terminates the RF signal to avoid reflections from the device. 8 NATURE PHOTONICS www.nature.com/naturephotonics