Title: Silicon Optical Modulators
1Silicon Optical Modulators
- Recent developments in fabrication of High Speed
Modulators - J
- ee290f
2Outline
- Motivation
- Png, Reed, et al. work from Surrey.
- Shows basic principle and gives one of the two
major design types - Intel device
- History making device, designed and fabricated in
alternate major design type. - Conclusions
3Motivation
- Very clear Si modulators means CMOS integration
and using all our experience in silicon
micromachining. - Unfortunately, prior to 2003 the fastest Si
optical modulator was 20MHz (Lithium Niobate
modulators are 10GHz).
4p-i-n Si Optical Modulators 1/4
- Follows the work presented by Png, Reed, et al
from Surrey University (UK).
5p-i-n Si Optical Modulators 2/4
VD is applied to Anode vs. Cathode. ? forward
biased p-i-n junction. ? e and holes injected
into guiding region ? changes refractive index
6p-i-n Si Optical Modulators 3/4
?n ?ne ?nh -8.8 x 10-22(?Ne) 8.5 x
10-18(?Nh)0.8 ?? ??e ??h 8.5 x 10-18 (?Ne)
6.0 x 10-18(?Nh) From R.A. Soref B. R.
Bennett Electrooptical Effects in Silicon Jour.
Of Quan. Elec. 1987.
7p-i-n Si Optical Modulators 4/4
- Simulation results show modulation can be
optimized to 1.3GHz and a 180? at .7 mA of
current. - However, performance is very dependant on doping
profile and a critical dimensions are not very
tolerant. - As of late 2003, fabrication is underway.
8MOS Si Optical Modulators 1/4
- Still uses the plasma dispersion effect, but
implements a MOS capacitor to induce change in
free carrier density instead of a - p-i-n device.
- Again designed for single mode 1.55?m.
9MOS Si Optical Modulators 2/4
- Apply VD to poly. Charge accumulation on both
sides of gate oxide. - ?Ne ?Nh ?/etoxt(VD VFB)
- ?ne -8.8 x 10-22?Ne
- ?nh -8.5 x 10-18(?Nh)0.8
- ?? (2?/?)?neffL
10MOS Si Optical Modulators 3/4
- Implemented phase shifter in both arms of a MZI.
- For VD 7.7V ? 16dB total switch.
- Loss is the big key 15.3dB insertion loss
(4.3dB from coupling 6.7dB from poly guides).
11MOS Si Optical Modulators 4/4
Switching test on psuedo-random time signal.
See 3dB roll off at gt1GHz from MOS cap.
12Conclusion
- MOS modulator has poor loss figures and still an
order of magnitude slower than commercial
modulators. Intel argues these can both be
theoretically fixed by decreasing the size of the
device and using Si in the guide region instead
of poly (still integrable?). - P-i-n modulator is still being fabricated and
depends on its optimal design for the high values
achieved, so potentially success or failure from
fabrication runs.