High speed, low driving voltage vertical cavity germanium-silicon modulators for optical

1.2 Optical Interconnect Systems 
In order to realize high-speed inter-chip or intra-chip interconnects, nanophotonics 
would be the best replacement for electrical interconnects if low cost and integration 
can be achieved. This has multiple advantages: (1) Low power: It is generally 
considered that power consumption in the data links will be critical factor for 
computation speed of computers. When devices scale down and circuit operation 
speed increases, the traditional electrical interconnects will consume more power than 
the system can handle. Light traveling in a proper medium, such as fibers or silicon 
waveguides, has nearly zero power loss over the distance ranges where electrical 
interconnects still exist (< 1 mile), (2) High bandwidth: An electrical wire is a low 
pass filter and will limit the bandwidth of data transmission. At high frequency, the 
current takes the path of least inductance which is primarily on the surface of a 
transmission line, which is known as the “skin effect” This will decrease the 
conductance and increase attenuation. This frequency-dependent attenuation due to the 
skin effect causes the signal pulse to spread out, and leads to inter symbol interference 
(ISI). An optical light-wave is an electromagnetic wave with bandwidth as high as 
200THz. It can carry a signal without changing its frequency or propagation and it is 
ITRS Roadmap 2005
Global
interconnects
Local connects
CMOS device

---

 
 
 
6 
immune to interference. (3) Cross talk free: An electrical wire acts as a good antenna 
at high frequencies, and it broadcasts its signal to adjacent wires through inductive and 
capacitive coupling between transmission lines. At high frequency, electromagnetic 
interference (EMI) will limit the density of electrical interconnects. Optical signals, on 
the other hand, are inherently immune to EMI. They don’t detect and generate RF 
signals and their wavelength multiplexing capability is strong [17]. 
Figure 1.4 Optical interconnect system and building blocks [11] 
Optical interconnect systems consist of four parts: light source, waveguide, 
modulation and photo-detection as shown in Fig. 1.4. The light source can be either an 
off-chip or on-chip laser. The modulator can be an electro-refractive Mach-Zehnder 
modulator, or an electro-absorptive surface normal or waveguide modulator. The 
carrier channels can be silica fibers, free-space air, or waveguides. Existing Si 
waveguide technology is based on SiGe/Si or silicon-on-insulator (SOI) [18, 19]. The 
detectors 
can 
be 
p-i-n 
diodes 
(low 
noise, 
unity 
responsivity), 
metal-semiconductor-metal (MSM) diodes (short response time), or avalanche 
photodiodes (APDs). Group-IV materials, such as silicon or germanium, have already 
been used as photodetectors [20]. There is also mature technology for optical carrier 
channels and receivers based on silicon-compatible technology. The key obstacle to 
realize optical interconnects is the transmitter. Prior to recent work [12], there was no 
efficient Si-based modulation mechanism and this function was only implemented by 

---

 
 
 
7 
expensive III-V compound semiconductor devices. Thus while the optical interconnect 
systems were promising, there was virtually no means to implement them for 
short-distance inter-chip or intra-chip communications. 
Modulators are favored instead of direct-driven lasers for several reasons. (1)There 
is no efficient light emission in group IV materials yet. Silicon and germanium are 
indirect band-gap and tin is a semi-metal. It is hard to see a path to find a reliable 
group IV laser device today. (2) In order to modulate lasers at high bit rates, they must 
be pre-biased and driven at current densities well above threshold, which consumes 
high power and leads to performance degradation and failure [21]. (3) The heat 
generation from lasers is undesired for CMOS chips. The temperature variation in 
CMOS chips also causes wavelength shifts and this instability can prohibit precise 
channel allocations for multiple wavelength multiplexing in the same medium, such as 
wavelength-division-multiplexing (WDM) schemes. (4) It is very hard to incorporate a 
large number of lasers on a single chip. The complexity of the fabrication process will 
increase the cost dramatically. So we prefer to use on-chip modulators as the solution 
for transmitters and modulate the light coming from an off-chip continuous-wave (CW) 
laser.