Vous êtes sur la page 1sur 3

Silicon photonics

(Nanowerk Spotlight) In its everlasting quest to deliver more data faster and on smaller
components, the silicon industry is moving full steam ahead towards its final frontiers of size,
device integration and complexity. We have covered this issue numerous times in previous
Spotlights. As the physical limitations of metallic interconnects begin to threaten the
semiconductor industry's future, one group of researchers and companies is betting heavily on
advances in photonics that will lead to combining existing silicon infrastructure with optical
communications technology, and a merger of electronics and photonics into one integrated dual-
functional device.
Today, silicon underpins nearly all microelectronics but the end of the road for this technology has
clearly come into view. Photonics is the technology of signal processing, transmission and
detection where the signal is carried by photons (light) and it is already heavily used in photonic
devices such as lasers, waveguides or optical fibers. Optical technology has always suffered from
its reputation for being an expensive solution, due to its use of exotic materials and expensive
manufacturing processes. This prompted research into using more common materials, such as
silicon, for the fabrication of photonic components, hence the name silicon photonics.
Although fiber-optic communication is a well-established technology for
information transmission, the challenge for silicon photonics is to manufacture low-cost
information processing components. Rather than building an entirely new industrial infrastructure
from scratch, the goal here is to to develop silicon photonic devices manufactured using
standard CMOS techniques. A recent review paper takes a look at the state of silicon photonics
and identifies the challenges that remain on the path to commercialization.
It is not only the big companies, like IBM and Intel, that are making headway in combining optical
and electronic elements on a single silicon CMOS chip. Smaller players such as Kotura, Lightwire
and Luxtera are already introducing silicon photonic components to the marketplace. IBM for
instance showcases its "Silicon Integrated Nanophotonics" project on its website. The ultimate
goal of this project is to develop a technology for on-chip integration of ultra-compact
nanophotonic circuits for manipulating the light signals, similar to the way electrical signals are
manipulated in computer chips.

Futuristic silicon chip with monolithically integrated photonic and electronic circuits. This
hypothetic chip performs all-optical routing of multiple N optical channels each supporting
10Gbps data stream. N channels are first demultiplexed in WDM (wavelength-division
multiplexing) photonic circuit, then rearranged and switched in optical cross-connect OXC
module, and multiplexed back into another fiber with new headers in WDM multiplexer. Data
packets are buffered in optical delay line if necessary. Channels are monitored with integrated Ge
photodetector PD. CMOS logical circuits (VLSI) monitor the performance. Electrical pads are
connecting the optoelectronic chip to other chips on a board via electrical signals. (Image: IBM)
The traditional arguments for silicon photonics have been based on its compatibility with the
mature silicon integrated circuit manufacturing. Silicon wafers have the lowest cost (per unit area)
and the highest crystal quality of any semiconductor material. This technology represents the most
spectacular convergence of technological sophistication and economics of scale. Another
motivation is the availability of high-quality silicon-on-insulator wafers, an ideal platform for
creating planar waveguide circuits.
The argument for silicon photonics can also be supported by the material itself: Silicon has
excellent material properties that are important in photonic devices. These include high thermal
conductivity, high optical damage threshold, and high third-order optical nonlinearities.
"The highest impact of silicon photonics may be in optical interconnection between digital
electronic chips" writes Dr. Bahram Jalali, a Professor of Electrical Engineering at UCLA. "This
technology addresses the communication bottleneck in VLSI electronics. A key finding is that for
photonic interconnects to be advantageous over their copper counterparts, wavelength-division
multiplexing (WDM) must be employed."

Comparison of metal and optical interconnects for onchip communication. The figure of merit is
in GBps/µm ps and represents the bandwidth normalized to wire width and latency time. Results
show that optical interconnects will be advantageous only if wavelength division multiplexing
(WDM) is employed. The reason is that the minimum width of optical waveguides is limited by the
optical wavelength. Plasmonic waveguides can in theory overcome this limit but the losses for
such waveguides are too high to be practical. (Reprinted with permission from Wiley Verlag)
In his review paper in physica status solidi (a) (Can silicon change photonics?), Jalali points out
that the benefits of integrated optics and electronics extend beyond computers: "For example, in
next-generation ultrasound medical imaging systems, the rate for signals generated by the array of
transducers will exceed 100 GBps, once digitized, and will continue to increase as radiologists
demand better image resolution. The size and power dissipation of conventional optical
transceivers prevent them from being used in the imaging probe. Silicon integrated circuits with
onchip optical interfaces can potentially solve this problem."
Other applications, such as disposable mass-produced biosensors or lab-on-a-chip devices are also
being considered. Jalali mentions that a near-term application will be low-cost components for
optical communication networks as well as bringing the power of optical networking to personal
computers. "With a computer’s copper networking cable replaced by an optical fiber, high-
definition video files can be effortlessly transferred through local area networks and between a
computer its peripherals."
The review is structured into major sections describing the following aspects of silicon photonics
and the technical breakthroughs and challenges associated with them:
 Light modulation
 Photodetectors
 Optical amplifiers and lasers
 Propagation and coupling losses, and optical filters
Finally, Jalali addresses the outlook for silicon photonics. He mentions the main challenge that
will determine if the promises of this technology will come to fruition: can it function within the
constraints of the chip industry? These constraints are twofold. One is economic – just being
CMOS-based doesn't mean it's low cost. Low unit cost can only be achieved if the economies of
scale work, i.e. if a sufficiently high-volume market can be created.
The other constraint has to do with the thermal-dissipation problem. Among photonic components,
lasers (and laser driver circuits) are the most power-hungry devices. Jalali explains that the lack of
an electrically pumped silicon laser, to date, dictates an architecture where the light source remains
off-chip: "By placing the “optical power supply” off-chip, this architecture is in fact preferred as it
removes a main source of heat dissipation."
Jalali's own belief is that silicon photonics is a technology whose time has come. "It stands to
impact a number of industries ranging from computing and communication to biomedicine. Fueled
by recent government and private-sector investments, the technological progress has been nothing
short of spectacular. Going forward, the technology’s faith will be governed not by technological
breakthroughs alone, but also by careful attention to the economics of chip manufacturing and the
power-dissipation issues that lie on the path to commercial success."