Ever read this?: It is interesting.
December 04, 1997, TechWeb News
Silicon-Opto Integration Nears Reality
ByChappell Brown
Insights into how photons and electrons interact with silicon are stimulating
new directions in optoelectronic research that could lead to a fully integrated
optical technology within a few years. But getting there will require additional
basic research along with mastering some difficult fabrication techniques. That
view of the field emerged from a symposium on materials for silicon
optoelectronics at the fall meeting of the Materials Research Society in
Boston.
"For most optical functions, we have viable approaches based on silicon, but
the one really weak point is silicon emitters," said Chris Buchal, a researcher
at Institute fur Schichtun Ionentechnic (ISI), in Jurlich, Germany, who
introduced the symposium with an overview of the state of the art. As to
when all the new insights will result in a practical silicon optoelectronics
compatible with CMOS VLSI, Buchal said, "Silicon technology is going to be
around for a long time to come, so whether it is two years or five years is not
going to really make any difference."
Work at ISI is partly funded by the European Esprit program under the
acronym "Scoop," for Silicon-Compatible Optoelectronic Program. Buchal
cited two new systems -- erbium-doped p-n junctions and iron disilicide --
that show promise for efficient light emission in the infrared region of the
spectrum.
Infrared is likely to be the first point where a fully integrated silicon
optoelectronics will emerge, he said. At IR wavelengths, all the components
for optical systems -- emitters, waveguides, modulators, and detectors -- are
easier to build in silicon.
"For example, you can easily define silicon waveguides over silicon dioxide
for infrared light, but at visible wavelengths, the light is absorbed," Buchal
said. At visible wavelengths, more complicated nonsilicon materials are
required to get efficient light transmission.
The use of iron disilicide in IR emitters is only just starting, but the fact that the
compound has a direct bandgap -- such as gallium arsenide, also an efficient
emitter material -- bodes well for building silicon lasers. So do some very
recent results that reveal strong light-generating capability in iron disilicide.
The first step in finding a practical application for the compound requires an
epitaxial growth process that is compatible with silicon. A joint project
between Hitachi's Power and Industrial R&D Division in Ibaraki, Japan, and
the Osaka Prefecture University is evolving an ion-beam synthesis approach
that looks promising. A three-step process using synthesis followed by
controlled annealing seems to create good-quality films, the team reported.
Work has progressed further on erbium-doped junctions, which are also
showing strong light-generating abilities. Erbium, a rare-earth element that is
usually employed in fiber optic systems to introduce optical non-linearities,
acts as a light-emission center when combined with oxygen in silicon.
By doping the n-type side of a p-n junction, a light-emitting device is created
when hot electrons are injected from the p-type side. The electrons are
trapped by a complex formed from four oxygen atoms surrounding an erbium
atom, which uses the extra energy to generate a photon. The effect is
enhanced by additional oxygen atoms implanted into the silicon. Several
presentations described different approaches to creating light-emitting
junctions using that basic strategy.
For creating silicon light emitters in the visible region of the spectrum, the
most promising technology so far is silicon nanoparticles. Philippe Fauchet, a
researcher at the University of Rochester, in New York, gave an overview of
that approach.
Insight has accumulated in the past few years into why tiny particles of silicon
-- only tens of nanometers across -- are so much better at generating photons
than bulk silicon. The new knowledge is being put to use in building silicon
lasers that may soon emerge in commercial applications, Fauchet said.
"The issue of light emission in silicon has always centered on the problem of
an indirect bandgap, but really, it turns out that it is a passivation issue," he
said. Extraneous factors such as dangling bonds at the surface of silicon and
defect centers in the silicon lattice itself sidetrack the formation of photons.
Perfect Nanoparticles
Thus, if silicon were perfect, with a perfect surface, it would be a good light
emitter. "But that is exactly the conditions inside of silicon nanoparticles,"
Fauchet said. "They are too small to have defects." The small volume also
means it is impossible for more than one electron-hole pair, which are the
photon generators, to coexist. That prevents another problem known as
auger recombination from sapping the particle's light-generating ability.
"At room temperature, silicon nanoparticles have a 10 percent conversion
efficiency, which is certainly good enough for practical applications," Fauchet
said. A viable commercial silicon LED with that level of efficiency would
represent "a revolution in electronics," he maintained.
Researchers are wrestling with the problem of how to make good electrical
contact with the particles. The promising efficiency figure only occurs when
the nanoparticles are optically stimulated. In that case, the electrical-contact
problem does not occur. Because of the difficulty of getting electrons into the
nanoparticle structures, efficiency now stands at around 0.2 percent for
optoelectron devices.
A promising form of nanoparticle silicon comes in the guise of porous films
fabricated by etching silicon wafers. In this case, the dangling-bond problem
reappears, since the etchant leaves hydrogen atoms attached to the surface of
the silicon. The Rochester group has found a means of replacing the hydrogen
with a very thin passivation layer of silicon dioxide. As in conventional
circuits, the silicon-dioxide layer also protects the films from environmental
degradation.
"If you expose untreated porous films to ambient conditions, the emission
wavelength moves toward the red end of the spectrum within an hour,"
Fauchet said. "Obviously, you can't have a device that changes its
characteristics over time."
Temperature changes, humidity, oxygen, and "just about everything else in the
environment acts to degrade the films." Tests of the treated porous films have
shown stable operation for weeks in ambient conditions.
The protective silicon-dioxide coatings are difficult to achieve because they
must be thick enough to protect the silicon underneath, but thin enough not to
block photons or prevent electrical contact. The Rochester group has devised
a lower-temperature anneal in a dilute oxygen atmosphere to control the film
thickness.
The protective coating, while helping to stabilize the films, does not cure many
of their problems. "Any nanoscale porous structure is basically a small amount
of silicon with mostly air in between," Fauchet said. "That means it is
mechanically fragile and offers a poor conduction path for electrons."
Two-Stage Solution
To address those problems, the project has created a two-stage film -- an
underlayer of porous silicon with a top layer that is less porous. Termed
mesoporous silicon, the top layer does not have good light-emission qualities.
But it is still transparent to photons while also offering mechanical strength and
much better conductivity.
An FET structure is completed by building a polysilicon layer over the
mesoporous silicon, establishing source, gate, and drain regions. Working
devices that use a bipolar transistor as a driver have been demonstrated in the
Rochester lab, and Fauchet said he believes a commercial process
incorporating the silicon LEDs will arrive within two years.
Not content to wait for the arrival of viable silicon light emitters, engineers at
Lucent Technologies' Bell Laboratories have developed a process for
flip-chip bonding arrays of compound semiconductor optical emitters and
detectors to VLSI chips. While the performance of these devices is not in
question -- LEDs and laser diodes have become established as a major
industry -- their power characteristics are incompatible with CMOS circuits.
"We would like to see an entire chip covered with a two-dimensional array of
VCSELs vertical cavity surface-emitting lasers, but no one has achieved that
with low enough power dissipation to avoid burnout at the silicon CMOS
level," said Jack Cunningham, a Bell Labs researcher who described the
strategy he and his colleagues are implementing to solve the problem.
A viable optical I/O scheme of that type would make it possible to hook
optical-fiber communications directly to desktop computers, creating a
bandwidth revolution. "It is clear from the Semiconductor Industry
Association road map that I/O performance is going to significantly lag behind
circuit performance," Cunningham added.
Wet-Oxide Process
Bell Labs is focusing on building a practical VCSEL with acceptable current
and threshold-voltage characteristics for CMOS integration. The strategy is
to use a recent innovation -- an oxide process for GaAs systems that
performs the same functions as silicon oxide in CMOS.
In the past few years, device research has evolved an aluminum-oxide
process, termed "wet oxide," that offers electrical and optical isolation
capabilities. The basic process is simple: AlGaAs is annealed in a
water-vapor atmosphere, which stimulates the growth of aluminum oxide.
The wet-oxide process has already been employed in building
high-performance VCSELs that are beginning to appear in commercial
applications. The Bell Labs team uses the technique at several points in the
development of a new type of flip-chip bondable VCSEL. The power
characteristics were attacked by using a wet-oxide process to build a highly
efficient top mirror that has greater than 99 percent efficiency.
In addition, the aperture of the active multiple quantum-well region was
narrowed with an oxide aperture that concentrates the generated photons.
Oxide was also used to improve the electrical contacts to the VCSEL by
creating a reliable ohmic contact.
The innovations will create a laser with a low-threshold current of less than 1
milliamp, along with a drive voltage in the 4-V range, making it compatible
with CMOS power characteristics. In addition, the contact scheme offers a
viable connection to integrated circuits in a flip-chip bonding configuration.
The first application of the new techniques will be in gigabit LANs. The
project goal is to create a linear array of optical I/O ports that can be
attached to CMOS chips. Viable lasers with the required characteristics
have been developed, according to Cunningham, and arrays of the devices
with the right pitch for chip I/O have been fabricated.
While such packaging schemes involving essentially incompatible materials
systems may solve some problems in the short term, full integration with
silicon is a tantalizing possibility that is fueling a broad-based research effort
into silicon-based systems. The symposium featured a variety of other
materials that different groups are trying in a concerted attack on the problem.
Silicon germanium/silicon heterostructures, various erbium-doped silicon
structures, and silicon-carbide approaches were all presented at the
symposium. The rich set of materials and processes, combined with a number
of structural innovations such as superlattices and nanostructures, promises a
wide number of choices for scientists and engineers working toward a
practical silicon optoelectronic technology.
techweb.com
I have no idea if the above elongated address will work for you but here it is..
Wow checked the URL...it works.... |
