Silicon Photonics
Silicon Photonics
Silicon Photonics
GENERAL ARTICLE
Theincreaseduseof optoelectronicdevicessuchas TVscreens
andcomputermonitors ineverydaylife, calls foramaterial on
which integration of all circuits may be performed. Silicon,
the most widely used material in electronics, has limited
application in optoelectronics due to its poor light emission
property. In this article, the need for silicon photonics in
present day ICs, and communication, the reason for its inef-
ficient light emission, and recent breakthrough inthe formof
Si Raman laser are discussed.
Optoelectronics in Todays Society
Citizens of todays world are exposed to many optoelectronic
equipments and gadgets. Take for example, the bar-code-reader
in front of cash machines in supermarkets. A red beam of light
scans the bar codes on the bill or price tags of the commodity
purchased and the price is displayed on the screens. The liquid
crystal displays on the cell phone, the monitor of the computer or
the television screen are examples of optoelectronic devices. CD
players, laser printers, the remote control of TV, etc., also use
light emitting devices. Students encounter many more optoelec-
tronic devices in their institutions and laboratories, such as a laser
pointer, a sophisticated projection system during lectures or
Light Emitting Diodes (LEDs) of different colours that glow
whenever an equipment is turned on, etc.
The above fewexamples illustrate the invasion of optoelectronics
into todays information-based society. The optoelectronic de-
vices rely on the conversion of optical signals into electrical
signals or vice versa. Today all the electronic and other instru-
ments and gadgets are lightweight, slim, sleek and compact,
thanks to the phenomenal growth of devices and circuits based on
P K Basu is a professor
and the coordinator of the
Centre of Advanced Study
in Radio Physics and
Electronics. His research
interests are semiconduc-
tor physics, devices,
photonics, optical
communication,
nanoelectronics and
nanophotonics
(visit: www.irpel.org).
Silicon Photonics
Silicon Raman Lasers
P K Basu
Keywords
Silicon photonics, Si Raman la-
ser, semiconductor laser, light
emitter, optical interconnect, op-
tical communication, Indirect gap
semiconductors.
38 RESONANCE October 2007
GENERAL ARTICLE
semiconductors. No wonder that almost all the optoelectronic
devices we see today are semiconductor based. Is there any ubi-
quitous semiconductor? We seek the answer in the next section.
Photonic Materials
The processor that handles the information or the memory ele-
ment that stores the information within most electronic gadgets is
made of silicon-based devices known as Integrated Circuits (ICs).
Since the demonstration of ICs by Jack Kilby (one of the three
Nobel Laureates in Physics, 2000) and the subsequent realisation
of a true IC by Robert Noyce in 1960 on a silicon chip, the
industry has considered no material as an alternative to silicon.
Today the semiconductor industry earns several hundred billion
dollars. The number of transistors, often Metal Oxide Semicon-
ductor Field Effect Transistors (MOSFETs), in a chip exceeds a
billion. The minimum feature size, which is approximately the
distance between source and drain in a MOSFET, is a fewtens of
a nanometer.
In contrast to electronic systems that invariably use Si based ICs,
no single material or single technology is nowleading the market
of optoelectronic devices and systems. The laser, Light Emitting
Diode (LED), photodetectors, and all other electronic and opto-
electronic devices are nowassembled on compound semiconduc-
tor substrates. Some attempts are being made to integrate differ-
ent optical devices and circuits using costly Indium Phosphide
(InP) as the substrate. The production technology for such a
monolithic optoelectronic circuit is still at a primitive stage.
Chip-scale integration of optical components, which offer low
cost and high reproducibility, has not yet been achieved.
Si Photonics: Challenges
In view of the above, the mature Si technology could be better
utilised to produce a truly monolithic Opto Electronic IC, which
is expected to offer low cost and high reproducibility. Unfortu-
nately, due to fundamental limitation arising out of peculiar band
structure of Si, no efficient light emitters and modulators can be
The processor that
handles the
information or the
memory element that
stores the information
within most electronic
gadgets is made of
silicon-based devices
known as Integrated
Circuits (ICs).
In contrast to
electronic systems
that invariably use Si
based ICs, no single
material or single
technology is now
leading the market of
optoelectronic
devices and systems.
39 RESONANCE October 2007
GENERAL ARTICLE
realized using Si and its related alloys. There have been many
attempts during the past four or five decades to overcome this
problem. Recently, some practical solutions have appeared in the
horizon. In the following section, we shall discuss the fundamen-
tal processes and the solutions to the problem.
Silicon: A Poor Light Emitter
In all practical optoelectronic systems, be it a fiber optic link or a
CD player, the optical devices are made of semiconductor junc-
tions, mostly pn junctions. First, consider the band diagram of
a bulk semiconductor shown by two horizontal lines in Figure 1.
The symbols E
C
and E
V
represent the edges of the conduction and
the valence band, respectively. The separation between the two is
the band gap E
G
, and thus
E
C
E
V
= E
G
.
When a photon of energy hv > E
G
is incident on the semiconduc-
tor, an electron in the valence band absorbs it and makes a
transition to one of the empty states in the conduction band,
leaving behind an empty state known as a hole in the valence
band. We now get a pair, an electron-hole pair, in excess of the
thermal equilibrium number of electrons and holes in the
E
C
E
V
a
e
phonon
Figure 1. Absorption and
luminescence indirect gap
semiconductors. (i) Con-
ventional band diagram,
absorption(a)andemission
(e) are indicatedby upward
and downward transitions;
the wiggly line represents
phonon emission (ii) E-k
diagram, absorption(a) and
emission(e) processes are
indicated.
(i)
(ii)
40 RESONANCE October 2007
GENERAL ARTICLE
semi-conductor. The electron usually lands in a higher state in the
conduction band. It loses a part of its energy by emitting a phonon
(quantized lattice vibration, see below) and comes to the conduc-
tion band edge. The excess electronhole pair now recombines,
i.e., the electron moves down to the valence band to occupy an
empty state there. The excess energy ~ E
G
is thus liberated in the
form of a photon. Its worth noting that the total energy is con-
served for electrons, holes and photons during both absorption
and emission.
Another condition to be satisfied for carriers and photons is the
momentum conservation, which cannot be explained with the
help of the more familiar band diagramas in left part of Figure 1.
For this purpose, we resort to the E-k diagramshown on the right
side of Figure 1. Here k is called the wave vector of the electron
and h k is the momentum of the electron ( h =h/2t).
Let us digress a moment to briefly discuss the meaning of the
E-k diagram. Wave-particle duality tells that an electron may be
treated as a particle as well as a matter wave. Asingle electron in
a crystalline solid encounters the periodic potential of the nucleus
and the core electrons and potential due to free electrons. The
wave-like character of the electron is determined from the solu-
tion of Schrdinger equation. The wave function characterizing
the electron may be written as
) exp( ) , ( r k r k
k
= + t U e , (1)
where U is a periodic function, k is the wave vector and r is the
position vector of the electron. The solution to the Schrdinger
equation can be obtained by a number of theoretical techniques.
The energy states near the conduction band minima E
C
and
valence band maxima E
V
are of interest. A representative E-k
diagram for semiconductors useful as light emitters is shown on
the right side of Figure 1.
Note that in the E-k diagram shown, the extrema of the conduc-
tion and valence bands occur at the same point, k = 0. The type of
41 RESONANCE October 2007
GENERAL ARTICLE
semiconductor whose band diagramis as shown is called a direct
band gap semiconductor. Common examples of such semicon-
ductors are GaAs, InP, ZnS, etc., and they are used as light
emitting materials.
Now consider the momentum conservation to be satisfied in the
absorption process, (see process (i) in Figure 1). Omitting the
common term h , the condition may be written as
k k
h e
+ = , (2)
where the subscripts e (h) stands for electron (hole) and k is the
photon wave vector (= 2t/). It may be proved easily that the
photon wave vector is negligible in comparison to wave vectors
for electrons and holes and thus, the momentum conservation
reduces to k
e
=k
h
. The absorption of a photon is therefore indica-
ted by a vertical line in the E-k diagram. In the reverse process of
recombination too, photons supply negligible momentum and
this process is shown as a vertical downward line in the figure.
The light emission, called luminescence, described above has
been illustrated for the case when the excitation mechanism is
photon absorption. The corresponding luminescence is termed as
photoluminescence. In usual emitters, like LEDs or Laser Diodes
(LDs), the excitation is by electrical means. The corresponding
luminescence is called electroluminescence. As the name im-
plies, the light emitting devices are junction diodes. When a
forward bias is applied to a pn junction, electrons are injected
fromn-side to p-side and similarly holes are injected fromp-side
to n-side. In the junction region, there are excess electron hole
pairs, where they recombine giving rise to a characteristic
emission. The recombination here is predominantly spontaneous.
Similar recombination also occurs at the junction region of LD.
However, in this case stimulated emission takes over spontane-
ous emission . To ensure population inversion, i.e., dominance of
stimulated emission, the p and n regions are heavily doped and
suitable cavity structures are incorporated to achieve self-
sustained oscillation of electromagnetic radiation.
42 RESONANCE October 2007
GENERAL ARTICLE
Consider now the E-k diagram of Si shown in Figure 2. The top
of the valence band is located at k = 0 as in the direct gap
semiconductor. But the difference lies in the position of the
bottom of the conduction band in this case it lies at k
0
, away
fromthe origin. The semiconductor is an indirect gap material, of
which Si and Ge are notable examples. An electron in the valence
band may absorb a photon, but in order to reach its destination in
the conduction band, it needs a momentum of value h q = h k
0
.
This momentum is usually derived from the quantized lattice
vibration, i.e., phonons. The absorption of light is nowa two-step
process, photon absorption first and then phonon absorption or
emission by the electron. Quantummechanically, the probability
of such transition is abnormally low. Indirect gap semiconduc-
tors are therefore poor absorber of light.
The opposite process of emission may be described similarly. An
electron in the conduction band must first absorb or emit a
phonon to conserve momentum and then fall directly into the
valence band to recombine with a hole. Again, the process is less
probable. Avast majority of electrons in Si recombine with holes
via defect or trap levels, which generate heat. The very feeble
light emission from silicon is not of any practical use at all.
The Changing Scenario
A Possible Solution: Stimulated Raman Scattering
Many efforts to enhance light emission from Si produced little
success until the breakthrough announcement came in 2005 by
using the effect named after the Indian legend, C V Raman.
Raman Effect was first observed in liquids as a result of their
interaction with light. When a beam of light is made to pass
through a liquid, the liquid either emits or absorbs a quantum of
energy corresponding to its vibrational modes and as a conse-
quence one observes Stokes and anti-Stokes lines, respectively
flanking the incident frequency. The effect is also observed in
crystalline solids. We use the terms phonons, Stokes and anti-
Stokes lines with reference to vibrational modes of atoms in
crystals.
Figure 2. Absorption pro-
cess in an indirect gap
semiconductor.
0
k
Indirect gap
semiconductors are
therefore poor
absorber of light.
43 RESONANCE October 2007
GENERAL ARTICLE
All textbooks use a linear one-dimensional atomic chain to ex-
plain the vibrational modes in solids, in which the atoms are
assumed to be interconnected by springs. Any movement of one
atomis therefore transmitted to its neighbours by the action of the
springs. If the neighbouring atoms move in the same direction, the
vibration is termed acoustic. If they move oppositely, the vibra-
tion generates optical mode. In terms of quantum mechanical
language, the vibrations are quantized. The quanta are called
phonons just as quantized electromagnetic radiation gives rise
to photons. Due to interaction, a photon gives up a part of its
energy to emit a phonon of energy h e
v
or absorb a phonon of the
same energy. The energy of the new photon then becomes
h e
o
= h e
L
h e
v
(3)
The new light of lower energy is called Stokes radiation, while
that with higher energy is called the anti-Stokes radiation.
Suppose nowa monochromatic light beamof photon energy h e
L
is allowed to fall on the solid. An electron in the ground state
absorbs a photon h e
v
and moves to a higher state (virtual state).
There may be interaction of photons with phonons. As a result,
the photon gives up a part of its energy to the phonon. The
resulting Stokes line of energy ( h e
S
= h e
L
h e
v
) is emitted as
the electron comes down fromthe higher virtual state to state 1, as
shown in Figure 3. The
emission of anti-Stokes line
may also be understood
from the transitions in the
same diagram.
The emission of Stokes or
anti-Stokes line represents
the spontaneous process.
An interesting situation oc-
curs when the intensity of
the original beam, calledthe
pump light, becomes quite
Excited state
Stokes
emission
L-v)
Absorption
L
Phonon emission
v
Ground state
Figure 3. Illustration of
Raman scattering and
Stokes emission.
44 RESONANCE October 2007
GENERAL ARTICLE
strong. Under favourable circumstances, the Stokes signal grows
at the cost of the pump signal. The amplificationof the light signal
is due to stimulated Raman scattering (SRS). There may be
amplification of anti-Stokes radiation also, but that is not signifi-
cant in comparison to Stokes line growth.
Use of SRS in photonics and fibre optic communication is not
new. In fact, SRS has been utilized to produce a newclass of fibre
amplifiers, that, in conjunction with the Erbium-doped fibre
amplifiers, are deployed in present day commercial communica-
tion networks using optical fibres. SRS occurs in the core layer of
the optical fibre, which is made of glass doped with some impu-
rities.
The amplification in a Raman active medium depends on the
properties of the material and the related coefficient is about a
thousand times stronger in crystalline Si than in amorphous silica
glass. A group led by B Jalali in the University of California at
Los Angeles, (UCLA), has been trying over the last few years to
achieve Raman amplification in Si. The UCLA group reported
laser action in silicon; but their device did not work continuously.
The laser light could be produced only in the form of pulses.
Continuous Wave Raman Laser
The continuous wave (CW) operation of Raman laser was first
reported by the workers at Intel Corp. [9]. The top view of its
structure is shown in Figure 4. SRS and amplification take place
in the S-shaped region labeled as SOI rib waveguide. The actual
material (SOI) is intrinsic silicon surrounded by two p and n-type
regions of Si on its two sides and a layer of silicon oxide
(insulator) beneath it.
There are three essential components in a laser:
1) An active medium in which amplification of lightwave takes
place due to stimulated emission.
2) A mechanism to provide positive feedback to amplified
radiation to enable emission of self-sustained oscillation.
45 RESONANCE October 2007
GENERAL ARTICLE
3) A pump source which lifts the electrons from the ground state
to an excited state to create population inversion between the
excited and ground states.
The active Si region (intrinsic, i-region) in which SRS occurs, is
surrounded on its two sides by lower refractive index p- and n-
type silicon. Light is then confined in the higher index Si region,
much in the same manner as light is guided in the core of an
optical fibre. The two plane parallel mirrors, R1 and Rb provide
the feedback. Note that the front mirror R1 has higher reflectivity
for the Stokes line of wavelength 1686 nm to provide more
feedback. Its reflectivity for pump light of 1550 nmwavelength is
smaller to allow more pump power to fall on Si active region.
Experimentally, the emission of Stokes radiation is quite feeble
for low pump power. However, as the pump power exceeds a
threshold, stimulated process dominates and as a result, the
Stokes intensity increases rapidly with increase in pump power.
Further Work
The breakthrough work reported above is only the beginning of a
Figure 4. Schematic dia-
gramshowingthestructure
of theCWRamanlaser. The
light is confined in the S-
shaped waveguide formed
by using Silicon-On-Insu-
lator (SOI) structure. The
intrinsic (i) Si layer in the p-
i-n structure guides the
light. External bias is ap-
plied to the p and n layers
surroundingtheintrinsic(i)
layer on its two sides.
46 RESONANCE October 2007
GENERAL ARTICLE
huge task. This laser is still an optically pumped source. How-
ever, in all applications related to electronics, computer and
communication, it is necessary that the control should be by an
external electrical bias. Asuitable modification, perhaps a drastic
change in the design involving a newconcept may be in order. To
further clarify the above statement, consider how a message or
signal, i.e., voice or picture or data, is transmitted in communica-
tion systems. The signal is impressed on a carrier wave; in other
words, the signal modulates the carrier. In present day fibre optic
telephony, the signal is impressed on the light source and the
intensity of light changes in accordance with the strength of the
impressed signal. This is called direct modulation of light. At
present, the light intensity of Si based Raman lasers cannot be
changed directly by electrical signal. Future work on Si Raman
lasers must try to accomplish this task.
Suggested Reading
[1] For use of Si in VLSI Technology, S MSze, VLSI Technology, McGraw-
Hill International, 1988.
[2] For optical links in computers, Neil Savage, Linking with Light, IEEE
Spectrum, pp.3236, August 2002.
[3] For details about fibre optic communication, JohnMGowar, Fibre Optic
Communication, Prentice Hall, India, New Delhi, 1993.
[4] For band structure, E-k diagram, B G Streetman and S Banerjee, Solid
State Electronic Devices, Prentice Hall, New Jersey, 5th Edn., 2000.
[5] Advanced theory of band structure, optical processes, etc., are treated,
in P K Basu, Theory of Optical Processes in Semiconductors: Bulk and
Microstructures, Clarendon Press, Oxford, 2003.
[6] For non linear optical processes and laser theory, Amnon Yariv, Quan-
tum Electronics, John Wiley, 5th Edn, 1998.
[7] Light emission in Si nanocrystals, Salvatore Coffa, Light from Silicon,
IEEE Spectrum, pp.3641, October 2005.
[8] For Raman laser, Mario Paniccia and S Koehl, The Silicon Solution,
IEEE Spectrum, pp.3035, October 2005,.
[9] Ramanamplificationinsiliconis reported in HRong et al, AContinuous
Wave Raman Silicon Laser, Nature, Vol.433, pp.725728, 2005.
Address for Correspondence
P K Basu
Institute of Radio Physics
and Electronics
University of Calcutta
92, Acharya Prafulla
Chandra Road
Kolkata 700 009, India.
Email: pkb.rpe@caluniv.ac.in
pkb_cal@yahoo.comCalcutta
Acknowledgements
Theauthor is thankful toSumitra
Ghosh and Bratati Mukhopadh-
yay for collaboration and help in
preparing figures.