29-02-2012, 10:30 AM
SILICON PHOTONICS
ABSTRACT
Silicon Photonics can be defined as the utilization of silicon-based materials for the generation, guide, control and detection of light to communicate over distances. Optical technology suffered from a reputation as an expensive solution, based on high cost of hardware components, as they are typically fabricated using exotic materials that are expensive for manufacturing.
These limitations prompted Intel to research the construction of fibre-optic components from other materials, such as silicon (suggested since 1980’s).Silicon Photonics has attained much attention in recent years owing to the maturity of silicon in the electronics industry and its possibility of monolithic integration of both photonic and electronic devices on one chip. It develops high-volume low cost optical components using standard CMOS process-the IC manufacturing process used today.
The various challenges as well as the milestones in the development of Silicon Photonic are discussed.The difficulty in fabricating optical devices such as laser source, modulators, detectors etc. on silicon for high switching speeds that provides high data rates for communication links as well as the solutions put forward by the Silicon Photonics research group at Intel are projected.With the developments up till now the devices available on silicon can form only a 40Gbps optical link.Tbps data rates has already been achieved in optics with Dense Wavelength Division Multiplexing technology. With further developments Silicon Photonics is expected to bring an optical revolution in Electronics and Communication industry with the realization of the above said Tbps data links using microelectronic silicon chips. The hopes and hurdles towards this development are discussed in detail
1. INTRODUCTION
Fiberoptic communication is well established today due to the great capacity and reliability it provides. . Fiber-optic communication is the process of transporting data at high speeds on a glass fiber using light. However, the technology has suffered from a reputation as an expensive solution. This view is based in large part on the high cost of the hardware components. These components are typically fabricated using exotic materials that are expensive to manufacture. In addition, these components tend to be specialized and require complex steps to assemble and package. These limitations prompted Intel to research the construction of fiber-optic components from other materials, such as silicon. The vision of silicon photonics arose from the research performed in this area. Its overarching goal is to develop high-volume, low-cost optical components using standard CMOS processing the same manufacturing process used for microprocessors and semiconductor devices. Silicon presents a unique material for this research because the techniques for processing it are well understood and it demonstrates certain desirable behaviors. For example, while silicon is opaque in the visible spectrum, it is transparent at the infrared wavelengths used in optical transmission, hence it can guide light. Moreover, manufacturing silicon components in high volume to the specifications needed by optical communication is comparatively inexpensive.
Researchers at Intel have announced advancement in silicon photonics by demonstrating the first continuous silicon laser based on the Raman effect. This research breakthrough paves the way for making optical amplifiers, lasers and wavelength converters to act as light source and also switch a signal’s color in low-cost silicon. It also brings Intel closer to realizing its vision of “siliconizing” photonics, which will enable the creation of inexpensive, high-performance optical interconnects in and around PCs, servers and other devices. There has also been developments which include the achievement of GHz range optical modulator and detector devices n silicon
Silicon’s key drawback is that it cannot emit laser light, and so the lasers that drive optical communications have been made of more exotic materials such as indium phosphide and gallium arsenide. However, silicon can be used to manipulate the light emitted by inexpensive lasers so as to provide light that has characteristics similar to more-expensive devices. This is just one way in which silicon can lower the cost of photonics. Intel’s silicon photonics research is an end-to-end effort to build integrated photonic devices in silicon for communication and other applications. To date, Intel has demonstrated laser production from external light source, tunable filters, optical modulators, photo-detectors and optical packaging techniques using silicon that can establish optical links with Gbps datarates. Even more is yet to achieve.
1.1 MOORE’S LAW AND SILICON TECHNOLOGY
It is an understatement to remark that we live in a world made possible by silicon technology. Modern life has been shaped and defined by innumerable products that rely on integrated electronic circuits fabricated in mind-boggling number and precision on silicon wafers. The grand success of silicon technology is not only the dramatic improvements that have been achieved in performance, but also the exponentially decreasing per-component manufacturing costs that have kept that performance affordable. In fact, Gordon Moore’s famous law(1962) describing progress in the semiconductor industry was originally stated in similar economic terms:
“ The complexity for minimum component costs has increased at a rate of roughly
a factor of two per year. . . , this rate can be expected to continue”
Complexity is usually equated to transistor count, and by that measure the exponential progress predicted by Moore’s Law has been maintained through the present day (figure1.1). It has become cheaper over time to pack more and more transistors into integrated circuits because each individual transistor is continually being made smaller. This scaling process allows more powerful chips with more transistors to be made for a reasonable price. Smaller transistors also drive down the price of previous generation chips of any given complexity, because more functionally identical copies can be simultaneously made on the surface of a silicon wafer for nearly the same cost. Scaling is the engine of progress in silicon micro electronics. It is sustained only by intensive research and development in the face of perpetual technology challenges always looming on the horizon
Goals and benchmarks for scaling are established and monitored in the International Technology Roadmap for Semiconductors (ITRS), a public document.
In Figure 1.1, transistor counts for integrated circuits showing the historical accuracy of Gordon Moore’s prediction of exponentially increasing integrated circuit complexity with year by a consortium representing the global semiconductor industry. The roadmap is intended “to provide a reference of requirements, potential solutions, and their timing for the semiconductor industry” over a fifteen -year horizon. For many years, the ITRS has highlighted one threat to continued scaling in particular that must be addressed in the short term future in order to avoid slowing down the pace of Moore’s Law.
The anticipated problem is often referred to as the “interconnect bottleneck.” As the number of transistors in an integrated circuit increases, more and more interconnecting wires must be included in the chip to link those transistors together. Today’s chips already contain well over one kilometer of wiring per square centimeter of chip area . Sending information along these wires consumes significant power in various losses and introduces the majority of speed-limiting circuit delay in a modern integrated circuit. Scaling exacerbates both of these problems by decreasing the cross sectional area of each wire, proportionately increasing its electrical resistance. With further scaling the RC capacitive charging delays in the wires will increasingly dominate the overall performance of future integrated circuits. The interconnect bottleneck has threatened Moore’s Law before. In the late 1990s, integrated circuits contained aluminum wires that were surrounded by silicon oxide. As interconnect cross sections decreased, mounting circuit delay in capacitive charging of these aluminum wires began to effect chip performance. A solution was found in a change of materials. Copper was introduced in place of aluminum, which cut the resistance of the wires nearly in half. Eventually low dielectric constant (“low-κ”) doped silica infill materials were also phased in to reduce the capacitance.
In Figure 1.2, according to the ITRS, there is no known manufacturable global or intermediate interconnect solutions for the 45 nm technology node. In the roadmap, such challenges are highlighted on a spreadsheet in red, forming the “red brick wall.”
Incorporating these new materials into existing fabrication processes posed significant integration challenges. Copper can diffuse quickly through silicon and create short circuits in the transistors of a chip unless care is taken to avoid contact between the copper wires and the silicon substrate. Additionally, the nonexistence of any suitable gas phase etching process for copper requires additive deposition techniques to be used. The silicon industry invested heavily in research and development to find diffusion barriers and to perfect “Damascene” deposition processes relying on chemical-mechanical planarization (CMP) . These technologies made copper interconnects possible and have allowed scaling to continue through the present day.
Further evolutionary progress through materials research in very low-κ dielectrics may postpone the return of the interconnect bottleneck, but a new approach to information transfer within integrated circuits will inevitably become necessary if transistors are to continue shrinking into the next decade. According to the latest update of the ITRS chapter on interconnects, traditional interconnect scaling is not expected to satisfy performance requirements after approximately 2010 (figure 1.2).
1.2 OPTICAL INTERCONNECTS
Many expect photonics to provide the long term solution. In so-called optical interconnect schemes, the copper wires between regions of an integrated circuit would be replaced by a system of lasers, modulators, optical waveguides and photo-detectors. The metal interconnects at all levels starting from those within the ICs to that between ICs on boards and that with peripheral devices are replaced with optical links. The potential benefits of this approach include the virtual elimination of delay, cross talk, and power dissipation in signal propagation, although significant new challenges will be introduced in signal generation and detection.
The current integration level of about 1.7 billion transistors is responsible for the high processing capability of todays processors. More transistors means more switching power. Since switching decides digital signal processing power the very high integration of transistors is responsible for the processing power of todays processors.But the maximum performance power of systems with these processors is limited by the heat loss in metal connections, inductive losses due to nearby conductors, proximity effect, i.e expeltion of current from inner conductor when conductors are in close proximity, skin effect i.e concentration of current flow to the surface of conductor due to its suppression at the interior due to the formation of eddy currents(loop currents) at the interior whose flux linkage opposes the flux of the main current that caused them. There are also losses due to metallic imperfections (impurities, lattice mismatch etc.). Due to all these a speed grater than 10Gbps has never been possible with metal interconnections. Even the core series processors from Intel has databus speed around 5Gbps
The integration density and data rate that can be achieved using conventional electrical interconnects set very high performance requirements for any optical interconnect system to be viable. We can anticipate that optical interconnects will make the chip-scale integration of the very best photonic technologies available today. Stable laser sources, interferometric modulators, dense wavelength division multiplexing (WDM), and low loss planar waveguides will all be necessary components of an optical interconnect system that can reach an acceptable per-wire information bandwidth-per-watt figure of merit.
These photonic technologies are now applied primarily in the long-haul telecommuting actions industry, where individual component cost and size do not drive the market. Data transfer rates and the cost per transmitted bit through optical fiber networks have improved dramatically in performance over the last few decades, following exponential progress curves that can compound even faster than Moore’s Law. These advances underlie the infrastructure of the internet and are responsible for fundamental changes in our lives, particularly in our experience of distance around the globe. However, while millions of miles of fiber optic cable now stretch between cities and continents, the photonic components they connect are still typically packaged separately. Obviously this must change if optical networks are to be replicated in microcosm within millions of future chips.
Micro photonics refers to efforts to miniaturize the optical components used in long-distance telecommunications networks so that integrated photonic circuits can become areality. Work in this field spans many subjects, including planar waveguides and photonic crystals, integrated diode detectors, modulators, and lasers. In more recent years, research focused on the sub wavelength manipulation of light via metal optics and dispersion engineered effective media has begun to explore the anticipated limits of scaling in future photonic integrated circuits. Advances in the related and often overlapping field of “nanophotonics” suggest the possibility of eventually controlling optical properties through nanoscale engineering.
Between the long-haul telecommunications industry and research in micro photonics lies a small market that will undoubtedly aid in driving the integration of on-chip optical networks: high performance supercomputing. Modern supercomputer performance is typically dominated by the quality of the interconnecting network that routes information between processor nodes. Consequently, a large body of research exists on network topology and infrastructure designed to make the most of each photonic component. This knowledge is ready to be applied to future optical interconnect networks that connect sub processor cores within a single chip.
If optical interconnects become essential for continued scaling progress in silicon electronics, an enormous market will open for integrated photonic circuit technology. Eventually, unimagined new products will be made possible by the widespread availability of affordable, high-density optical systems. Considering the historical development of computing hardware from the relays and vacuum tubes of early telephone networks, it is possible that optical interconnects could someday lead to all-optical computers, perhaps including systems capable of quantum computation.
Unfortunately, there is at present no clear path to practical on-chip optical data transfer and scalable all-photonic integrated circuits. The obstacles that currently stand in the way of optical interconnects are challenges for device physics and materials science. Break through are needed that either improve the set of materials available for micro photonic devices or obviate the need for increased materials performance through novel device designs.
1.3 ENTER OPTOELECTRONICS
Fiber optics use light to transmit data over a glass or plastic fiber(silica),and a seed of about 1.7Gbps was achieved in 1980s itself. Though plastic fibres are also used silica(glass fibre), i.e Silicon dioxide, that we use as insulator in CMOS fabrication is most commonly used. The primary benefit of using light rather than an electric signal over copper wiring is significantly greater capacity, since data transmission through fibres is at light speed.But this alone cannot make high speed transmission possible, it aso requires the end devices like modulators, demodulators etc where convertion between optical and electrical data takes place, also to work at such high speeds. The Bell Labs in France currently holds the record of transmission with muxing of about 155 different data streams each on its own light wave and each with a capacity of about 100Gbps that constitute in total a 14Tbps data link using a fibre pair with Dense WDM technology. In addition, glass fiber has desirable physical properties: it is lighter and impervious to factors such as electrical interference and crosstalk that degrade signal quality on copper wires. Hence optic fibres can be used even at places of high lightning with all dielectric cables.The high electrical resistance of fibres makes them usable even near high tension equipments. Hence repeaters are placed at ranges over 100Kms.
Photonics is the field of study that deals with light, especially the development of components for optical communications. It is the hardware aspect of fiber optics; and due to commercial demand for bandwidth, it has enjoyed considerable expansion and developments during the past decade. During the last few years, researchers at Intel have been actively exploring the use of silicon as the primary basis of photonic components. This research has established Intel’s reputation in a specialized field called silicon photonics, which appears poised to provide solutions that break through longstanding limitations of silicon as a material for fiber optics. In addition to this research, Intel’s expertise in fabricating processors from silicon could enable it to create inexpensive, high-performance photonic devices that comprise numerous components integrated on one silicon die.
1.4 COMPONENTS OF AN OPTICAL SYSTEM
To understand how optical data might one day travel through silicon in your computer, it helps to know how it travels over optical fiber today. First, a computer sends regular electrical data to an optical transmitter, where the signal is converted into pulses of light. The transmitter contains a laser and an electrical driver, which uses the source data to modulate the laser beam, making beam on and off to generate 1s and 0s. Imprinted with the data, the beam travels through the glass fiber, encountering switches at various junctures that route the data to different destinations. If the data must travel more than about 100 kilometers, an optical amplifier boosts the signal. At the destination, a photo detector reads and converts the data encoded in the photons back into electrical data. Similar techniques could someday allow us to collapse the dozens of copper conductors that currently carry data between processors and memory chips into a single photonic link.
The core of the internet and long-haul telecom links made the switch to fiber optics long ago. A single fiber strand can now carry up to one trillion bits of data per second, enough to transmit a phone call from every resident of New York City simultaneously. In theory, you could push fiber up to 150 trillion bits per second—a rate that would deliver the text of all the books in the U.S. Library of Congress in about a second.
Today’s devices are specialized components made from indium phosphide, lithium niobate, and other exotic materials that can’t be integrated onto silicon chips. That makes their assembly much more complex than the assembly of ordinary electronics, because the paths that the light travels must be painstakingly aligned to micrometer precision. In a sense, the photonics industry is where the electronics industry was a half century ago, before the breakthrough of the integrated circuit.
The only way for photonics to move into the mass market is to introduce integration, high-volume manufacturing, and low- cost assembly—that is, to “siliconize” photonics. By that we mean integrating several different optical devices onto one silicon chip, rather than separately assembling each from exotic materials. In our lab, we have been developing all the photonic devices needed for optical communications, using the same complementary metal oxide semiconductor (CMOS) manufacturing techniques that the world’s chip makers now use to fabricate tens of millions of microprocessors and memory chips each year.
A source that can produce narrow coherent beam of light is the prime necessity in optical communication. Hence lasers are the first choice. However LEDs are also used for some low cost applications. Also for lasers a 1000 times more power output may be obtained compared to LEDs, based on how we set the gain medium. However optical communication has limitations due to scattering effects at discontinuities or imperfections in fibre and also very slight variations in refractive index along the fibre that can affect the wavelength of signal transmitted. When such limiting factors persist a coherent narrow beam from source, i.e a beam of light with each photon at equal lengths along the fibre as well that at a singe cross section showing the same wave properties(frequency, phase, polarization etc.) , is a must otherwise the dispertion and diffraction phenomenon may occur in a different way to each photon in the beam and this can severly distort or destroy the light signal.
Optical communication operates on the short wave or IR region of EM spectrum (i.e from 1260-1675nm). The operating range of wavelength is divided into 6 bands. Among them the C-band(Conventional band), i.e from 1530-1565nm is most commonly used, since it has showed the least scattering. Most optical devices have been developed to work in this range. For communication single mode fibres are preferred, where mode represents the angles of incidence at the core-cladding interface for which transmission is possible. Multimode fibres (cross-section diameter>50um) are avoided due to intermodal dispertion, and even LEDs can b used. However single mode fibres require high stability for the light source used.
WDM started with muxing of 2 channels and now you can pack dozens of channels of high-speed data onto a single mode fiber with cross-section as low as 9um,separating the channels by wavelength, a technique called wavelength-division multiplexing, similar to frequency division multiplexing in radio communication. Arrayed waveguide grating structures that can perform both muxing and demultiplexing are used to implement WDM(Figure 1.3).
In AWG shown in Figure1.3, from 1 to 5, it acts as mux and demux the otherway. Regions 2 and 4 are free space segments and section 3 forms the array of waveguides with a constant length increment. Whereever light comes out of waveguide to free space, it diffracts, i.e spreads. A multiwavelength beam coming from section 1 after diffraction at section 2 passes to each of the waveguides of the array. The phase shift between the waves coming to section 4 will be such that the waves after diffraction constructive interference of the composed waves occur where they are received by different waveguides as shown in It has the advantage of integrated planar structure, low cost, low insertion loss and ease of network upgradation, since with increasing demand for bandwidth instead of laying new fibres, it only requires this device replaced with a higher capacity structure. As in any optical devices, changes in refractive index with temperature that can affect wavelength is a problem, and hence precision temperature control within +/-2 degree Celcius is required.
ABSTRACT
Silicon Photonics can be defined as the utilization of silicon-based materials for the generation, guide, control and detection of light to communicate over distances. Optical technology suffered from a reputation as an expensive solution, based on high cost of hardware components, as they are typically fabricated using exotic materials that are expensive for manufacturing.
These limitations prompted Intel to research the construction of fibre-optic components from other materials, such as silicon (suggested since 1980’s).Silicon Photonics has attained much attention in recent years owing to the maturity of silicon in the electronics industry and its possibility of monolithic integration of both photonic and electronic devices on one chip. It develops high-volume low cost optical components using standard CMOS process-the IC manufacturing process used today.
The various challenges as well as the milestones in the development of Silicon Photonic are discussed.The difficulty in fabricating optical devices such as laser source, modulators, detectors etc. on silicon for high switching speeds that provides high data rates for communication links as well as the solutions put forward by the Silicon Photonics research group at Intel are projected.With the developments up till now the devices available on silicon can form only a 40Gbps optical link.Tbps data rates has already been achieved in optics with Dense Wavelength Division Multiplexing technology. With further developments Silicon Photonics is expected to bring an optical revolution in Electronics and Communication industry with the realization of the above said Tbps data links using microelectronic silicon chips. The hopes and hurdles towards this development are discussed in detail
1. INTRODUCTION
Fiberoptic communication is well established today due to the great capacity and reliability it provides. . Fiber-optic communication is the process of transporting data at high speeds on a glass fiber using light. However, the technology has suffered from a reputation as an expensive solution. This view is based in large part on the high cost of the hardware components. These components are typically fabricated using exotic materials that are expensive to manufacture. In addition, these components tend to be specialized and require complex steps to assemble and package. These limitations prompted Intel to research the construction of fiber-optic components from other materials, such as silicon. The vision of silicon photonics arose from the research performed in this area. Its overarching goal is to develop high-volume, low-cost optical components using standard CMOS processing the same manufacturing process used for microprocessors and semiconductor devices. Silicon presents a unique material for this research because the techniques for processing it are well understood and it demonstrates certain desirable behaviors. For example, while silicon is opaque in the visible spectrum, it is transparent at the infrared wavelengths used in optical transmission, hence it can guide light. Moreover, manufacturing silicon components in high volume to the specifications needed by optical communication is comparatively inexpensive.
Researchers at Intel have announced advancement in silicon photonics by demonstrating the first continuous silicon laser based on the Raman effect. This research breakthrough paves the way for making optical amplifiers, lasers and wavelength converters to act as light source and also switch a signal’s color in low-cost silicon. It also brings Intel closer to realizing its vision of “siliconizing” photonics, which will enable the creation of inexpensive, high-performance optical interconnects in and around PCs, servers and other devices. There has also been developments which include the achievement of GHz range optical modulator and detector devices n silicon
Silicon’s key drawback is that it cannot emit laser light, and so the lasers that drive optical communications have been made of more exotic materials such as indium phosphide and gallium arsenide. However, silicon can be used to manipulate the light emitted by inexpensive lasers so as to provide light that has characteristics similar to more-expensive devices. This is just one way in which silicon can lower the cost of photonics. Intel’s silicon photonics research is an end-to-end effort to build integrated photonic devices in silicon for communication and other applications. To date, Intel has demonstrated laser production from external light source, tunable filters, optical modulators, photo-detectors and optical packaging techniques using silicon that can establish optical links with Gbps datarates. Even more is yet to achieve.
1.1 MOORE’S LAW AND SILICON TECHNOLOGY
It is an understatement to remark that we live in a world made possible by silicon technology. Modern life has been shaped and defined by innumerable products that rely on integrated electronic circuits fabricated in mind-boggling number and precision on silicon wafers. The grand success of silicon technology is not only the dramatic improvements that have been achieved in performance, but also the exponentially decreasing per-component manufacturing costs that have kept that performance affordable. In fact, Gordon Moore’s famous law(1962) describing progress in the semiconductor industry was originally stated in similar economic terms:
“ The complexity for minimum component costs has increased at a rate of roughly
a factor of two per year. . . , this rate can be expected to continue”
Complexity is usually equated to transistor count, and by that measure the exponential progress predicted by Moore’s Law has been maintained through the present day (figure1.1). It has become cheaper over time to pack more and more transistors into integrated circuits because each individual transistor is continually being made smaller. This scaling process allows more powerful chips with more transistors to be made for a reasonable price. Smaller transistors also drive down the price of previous generation chips of any given complexity, because more functionally identical copies can be simultaneously made on the surface of a silicon wafer for nearly the same cost. Scaling is the engine of progress in silicon micro electronics. It is sustained only by intensive research and development in the face of perpetual technology challenges always looming on the horizon
Goals and benchmarks for scaling are established and monitored in the International Technology Roadmap for Semiconductors (ITRS), a public document.
In Figure 1.1, transistor counts for integrated circuits showing the historical accuracy of Gordon Moore’s prediction of exponentially increasing integrated circuit complexity with year by a consortium representing the global semiconductor industry. The roadmap is intended “to provide a reference of requirements, potential solutions, and their timing for the semiconductor industry” over a fifteen -year horizon. For many years, the ITRS has highlighted one threat to continued scaling in particular that must be addressed in the short term future in order to avoid slowing down the pace of Moore’s Law.
The anticipated problem is often referred to as the “interconnect bottleneck.” As the number of transistors in an integrated circuit increases, more and more interconnecting wires must be included in the chip to link those transistors together. Today’s chips already contain well over one kilometer of wiring per square centimeter of chip area . Sending information along these wires consumes significant power in various losses and introduces the majority of speed-limiting circuit delay in a modern integrated circuit. Scaling exacerbates both of these problems by decreasing the cross sectional area of each wire, proportionately increasing its electrical resistance. With further scaling the RC capacitive charging delays in the wires will increasingly dominate the overall performance of future integrated circuits. The interconnect bottleneck has threatened Moore’s Law before. In the late 1990s, integrated circuits contained aluminum wires that were surrounded by silicon oxide. As interconnect cross sections decreased, mounting circuit delay in capacitive charging of these aluminum wires began to effect chip performance. A solution was found in a change of materials. Copper was introduced in place of aluminum, which cut the resistance of the wires nearly in half. Eventually low dielectric constant (“low-κ”) doped silica infill materials were also phased in to reduce the capacitance.
In Figure 1.2, according to the ITRS, there is no known manufacturable global or intermediate interconnect solutions for the 45 nm technology node. In the roadmap, such challenges are highlighted on a spreadsheet in red, forming the “red brick wall.”
Incorporating these new materials into existing fabrication processes posed significant integration challenges. Copper can diffuse quickly through silicon and create short circuits in the transistors of a chip unless care is taken to avoid contact between the copper wires and the silicon substrate. Additionally, the nonexistence of any suitable gas phase etching process for copper requires additive deposition techniques to be used. The silicon industry invested heavily in research and development to find diffusion barriers and to perfect “Damascene” deposition processes relying on chemical-mechanical planarization (CMP) . These technologies made copper interconnects possible and have allowed scaling to continue through the present day.
Further evolutionary progress through materials research in very low-κ dielectrics may postpone the return of the interconnect bottleneck, but a new approach to information transfer within integrated circuits will inevitably become necessary if transistors are to continue shrinking into the next decade. According to the latest update of the ITRS chapter on interconnects, traditional interconnect scaling is not expected to satisfy performance requirements after approximately 2010 (figure 1.2).
1.2 OPTICAL INTERCONNECTS
Many expect photonics to provide the long term solution. In so-called optical interconnect schemes, the copper wires between regions of an integrated circuit would be replaced by a system of lasers, modulators, optical waveguides and photo-detectors. The metal interconnects at all levels starting from those within the ICs to that between ICs on boards and that with peripheral devices are replaced with optical links. The potential benefits of this approach include the virtual elimination of delay, cross talk, and power dissipation in signal propagation, although significant new challenges will be introduced in signal generation and detection.
The current integration level of about 1.7 billion transistors is responsible for the high processing capability of todays processors. More transistors means more switching power. Since switching decides digital signal processing power the very high integration of transistors is responsible for the processing power of todays processors.But the maximum performance power of systems with these processors is limited by the heat loss in metal connections, inductive losses due to nearby conductors, proximity effect, i.e expeltion of current from inner conductor when conductors are in close proximity, skin effect i.e concentration of current flow to the surface of conductor due to its suppression at the interior due to the formation of eddy currents(loop currents) at the interior whose flux linkage opposes the flux of the main current that caused them. There are also losses due to metallic imperfections (impurities, lattice mismatch etc.). Due to all these a speed grater than 10Gbps has never been possible with metal interconnections. Even the core series processors from Intel has databus speed around 5Gbps
The integration density and data rate that can be achieved using conventional electrical interconnects set very high performance requirements for any optical interconnect system to be viable. We can anticipate that optical interconnects will make the chip-scale integration of the very best photonic technologies available today. Stable laser sources, interferometric modulators, dense wavelength division multiplexing (WDM), and low loss planar waveguides will all be necessary components of an optical interconnect system that can reach an acceptable per-wire information bandwidth-per-watt figure of merit.
These photonic technologies are now applied primarily in the long-haul telecommuting actions industry, where individual component cost and size do not drive the market. Data transfer rates and the cost per transmitted bit through optical fiber networks have improved dramatically in performance over the last few decades, following exponential progress curves that can compound even faster than Moore’s Law. These advances underlie the infrastructure of the internet and are responsible for fundamental changes in our lives, particularly in our experience of distance around the globe. However, while millions of miles of fiber optic cable now stretch between cities and continents, the photonic components they connect are still typically packaged separately. Obviously this must change if optical networks are to be replicated in microcosm within millions of future chips.
Micro photonics refers to efforts to miniaturize the optical components used in long-distance telecommunications networks so that integrated photonic circuits can become areality. Work in this field spans many subjects, including planar waveguides and photonic crystals, integrated diode detectors, modulators, and lasers. In more recent years, research focused on the sub wavelength manipulation of light via metal optics and dispersion engineered effective media has begun to explore the anticipated limits of scaling in future photonic integrated circuits. Advances in the related and often overlapping field of “nanophotonics” suggest the possibility of eventually controlling optical properties through nanoscale engineering.
Between the long-haul telecommunications industry and research in micro photonics lies a small market that will undoubtedly aid in driving the integration of on-chip optical networks: high performance supercomputing. Modern supercomputer performance is typically dominated by the quality of the interconnecting network that routes information between processor nodes. Consequently, a large body of research exists on network topology and infrastructure designed to make the most of each photonic component. This knowledge is ready to be applied to future optical interconnect networks that connect sub processor cores within a single chip.
If optical interconnects become essential for continued scaling progress in silicon electronics, an enormous market will open for integrated photonic circuit technology. Eventually, unimagined new products will be made possible by the widespread availability of affordable, high-density optical systems. Considering the historical development of computing hardware from the relays and vacuum tubes of early telephone networks, it is possible that optical interconnects could someday lead to all-optical computers, perhaps including systems capable of quantum computation.
Unfortunately, there is at present no clear path to practical on-chip optical data transfer and scalable all-photonic integrated circuits. The obstacles that currently stand in the way of optical interconnects are challenges for device physics and materials science. Break through are needed that either improve the set of materials available for micro photonic devices or obviate the need for increased materials performance through novel device designs.
1.3 ENTER OPTOELECTRONICS
Fiber optics use light to transmit data over a glass or plastic fiber(silica),and a seed of about 1.7Gbps was achieved in 1980s itself. Though plastic fibres are also used silica(glass fibre), i.e Silicon dioxide, that we use as insulator in CMOS fabrication is most commonly used. The primary benefit of using light rather than an electric signal over copper wiring is significantly greater capacity, since data transmission through fibres is at light speed.But this alone cannot make high speed transmission possible, it aso requires the end devices like modulators, demodulators etc where convertion between optical and electrical data takes place, also to work at such high speeds. The Bell Labs in France currently holds the record of transmission with muxing of about 155 different data streams each on its own light wave and each with a capacity of about 100Gbps that constitute in total a 14Tbps data link using a fibre pair with Dense WDM technology. In addition, glass fiber has desirable physical properties: it is lighter and impervious to factors such as electrical interference and crosstalk that degrade signal quality on copper wires. Hence optic fibres can be used even at places of high lightning with all dielectric cables.The high electrical resistance of fibres makes them usable even near high tension equipments. Hence repeaters are placed at ranges over 100Kms.
Photonics is the field of study that deals with light, especially the development of components for optical communications. It is the hardware aspect of fiber optics; and due to commercial demand for bandwidth, it has enjoyed considerable expansion and developments during the past decade. During the last few years, researchers at Intel have been actively exploring the use of silicon as the primary basis of photonic components. This research has established Intel’s reputation in a specialized field called silicon photonics, which appears poised to provide solutions that break through longstanding limitations of silicon as a material for fiber optics. In addition to this research, Intel’s expertise in fabricating processors from silicon could enable it to create inexpensive, high-performance photonic devices that comprise numerous components integrated on one silicon die.
1.4 COMPONENTS OF AN OPTICAL SYSTEM
To understand how optical data might one day travel through silicon in your computer, it helps to know how it travels over optical fiber today. First, a computer sends regular electrical data to an optical transmitter, where the signal is converted into pulses of light. The transmitter contains a laser and an electrical driver, which uses the source data to modulate the laser beam, making beam on and off to generate 1s and 0s. Imprinted with the data, the beam travels through the glass fiber, encountering switches at various junctures that route the data to different destinations. If the data must travel more than about 100 kilometers, an optical amplifier boosts the signal. At the destination, a photo detector reads and converts the data encoded in the photons back into electrical data. Similar techniques could someday allow us to collapse the dozens of copper conductors that currently carry data between processors and memory chips into a single photonic link.
The core of the internet and long-haul telecom links made the switch to fiber optics long ago. A single fiber strand can now carry up to one trillion bits of data per second, enough to transmit a phone call from every resident of New York City simultaneously. In theory, you could push fiber up to 150 trillion bits per second—a rate that would deliver the text of all the books in the U.S. Library of Congress in about a second.
Today’s devices are specialized components made from indium phosphide, lithium niobate, and other exotic materials that can’t be integrated onto silicon chips. That makes their assembly much more complex than the assembly of ordinary electronics, because the paths that the light travels must be painstakingly aligned to micrometer precision. In a sense, the photonics industry is where the electronics industry was a half century ago, before the breakthrough of the integrated circuit.
The only way for photonics to move into the mass market is to introduce integration, high-volume manufacturing, and low- cost assembly—that is, to “siliconize” photonics. By that we mean integrating several different optical devices onto one silicon chip, rather than separately assembling each from exotic materials. In our lab, we have been developing all the photonic devices needed for optical communications, using the same complementary metal oxide semiconductor (CMOS) manufacturing techniques that the world’s chip makers now use to fabricate tens of millions of microprocessors and memory chips each year.
A source that can produce narrow coherent beam of light is the prime necessity in optical communication. Hence lasers are the first choice. However LEDs are also used for some low cost applications. Also for lasers a 1000 times more power output may be obtained compared to LEDs, based on how we set the gain medium. However optical communication has limitations due to scattering effects at discontinuities or imperfections in fibre and also very slight variations in refractive index along the fibre that can affect the wavelength of signal transmitted. When such limiting factors persist a coherent narrow beam from source, i.e a beam of light with each photon at equal lengths along the fibre as well that at a singe cross section showing the same wave properties(frequency, phase, polarization etc.) , is a must otherwise the dispertion and diffraction phenomenon may occur in a different way to each photon in the beam and this can severly distort or destroy the light signal.
Optical communication operates on the short wave or IR region of EM spectrum (i.e from 1260-1675nm). The operating range of wavelength is divided into 6 bands. Among them the C-band(Conventional band), i.e from 1530-1565nm is most commonly used, since it has showed the least scattering. Most optical devices have been developed to work in this range. For communication single mode fibres are preferred, where mode represents the angles of incidence at the core-cladding interface for which transmission is possible. Multimode fibres (cross-section diameter>50um) are avoided due to intermodal dispertion, and even LEDs can b used. However single mode fibres require high stability for the light source used.
WDM started with muxing of 2 channels and now you can pack dozens of channels of high-speed data onto a single mode fiber with cross-section as low as 9um,separating the channels by wavelength, a technique called wavelength-division multiplexing, similar to frequency division multiplexing in radio communication. Arrayed waveguide grating structures that can perform both muxing and demultiplexing are used to implement WDM(Figure 1.3).
In AWG shown in Figure1.3, from 1 to 5, it acts as mux and demux the otherway. Regions 2 and 4 are free space segments and section 3 forms the array of waveguides with a constant length increment. Whereever light comes out of waveguide to free space, it diffracts, i.e spreads. A multiwavelength beam coming from section 1 after diffraction at section 2 passes to each of the waveguides of the array. The phase shift between the waves coming to section 4 will be such that the waves after diffraction constructive interference of the composed waves occur where they are received by different waveguides as shown in It has the advantage of integrated planar structure, low cost, low insertion loss and ease of network upgradation, since with increasing demand for bandwidth instead of laying new fibres, it only requires this device replaced with a higher capacity structure. As in any optical devices, changes in refractive index with temperature that can affect wavelength is a problem, and hence precision temperature control within +/-2 degree Celcius is required.