The past 2007 was a very successful year for the development of many Intel technologies, including those in the field of silicon photonics. The latest breakthrough achievements of Intel in this area were compared by the MIT Technology Review magazine with a triple win at the races - this is how the reviewers of the leading publication assessed a series of official announcements by the corporation. As Justin Rattner, chief technology officer and head of Intel's Corporate Technology Group, said, "We have demonstrated empirically that manufacturing technologies compatible with silicon CMOS development technology enable the creation of semiconductor optical devices.
The proof of this fact was a huge achievement, but no less significant steps are needed for the further development of this technological direction. Now we need to learn how to integrate silicon photonics devices into standard computer components; we are not yet able to do this. But at the same time, we continue to work actively together with the development departments various kinds products to offer manufacturers models for using semiconductor photonics in Intel solutions."
Silicon photonics as a means to eliminate bottlenecks on the way to the era of tera computing
Silicon photonics is the most important component long-term development strategy of the Corporate Technology Group, aimed at accelerating the transition to tera computing. The point is that as development multi-core processors with enormous computing power, engineers face new challenges. For example, the need for data transfer speed between the memory and the processor will soon exceed the physical limits imposed by copper conductors, and the transfer rate of electrical signals will become less than the speed of the processor. Even now, the performance of powerful computing systems is often limited by the speed of data exchange between the processor and memory. Today's data transfer technologies are designed for much lower bandwidth than photonics, and as the distance over which data is transferred increases, the transfer rate becomes even lower.
Tests of a prototype optical memory module showed that not electricity, but light can be used to access the server's memory
"It is necessary to bring the speed of data transfer between the components of the computing platform in line with the speed of processors. This is indeed a very important task. We see silicon photonics as a solution to this problem, and therefore we are pursuing a research program that allows us to occupy the forefront in this area" , said Kevin Kahn, Distinguished Research Engineer at Intel Corporation.
A team led by Intel Lead Optics Researcher Drew Alduino is building an optical communication system between processor and memory for Intel platforms. A test platform has already been created based on a fully buffered FB-DIMM memory, on which it boots and runs Microsoft Windows. The working prototype is proof of the possibility of connecting memory to the processor using optical communication lines without compromising system performance.
Creating a commercial version of such a solution has huge benefits for users. Optical communication systems will eliminate the bottleneck associated with the difference in memory bandwidth and processor speed, and improve the overall performance of the computing platform.
From research to implementation
The Photonics Technology Lab, run by Distinguished Intel Research Engineer Mario Paniccia, has proven that all components for optical communications—laser, modulator, and demodulator—can be fabricated from semiconductors using existing manufacturing techniques. The PTL has already demonstrated key silicon photonics components operating at record performance, including modulators and demodulators capable of data rates up to 40 Gbps.
Semiconductor photonics technology requires six main components:
- a laser that emits photons;
- a modulator for converting the photon stream into an information stream for transmission between elements of the computing platform;
- waveguides that act as "transmission lines" to deliver photons to their destinations, and multiplexers to combine or separate light signals;
- case, especially necessary for the creation of assembly technologies and low-cost solutions that can be used in the mass production of PCs;
- a demodulator for receiving photon streams carrying information and their inverse - converting into an electron stream available for processing by a computer;
- electronic circuits to control these components.
The implementation of all these components of optical communication based on semiconductor technologies is widely recognized as the most important research problem, the solution of which will lead to a huge technical breakthrough. PTL has already set a number of world records by developing high performance devices, modulators, amplifiers and demodulators that deliver data rates up to 40 Gbps. Over the next five years, Intel will look for ways to integrate these components into actual products.
In the field of semiconductor photonics, Intel has already reached the finish line. Research in the field of integration of optical elements has already moved from the stage of scientific or technological development to the stage of creating commercial products. The research team is now in the process of identifying the capabilities and specifications for designing innovative products based on this revolutionary technology. Ultimately, Intel creates prototypes and works closely with product development teams to accelerate adoption. new technology.
In addition to its own activities, Intel Corporation is funding some of the most promising research in this direction outside of CTG - in particular, it is collaborating with the University of California at Santa Barbara, which is developing a hybrid semiconductor laser. Talented graduates from various universities from other countries are also trained in the PTL laboratory.
According to Intel Lead Optics Researcher Richard Jones, "There are two critical challenges ahead of us in our hybrid semiconductor laser project right now. secondly, we have to combine a hybrid laser, a high-speed semiconductor modulator and a multiplexer to prove that we can create a single optical transmitter based on production technology CMOS compatible".
The introduction of silicon photonics technologies will include the development of new manufacturing processes for manufacturing lasers on a large scale. The success of Intel Corporation in the field of photonics will allow it to significantly outperform potential competitors. PTL has already registered about 150 patents. The most prestigious publications, such as Nature, have noted the unprecedented achievements of Intel specialists. In addition, in 2007, Intel was awarded the EE Times ACE Award for Most Promising New Technology.
In pursuit of photons
Unlike the existing well-established and decades-old processes for manufacturing transistors, the technology for creating elements for semiconductor photonics is completely new. Certain problems stand in the way of its implementation: optimization of devices, increasing the reliability of the design, testing methodology, ensuring energy efficiency, and developing subminiature devices.
In order for new components to be put into practice, PTL must ensure that optical components meet the extremely high reliability criteria used in the computer industry. In the optical industry, stringent reliability standards have been developed over decades. In accordance with them, months of testing are required before the start of serial production of new products. If problems are identified during these lengthy tests, fixing them and retesting them can significantly delay a product's time to market.
One of the most important problems is optimization, because the PTL laboratory develops optical devices for mass computing. While there are no other similar products, standards, or other benchmarks, it is up to engineers developing a new process to find solutions that best meet the needs of computer applications.
Currently, the PTL research team, relatively small by the standards of photoelectronics, is gradually switching to the commercialization of semiconductor photonics solutions and expects that the mass implementation of this incredible technology can begin as early as 2010.
A team of optics specialists from the Digital Enterprise Group (DEG), led by Victor Krutul, is developing applications that will provide the foundation for the emerging technology. "We believe that Intel's products will continue to conform to Moore's Law through the development of optical communications," says Krutal.
When to transfer information between components of the same computing platform and between different systems not electrons, but photons will be used, the next computer revolution will be accomplished. Leading electronics manufacturers around the world are already joining this race to gain competitive advantages. The significance of a new technology can be compared to an invention integrated circuits. Intel is leading the way in this research and in the development of semiconductor photonics components.
News Electronics newsSilicon photonics: will light replace electricity?
All-semiconductor CW laser solves previously insurmountable problem of two-photon absorption
Microelectronics already faces physical limitations (at the atomic level) in transmitting electrical signals between microcircuits. Possible Solution This problem may be the development of non-traditional technologies, in particular, silicon photonics.
Intel has already created many of the structures needed to make signaling between chips using light as easy as electrons now do. The main problem for this was the lack of a suitable light source. Recently, Intel announced a new breakthrough in this area, the first all-semiconductor continuous wave laser using a physical phenomenon called the Raman effect (in quantum mechanics, the Raman effect is described as the exchange of energy between scattering molecules and incident light), and built using standard commercial CMOS -crystals.
Using the power of semiconductors, Intel researchers were able to realize the functionality of a traditional, bulky Raman laser that uses glass and is typically the size of a suitcase by shrinking it down to the thickness of a single track on a silicon wafer.
This breakthrough in silicon photonics will lead to practical and affordable solutions for communications and computing, to the creation of new medical equipment and sensors, and the tunable semiconductor laser can replace its predecessors that cost hundreds and thousands of dollars. This achievement may also lead to the acceleration of the creation of new optical interconnects between microcircuits and external devices, because thin optical fibers take up less space than electrical cables and will provide better cooling conditions for computers and servers.
The semiconductor laser demo wafer was manufactured using standard CMOS technology on an existing production line. This means that for these new technologies, the path from laboratory to production may not be long and complicated, as is the case for some non-traditional technologies, but rather direct and fast.
The silicon-photonic chip, the result of ten years of research, is capable of transmitting data using light pulses at speeds up to 100 Gbps. During testing, the transmission distance reached two kilometers.
Light allows you to transfer data faster than the copper cables that connect storage systems, network equipment and servers in data centers. The silicon-photonic chip will make it possible to connect servers and supercomputers of future generations with high-speed fiber-optic connections, in which huge amounts of data must be transferred between computing nodes.
IBM is developing its technology with the expectation of advancing in the data center, and in PCs or handheld devices, you should not expect it soon, said Wilfrid Hensch, senior manager of IBM's silicon photonics division.
Silicon photonics technology has the potential to revolutionize how servers are deployed in data centers by separating processing, memory, and storage blocks from each other. As a result of this decoupling, applications can run faster and component costs can be reduced by consolidating fans and power supplies.
Due to the growth in the use of machine learning systems and Big Data processing, today the need for computing power servers. With optical interconnects, dozens of processors could interact within one server rack, which would make it easier to distribute tasks for multi-node processing, said Richard Doherty, director of research at The Envisioneering Group.
With optical interconnects, servers could be easily swapped out, like storage drives, without interruption, depending on processing power needs, he added.
Light is already being used for long-distance data transmission in communication networks, but fiber-optic technologies are not cheap. Optical cables are also supported by the Thunderbolt interface, which is used in Macs and PCs for high-speed communication with peripherals.
IBM's silicon photonics technology is cheaper and designed for shorter distances than optical telecommunications equipment, Hensch says.
Intel also created silicon-photonic chips for the data center, but the corporation failed to meet the announced release dates. IBM may not be the first to offer a silicon-photon transmitter, but its technology is more viable and less complex than Intel's, Doherty said.
According to him, the IBM chip is simpler and cheaper to manufacture and has simple structure, while the Intel solution requires additional physical components.
Intel itself, however, claims that its optical modules are integrated and have advantages in terms of testing and cost.
The chips of the two companies transmit data in completely different ways, and each has its own advantages. The IBM chip is designed to transmit on a single fiber over four channels with different wavelengths, while Intel technology scales better, allowing more strands in the cable, Doherty pointed out.
Intel has MXC optical cables with up to 64 strands, each with a transfer rate of 25 Gbps. But increasing the number of fibers can come at a cost, and IBM's single-fiber option at a lower cost could meet the demands of many data centers in terms of speed and distance, Doherty added.
IBM did not specify when its silicon-photonic chips could enter the market.
On September 18 this year, Intel, together with the University of California, Santa Barbara, demonstrated the world's first electrically pumped hybrid silicon laser, which combines the ability to emit and propagate light through a silicon waveguide, and also takes advantage of the low cost of silicon production. . The creation of a hybrid silicon laser is another step towards obtaining silicon chips containing dozens and even hundreds of cheap lasers, which will form the basis of computer electronics in the future.
History of silicon photonics
Silicon photonics is one of the main directions in the research work of Intel Corporation. The next breakthrough of the company in this area was the creation of the world's first electrically pumped hybrid silicon laser.
Now, in fact, the way has been opened for the creation of optical amplifiers, lasers, and light wavelength converters using the well-established technology for the production of silicon microcircuits. Gradually, the "siliconization" of photonics is becoming a reality and in the future will make it possible to create low-cost high-performance optical circuits that allow data exchange both inside and outside the PC.
Optical communication systems have certain advantages over traditional cable systems, chief among which is their enormous bandwidth. For example, optical fibers used today in communication systems can simultaneously transmit up to 128 different data streams. The theoretical limit for data transmission over fiber is estimated at 100 trillion bits per second. In order to present this huge figure, let's make a simple comparison: this bandwidth is enough to provide transmission telephone conversations simultaneously all the inhabitants of the planet. Therefore, it is quite understandable that optical communication systems attract the close attention of all research laboratories.
To transmit information using light radiation, it is necessary to have several mandatory components: radiation sources (lasers), light wave modulators, through which information is embedded in the light wave, detectors and optical fiber for data transmission.
With the help of several lasers emitting waves of different wavelengths and modulators, it is possible to transmit many data streams simultaneously via a single optical fiber. On the receiving side, for information processing, an optical demultiplexer is used, which separates carriers with different wavelengths from the incoming signal, and optical detectors, which allow converting optical signals into electrical ones. Structural scheme optical communication system is shown in fig. one.
Rice. 1. Structural diagram of an optical communication system
Research in the field of optical communication systems and optical circuits began back in the 1970s - then optical circuits were presented as some kind of optical processor or super-optical chip, in which a transmitting device, a modulator, an amplifier, a detector, and all the necessary electronic components. However, the practical implementation of this idea was hampered by the fact that the components of optical circuits were made of different materials, therefore, integrate into a single platform (chip) based on silicon all necessary components was impossible. Despite the triumph of silicon in the field of electronics, its use in optics seemed highly doubtful.
The study of the possibility of using silicon for optical circuits has been going on for many years - since the second half of the 1980s. However, little progress has been made during this time. Compared with other materials, attempts to use silicon to build optical circuits did not bring the expected results.
The fact is that due to the structural features of the band gap of the crystal lattice of silicon, the recombination of charges in it leads mainly to heat release, and not to the emission of photons, which does not allow it to be used to create semiconductor lasers that are sources of coherent radiation. At the same time, in semiconductors such as gallium arsenide or indium phosphide, the recombination energy is released mainly in the form of infrared photons; therefore, these materials can serve as photon sources and be used to create lasers.
Another reason preventing the use of silicon as a material for creating optical circuits is that silicon does not have a linear electro-optical Pockels effect, on the basis of which traditional fast optical modulators are built. The Pockels effect consists in changing the refractive index of light in a crystal under the influence of an applied electric field. It is due to this effect that light can be modulated, since a change in the refractive index of a substance in a corresponding way leads to a change in the phase of the transmitted radiation.
The Pockels effect manifests itself only in piezoelectrics and, due to its low inertia, theoretically allows light modulation up to a frequency of 10 THz. In addition, due to linear dependence between the refractive index and the electric field strength, the nonlinear distortions during light modulation are relatively small.
Other optical modulators are based on such effects as electro-absorption or electro-refraction of light under the influence of an applied electric field, however, these effects are also weakly expressed in silicon.
The modulation of light in silicon can be obtained on the basis of the thermal effect. That is, when the silicon temperature changes, its refractive index and light absorption coefficient change. Nevertheless, due to the presence of hysteresis, such modulators are rather inert and do not allow obtaining a modulation rate higher than a few kilohertz.
Another method of radiation modulation based on silicon modulators is based on the effect of light absorption on free carriers (holes or electrons). This modulation method also does not allow one to obtain high speeds, since it is associated with the physical movement of charges inside the silicon modulator, which in itself is an inert process. At the same time, it should be noted that silicon modulators based on the described effect can theoretically maintain a modulation rate up to 1 GHz, but in practice, modulators with a rate of up to 20 MHz have so far been implemented.
Despite all the difficulties of using silicon as a material for optical circuits, significant advances have been made in this direction recently. As it turned out, the doping of silicon with erbium (Er) changes the structure of the band gap in such a way that the charge recombination is accompanied by the emission of photons, that is, it becomes possible to use silicon to obtain semiconductor lasers. The first commercial doped silicon laser was developed by ST Micro-electronics. Also promising is the use of tunable semiconductor lasers, demonstrated by Intel back in 2002. Such lasers use a Fabry-Perot interferometer as a resonator and emit at several frequencies (multimode). To isolate monochromatic radiation, special external filters based on diffraction gratings (dispersive filters) are used - fig. 2.
Rice. 2. Tunable lasers with filters
based on dispersion gratings
The resulting laser system with an external dispersive resonator makes it possible to tune the radiation wavelength. Traditionally, to obtain the required wavelength, the filters are finely tuned relative to the resonator.
Intel has been able to create a tunable laser that has no moving parts at all. It consists of an inexpensive multimode laser with a grating embedded inside a waveguide. By changing the grating temperature, it is possible to tune to a certain wavelength, that is, to switch between individual laser modes.
Silicon Optical Modulators
In February 2004, Intel made another breakthrough in silicon photonics by demonstrating the world's first silicon optical phase modulator at 1 GHz.
This modulator is based on the effect of light scattering on free charge carriers and in its structure resembles in many ways a CMOS transistor based on SOI (silicon on insulator) technology. The structure of the optical phase modulator is shown in fig. 3.
Rice. 3. Structural diagram of an optical silicon phase modulator
On a substrate of crystalline silicon with a layer of insulator (silicon dioxide) there is a layer of crystalline silicon n-type. This is followed by a layer of silicon dioxide, in the center of which is a layer of polycrystalline silicon p-type, which performs the function of a waveguide. This layer is separated from the crystalline silicon n-type the thinnest layer of insulator (gate dielectric), the thickness of which is only 120 angstroms. In order to minimize light scattering due to metal contact, the metal contacts are separated from the silicon oxide layer by a thin layer of polycrystalline silicon on both sides of the waveguide.
When a positive voltage is applied to the gate electrode, a charge is induced on both sides of the gate dielectric, and on the waveguide side (polycrystalline silicon p-type) these are holes, and from the side of silicon n-type - free electrons.
In the presence of free charges in silicon, the refractive index of silicon changes. A change in the refractive index causes, in turn, a phase shift of the transmitted light wave.
The modulator considered above makes it possible to produce phase modulation of the reference signal. In order to turn phase modulation into amplitude modulation (a signal modulated in phase is difficult to detect in the absence of a reference signal), the optical modulator additionally uses a Mach-Zender interferometer (MZI), which has two arms, each of which integrates a phase optical modulator (Fig. . four).
Rice. 4. Block diagram of the optical modulator
The use of phase optical modulators in both arms of the interferometer makes it possible to ensure the equality of the optical lengths of the arms of the interferometers.
The reference light wave propagating along the optical fiber is divided by a Y-splitter into two coherent waves, each of which propagates along one of the arms of the interferometer. If both waves are in phase at the junction point of the interferometer arms, then as a result of the addition of these waves, the same wave will be obtained (losses in this case are neglected) as before the interferometer (constructive interference). If the waves are added in antiphase (destructive interference), then the resulting signal will have zero amplitude.
This approach allows for amplitude modulation carrier signal - applying voltage to one of the phase modulators, the phase of the wave in one of the arms of the interferometer is changed to n or do not change at all, thus providing a condition for destructive or constructive interference. Thus, applying a voltage to the phase modulator with a frequency f, it is possible to carry out amplitude modulation of the signal with the same frequency f.
As already noted, Intel's silicon optical modulator, demonstrated in February 2004, was capable of modulating radiation at a speed of 1 GHz. Subsequently, in April 2005, Intel demonstrated a modulator operating at a frequency of 10 GHz.
Raman continuous silicon laser
In February 2005, Intel announced another technological breakthrough - the creation of a continuous-wave silicon laser based on the Raman effect.
The Raman effect has been used for quite a long time and is widely used to create light amplifiers and lasers based on optical fibers.
The principle of operation of such devices is as follows. Laser radiation (pump radiation) with a wavelength is injected into an optical fiber (Fig. 5). In an optical fiber, photons are absorbed by atoms of the crystal lattice, which, as a result, begin to "swing" (vibrational phonons are formed), and, in addition, photons with lower energy are formed. That is, the absorption of each photon with a wavelength l=1.55mm leads to the formation of a phonon and a photon with a wavelength l=1.63mm.
Rice. 5. The principle of operation of a light amplifier due to the Raman effect
Now imagine that there is also modulated radiation that is coupled into the same fiber as the pump radiation and results in stimulated emission of photons. As a result, the pump radiation in such a fiber is gradually converted into signal, modulated, amplified radiation, that is, the effect of optical amplification is achieved (Fig. 6).
Rice. 6. Using the Raman effect to enhance
modulated radiation in optical fiber
The problem, however, is that such a conversion of the pump beam into signal radiation and, accordingly, amplification of the signal radiation requires that both the signal radiation and the pump radiation travel along the fiber for several kilometers. Of course, amplification schemes based on multi-kilometer optical fiber cannot be called simple and cheap, as a result of which their application is significantly limited.
Unlike glass, which forms the basis of an optical fiber, the Raman effect in silicon is 10 thousand times stronger, and to achieve the same result as in an optical fiber, it is enough that the pump radiation and signal radiation propagate together only a few centimeters . Thus, the use of the Raman effect in silicon makes it possible to create miniature and cheap light amplifiers or optical lasers.
The process of creating a silicon optical amplifier, or Raman laser, begins with the creation of an optical silicon waveguide. This technological process is no different from the process of creating traditional CMOS chips using silicon substrates, which, of course, is a huge advantage, since it significantly reduces the cost of the manufacturing process itself.
The radiation fed into such a silicon waveguide travels only a few centimeters, after which (due to the Raman effect) it is completely converted into signal radiation with a longer wavelength.
In the course of the experiments, it turned out that it is advisable to increase the pump radiation power only up to a certain limit, since a further increase in power does not lead to an increase in the signal radiation, but, on the contrary, to its weakening. The reason for this effect is the so-called two-photon absorption, the meaning of which is as follows. Silicon is an optically transparent substance for infrared radiation, since the energy of infrared photons is less than the band gap of silicon and it is not enough to transfer silicon atoms to an excited state with the release of an electron. However, if the density of photons is high, then a situation may arise when two photons simultaneously collide with a silicon atom. In this case, their total energy is sufficient to transfer the atom with the release of an electron, that is, the atom goes into an excited state with the simultaneous absorption of two photons. This process is called two-photon absorption.
Free electrons produced as a result of two-photon absorption, in turn, absorb both pump and signal radiation, which leads to a strong weakening of the optical amplification effect. Accordingly, the higher the pump radiation power, the stronger the effect of two-photon absorption and absorption of radiation on free electrons. The negative consequence of two-photon absorption of light for a long time prevented the creation of a continuous-wave silicon laser.
In a silicon laser created in the Intel laboratory, for the first time, it was possible to avoid the effect of two-photon absorption of radiation, more precisely, not the phenomenon of two-photon absorption itself, but its negative consequences - the absorption of radiation on the resulting free electrons. The silicon laser is a so-called PIN structure (P-type - Intrinsic - N-type) (Fig. 7). In such a structure, a silicon waveguide is embedded inside a semiconductor structure with a P- and N-region. Such a structure is similar to a planar transistor circuit with a drain and source, and a silicon waveguide is integrated instead of a gate. The silicon waveguide itself is formed as a region of silicon rectangular in cross section (refractive index 3.6) surrounded by a silicon oxide shell (refractive index 1.5). Due to this difference in the refractive indices of crystalline silicon and silicon oxide, it is possible to form an optical waveguide and avoid radiation losses due to transverse propagation.
Rice. 7. PIN structure of a continuous-wave silicon laser
Using such a wave structure and a pump laser with a power of fractions of a watt, it is possible to create radiation in the waveguide with a density of about 25 MW/cm 2, which is even higher than the radiation density that can be obtained using high-power semiconductor lasers. Raman amplification at such a radiation density is not too high (on the order of several decibels per centimeter), but this density is quite sufficient for the implementation of a laser.
In order to eliminate the negative effect of absorption of radiation on free electrons formed in the waveguide as a result of two-photon absorption, a silicon waveguide is placed between two gates. If a potential difference is created between these gates, then under the influence of an electric field, free electrons and holes will be “pulled out” from the silicon waveguide, thereby eliminating the negative consequences of two-photon absorption.
In order to form a laser based on this PIN structure, it is necessary to add two mirrors to the ends of the waveguide, one of which must be semitransparent (Fig. 8).
Rice. 8. Scheme of a continuous silicon laser
Hybrid silicon laser
A continuous-wave silicon laser based on the Raman effect basically assumes the presence of an external source of radiation, which is used as pump radiation. In this sense, this laser does not solve one of the main problems of silicon photonics - the ability to integrate all structural blocks (radiation sources, filters, modulators, demodulators, waveguides, etc.) into a single silicon chip.
Moreover, the use of external sources of optical radiation (located outside the chip or even on its surface) requires a very high accuracy of laser alignment relative to the silicon waveguide, since a misalignment of several microns can lead to the failure of the entire device (Fig. 9). The requirement of precise adjustment does not allow bringing this class of devices to the mass market and makes them rather expensive. Therefore, the alignment of a silicon laser with respect to a silicon waveguide is one of the most important problems in silicon photonics.
Rice. 9. When using external lasers, precision laser alignment is required
and waveguide
This problem can be solved if the laser and the waveguide are created in the same crystal within the same technological process. That is why the creation of a hybrid silicon laser can be considered as bringing silicon photonics to a new level.
The principle of operation of such a hybrid laser is quite simple and is based on the emitting properties of indium phosphide (InP) and the ability of silicon to conduct light.
The structure of the hybrid laser is shown in fig. 10. Indium phosphide, which acts as the active substance of a semiconductor laser, is located directly above the silicon waveguide and is separated from it by the thinnest layer of dielectric (its thickness is only 25 atomic layers) - silicon oxide, which is "transparent" for the generated radiation. When a voltage is applied between the electrodes, a flow of electrons occurs in the direction from the negative electrodes to the positive. As a result, an electric current passes through the crystal structure of indium phosphide. When passing electric current through indium phosphide, as a result of the process of recombination of holes and electrons, photons arise, that is, radiation. This radiation directly enters the silicon waveguide.
Rice. 10. Structure of a hybrid silicon laser
The described structure of the silicon laser does not require additional adjustment of the laser relative to the silicon waveguide, since their mutual arrangement relative to each other is implemented and controlled directly during the formation of the monolithic structure of the hybrid laser.
The production process of such a hybrid laser is divided into several main stages. Initially, in a “sandwich” consisting of a layer of silicon, an insulator layer (silicon oxide) and another layer of silicon, a waveguide structure is formed by etching (Fig. 11), and this technological stage of production does not differ from those processes that are used during production microchips.
Rice. 11. Formation of a waveguide structure in silicon
Next, on the surface of the waveguide, it is necessary to form a crystal structure of indium phosphide. Instead of using the technologically complex process of growing an indium phosphide crystal structure on an already formed waveguide structure, an indium phosphide substrate along with a semiconductor layer n-type is formed separately, which is much simpler and cheaper. The challenge is to connect the indium phosphide to the waveguide structure.
To do this, both the structure of silicon waveguides and the indium phosphide substrate are subjected to the oxidation process in a low-temperature oxygen plasma. As a result of this oxidation, an oxide film with a thickness of only 25 atomic layers is formed on the surface of both materials (Fig. 12).
Rice. 12. Indium phosphide substrate
with formed oxide layer
When two materials are heated and pressed against each other, the oxide layer acts as a transparent glue, ensuring their fusion into a single crystal (Fig. 13).
Rice. 13. "Gluing" the structure of silicon waveguides
with indium phosphide support
It is precisely because the silicon laser of the described design consists of two materials glued together that it is called a hybrid laser. After the bonding process, the excess indium phosphide is removed by etching and metal contacts are formed.
The technological process for the production of hybrid silicon lasers makes it possible to place dozens and even hundreds of lasers on a single chip (Fig. 14).
Rice. 14. Scheme of a chip containing four
hybrid silicon laser
The first chip, demonstrated by Intel together with the University of California, contained seven hybrid silicon lasers (Fig. 15).
Rice. 15. Radiation of seven hybrid silicon lasers,
made on a single chip
These hybrid lasers operate at a wavelength of 1577 nm at a threshold current of 65 mA with output power up to 1.8 mW.
Currently, the hybrid silicon laser is operable at temperatures below 40 °C, but in the future it is planned to increase the operating temperature to 70 °C, and reduce the threshold current to 20 mA.
The Future of Silicon Photonics
The development of a hybrid silicon laser could have far-reaching implications for silicon photonics and serve as Starting point for the era of high performance computing.
In the near future, dozens of silicon lasers, modulators and a multiplexer will be integrated into the chip, which will make it possible to create optical communication channels with a terabit throughput(Fig. 16).
Rice. 16. Chip of the optical communication channel,
containing dozens of silicon lasers,
filters, modulators and multiplexer
“Thanks to this development, we will be able to create low-cost optical data buses with terabit bandwidth for the computers of the future. By doing so, we can bring a new era of high-performance computing closer,” said Mario Paniccia, Director of the Photonics Technology Lab at Intel Corporation. “Despite the fact that commercial use of this technology is still very far away, we are confident that dozens and even hundreds of hybrid silicon lasers, as well as other components based on silicon photonics, can be placed on a single silicon chip.”