Silicon Photonics: The Next Silicon Revolution?

silicon photonics what a cool sounding word it sounds like something from the age of the jetsons but behind that futuristic phrase is a simple need if mems is the result of applying modern nanoscale cmos processes to the mechanical world then doing the same for the optical realm gives us silicon photonics in this video i want to talk about another magic silicon technology one that's starting to make a splash in the contemporary technology world but first the asian armature newsletter check out this one about the semiconductor water problem the video did quite well and it's worth another read considering today's issues the link is in the video description below i try to put one out every week maybe two alright back to the show let us break it down starting with photonics photonics what does that mean lasers sharks with lasers eyeglasses lithography machines how does that jive with the silicon world in the context i am talking about here photonics technologies transmit and manipulate light in the form of light particles or photons it is related to the world of optical data transmission previously networking companies communicated using electrical signals sent through copper wire the issue is that the electrons traveling through such wires interact with other atoms which slows them down and generates heat by the 1990s networking companies found themselves struggling to deal with exploding data traffic network demands they revolutionized long distance communications by switching to using light center via optic fiber to efficiently transmit data across great distances light moves through optical fiber at the speed of light nothing is faster than the speed of light this lets us transmit optical signals at super high frequencies which means higher data volume transmissions later on engineers took the concept yet another step forward sending multiple signals through the same fiber by using different light wavelengths that don't interfere with each other today optical fiber technologies dominate the long distance communication space over 2 billion kilometers of optical fiber have been deployed enough to wrap around the world over 50 000 times now silicon the nano electronics industry has been using silicon for decades the element is the single most widely studied element in human history it is plentiful low cost and allows for massive scale people have been able to make silicon do amazing things billions of transistors on one wafer entire systems on chips that are faster smaller and cheaper than their priors and for that reason engineers have wanted to bring the titanic scale of modern cmos manufacturing with its edas pdks and all the other tools and processes used for it over to the photonics world furthermore transistors on traditional chips right now are still communicating using electrons through wires what if you were to replace those wires and electrons with optical fibers and light more on that later dreams of a monolithic silicon chip meaning everything on it being made of a single material system transmitting and manipulating light date back to the 1970s such a system would have five components one a light source usually a laser two routes and pathways to manipulate light bend it guide it filter it couple it split it and combine it kind of like optical fiber does but just within the integrated chip these are broadly referred to as passive structures but as with legos there are different shapes and structures with their own names three and four ways to convert digital electronic signals into digital optical signals and vice versa the former is called a modulator three the latter is a photo detector four which makes sense if you think about it a component that can do both modulator and photo detector work is called a transceiver in data centers a transceiver sits at each end of the optical cable converting light data to electrical data back and forth and finally number five we need traditional cmos electronics to accompany the various aforementioned photonics components these serve support functions like encoding and decoding certain data items anyway so there you have it a complete photonic system that researchers are trying to incorporate all onto a single monolithic chip but there's one big issue actually two and they have to do with the light source and the modulator the first issue is that on its own silicon cannot emit light crystalline silicon has what is called an indirect band gap this prevented it from being used for the first light emitting diodes or leds and also means that it cannot lase without lazing we cannot use silicon to make light no pure silicon light source researchers and engineers have to modify the silicon structure to force it to emit light for instance they implanted boron into the silicon to finally create efficient room-temperature silicon-based leds but commercial silicon-based lasers still aren't there yet second silicon's crystal structure means that it does not exhibit the pokel's effect the poco's effect describes a phenomenon where you can use an electric field to control how fast light goes through a certain object in other words its refractive index lasers typically lays continuously the modulator's purpose is to convert its continuous lasing into a digital signal the preferred way to do this is with a material sitting in front of the light changing its intensity by absorbing it if we cannot control light's progress through silicon using electrical fields then we cannot convert digital electrical signals into digital light signals no modulator these two big issues make it that much harder to produce a fully integrated photonics device out of pure silicon using the same methods we used to make an apple a15 intel core or mems accelerometer part early optics researchers instead focused on other materials and made a great deal progress with gallium arsenide and indium phosphide silicon photonics as we know it today starts in the mid-1980s with the work of richard soroff in 1987 he co-authored a paper discussing how silicon can be manipulated into adjusting its refractive index from there the industry was able to replicate one of the elementary building blocks of a semiconductor electronic device called a p end junction using a type of photonic passive structure called a waveguide this kickstarted the silicon photonics industry which began work on building out a photonic system with a practical light source and modulator the light source is where silicon faces its greatest challenge scientists have tried a lot of things in order to make it laze a silicone-based laser is considered the holy grail of the silicon photonics space the final piece of the puzzle but with that still far off engineers settled on workarounds the most pragmatic workaround is to use an external laser positioned outside the chip itself this has the added benefit of keeping the chip from getting overheated another is to bond a pre-made laser made from a different material like indium phosphide something known as hybrid integration today most commercial silicon photonic providers do one of the two silicon modulators had been studied from the very beginning with slow but steady breakthroughs throughout the 1980s and 1990s first in shrinking the device and then in speeding up its throughput in 2004 intel announced the first silicon based high-speed optical modulator meaning to have a bandwidth over 1 gigahertz this attracted huge media attention and represents a big breakthrough that 2004 system used a mach zedner interferometer shortened to mzi these modulators work by splitting light into two wavelengths and then recombining them to replicate a one or zero signal then in 2012 intel announced their first fully integrated cmos silicon photonics transceiver with four channels each at 25 gigabit per second it was fab with a 90 nanometer process this used a different type of modulator a ring modulator device that offers size benefits over the mzi with these advancements the silicon photonics industry managed to progress out of the laboratory despite still lacking a pure silicon based laser the industry quickly found its first big commercialization opportunity inside the data center a hyperscaler is a term that describes alibaba aws google and microsoft companies offering immense cloud computing scalability to their customers to do this they are building out titanic data centers spending tens of billions a year in capital expenditure there is more data transmitting between a couple hundred servers within a single hyperscaler data center than what goes between the east and west halves of the united states public internet thus the hyperscalers are always looking for new ways to improve their internal performance if you might recall from earlier transceivers are products that convert between digital optical and digital electrical signals they are a separate item that is plugged into the switch gear at the top of each server rack data flows through optical fiber into the servers through this equipment with silicon photonics you can now integrate transceiver functionality right onto the chip replacing the legacy component using them saves on cost power and labor and cracks a bandwidth bottleneck today companies like intel cisco and macomb are selling millions of units a year and photonic components will continue to take share from legacy optics and copper wire in this expanding space silicon photonics biggest market in the short term will likely be in the data center but there is some potential in the sensor and lidar markets lidar uses light to help acquire a 3d picture of a particular environment as the name implies it works similar to radar but since optical light waves are so much smaller than radio waves you can get higher resolutions it is seen as an important part of the autonomous driving puzzle the problem is that lidar setups are rather expensive depending on what you are using a single system can cost up to seventy thousand dollars they are also quite bulky which has its own issues a silicon photonics-based lidar system offers the possibility of integrating many discrete optical components right onto the chip this would drastically bring down lidar costs in addition to shrinking it to a more manageable size there are a number of companies pursuing this space until subsidiary mobileye recently presented a small lidar system on chip with integrated lasers this is a pretty crowded space with notable other players including point cloud aeva voyant photonics and analog photonics lidars are so hot right now silicon photonics products use silicon on insulator or soi wafers these are wafers with special layers usually silicon dioxide in addition to the silicon the layers contrasting refractive indices help confine light because of this silicon photonics require a slightly different fabbing process a couple years behind the leading edge this makes it a specialty node with special opportunities for foundries not looking to compete for silly things like three nanometers global foundries in particular has done a lot in the space they had acquired a great deal of relevant ip from ibm when they acquired big blue's micro electronics division back in 2014. dylan patel of semi-analysis which i recommend seems to believe that they are the market leader it's great for them after they abandoned going for the leading edge a few years back intel has been an r d pioneer for silicon photonics for a very long time they're also setting up a foundry and i feel that they would be missing out if they didn't offer some of that ip and capability to outsiders and finally tsmc they have not offered much in this space an absence that more than a few have noticed recently it seems like their corporate strategy has been building up integration schemes that allow silicon photonics chiplets to work together seamlessly with traditional semiconductors i've been wondering about this for a while and here's my thinking the issue seems to be one of volume the single biggest market is transceivers a market research publication estimates about 50 to 75 million transceiver units sold annually over the next few years assuming a 200 millimeter soi wafer a 25 millimeter die and 100 yield then that's about 40 to 60 000 wafers for an entire industry less than a month's production for a typical mega fab broken down to any one customer that's just a few days run at a foundry i guess for tsmc the juice isn't worth the squeeze it illustrates one of the challenges the industry now faces even if it takes over the entirety of the transceiver and lidar markets what other big markets are out there investors and enterprises have poured hundreds of millions of dollars into silicon photonics r d over the years and it's yielded some fruits but not enough to go mainstream thus some startups have sought to achieve the dream of the 1970s replacing the copper wiring on chips with optical fiber to create a silicon photonics microprocessor capable of disrupting traditional semiconductors but there are challenges here too today's leading edge transistors now have feature sizes only a few nanometers large but photonics components cannot be made smaller than the wavelengths of the light they carry this tops out at about one micrometer electrons on the other hand have wavelengths of just a few nanometers at the seven nanometer node one square micrometer of real estate can house over a hundred transistors a very high opportunity cost it pushes the industry away from highly integrated silicon photonics monoliths and towards packaging solutions that pair photonics and traditional chiplets together which might explain what tsmc is doing right now is not a deal breaker like i said there are a few startups out there working on it but it is something to think about i discussed in an earlier video the situation faced by the mems industry that industry sells millions of units each year for everyday items like accelerometers and sensors units-wise it's a legitimate hit however each mems die sells for cents with the majority of value accruing to the packaging this has stifled innovation and made mems commercialization extremely difficult as a result the technology has yet to really hit its financial potential as the next silicon revolution silicon photonics is a technology from the future vying to change the way things are done but it also happens to be a technology in search of a commercial market big and valuable enough to fulfill its potential without that it risks suffering the same fate as its elder sibling alright everyone that's it for tonight thanks for watching subscribe to the channel sign up for the newsletter and i'll see you guys next time

Channel: Asianometry

Views: 408,671

Keywords:

Id: 29aTqLvRia8

Length: 15min 45sec (945 seconds)

Published: Thu Jun 16 2022