Photonic ICs, Silicon Photonics & Programmable Photonics - HandheldOCT webinar

what i'm going to talk to you about today uh as gunai asked me is photonic integrated circuits but i'm gonna dive a little bit more deeper into silicon photonics and silicon nitride photonics and then finally into programmable photonics so just a few words about myself so as already introduced i'm part of the photonics research group again so that's a fairly large research group of about almost 100 people nine professors we do make photonic systems on a chip for a variety of applications so telecommunication data com biomedical applications sensing etc our key technology is silicon photonics where we make use of imix silicon photonics platforms but we also look into different material systems especially where silicon is lacking some functionality so in this presentation i'm going to start very basic on photonic integrated circuits then a move towards larger scale circuits and silicon photonics and then finally go end with the the topics that i'm working on myself mostly today is programmable photonics so the basic rationale is manipulation of light i mean if you uh if you look at light light is extremely useful it contains a lot of information so you can carry information in the power of the face uh the beam profiles and the polarization the wavelength of the light and that's why we want to process light because we can use it for sensing we can use it to carry to transport information now if you look in more detail to what light is we guess all know that light is an electromagnetic wave with a time oscillation and a spatial oscillation that propagates and that you also have an electrical and magnetic field in there at the same time you can represent light as a flux of photons so where you where each photon carries an energy that is proportional to the frequency now light finds its its use in a lot of different applications the the largest application of of lights or of lasers for instance today is in material processing laser cutting laser welding uh laser scribing etc but then displays telecommunication sensing are also very very important applications of light and then recently there's been more and more talk about using light for computing that was actually the cover of nature this week was about an optical brain inspired computer now the key driver today for photonics and if we talk about photonics we're usually talking about manipulating light on a small scale is their communication essentially the entire internet traffic runs over optical fibers and without those optical each of these optical fibers needs to have a transceiver at the beginning and the end now optical communication unlike what most people think optical communication is not new optical communication has been around for ages even since the early days of mankind and the first version of the internet was actually an optical uh communication network installed in france in the 18th century which was called the semaphores and it was an optical network it could reach you could send messages from uh from the south of france to paris in a matter of hours which was much more efficient than any other communication method now okay optical communication got displaced by electrical communication only actually for a fairly short time the introduction of telegraph introduction of the wireless was uh the transition from optical to electrical communication but that quickly changed from the 1960s on with the introduction of a laser and since the introduction of the laser and the introduction of the optical fiber optical communication has again dominated the amount of data traffic now just to understand what is needed to make an optical link you need a light source which can for instance be a laser because that gives you the highest quality of light then you need to encode a signal onto that onto that light source and that can be done by either switching the laser itself on and off or by using an extra modulator which for instance modulates the amplitude or the phase or the polarization of the light and then you have to transmit that signal over a medium and at the end your output signal comes out at the detector on the right side now optical links can scale quite well they can scale to hundreds of gigabits per second on just a simple schematic like this now there's a problem with lights if you want to use it for communication is that light has the tendency to travel in a straight line so if you want to route light to whatever place you want you need some mechanism you need a kind of guides that helps you guide light to the point where you want it so optical wave guides are essential in photonics now if you have your diffraction of light which you can describe as light rays you need some way to confine them and the way to do that in optics is with a dielectric waveguide in microwave signals you can use metallic waveguides but in optics metals have too much loss so you need some form of dielectric confinement with materials that do not absorb the line and the trick is actually very simple you take a material with a high refractive index and you surround it with a material of lower refractive index and then the light the beams of light or the rays of light are confined in the high index material now this is not a human invention nature nature was there first nature always is there first so you find the first or the some natural form of optical fibers you find that in certain rocks and minerals but the breakthrough was with optical fibers was to be able to make glasses that really had extremely low optical losses and that enabled the use of wave guides throughout the world and essentially powering the internet now why are optical fibers so useful for optical communication well they first of all they have been engineered to have extremely low loss but they always also have almost unlimited bandwidth if you if you look at electrical signals we're talking about signals that are on the order of 10 10 gigahertz radio signals and maybe maybe we can push that up to 60 gigahertz radio signals but if you look to the optical domain optical weights are also electromagnetic waves they're similar to radio waves but now you have frequencies of the order of 200 terahertz and you have a useful bandwidth of 40 terahertz that is an enormous amount compared to what you have for a simple radio communication now it's not that easy to use that 40 terahertz that you see here in one go it's a that's quite a challenge so what people usually do is chop it up into wavelength channels and frequency and wavelength are a kind of interchangeable quantity so you can chop up this 40 terahertz into a lot of different channels with each a bandwidth that can be modulated with an electrical signal and this type of wavelength division multiplexing allows you to extend your optical links not just with one channel but with multiple channels each carrying its own wavelength on the fiber so and for that you need what we call a wavelength multiplexer and demultiplexer now this function of wave like multiplexing and demultiplexing it's essentially separating a wave uh separating uh a broadband optical signal into many different wavelength bins so you can see that multiplexing or demultiplexing system as a similar function as for instance a spectrometer which you could use for sensing now this allows you to scale your signal bandwidth to petabits per fiber or better bit per second per fiber and that's this kind of technique this wavelength division multiplexing is what today powers the internet now optical communication is actually really good for this type of long-range communication because these links allow you to very efficiently connect boxes together and boxes started out being continents like in the picture i just showed but they are slowly decreasing in size so we then find boxes that are the size of cities and nowadays we're looking into data centers where people are connecting racks and shelves and even within a rack uh different blades together with optical fibers so we're now down to the limit where boxes are in size on the order of 10 centimeter to one meter now if we have this this optical fibers and we have these photonic systems if we want to build larger systems with it we usually start off with building things like this on an optical table but this is not necessarily very scalable even if you use optical fibers and you see in this case all the yellow lines that you see in this picture are optical fibers but still this is quite a convoluted system and this brings me to the core of my talk what if we integrate this all this functionality on a chip what can we manipulate the light on chips bringing information or bringing a systems down to a chip has a lot of benefits we've seen that in electronics you can make much more complex circuits your performance goes up your power consumption goes down in the end your cost goes down and you can make these chips in very large volumes electronics has enabled an entire revolution of new systems just because you can embed an enormous complexity inside a very small cheap package can we do the same thing for photonics well if we want to do that we need to be able to implement all the essential functions on the surface of a chip so we're talking about the transport of light filtering of wavelengths a light source like a laser signal modulation and signal detection and if you can do all that if you can bring these together you get circuits like we see here for instance like a very simple uh simple photonic integrated circuits where you combine a light source a modulator you filter the light you finally send it to a photo detector and you send it back and forth through optical fibers now the key to making this work is that you treat this as a circuit so the difference between treating things as a bulk element or as a system where you model the entire thing as one geometric entity it towards a circuit where you treat things as building blocks that you connect together is very it's a very big difference so if you want to scale up in complexity in photonics you also need to treat your optical signals your optical functions as a circuit and then connect those functional elements those building blocks together with waveguides so essentially the basic element of such a photonic integrated circuits is the waveguide so you need as you need on your chip you need some way to define a strip of higher refractive index surrounded by lower refractive index that can confine the light and you can guide it where you want it on the chip just like you would put an optical fiber on a planar surface but now you need to be able to manufacture a lot of these waveguides together on the same chip now inside the waveguide the light propagates essentially along the path of the wavekite and we can abstract that light as a single guided mode basically it's a blob of light which is centered around the core of the waveguide but we also see that some of that light is actually extending slightly as outside of that core still the entire blob of light the mode is guided or is confined or is bound to the waveguide and that mode propagates at a fixed velocity so that an entire packet of light moves at a fixed velocity through the waveguide we can identify two types of velocities we can identify a phase velocity which is kind of the wave fronts or the phase fronts that propagate but if you send a modulated package of light like uh like a bit on and off or like some some modulated signal that will usually propagate at a slightly different velocity called the group velocity so that's something to keep in mind if you design photonic chips is that there is a difference between what we call the phase velocity and what we call the group velocity and depending on the functionality either one or the other is important now if you want to build a circuit you need more than just a waveguide you need components like for instance a splitter where you can split light in two parts or more commonly used than splitter is a two by two coupler where you also split light in two parts but you basically instead of just splitting one input port you're mixing two input ports into two output ports and this get this type of two by two coupler can be implemented in such a broad rectangular waveguide which we call a multi-mode interferometer or it can be implemented by coupling two waveguides close together side by side which we call a directional coupler now with these building blocks we can now start making more complicated circuits for instance we can start thinking about making wavelength filters that separate out wavelengths now wavelength filters on the chip there's different ways to implement them and there's different types of wavelength filtering functions that you might want depending on your type of application for instance you could talk about what we call a channel drop filter where you have a lot of wave got wavelengths coming in and you filter out one particular small band of wavelengths but you can also build something like an interleaver which filters out even an odd wavelength bands into the output channels and then the last an often quite useful a function is called a multiplexer which in one operation separates all the wavelength channels in its own output waveguides now these type of wavelength filter functions can be used for separating channels in optical communication can be used for sensors can be used for spectrometry but all the all these filters are essentially built on the same operational principle you basically split your light up in one or more parts and then you delay that light along some of those paths and then you bring everything together and because light is a wave you will now depending on the path length difference you will have some constructive and destructive interference at the output which means that you will get some wavelengths going into one output and some wavelengths going into the other output and the most simple version of such a wavelength filter is a maxillary interferometer which we see here which splits light in two parts and that with a certain delay delta l and at the output we see that the wavelength transmission in the outputs is a sinusoidal which with a period that's inversely proportional to that path length difference if you want to make it more complicated you can just add more of these pot links a second type of wavelength filter which also is based on the fact that you have certain delay is called a resonator and this is an example of what we call a ring resonator where you have your light coming in you have a coupling section where a part of the light is coupled to your ring and then it starts circulating now because light is a wave again it will only work for the circulation will only work if the wavelength of the light fits inside the ring an integer number of times so that the waves nicely add up now if that happens for those wavelengths your light is coupled out to the output while for the other wavelengths it basically cancels out and your light just passes through so these type of ring resonators and maxender filters are essentially the basic building block for all types of spectrometry or wavelength division multiplexing applications they can all be described as some combination of those devices another important element that you need in a photonic integrated circuit is some way to interface with the fiber and as we will see further that's not always easy because the waveguides on a chip are not necessarily they don't necessarily have the same size as an optical fiber so you need to squeeze the light from the fiber into the waveguide and you need to be able to convert it back from the waveguide into the fiber so you need some form of fiber coupling device now all the devices that i showed you now that make makeup photonic integrated circuits are what we call passive devices these are devices that do not take any an electrical signal to change their behavior essentially just manipulating light on its own if you want to do more complicated things you need active devices devices that take that that use some electrical signal or make some interface between the electrical domain and the optical domain for instance the most straightforward that we device that we have there is what we call an electrical modulator or electro-optic modulator it takes an electrical signal it takes an unmodulated continuous wave beam of light and at the output you get some modulation you get a signal imprinted on that beam of light and that the technique to do that electrical modulation can be based on different physical mechanisms it can be based on temperature it can be based on the presence of electrical carriers and like electrons and holes it can be based on intrinsic electro-optic effects of the material and depending on the speed at which you operate these this can be this effect can be used for tuning or can be used for switching or for fast signal modulation also it's important to remember that these modulators there there are a lot of different types of modulators so some modulators modulate the amplitude or the intensity of the light so modulate the phase for instance now on the opposite side of the modulator we need some way to translate our optical signal back into an electrical signal and for that we typically use something like a photo detector so where you get modulated light in and the in the incident photons are converted into an electrical current or a voltage so in the photodiode the absorbed photon creates an electron hole pair which can be collected by a readout circuit and amplified now a last function that is very important in an optical circuit is a light source in some way you need to be able to get some light into your circuit or to have some light to process and so the typical elements that you use there are either a light source a laser or an led or some amplifier which combined with your circuit can be made can be turned into a laser now to make these light sources you need materials that are really good at emitting light so there's a certain selection of what you can do with your photonic integrated circuits depending on the materials that you use and that brings me to one of the key points in the landscape of photonic integrated circuits unlike in electronics where 99.9 percent of the electronics is all made in silicon in photonic integrated circuits we see a really wide variety of materials for instance if we want to make a laser or a light source we almost always end up with 3-5 semiconductors because these have a direct band gap that is really good at emitting light while if we want to guide light we want preferably a light with a material with low loss and so we end up with glasses or polymers three five semiconductors work but silicon works but they're not as good and low laws as for instance glass and polymers if you want to do fast signal modulation you need a material that has a fast electro-optic effect and then you end up with exotic materials like lithium niobate or some exotic polymers but you can also look at three five semiconductors for instance but silicon is like not a very good material for that and then for detection you have the same problem you also need a material that can absorb light and efficiently convert it to photons or two electrons sorry so that means that if we look at the landscape of photonic integrated circuits we can we have a table like this where you have you can set off all your different material systems and look at all the pros and cons and you will find that they all have their strengths and weaknesses and that's why if we look at the photonic integrated circuit market today we see that it's still quite largely fragmented we have a large section actually the majority of the photonic integrated circuits today are still indium phosphites so three five semiconductor uh you have but you're also depending on the wavelength you might prefer gallium arsenide now you also see that there's silicon there's lithium niabate there's silica on silicon so there's there's a lot of there's a lot of different material systems that can be used now one that i want to focus a bit more on in particular is what we call silicon photonics because if you if you look at what's happening online and people talking about photonic integrated circuits silicon photonics gets really the largest part part of the attention so what is so special about silicon photonics it's really according to many analysts it's the one technology that's really pushing the growth of photonic integrated circuits nowaday so why is that there's no good reason actually if you look at it on first sight silicon is really a bad material for photonic integrated circuits it has no indirect band gap so making a laser source out of it is really not that easy if you want to use it as a waveguide okay the material itself is quite low loss but fabricating it in such a way that you don't get a lot of absorption or that you don't get a lot of losses is not that easy there's no efficient modulation mechanism in silicon and also if you if you're using it for telecom wavelength silicon is transparent so it has poor absorption for telecom wavelengths so why would you want to use it still you see that this if you look around the world there's a ton of companies that are really betting heavily on silicon photonics and nowadays again with the driver of telecom and datacom we see all these companies pushing silicon photonic solutions for really high-speed data center and metro transmitters so what makes silicon photonics so unique that people are so enthusiastic about this well if you can look at the definition of silicon photonics as this one silicon photonics allows you to do really high density integration of photonic integrated circuits and now comes the catch using cmos process technology in a cmos fab and that's the unique property of silicon photonics here is that it you can you can use these materials you can process these materials in the same manufacturing infrastructure as cmos electronics so all the investments that have been made to build these multi-billion dollar fabs to make electronics could be leveraged to make photonic integrated circuits and as a result by using by using all this state-of-the-art technology you can integrate you can integrate really a lot of complex functionality into small chips and potentially fabricate them at high volumes and low cost now silicon photonics comes with a lot of flavors i mean if you just look at this definition there's a lot of materials that really meet these requirements for instance we have the standard what we call the standard submicron silicon photonics which operates between a wavelength of like 1.2 and 3 or 2.5 microns which is i think the most popular uh brand of silicon photonics because it fits squarely in the telecom wavelength range which people like to use for optical communication but if you want to move to shorter wavelengths you can for instance start using materials like silicon nitride which is transparent for much shorter wavelengths the index contrast is not as high as in silicon but it's quite good if you want to go to longer wavelengths you can start using germanium which is since 15 years an accepted material inside a cmos fan so silicon photonics allows you to scale it has this large scale manufacturing capability but at the same time the use of materials with a high refractive index allows you to scale your waveguides down to sub micrometer scale to understand why this is let's look at waveguides that are like kind of an optical fiber you see when you make a waveguide out of glass like an optical fiber which is surrounded by another type of glass with a slightly smaller refractive index you get fairly large waveguide cores on the order of 10 micrometer in order to guide the light efficiently now if you go to materials with a higher refractive index contrast like for instance 3 5 semiconductors where you can have an index contrast of 10 you can scale your waveguides down to a scale of a few micro if you go to silicon you can get an index contrast of more than 200 percent your waveguide now can be scaled down and shrunk down to an area a cross-section area of like half a micron by half a micron or even less so you end up with wave guides like this with a high effective index core and a low effective index cladding where your light is confined in a mode that is only 500 by 200 nanometer in size and this is a really strong enabler not only does it make your circuit smaller it allows you on the same footprint to make more complex circuitry just as an example these are three circuits of an arrayed waveguide grating it's a type of wavelength demultiplexer but in three different technologies and as we go from very low contrast silica and silicon here all the way to silicon on insulator over there we see that we get like a factor of 10 000 reduction in footprint which means that on the same chip area you can put ten thousand times the functionality and that's what enabled electronics to become so successful by shrinking down the transistors you could get more complex functionality the complexity did not come from the fact that the transistor was better the complexity comes from the fact and the functionality from the fact that you put so many transistors together on a chip and so what we see now in silicon photonics is something emerging like a moore's law for silicon photonics where we see on the order of thousands and going to ten thousands of optical components on a chip scaling up the complexity of the circuits and so we already we've already seen uh published like really complex circuits uh where we have like these uh these thousands of optical elements together on a chip and a lot of these elements are also electrically controlled so you can have kind of switching circuits now let's dive a little bit as an introduction and how these circuits are fabricated so what is needed to make a photonic circuits and let's focus on uh first on how to make the waveguides themselves so if you look at silicon photonics it's usually based on this on a layer stack that looks a bit like this you have a silicon substrate which is the wafer and then you have two or three microns of uh insulator layer it's not an electrical insulation layer or it not doesn't act as an electrical insulation layer it acts as a low index layer that separates the silicon substrate from the guiding layer in silicon at the top which is on the order of 220 nanometers now if you want to create patterns in these you essentially start with coating a photosensitive resist you imprint a pattern in the photosensitive resist this can be done with an electron beam or this can be done with an optical stepper then you develop that resist and then you use an iron etching a reactive iron etching to etch away the silicon in the exposed areas and in this case for instance you do a pattern of a grating which is edged 70 nanometer deep and then you can repeat this kind of steps again you can deposit resist again now you pattern some other patterns in this case it would be trenches that define the waveguides and then you edge these trenches away and what you see now is become a circuit that consists of two grating couplers to interface to a fiber and a ring resonator connected with wave guides now these passive functions essentially just require cross-sections that look a bit like this so with multiple edge levels you can define optical wave guides you can define great things etc if you now want to introduce new functionality like active functions like modulators you have a bit of a problem in the end you just have silicon there it has no electro-optic effects it has no photo detection for telecom wavelengths and it has no optical gain so how do you add these active functions to this cross section well for the modulation people in silicon photonics came up with a clever trick that bypassed the problem of silicon as a not really good electronic material they basically use the presence of carriers in silicon and the presence of electrons and holes affects the optical behavior so what they did is using all the tricks they know that they were already developed for electrical engineering is to make pn junctions like diodes inside your waveguide and by applying a voltage over that diode you can pull out the carriers or you can inject carriers there which modulates the refractive index note that that also modulates the loss because these carriers can be quite can introduce quite some optical losses so that allows you to do electrical signal modulate modulation and you can do that in different geometries you can do that in like in the in the left one you can do that by using a traveling wave electrode and a maxender interferometer so you embed this modulator inside an interferometer or you modulate it in a ring resonator which we see on the right where you put your diode inside a resonance structure and then you change the wavelength the resonant frequency of the wavelength and we see that that works uh pretty nicely now if you want to do modulation you can use these carriers if you want to do detection you also have to come up with some new trick and there the the trick is to do epic actually grown germanium into your silicon photonics so you basically grow a germanium layer on your silicon photonics the germanium has a much smaller band gap in energy which means it absorbs these telecom wavelengths very nicely but of course you need to go through a process where you have a very high quality epitaxial growth of germanium which is not that trivial so it took some time to develop this kind of processes as a result you see something like this cross sections like this where you see that you have a germanium which is properly contacted with the metals now that's photo detection the biggest challenge in silicon photonics is its lack of a light source and there's no really good trick to pull in a cmos fab to make silicon emit light especially at telecom wavelengths so the end in the end the approach is now to try and bring 3-5 materials to your silicon to make a laser and the first approach that was that's still quite popular and that that's also part of commercial products for instance at intel is to basically just glue or bond three five layers on top of your silicon photonics and then process a laser in it that works quite well but it's a quite convoluted process the holy grail where people are moving towards at the moment is to epitopically grow the 3-5 semiconductor in your silicon photonics just like they epit actually grow the germanium now grow 3-5 semiconductors that's far from trivial and there's already a massive amount of research invested in this but we're going towards in the right direction with already first demonstrations of lasers in the silicon photonic platform epitaxally chrome now imac silicon photonics platform embodies most of these functionalities at this point all of those except for the laser so if you look at how that platform looks in cross-section and this is a schematic cross-section of imix platform but it's also quite representative of what you see with other technology platforms uh in silicon photonics so you see that you have your optical waveguide you you have high efficiency grating couplers you have high speed modulators and you have detectors and then everything is electrically connected with mod with two or more layers of metal in the same way as you interconnect optical elements electrical elements in an electronics chip now on top of that you also have integrated heaters these heaters do not really are not really used for fast modulation but they are used to tune your optical behavior of your waveguides so you can for instance implement a phase shifter by putting a heater on top of it and because all these materials are temperature sensitive uh the optical index the effective index of the waveguide will change you can also embed a heater for instance next to the waveguide now heaters have the advantage of the fact that they're really easy to make i mean you just need an electrical resistor somewhere close to your waveguide the big disadvantage of heaters is that they are hot they burn power continuously and we're talking about depending on the type of material we're talking about 20 to 200 milliwatts of power for just driving your heater and if you need a lot of those that's not going to work very well so people have been looking at alternative mechanisms to implement these phase shifters and already we have been showing carriers they work quite well but it's a much weaker effect and it also introduces loss but for instance other material introducing other materials like liquid crystals or even introducing moving elements like mems can bring a solution now all of these type of integrations require again some careful thinking because you have this cross-section of this platform you now need to introduce liquid crystals which is not that trivial so you need to etch a hole in your device uh which where you can then deposit your liquid crystal now once you have done that if you can do that that works that works a treat i mean you can you can really get very high efficiency phase shifters and actuators in your photonics this is an example of where we embedded a drop of liquid crystal inside our silicon photonic circuit and everything else in the circuit still works you can do a similar approach with mems for instance trying to make movable elements in your silicon photonics not that trivial because you now have to undercut your waveguides but again if you are careful in developing your processes you can get really nice results where you have these freestanding wave guides uh which you can then modulate and what you see here at the at the bottom is a phase shifter that modulates the face of the light and this is part of what of a european project called morphic which we do together with epfl and kth amongst others to design these high performance mems devices inside an existing silicon photonics technology and again that works quite nicely you get very high efficiency phase shifters now if you have all these elements together and bring them together in a circuit like this for instance this really becomes more complex so if you want to design this you don't have to design just the geometry of the circuit you have to abstract the circuit down to building blocks which you can connect together so very much like you design electronic circuits and that means that on the design side you also need certain functionality you need a way or a mechanism that abstracts all the geometric and process knowledge that is being used by the fab and hand that over to the designer and in electronics this is called a pdk or process design kit or physical design kit for photonics these pdks are slowly emerging or slowly coming to maturity pdks for photonics have been around now for like 10 15 years but they were fairly rudimentary nowadays we see more and more complexity being embedded into these pdks and that allows you to more and more adopt the same design methodology that is used for photonic for electronic integrated circuits so building on the experience of electronic design automation tool to get first time right designs for photonics we're not yet there but we're moving slowly in that direction because in the end the whole idea of such a design flow is to translate the idea into a working chip so you start by making an abstract description of your circuit and then gradually you implement it as an actual geometry that can be fabricated now just an example of how such a circuit design works well let's say that we build an optical transceiver well we need to design it fabricate package and test it before we know it actually gonna work so you take these building blocks like modulators grating couplers etc and you integrate them together uh like for instance in four channels of 25 gigahertz with a modulator and then four receivers that coupled to a photodetector but now if you want to change or upgrade your link you can use these same building blocks to design another type of circuit like for instance a coherent transceiver or you can with which has an iq modulator and a 90 degree hybrid at the receiver side but if you want to use it for instance in yet another type of modulation scheme for instance wavelength division multiplexing you can again take the same type of building blocks as just reorganizing your circuit so now you have four modulators that each carry its own their own wavelength you multiplex them together in a wavelength multiplexer and then at the return signal is again demultiplexed and sent to four different photodiodes now the key thing is that such an iteration cycle every time you want to build a new circuit it takes you like nine months to a year to get it done and it also costs a lot of money so your feedback cycle in your design is nine months if you have a good predictive design process you should be able to do that in hours days or weeks at this point photonic integrated circuit design is not yet at the point where you can do it so it's a kind of costly proposition you need something that allows you to predict in advance what is going to happen after you have fabricated your circuit the difficulty in this kind of predictive capability is variability these silicon wave guides that i showed before they are only like a few hundred nanometers across but they're so sensitive to even nanometer scale variations the high index contrast means that if you remove a cubic nanometer of material you will already be able to observe it that means that you need really really good process technology and even then you will be suffering just as an example if you change your width of your waveguide with one nanometer your filter response will shift with one nanometer which is already a few communication channels in a telecom network and then if you look at actual variations on a wafer this is a wafer of a couple years ago you'll see that we can within a way for c 16 nanometer of line with variations so that's that's an enormous number of uh that's an enormous shift so how do we cope with such variability well there's three ways to do that first of all you can keep on improving your fabrication technology and design your building blocks so that they as they're as tolerant as possible the second technique is coming up with a circuit design methodology that allows you to predict how this variability will affect your circuit and then allows you to optimize your circuit for lower variability the last technique to do that to address variability is to go towards more tunability and even programmable circuits where you can where you can reconfigure or retune your circuit for a certain functionality after the fact so if you want to cope with variability first of all improve fabrication this is something that's continuously ongoing so you can go to better fabrication tools which typically means moving from 200 millimeter to 300 millimeter wafer fabrication which has better lithography better etching techniques you get higher quality substrates with better uniformity of your layer thicknesses and you really introduce very stringent statistical process control now the second technique is going towards more variability aware design and that's not that easy because you need very accurate models of your devices but you also need to know how they are being affected by the by different effects that happen during fabrications like how sensitive are my components to line with variations and thickness variations and you need to know how these quantities are usually distributed over the wafer so what's the typical distribution pattern of thickness variation or line with variation you can combine all that together using monte carlo techniques or stochastic techniques to get some form of yield prediction but this is still in a very rudimentary stage in photonics so the last part or the last technique to cope with variability is tuning so you can actually start compensating on improving and optimizing all the elements in the photonic circuits dynamically so instead of building a circuit like this where essentially all the elements are fixed you can now for instance replace the filter which is the most sensitive part with a tunable filter where all the elements are tunable and that opens up an old avenue to making circuits that can actually be reconfigured or slightly modified to do multiple functions and what we're seeing today now is circuits like this where you have this is called a multipath interferometer where you have circuits where which can be reconfigured to do different types of operations or and this is one step further or can be really like operate as a multifunctional processor that can process not just electrical but not just optical data but also electrical data and such a circuit just as an example you can by having all these tunable elements you can route your light wherever you like and this is the kind of programmable photonic circuits that that most of my research is now focusing on you can do multiple routes of light you can even route light through both the same coupler and cross and introduce crossings and then then it becomes interesting because these couplers are not really switches you could everything is analog on such a circuit so you can do partial splitting and you can combine it again to form an interferometer which can then be used to make a programmable wavelength filter you can even introduce resonances in such a circuit that allows you to make really very sharp notch filters and so the the first demonstration of such a circuit was done by the university polytechnic of valencia by the group of jose carmani and they are they actually introduced this concept of a general purpose photonic processor which could not just have optical signals going in and out but also process high speed and high speed electrical signals uh by modulating a high speed electrical radio frequency signal or microwave signal onto an optical carrier and then process it in this optical mesh of waveguides in the middle now inside of this mesh is an electrically tunable waveguide mesh so and this is key it's electrically tunable so it becomes essentially programmable the challenge here is that you need to interface all these electrical couplers and and phase shifters with a driver and a controller and then you need to read out what their status is which means you need very tight integration of photonics and electronics and this is one of the key challenges for silicon photonics today is really good integration of photonics and electronics and you can do that by trying to integrate the silicon photonics and electronics on the same chip which we see in the top right corner but a more common approach is using flip chipping or 3d stacking or even just wire bonding here now you'll hear more about these packaging techniques i guess in the next presentation so if you want to build a photonic circuit or a programmable photonic circuit you don't just need the photonic chip you need to bring everything together fiber interfaces microwave interfaces driver electronics everything needs to be properly packaged and then you need on top of that you need software layers to make everything work so the complete technology stack for photonic integrated circuits and programmable photonic integrated circuits is not just a photonic chip the photonic chip is just the bottom layer you need everything else on top of that to make everything work now if you have a photonic chip or a general programmable processor you can use for these in many different applications so just as some examples where you could use photonic chips or photonic processors optical communications is an obvious one that's been the mainstay driver for photonic circuits since decades but you can now see them being deployed in a much wider scale in buildings for fiber to the home but also for converting legacy rf signals like adsl signals to optical signals and one of the key drivers for this type of architecture is also the need for 5g antennas to be very widely distributed so you need lots of these antennas and these are going to be optically connected to the central office apart from communication you can also see a lot of sensor systems which will make use of these optical chips like for instance fiber sensors are already being used in constructions for buildings and bridges and airplanes but you need efficient readout circuits for these fiber sensors so if you want to make a really widespread use of optical fiber systems sensors you need these photonic integrated circuits also in cars you can use these photonic integrated circuits to do multiple sensor readouts and even go as far as using them as an engine for a lidar for self-driving cars or a microwave engine for lightweight drone radars and these and sensors can be extended to the biomedical community as well so you have biosensors optical biosensors that can do specific molecular analysis or i think this is useful to this audience you have also optical coherence tomography which can be very uh useful in dentistry but also in ophthalmology and then finally some other applications you can find for instance security quantum communication for instance or optical hashing or like doing bitcoin like of calculations uh blockchain calculations in optics could be extremely useful so that brings me to the end of my presentation [Music] so overall photonic integrated circuits have a lot of benefits just by bringing everything to the chip similar benefits as you find in electronics so better performance lower size weight and power one of the key technologies driving this is silicon photonics in all its flavors which also includes silicon nitrite waveguides germanium waveguides etc and that's actually driving the fact to more complex circuitry which is slowly uh creating the kind of scaling that we see in electronics where the functionality is no longer determined by the individual building blocks but by the complexity of the circuit and programmable photonics is one exponent of that evolution so with that i thank you and feel free to ask questions

Channel: Photonics Research Group - UGent-imec

Views: 122,023

Keywords: photonic integrated circuits, silicon photonics, programmable photonics

Id: CBhdLTTbYoM

Length: 53min 53sec (3233 seconds)

Published: Mon Jan 11 2021