[00:00:05]
>> It's my great privilege to use my work like this associate professor with chemical weapons by a lack of engineering your back to use the trailer. On his research you'll be a little bit nicer but not everything he's done is that he's a national scale material to work with you know electronics a foot harness integrity version of them are based on silicon for reasons that I'm sure he will explain He's also the co-director of a group that I enjoy the Q.B. for research but it surfaces in interfaces coming out it's crazy and we get together very nicely and try to come up with many ideas on things and services and a lot of silicon supply for us to do their services and services one along with all.

[00:00:55]
My guest host of national nation a wonderful high caste about now somebody in the technology and manufacturing society is a recipient of the most prestigious National Foundation Award for early career scientists are really worried users of the drug accident I mean back you wore the small junior back peachy Excellence Award and years and years if you win you're right because the mining industry.

[00:01:25]
And before everyone rushed to the rescue I want to read you a paragraph from my one of my favorite books call the periodic table it is not one of my favorite books because of the title only because of because of the gorgeousness of the right it is by the town chemist primo Betty who with more famously a survivor of Auschwitz and wrote 2 of the most powerful books about that but Levy was a practicing tennis throughout his life and wrote this amazing collection of essays which I really put it all to read it's not technical in the nature of the way but it's it's about aspects of his wife and he just.

[00:02:05]
As the title you know there is no character say what it was so it would come isn't it in a couple places I want to be doing a paragraph from my favorite. Title You just heard recognition still looking but more importantly he captures the essence of mind. And it's an out of character in this real life characters and Codell not always just her.

[00:02:32]
Sandro climb the rocks talking about looking for. More by instinct than we trust into the strength of this solution I don't believe in the project right well in the silicon house you know he's going. To feel that you've wasted the day Vietnam in some way the bottom is reserved.

[00:02:57]
And then even his eyes became brighter and the slaves of the sedentary life a deposit of forms. Which is not by working hard the fact is consume the highs back there so. I give you my. Thank you. Good evening everyone hope with a microphone as on. Thank you to M.G. That is clearly the best introduction I've ever had so thank you thank you to Maureen Ruby for helping to coordinate and the invitation also to Matt Baker for the invitation and for organizing this really great series I'm here today to share a technological story you thought you knew and when to tell it from my perspective and I hope you find it interesting so we're going to celebrate so I can today we're going to celebrate how successful it's been as a technological material we're going to talk about its hidden history and what that history tells us about what might happen next.

[00:04:04]
I'm an associate professor in the school of chemical and biological engineering as Energy said and while I'm a chemical engineer. To use the periodic table and I figure we'd start this celebration of silicon by showing you where it is in the periodic table to people in my community we call this the group for column of the periodic table so look in turns out to be very related because it's in the same column as carbon and germanium the name silicon comes from the Latin Silex for Flintstone S I O 2 that's the naturally occurring form of silicon it's bonded to 2 oxygens because silicon loves oxygen and you'll see that in the 2nd.

[00:04:43]
Tell us was the 1st person to isolate more or less pure silicon that took until $823.00 because silicon loves oxygen so much it's hard to break those bonds the 1st commercial use of silicon was in began the late 800 to produce something called faro silicon which I'll talk about in a minute so we have a lot of silicon on this planet in the upper crust of the Earth this is the abundance of all the elements as a function of their atomic number and you can see here circled white is silicon it is only 2nd to oxygen because it's S I O. 2 and for every silicon atom there are 2 oxygen atoms if you don't believe me that we have a lot of silicon just go to the beach go dig up some soil outside which is not necessarily the purest form but it is largely silicate based So that's the form that's all around us 33 percent silicon effectively an infinite quantity we use it in construction and glass manufacture and among other things.

[00:05:44]
Now when we want to use silicon for silicon sake we have to work at it because those bombs are so strong with silicon oxygen we have to work hard to break it we for what's known as metallurgical grade silicon that is the 1st upgraded version of Silicon it is 98 to 99 percent pure we would call this $1.00 to $2.00 N. and median how many 9 are in the percent OK So 99 percent is to end pure you'll see that number later we make a decent amount of this because it goes into steel and it's used in the Lumina manufacturing and in those materials are all around us how do we make it well we take silica the naturally occurring form and we mix it with carbon and we do this at very high temperature to drive this chemical reaction where carbon will grab the oxygen from the silica and be emitted as C O 2 leaving behind something that's a much purer form of silicon we collect that and will do other things with it that effectively as metallurgical grade silicon is done in very high temperature to break that SCI Bomb that's what it looks like it's a bunch of stuff that we then melt down and into steel manufacturing as I said aluminum manufacturing to do this so it's a it's a minor component in steels but it's a significant one in that it helps improve their processing as well as their durability.

[00:07:15]
This wouldn't be a celebration of silicon without telling you what else silicon is used for this is going to be a quick side point that if you do a little bit more work and you take metallurgical grade silicon and you process it a little bit harder with similar techniques you can get up to 6 N pure silicon and we call that solar grade silicon this is actually a relatively new route to this purity driven by the solar industry so they need purities that are better than steel but not as good as electronics which will be the majority of my talk today and in terms of the amount we produce this number is rapidly increasing as we use silicon for this right this is another story it's equally as fascinating in my opinion about the use of silicon in solar technology we are revolutionizing our energy infrastructure and silicon is a key driver in that I'll point out right here to come back to it Silicon is a horrendous light absorber right so it's kind of ironic that we use silicon to produce the large majority of solar cells that are made but that's not more here about that what I'd like to talk about is the story of silicon electronics and that story is often told as a story of materials devices circuits and systems and let me talk about each one of those for a 2nd and tell you what I mean by that so we sometimes talk about the material silicon and this is silicon structure it's the diamond cubic structure it is identical to diamond that you might have in a ring.

[00:08:49]
Germanium except the atoms are silicon and not those other atom this arrangement of atoms dictates this electronic structure that's captured in what physicists call a band structure these curved lines on here more or less tell you everything you need to know about the electronic properties of silicon when it's arranged like that.

[00:09:08]
One thing I'll point out I've highlighted in green here is what we call the bandgap that's a region where electrons don't go they are forbidden to be in that gap and that's good because it means you can turn silicon off and that's an important capability of a semiconductor it conducts some of the time we can control when it conducts.

[00:09:28]
This band structure will tell you how good electrons move through Silicon when you allow them to move and that's going to be really important in electronics because we're going to create switches that control the flow of electrons This is the part of what we call mobility as a function of a couple different materials people have studied mobility is a measure of the speed of motion so the higher you are in this chart the faster electrons can move and the faster switches can switch and the better performing circuit you might yet if Silicon the best know it's not the silicon the worst certainly not it's pretty good and you'll talk more about that later that silicon the material silicon the device is most best embodied in my opinion by the transistor So this is a transistor this is a cross-section of a transistor a transistor for those who aren't familiar with it is literally a switch.

[00:10:24]
Current in a switch in your home goes from one terminal to the other and it's modulated when you mechanically move that switch we do the same thing in this device and I've highlighted the different components of the device for those who aren't used to looking at images like this so a current is moving down here from the source through the channel and into the drain that's like the entrance lead in the exit lead in the switch in your home and the switch that we physically move in our house is substituted by this electrical switch we call the gate and we apply a voltage to the gate we modulator how current flows down here.

[00:10:59]
The silicon is all down here this part is built on top of the silicon but the silicon is central in terms of the behavior because that's where the charge is being transported there are the 3 leaves OK that's the device that will come back to the devices when I take many devices and I interconnect them together to form a circuit this is what it would look like in cross-section So all the transistors are down here and all these bright lines you're seeing are are interconnects these are wires that connect the transistors to each other to form a circuit that does logic or.

[00:11:38]
Stores data or what have you OK. This is what's in your cell phone this is an elephant bionic chip produced by was produced by Samsung for Apple if you were to look top down at a chip and look at those interconnections it would look something like this this is kind of a late model Intel chip really remarkable level of complexity and modern chips today the ones in your cell phone have billions of transistors that are all interconnected so it's really quite remarkable that is at the chip level so then we take chips and we build systems out of those chips this is a system we have with all of us expect many of us have in our pockets the cell phone we have processors we have image images cameras we have gyroscopes of all sorts of things in here made out of silicon if you take many computers and you put them together you might get a data center like this which is what Google has right $50000.00 processors together create a data center that's the system level story of silicon and this has changed the world it's material the device is built from it the circuits built from those devices and the systems built from the circuits it's a remarkable story but turns out it's getting harder and harder to fabricate the circuits.

[00:13:02]
Let me show you a little bit about that so there's something called Moore's Law which I put in quotes because it's not a real law physics it's just something that we ascribe to a trend and that trend has been that over time we can put more and more transistors into chips that's what that blue line shows you and that's really great and from that perspective we're doing OK from a couple other perspectives we're not doing OK these 2 curves that seem to have plateaued a decade or a decade a half and a half ago.

[00:13:32]
Are describing the frequency at which we can switch transistors the faster the better the faster you can switch the faster you can do computation but that's plateaued and that's largely plateaued because we've stopped it from growing because these transistors emit too much heat and if we kept growing the chip would melt so we stopped letting that grow but we've been limited in how fast devices can run for those of you pay attention you know the chips are always gigahertz and have been gigahertz for the last decade or 2 in their speed whereas before it was slowly and steadily increasing but probably the most damning thing is this plot and I love that in this Economist article it was this little thing in the corner but to me it's the most important one is the economics of making these circuits we had a steady increase the number of transistors you could buy over time for many many many years until relatively recently it's been very hard in recent years to give me more transistors per the same amount of money Intel cannot make a new chip and say hey I put twice as many transistors on but it will cost you twice as much that's not going to fly economically so there's an issue there there are other issues this is a fab this is where chips are manufactured ultra clean because you have to prevent dust and things from falling on these chips when they're manufactured a modern fab is $10000000000.00 or more.

[00:14:57]
OK fine there are certain companies that can afford that but there aren't that many from an innovation perspective it's really problematic in my view and others views that we have 3 or 4 companies worldwide that can invest at this scale of capital and when they do so they are unlikely to want to be risky with their innovations and that's a real problem the cost to design these chips in recent years has started to skyrocket and it's really becoming problematic to continue making more transistors per unit area people have recognized this and they've recognized this for many years and so this is trying to show you a smattering of the types of different things people have tried to try and overcome what was this impending challenge of the end of Moore's Law Now of course people predicted this for many many years it really seems like we're kind of getting there at this point but people have tried different state variables instead of electric charge maybe spin of electrons instead of silicon there's all sorts of other things that have been studied there's always a material does your OK so it's whether it's graphene or some other material you've heard about maybe it's not that transistor maybe it's something that depends on the magnetic moment of material that I can switch that on and off that's another kind of switch How do I represent the data it goes on and on and on we've tried everything right and great science has come out of these studies but it feels like this.

[00:16:30]
Feels like we keep running into the wall we have advances and we try and bring them into manufacturing and we can't and I'm going to talk a bit today about what's going on here in fact it's so troubling to DARPA that very recently they've decided to invest another $1500000000.00 into trying to overcome these challenges and that's on top of the money the semiconductor industry already invests in R. and D..

[00:16:59]
So to me this is quite striking that this initiative has been started and it gets me thinking the story we usually tell us about materials devices circuits and systems but what if that way of telling the story doesn't tell the whole story what if it blinds us to future possibilities about ways to get around these roadblocks we're facing and we're missing them because of the way we tell the story what might be a better way to tell silicon story or an additional way to tell the story and what might that story tell us about how we can invent the future turns out clues are all around us you'll have clues in your pockets to go back to the phone there is a microprocessor but there's also a bunch of other chips in your phone that if you think about it carefully you might say why is that made out of silicon cameras gyroscopes memory this is a cross-section of the camera in relatively recent Samsung Galaxy S 8 phone those big chunky things on top are pixels OK the light comes in from the top and is absorbed by those pixels this is a bit of circuitry underneath there's nothing special about silicon in fact silicon to really horrendous light absorber so why are we using it for the camera this is a gyroscope from the i Phone 4 a gyroscope tells you about orientation in which you're holding the phone and it's doing that by measuring the forces when you move the phone something special about silicon in terms of the way it measures force so why do we use silicon.

[00:18:42]
And this one I love these images this is a modern cross-section of a flash chip made by Samsung. And what you're seeing here I love this is entirely analogous to a skyscraper they have built vertically $55.00 layers of memory bits so each of these kind of. Bright lines are seen as another bit of memory.

[00:19:04]
But there's nothing special about silicon in its ability to store data. There's materials that do it with less energy required that do it on smaller scale so you can pack more bits per unit volume so why do we use silicon these things to me just that we're missing a critical chapter in the story we tell about silicon and so this question I like to ask and continue this talk tonight is what if silicon story was told as a story of process innovation manufacturing innovation and I think that story is critical and is often missing what if there was this level in this previous chart that we haven't paid enough attention to in thinking about how we can invent the future of electronic let me talk about process as I dive into this part of the talk what is process process is an interconnected set of steps that transform an initial state to a final state that's a fancy Professor way of saying it's how you make something OK is the process what is process innovation it's changing the steps that you're doing and hopefully you're getting economic or some kind of performance benefit by changing the way you manufacture so now we need to stop and take a brief aside and tell you that process innovations are often hit and I would argue because they are hidden they're often underappreciated think about the Stone Age The Bronze Age and The Iron Age right we have these names for these periods of time because of the tools that were used during those periods of time.

[00:20:40]
But I would argue that we probably should think about naming those ages after the process that was used to manufacture those tools because without grinding without smelting we would have never had the Iron Age or the stone age it should be Iron Age or the Bronze Age the product embeds the process that process is absolutely critical to enabling the product there's another one that you don't think about ever but it's why it's hidden 100 years ago the average store in the U.S. might have looked something like this where a customer would come in and interact with a single shop clerk and that person would take your list of items and go pull them off the shelf and hand them to you that is remarkably inefficient compared to today's paradigm what happened was we traded the direct interaction with the shopkeeper and we made it harder on you we said go up and down aisles yourself and go find things you don't think anything of it today but you're being asked to do something that was more difficult to you personally than it was in the past but what this enables by asking you to go get the items off the shelf is dramatic efficiency gains right very few people compared to how it was 100 years ago we're required to service the same number of people right and so you don't notice this it's there it's embedded in the shopping experience that is a process innovation I'm going to talk about for process innovations that are embedded in silicon integrated circuits the player process monolithic integration what's called the Siemens process and across the crystal and I'll kind of give you a flavor for each of those I've picked these 4 out as I think they are the most important.

[00:22:27]
In the in the current process of silicon manufacturing there are a variety of supporting innovations that I'm not going to touch on hey if you're interested in learning more you can check out a website like computer history dot org It's a really phenomenal site in the early days.

[00:22:43]
People built things called Mesa transistors and they're called Mesa transistors because if you've ever been to Arizona and you've seen the mesas this is what they looked like where you had this layer of semiconductors rising off of the surface in a mesa these were not very good these bases had a lot of surface area that was exposed that degraded the performance of the transistor and the other thing that was really difficult to deal with from a manufacturing perspective was that you had a contact this 3 terminal device on the top the side and the back and that was really a pain in the rear to figure out how to put electrodes in these different orientations who seen this picture see all my students raising their hands.

[00:23:30]
This is a picture of the so-called traitorous 8 the founders of Fairchild Semiconductor. I think this is one of the most important pictures in the history of Silicon there are some famous people in this picture that you probably know Gordon Moore and Robert Noyce went on to found Intel after changing the world really is Fairchild.

[00:23:52]
I want to start though by talking about this gentleman here John Arnie who you don't hear about too much he invented a process called the planar process right before we had these mesas they were built on top he recognized that there was value of putting the transistor into the silicon so this is his patent filing and here was that piece of silicon and he realized through certain processes you could put.

[00:24:18]
A source let's say and then through a 2nd process you could put. A bass region and now you can see the basic structure of a transistor right 3 components here 123. The other thing that happened by part by in addition to putting these things into silicon was that he had a top surface that was flat it became much easier to go and connect these 3 points with this top flat surface rather than having to make all these weird maneuvers with the wiring so not only did he find a way to make the silicon perform better in the devices because now they're protected from the environment those mesas were above the silicon now we are in the silicon but he created a plainer top surface that made interconnection easier you can go to computer history dot org And you can see the lab notebooks that were written the patent filings of all of these folks really a fascinating read they recognize the value of doing this at the time and it's all written out there in their own handwriting this innovation led Fairchild to have significant performance advantages relative to competitors in the days of single transistors back when they would sell single transistor there are great great ads but I like this ad not for the fact that it's from 158 in a magazine called electronics like you never see the thin toil magazines and more whom I can't so but the fact that this was this was selling single transistors Bob Noyce who worked at Fairchild as I mentioned helped found it he realized their knees the value of Annie's innovation in connecting many devices so much as one but putting several devices so here's one device and here's another one now they're embedded multiple things in the Silicon.

[00:26:16]
OK And not only did he realize that you could do that it was fairly obvious he had to find methods to prevent them from talking to each other through the silicon and he had to find ways to isolate them and that was a key innovation that Bob Noyce brought forward in addition to leveraging a technique known as full of thought or a fee to start to put down the metal wiring to create the circuitry so what you're seeing here is the top view of a silicon wafer the transistors are embedded inside it and this is all the circuitry that is connecting the transistors that he put on top so a nice planer process made the device is good made the top surface flat allows you to put many in a surface and now interconnect them and this was the 1st monolithic integrated circuit we still today make circuits this way you can read his lab notebooks as well so the next part of this that is really important to recognize is that integrated circuit manufacturing is an entirely integrated process of course of this it's called an integrated circuit.

[00:27:27]
But this is important because of the way air NE and noise decided to start building things where they put things into the silicon there now became no way to remove any defective transistors there's no way to carve them back out when they don't work so you can't remove them so what does that mean that means that the processing of every single one better be perfect you can't make mistakes along the way and not only does the processing have to be perfect but that silicon layer where the charge is being transported also better be perfect Let me take away those labels are the charges moving through here if you had a defect in the silicon and impurity atom sitting there electron would see that in purity and scatter off of it and your device would be degraded if you had a defect in the crystal where the silicon atoms weren't aligned properly that would also be another scattering center where electrons would be prevented from reaching the other side of the device all of these things lead to degraded performance.

[00:28:32]
So what do we do we had to make this a look at more perfect and we started by reaching electronic grade silicon before we had 6 and that sounded pretty good now we're going all the way to 9 M. to remove all those excess impurities that would cause scattering of the electrons as they move through the channel I also point out since we're celebrating silicon that this is where we get silicones from this part of the process is where we take part of silicon and use it for silicones I'm not going to talk more about that but just to recognize that that is the non-negligible part or use of silicon today this is the so-called Siemens process and you're all are glazing over so I'm going to simplify it into a couple of pictures what we do in the same is process is fairly straightforward it's a lot like distilling water except we do it with silicon and it's a very extreme version of distilling water so here is that metallurgical grade silicon to N pure if we react it with Clark acid at a moderate temperature we can turn it into chloro silents we turn into a gas just like we do we distill water we put water molecules into the gas face it leaves behind the impurities.

[00:29:40]
So these red dots are meant to be impurities they don't go into the gas phase as readily as the silicon atoms do there is another purification step that happens here but bottom line is we then take that gas where we left the impurities behind and we re deposit it down to form the solvent that is 9 and pure electronic grade silicon it's kind of a crazy number when you think about it parts per 1000000000 or better that's all true pure silicon but that says there's no impurity atoms but what about those defects so now we need to align all the silicon and M's that have been purified in exactly the right locations and none out of place and this is where the 4th process innovation across the crystal growth comes in.

[00:30:23]
This technique was invented in 1000 916918 at a metals company they had no interest in silicon it was far too slow far too expensive to be used for metals at the time no one needed this kind of purity but in the forty's and fifty's people started to recognize the value of this growth process to create really perfect ordered materials basically what you do and it's been adapted for silicon things to call for time if you take that really pure electronic grade silicon you melt it down at high temperature you add a seed particle that is where silicon starts to deposit and you do this at a temperature really close to equilibrium very high temperature and so it's a very slow deposition process and if you do that really slowly the silicon atoms will go only where they're most favored to go and that if you keep pulling the crystal you can create these bulls of silicon I mean that are really remarkable this is the purest material man can make both in terms of its compositional purity and its crystal graphic perfection has also the electrons flow uniformly because we can't get rid of bad transistor that's what a process looks like today it's been highly engineered since it was 1st use and in fact we get a bump in purity up to 11 N. and that's because the impurities tend not to want to go into the silicon lattice when.

[00:31:52]
We take out these boules and we slice them into the wafers and these are the wafers on which we build the transistors and then the interconnection So these were the 4 process innovations the planar process to make things flat and protect the devices monolithic integration to make a circuit the same process to purify silicon and the growth process to make the silicon atoms all where they're supposed to go these 4 processes that underpin integrated circuit manufacturing for 60 years all my conductor R. and D. spending over 65000000000.

[00:32:29]
From industry in 2018 assumes assumes these 4 processes and these 4 process he's incentivized certain kinds of circuitry but we know we have these problems we know there is a strain in the system we know we keep running up against the wall and we try new materials new device physics etc and we know there are certain things that silicon can't touch wouldn't be great if when we pulled the tissue out and blew our nose we could have a sensor that would detect whether we were sick and would wirelessly communicate that data to our doctor so we can can't do that today wouldn't be great if in every pill we had silicon a very small amount of silicon that hopefully didn't harm our bodies that when you took the pill would wisely tell your doctor you took the medication you were supposed to take today what would be great if we had sensors on many many seeds if not all the seeds in this field to detect when they germinate to improve the efficiency of farming would be interesting to think about buildings and structures as their own data centers to kind of fill them with electronics rather than these data centers that are very localized today there's a lot of things silicon can't do for all the things that it's amazing at today's processing can't do many things that would be really great.

[00:34:02]
So here are 2 axes. This is kind of engineering want to one this is the manufacturing scale so if you want to do clean actually got to be way up here at large scale and this is performance if you want to do wireless communication you have to be a high performing device silicon is a good performing material makes the device a good circuits we know that we don't produce it on very large scales compared to the scale of buildings the scale of farming We've tried all sorts of other materials over the years and we seem to be stuck on this line we can't seem to find a combination of materials and performance that gets us to this space where these new technologies might become possible.

[00:34:44]
So why is silicon still king it's not good at some things the system is straining under the load it's King because it's processing is perfect the processing has to be perfect because of the way it's processed other materials fundamentally can't compete with the perfection that has been developed over 60 years and I would argue that any attempt at perfection is likely to require a similar investment over a similar timeframe so what's needed how do we get around this problem I don't want to wait 60 years for the replacement to silicon let's look at how other industries manufacture things like the food industry for example right ever wonder how the grocery store shelves look that good partially because there's employees who stack the vegetables like this I don't think my refrigerator like that.

[00:35:35]
But it's also not because nature is perfect in making food right nature's really not great at making all of the tomatoes the same size so we have methods to separate the tomatoes that are of incorrect size basically right and your definition of incorrect will vary right removed tomatoes that have.

[00:35:56]
Some bite chewed out of it by an animal or what have you you never see that in a store but we have methods to remove it take rivets particularly for high value applications like automotive or aerospace we have ways to sort bad rivets from good rivets my industry the chemicals industry this is how we operate we make things and then we separate stuff this is probably the most important process you never knew existed this is.

[00:36:26]
Oil refining basically. Also known as fluid catalytic cracking where we take crude oil and we break it apart into a whole bunch of pieces and they're all mixed together we don't care and then we separate them we take gases like propane we take gasoline which is a little bit longer chain hydrocarbon even longer still becomes the oils except for a we don't waste anything.

[00:36:50]
But we have methods to purify what we've made. All other industries assume imperfection all other industries have methods to separate imperfect components and those components I'll point out are modular the tomato is a modular piece of food the rivet is a modular item and a molecule is also modular also to separations consume 15 percent of worldwide energy so if you are ever wondering whether this was a big deal or not it's a huge deal right we separate everything in pretty much every industry except silicon this is what we've done with silicon it's a tremendous success over decades of work by many engineers and scientists and we have dropped the cost of silicon by orders of magnitude right that's the story of materials devices circuits and systems but if you compare this to other things that we process in different ways and now I'm plotting it not per transistor but per a mole of transistors 10 to the 23 of them doesn't look as good in fact when I compare it to molecules even advance things like a pharmaceutical lift a tour of the cholesterol lowering drug which is a multi-step synthesis and separation.

[00:38:10]
10 orders of magnitude cheaper to produce that pharmaceutical these are modularly manufactured with separations and removal of bad components here it's a fully integrated manufacturing we have to be perfect because we're embedding things in the silicon So our question the thing that my lab has been working on in collaboration with others for the last several years is asking the question can we introduce modularity into processing of high performance electronic devices and circuits and if we can what technologies might be possible with this more modular circuit processing paradigm and it's going to be some science and engineering needed to do this so we 1st have to modularize the transistor we have to take it out of the silicon so it can be a component that we can throw this way or that way we have to be able to move it around effectively.

[00:38:59]
If it's in a way for we can't move it around we're going to need to be able to separate transistors and ultimately to make circuits you're going have to connect those modular transistors so let's talk about those 3 processes for a minute or 2 I'll show you some data from the lab how do we modularized transistors another way of saying that or something that's related to this is process is to make transistors make themselves and we want to grow transistors just like we coax molecules to make themselves and chemical reactors this is the device again I'll focus here for a 2nd we've got to make source channel drain my group spends a lot of time growing crystals and we use a technique that's called the vapor liquid solid mechanism there's a vapor out here there's a liquid seed particle that collects material and injects it into the solid semiconductor and the great thing about this technique is that we can change what's in the gas phase so we can make silicon and then we can add in a dope and and we can make a region that's doped and then we can take that dope out and make a region that's undocked and you can see some work here from my lab exploring the ability to module 8 the composition or the structure along the length of nano wires using this capability but you can look at this and you can kind of see if you turn your head to the side source channel and drain right so we're we're doing that.

[00:40:23]
Great post-doc in the lab has been working on advancing this technique in our group. That's a wire she grew and it's in coded with a structure that looks like source channel drain but you should believe me because it's just an S.C.M. image of the exterior but if you put this in a selective etchant that is selective for certain regions and not others you can remove some of the silicon in the channel region and leave the silicon in the other regions to show that it's compositionally modulator just like a transistor would be she can do this to the cows come home so we can grow crystals from the what we call the bottom up with source channel drain source channel drain until you're blue in the face that's that part.

[00:41:05]
And that's a key part of a transistor but there's this part right and normally in the field of nanoscience what people often do at this point is resort to kind of the older not older but the techniques that we use traditionally in the industry to pattern that gate but to make that modular component and to make it cost effective you want to eliminate that process which is known as FOTA lithography So how do we do that how do we take what God has done so beautifully and go the next step to create the gate that we're not entirely there yet but this is what we're doing here is that same kind of modulating structure.

[00:41:39]
There it is to prove to you that the module and structure but we don't we don't want the modulation that's just really to reveal that we have the source channel drain structure Amar has been doing this work and we're really excited about this he's coated a special resist material around the wire and then he's used etchant that comes in and doesn't X. the wire but it removes this resist only from certain regions this is actually to our knowledge the 1st time this kind of thing has been possible with the kind of precision a more has the programmability the cool thing is there's still source channel drain in these structures and now that we have this mask that's protecting the source and the drain we're working on depositing the gate but we have a pattern with which to do so and we haven't had to use photo with a graphic we now need prosthesis separate transistors and it's not enough just to make them have a batch of them I know I'm going to assume they will be imperfect they won't all be the same in their properties and so what I really like to do is not throw some stuff out but fractionated just like we do in oil manufacturing where we have some transistors that turn on and defaulted some internal and that voltage and some that turn on at this voltage This is a really challenging problem to figure out how to purify electronic components based on the electronic components property but we're making strides with our collaborators that rockers.

[00:43:07]
We're they've developed this really beautiful technique for using electric fields to rotate materials and through their rotation they can back out what the conduct of it is of different wires which is really phenomenal and it shows us in fact that in the wires we grow there's a distribution of properties which is something that we can hopefully narrow That's always a goal is to narrow that distribution but we can also now that we know there's a distribution we can separate them.

[00:43:37]
And so we're working on techniques to do this and finally we're working on ways to take these modular components when we have them and interconnect them to make circuitry. By the way to do that you would think about something like a transistor ink or a set of printing black material to form the letters on a page there's actually transistors in that ink and you spit them out in a printer and then interconnect them we work with a great mechanical engineer here a bargain at the University of Michigan she is an expert in this technique called electro hydrodynamics printing or Egypt printing it is kind of like inkjet inkjet uses of a pressure differential to force ink out of a nozzle they use an electric field to pull ink inks or whatever material they're working on the positive and by pulling you can get much finer features and this is just an example of some of the features that are possible with this Egypt technique and it's really critical to have fine features because as we build better and better transistors with the process that goes on them are working on we're going to contact finer and finer things so this is a technique that should allow us to achieve that interconnection this is overall the modular process we're pursuing make these modular devices make many of them enough talk about that before wrapping up transistor purification and then print and interconnect.

[00:45:04]
None of these are easy I have don't have the answers for all of this yet it's going to take a few years but this is where we're going and we think by rethinking the process that there's exciting new things you can do so what things some of you may be sitting there saying I don't see how you're going to get 5000000000 transistors interconnected with your approach and you would be right.

[00:45:25]
So this is another one of these actually this is again it's manufacturing scale here and this is the level of integration right how many transistors can we interconnect and the way we do it today is down here relatively small scale highly integrated what we're trying to do is not going to be there at all it's going to be up there going to be a large manufacturing scale and much less integration but that's OK There's really exciting things you can do when you rethink what you should make with devices that you have at our that are really low cost and effectively disposable this is work done by our colleagues Gregory a bout in fab Starner really that into the work who am I kidding invented maybe in the audience did this really I think groundbreaking work to show that you could make a very simple autonomous wireless sensor with a single transistor we're all used to Bluetooth we're all used to the communications with our cell phones that requires hundreds and thousands of transits they developed a very ingenious way to use a single transistor the transistor they use though is one of these expensive single package transistors and would be great if you could just print out that transistor while you're making this little sensor they can detect by the way audio.

[00:46:40]
A kilometer away with a single transistor using this technique called backscatter So I think if we think differently about the systems and the circuits we could design there's lots to do now will it always be one probably not we think we can get better we can think about things that have tens of transistors like ring oscillators or amplifiers.

[00:46:59]
In the long run hundreds thousands who knows that there's a lot of things that Bob Noyce and Gordon Moore and John Arnie did not predict would happen when they started developing their profit the last thing a kind of leave you going to get this because the time. Now I'm not going to be sorry but this is an important piece of the story so even though individually it may be a relatively simple circuit if you put many many many many circuits together the system can be significant and I think that's something that's generally missed when we talk about simple circuitry if you think about the system which may be a data center or it may be a whole bunch of interconnected nodes that permeate your house right it's going to be the number of transistors on each node and then the total number of nodes that you can deploy today what we do in integrated manufacturing we have want to transistors transistors pursuit that and relative to producing.

[00:47:59]
Relative boosting chemicals a low number of circuits that we deploy what we're arguing for a more modular manufacturing approach is that yes we know we'll have fewer transistors per node but we think we can potentially produce these at very large scale and they can talk to each other if they can wirelessly communicate and you can think about billions of sensor nodes that then communicate to form a system and now this is looking way out but we don't think we should be living it ourselves to simple circuitry or at least simple systems in the long run but if you're going to put these things in buildings you could how do you produce enough and this is great work done by the group he's been developing a process we call the Geo process and that's a geode and what we want to do is grow transistors just like these crystals grow in this rock.

[00:48:50]
Our Geos are not that big they're not macroscopic they're more the size of the human hair this is a dispersion of the geode that merits it produces in the lab and what goes on in each one of these spheres is wire growth and ultimately transistor formation the bottom line is we take the little seed particle that drives growth and put it on the interior of the capsule kind of looks like this we allow the gases that do the growth of the transistor to go through that wall but we take that powder I just showed you we load it into a more traditional chemical reactor we process it as we would prosecute a process a chemical I don't view transistors any different than chemical we pull them out there now Geos because there are crystals but not any old Crystal there are silicon crystals and ideally silicon devices like P and junctions and perhaps more advanced things to give you a sense of the scale we think we can achieve simple back of the I Will calculation that in an average one meter cube reactor we can produce as many devices in one day as the entire semiconductor industry produces in one year or so in terms of can we hit high manufacturing scale with these processes we think we can so the last thing I'll leave you with is this story might have seen this movie The Social Network.

[00:50:10]
That's Justin Timberlake playing Sean Parker the founder of Napster who was an early investor in Facebook and he's talking in this scene to the founders of Facebook and they're all excited because they're going to make a 1000000 dollars and he says to them a $1000000.00 isn't cool you know what's cool a $1000000000.00 and it's not won't write.

[00:50:36]
Current state of the art integrated circuit 4300000000 transistors in your pocket but a 1000000000 devices is cool. You know what's cool a septillion device thank you for attention I'd be happy to answer that question. Yet right. This this story is a hard story to tell this is this hidden story all right the usual intros to manuscripts that people write are much shorter right it's hard to tell the story process innovation is hidden so particularly in this case it can be really challenging to be challenging when people view the world through the integration that's possible with silicon which is truly remarkable amount you know nothing I've said here take away from the success of the semiconductor industry.

[00:51:38]
But yes so we try and find ways to convince people who fun things in certain areas that this can be applied to that and we come up with ways to market the material in a way that would resonate with them I would never present this this way to many program managers I talk to right because it takes too long and they're biased these are often not in this direction right one of the reasons I accepted this was beyond just that it was a great opportunity to to share what we're doing with a bunch of people from the community was also to put together this broader story to tell people how we think about the world.

[00:52:22]
So I don't have any really great answers just to say that it's challenging to start trying to convince people everyone wants you to show them that you've done it already and none of this is easy. So it's not perfect but we keep trying we've tried different angles and as of late we're starting to see a bit more progress which And I think and the fact that Moore's Law is kind of ending and people are trying all these different things there's a ripe opportunity to try this you know.

[00:53:11]
Yeah we have a lot of launches. So for every one person I collaborate with I had lunch with 10 people. Right it's one of those things it's just like any personal relationship it's finding people who are doing synergistic science that you get along with and that you have similar values in terms of what's important in science and how you conduct science we're lucky at Georgia Tech Georgia Tech is a really phenomenal institution because we have people literally doing everything because we're so large so whereas someone in a small school would have to immediately go off campus to find expertise there's a lot of expertise right here when we don't have that expertise we go off campus the story of cure Barton is a really interesting one she I had her on the podcast that she was talking about had no idea that she would have a skill set that was so synergistic with what we were trying to do and in the after show we started talking and realize I realized really quickly that this person needed to be part of the team and so then invited her and convinced her we were good to collaborate with and we're doing cool things and she's part of the team so it's really kind of a kind of stochastic diffusion kind of process like most things.

[00:54:23]
But we're lucky at Georgia Tech because we have a large number of really great people doing lots of different things and. Yeah it's a great question so there's definitely no shortage we have saying. There was so we could we could build for the next 1000 years and we'll be fine sand wise the sand that's used in silicon manufacturing comes from a couple specific locations it may be just one and I'm not an expert on this but I've heard before that it comes from a very pure form of sand that you can get in like Minnesota or something like that.

[00:55:15]
Right I got yeah it's a great question and I can't remember the exact place but they do look for things that are more pure right they're not going to go down to Home Depot and get the sand for the sandbox it's a different kind of sand that's used.

[00:55:58]
So the question I think I heard you ask is What's the downside of all this OK So I think with every technology there is an upside and the downside and you have to manage the downside risk because if you just choose not to do any technology then we're kind of back in caves right so but yeah so there's a lot of good that can come from manufacturing efficiency having sensors to you know on the crops and all this kind of stuff but there's the down side right when I said you know would be great if that building was the data center half you were saying man I would not want to put my data in my friend's house and I know how he keeps his house.

[00:56:39]
So there are going to be concerns about this I think there are already plenty of concerns you could even argue that this approach may be used to mitigate some of the security risks we have today you know if you have disposable stickers that have circuits in them there may be ways to do certification and authentication in entirely new ways that are not software based there's these kinds of things that could emerge from a technology like this it's hard to predict where it goes but I'm one of these folks that says you just have to manage the downside risk over.

[00:58:04]
So I think the way we started isn't so bad and serves of picking a place to start. Yeah so I think you know technology that's my group studies these crystal growth process these and these surface processes to pattern those crystals that's that's my group's main expertise and I'm a chemical engineer so I love process in my heart also so I see I feel like I you know the the kid in The 6th Sense I see dead people I see process that's.

[00:58:39]
But I think students in my class right now appear laughing. So. I think nanotechnology and then the science has gotten to a point where it's time to take it out of the lab and try and use it particularly the very functional things like terms of stars we've demonstrated these things so the wires I showed you we we've made those for a decade or so.

[00:59:04]
The community has and we've done largely the old techniques in terms of building stuff but we know they're really good when you make them the old fashioned way or at least take a wire and then use the old fashioned process and so we have we have struggled to make it the next step and so part of this is recognizing though that we know enough about surface chemistry you know enough about nano materials growth we know enough from the chemicals industry to think about separations that we can start to put it together in fact if you look at some of the road maps for Nanotechnology a lot of them say we need a compelling application.

[00:59:42]
Now we've done all sorts of amazing science and we've understand materials way more than we did 102030 years ago when the Nano push started but finding that compelling application space so that we can drive directly to the innovation to get industry on board has been difficult to materialize particularly for the functional things we put graphene in lots of tennis rackets and mattresses and things like that but in terms of this kind of space we've struggled to do that and so we're we're trying to do that.

[01:00:12]
It's going to be one of these diffusion processes but once we can demonstrate a few of the key pieces and we've already made nice progress hopefully people. Will say. Thank you for you from the College of some Thank you thank you very much. So that. Our. Callers. There.

[01:01:06]
Are very good. And. The College of. Organizing. The event. Thank you. Everybody. So. That. There are. All.