Lessons from the Early Days of Semiconductors - Carver Mead - 4/24/2019

presented by Caltech welcome to the electrical engineering distinguished lecture series as you know today's lecture is a very special one it will be given by an extraordinary researcher who's fearless efforts have greatly impacted academia industry and even our everyday lives as someone who continues working in the fields that he pioneered it is my great pleasure and distinct honor to introduce professor Carver Mead Carver is known for electron tunneling semiconductor interface energies the first working method scaling a very large scale integrated circuit or VLSI technology structured VLSI design the first VLSI design course physics of computation neuromorphic VLSI systems and collective electrodynamics he pioneered the silicon foundry concept and the fabless semiconductor business model his honors and awards include National Medal of Technology BBVA frontiers of knowledge Award National Academy of Engineering Founders Award I Triple E John von Neumann medal Walter Wriston public policy award ACM Allen Newell award I Triple E centennial medal and lemelson-mit prize he is a member of the National Academy of Science and the National Academy of Engineering he is also a fellow of I Triple E Computer History Museum and National Academy of inventors among others he holds BS MS and PhD degrees in electrical engineering from Caltech as well as honorary doctorates from USC and the University of lund he has been a member of the Cal Tech faculty since 1958 and the Gordon and Betty Moore professor of Engineering and Applied Science emeritus since 1999 at this point I would like you all to turn off your cell phones and join me in welcoming Carver Mead to the podium thank you great to be here with old friends and younger friends and a great time in old times electricity was Sparks since antiquity and sparks were fun and you could make magic with them and do all kinds of things but there wasn't much quantitative you could do the qualitative stuff got worked out pretty well by Franklin 130 years before our story starts so our story is not about sparks our story is about controlled electrical current and that all started when Volta came up with a voltaic cell and here we are with what Volta patent said if you take a copper and a zinc chunk and separate them by a blotter soaked in salt water you get a voltage around a volt and if you want more than that you can stack them up and the minute he did that electricity went from sparks to the controlled flow of electrical current with huge Faraday wouldn't have done essentially any of his remarkable findings if he didn't have voltaic cells here's one over here it's just one cell it's got salt water in it the inner electrodes hooked to one wire and the outer electrodes hooked to the other wire of this coil and so there'd be a current flow which is shown by the arrows and he was interested in whether when he manipulated the little coil anything would happen in the outer one so he hooked it to a galvanometer well what's a galvanometer well people had observed shortly before that if you had a current and you put it near a compass needle it would turn the compass needle and of course that got turned into an instrument always happens any phenomena gets turned into an instrument and that's a good thing so that the galvanometer then became the thing that allowed us to do electricity quantitatively instead in a normal compass you have a little magnetic needle and it gets balanced on a pin point so it can rotate freely and it rotates until it points magnetic north but if you suspend it instead of on a needle by our silk thread then it will have a place a preferred place where it comes back and so if you put a current near it it will rotate one way or the other depending on which way the currents going and so you'll be able to measure how much the current is and that's the galvanometers that made all these experiments possible Ferriday used that galvanometers Abul looked up in a way that hoot stone had figured out he said okay if we have resistor set up this way by then of course there was people that made resistors since ohm figured out that currents got limited by resistors then people said well resistors have to have a resistance and some had more resistance some had less if you set them up in this situation then the ratio of R 1 to R 2 is the same as R 3 to rx then the voltage at D would be the same as the voltage at B and there'd be no current through the galvanometer so by using this null principle you didn't have to have an absolute calibration of the galvanometer and you could still do precise measurements as long as the people are selling of the resistors were were calibrating them right so that became the art form from then on and Faraday used that principle to measure the resistance of lots of things and he measured the resistance of copper wire and of iron wire and he found out that the resistance of iron wire was a lot more than the resistance of copper wire but both of them when you heated up the material the resistance got bigger so this was this was a generalization conductors metallic conductors the resistance got bigger when you heated him up well known lots of theory about it then he started measuring the resistance of some natural minerals and the particular one he he liked because it was easy to work with a silver sulfide and he found a remarkable thing the resistance of the same length to area ratio hundred times more resistance hundred times less conductive than normal metals were so it was kind of a conductor but it was really only kind of a conductor so it was a semi conductor and that was the first time that had been observed well Faraday's very smart guy so he kept on it and as he heated up the silver sulfide the resistance got less instead of more so exactly the opposite of metals that was a new deal and so he reported that and that became a class of materials that a lot of people studied one of the people that studied it later was Felix Braun in Germany and he was interested in the resistance of these minerals there a lot of minerals have behaved that way copper sulphide lead sulphide sulfides tended to be like that iron sulfide so he was measuring there's a sense more precisely and to measure a resistance you have to be sure that you have good contacts to the material so that the resistance you're measuring is the resistance of the stuff and not the contact to the stuff and in the process trying to figure out what contacts were good and what weren't he noticed a very interesting thing that he could make good contacts by embedding the mineral and a big molten metal by then there were metals that would that would melt how about the temperature of boiling water so it was easy to have molten metal and scrunched-up mineral around in it and then let it freeze that made a good contact the most doesn't and then he found out that if the other contact was a little wire that you put down on the surface a very remarkable thing happened and that is the voltage that it took to make a given current was a lot less in one direction and the other or said another way you get a lot more current in one direction for a given voltage than you do in the other direction so this was the beginning this is the very first electronic solid-state electronic device a little later a graduate student here in the US by the name of Hall was trying to understand things he was reading it was known by them that if you have a magnetic field by a magnet like this and you put a wire in it that that wire won't if it's say a copper wire it won't have any force on it but if you start running current through the wire it'll have a force on perpendicular to the direction of the current and Hall said well if it depends on the current it must be interacting with whatever stuff it is in there that makes that moves to make the current and if that's true since the force is in a direction perpendicular to the current in the wire you'd ought to be pulling on the stuff that's making the current and pushing it to the side and if it does let's say that the stuff that's making the current it's what we call negative charges they'll push the negative charges over to the to the left on this picture and that should make this left side of this picture of the sample in the magnetic field it should make it more negative than the other side we take the electrical stuff and push it to the left so it'll be less on the left side more on the right side are you looking at it so he figured this out well it looks just like a Wheatstone bridge if you put the two contacts for that full Vader on opposite sides of the sample there ought to be a place where the galvanometers want register anything and he found out sure enough if you if you didn't run any current through it it didn't register anything but there wasn't any voltage longitudinally either and if you didn't have a magnet to put the thing in you could run current through it and he could find a place along the edge where the voltmeter VH standing for hall I didn't register anything so it wasn't all experiment you could get a zero point and then you could go from there when you turn the magnet on he got a voltage first time that was observed well if you think about it if the stuff that's carrying the current has a negative charge it'll move it in wonderful direction and if it's positive it'll move in the opposite way and that means that this voltage V H will change side depending on the stuff the charge on the stuff that's carrying the current so you could tell if it was positive stuff or negative stuff that was carrying the current remember this is a long time before we knew anything about electrons so electricity was just a fluid still a pretty good way of reasoning he measured a bunch of things and funny enough some of had positive stuff that carried the current and some of them had negative stuff that carried that crime all of a sudden a picture got more complicated and that was a puzzle for a long time well by then we had radio Hertz had done his thing and had shown that you could communicate radio waves over a distance wasn't depended on it just coupling by being close and that you could steer them the way you steered light you could make a reflector and they would go in a direction just like a light reflector for a light bulb and then Marconi came along and said I know what to do with that that's a way to communicate between a ship and the shore before that there hadn't been a way unless you could see the ship well a lot of times bad things happen to ships and you can't see them from anywhere and it'll be really really good don't what they were where they were and what had happened so that became the first application of radio so the ship's got equipped with radio transmitters and radio receivers how do you make a receiver well they tried lots of stuff and the original ones were really crude and didn't they took a lot of radio energy to hear anything or make any marks on tape Marconi's first thing had a thing that made marks on tape and it took a lot of radio power very close to go run a thing that made marks on tape and then they figured out the human ear was a lot more sensitive than that but you had to be able to hear it well what's a radio wave well a radio wave is a oscillating signal at a frequency and the megahertz range and you can hear up to maybe 20 kilohertz so even if you could translate the radio wave into the oscillation of something like an earphone you couldn't hear it to be pushing in one direction the same amount as it pushing on the other direction on average and you couldn't hear anything but if you only looked at one half of the waveform he rectified the signal then you could turn the radio wave on and off or change its value and the average would pull the little diaphragm in your earphone and you could hear it so this was an application for semiconductors the first application for semiconductors and in the beginning it's like the lower picture there they had a little wire on the end of a little on the end of a little knob that you could move around and try to find a sensitive spot where the wire made a good rectifier on that piece of mineral that's down in that cup and you could poke around a little and occasionally get to hear a radio broadcast that was a lot better than not being able to hear it but it took a lot of fiddling around and it wasn't reliable until this guy Picard came along in 1906 and he explored contacts between different kinds of wire and different kinds of minerals and he went through about 30,000 combinations of minerals and and wires and finally he was able to get hold of a crystal of silicon actually it was a polycrystalline mass of silicon and he found that he could make a sharp pointed metal contact to the silicon and it was stable and you could even put it on a ship that was rolling around and bumping and thumping and it would keep working so it was the first time there was a really reliable detector for radio waves and the Navy got interested in that the US Navy and sponsored a lot of development of the use of his detector with various radio configurations so by the time World War one came along the US Navy was equipped with semiconductor detectors so that was the first real market for semiconductors meanwhile Bittaker in germany was interested in the physics of the semiconductor it was known all the people that dealt with minerals knew that every time you got a new mineral deposit the mineral specimens would have different conductivity and that would varied all over the map and everybody believed that that was because there were impurities in the semiconductor and they were responsible for it was just dirty stuff and so it was bound to have different properties for different samples and that's sort of the way they thought about it and bedecker is a very careful guy and he said there has to be a way of figuring out what's going on with semiconductors so perhaps I can find a way to bake controlled compositions of semiconductors and he did that for a lot of materials many of the materials which were like copper sulfide was a favorite of his and that was he could make that and the way he made all these samples he was copper sulfides and oxides were his favorites the way he made them as he'd make samples like the shape shown here and the two big pads on the ends were where he put down the clamps shown in the side view on the top and wow I can make great big contacts and then I can make narrow samples by for example in the B figure there you see it's a serpentine he got kind of scribe openings in the liens where the semiconductor had been scratched away and so you had a path that was semiconductor there was a wandering path so it was longer so it was easier to measure the resistance and he developed a whole bunch of techniques his most favorite one was copper iodide you see that in the bottom picture and he made contacts out of platinum it had really good properties and he found a remarkable thing in that with copper iodide the way he made of course is he vacuum deposited a copper layer a thin film which he could control a thickness by how long you evaporated the copper on the glass microscope side and then he treated the copper film with iodine either in a vapor or in a solution until it was copper iodide beautiful clear semiconductor film was a precursor to the to the conductors we now have on glass in all of our displays the conductive transparent conductors he was the the originator of all those and he found the plot that he sewed there that as you put more iodine in the iodide film it conducted better so if you had a low density of iodine in the solution you treated this film with you got a high resistance it didn't conduct very well but the more I had and you put in by making the concentration in the fluid that you are treating the copper in making it have more iodine in it or you waited longer for the idn't to go into the film then you got lower and lower resistance the film was able to carry current better so it's the first time that people had a controlled way of making a semiconductor of a given conductivity was huge it's a conceptual leap where you now could control the density of things that carried current by controlled chemical treatment and that's been with us ever since that process underlies every semiconductor device well by then there was radial we're now after the first world war a lot of development had taken place it was forbidden in the United States to even listen to radio during World War one most people don't know that but afterwards all of the people who had been doing amateur radio of which there were a large number were chomping at the bit because the vacuum tube had gone from a very early one in the early teens to something which was by the 1920 was a reliable source of radio transmission and so it's very interesting Irving Langmuir at General Electric really figured out how to make the modern high power transmitting to because for transmitting tube at high power you had to have high currents and high voltages this is one of the very earliest commercial transmitting tubes it's called a GE Kemna Tron and this this was rectifying voltages up to a hundred thousand volts and here's some of the other early tubes this is the first transmitting Tim that was manufactured here on the west coast and those vacuum tubes made broadcast radio possible because you could get enough power to cover enough area that people of substantial number of people people could listen to it well if you have the ability to put something out that people will listen to you have to make it interesting so they started with news and entertainment music and it went from there and broadcast radio that meant people wanted to listen and here's a picture of these kids listening their favorite radio broadcast you can imagine what that must have been in the 20s and it's a semiconductor detector in this little crystal set so it's very interesting this was the type of receiver that most people could afford they were making receivers with tubes in them but they were awful things they were great big and heavy and they used batteries they used a big battery to light the filaments inside the tubes and then they used a another battery to make the what was called the plate that would collect the electrons off the the filament in the tube and then they had to have another bias thing called a si battery so you had three batteries that you had to tend and they'd either go dead if they were if they were ordinary drive cells or you'd have to keep putting acid in them and checking and it was just a giant pain to own a radio that ran on batteries so that was the 20s and it wasn't until late 20s when the manufacturers figured out how to make a radio that you could plug in the wall an ordinary family could afford so in the 20s most people could afford crystal sets and there four of them here they're two home-brewed ones and they're two commercial ones and the commercial one this little commercial one was touted as a portable one now of course you had to have an antenna to make it work so that wasn't very portable but anyway if you were close enough to radio station it would work with a smaller antenna so people enjoyed carrying their radio around others this was a whole art form and whole families or groups of kids would listen to the radio and made possible by these big vacuum tubes the way it went during that period that was a very smart guy by the name of lillienfield he was a member of a group that was doing research on the brand-new quantum theory de Broglie had this crazy idea that electrons were really waves not little hard balls and that if you treated it that way you got a lot of the physics that was observed then Schrodinger came along and write down some wrote down some equations for that and Heisenberg made a fancier batch of mathematics for that and they were able to calculate the energy levels in the hydrogen atom and it was a huge effort from 1925 on to try to understand what was going on in the world from this theoretical point of view and lillienfeld was an experimental guy and he had already made some really interesting experiments that challenged the theories of the day and in 1925 he was watching here was radio and radio was going from semiconductor crystal sets where each one of these little things each one of these radio sets has a little cat's whisker detector just like the one I showed you and you can come up afterwards and look at them right there where you could fiddle with it because you had to and he watched that going from this passive device which just rectified the power that came in from the radio wave but it didn't amplify it and make it bigger and you really wanted it bigger so you could hear it better and or you could put it on a loudspeaker and that wasn't around there well about nineteen oh six or seven DeForest had come up with this vacuum tube thing didn't work very well and wasn't reliable and and it wasn't until 1913 that Irving Langmuir at GE came up with this kind of tube which was reliable and could be made in volume and lasted a long time and they made these big ones they also made smaller ones for receivers ordinary radios and Lilienfeld was watching this and it's a remarkable thing happening there when you had amplification you could do marvelous things and these two radios although they were expensive and heavy they gave remarkable results now Lillian said Lillian felt said I can do that with a semiconductor that was a fantastic thought and what Lillian said Lilienfeld said was okay I'm going to take a microscope slide and I'll break it in half that's what that little wiggly line is there in Figure one and then I'll take a piece of gold leaf and put it in between the two halves and squeeze them back together so now I have gold very thin layer of gold between the two sides of the glass and then I can polish down the gold to where only the edge is left and then I'll make one of my slides covered with semiconductor in his case he used copper sulphide I'll make that slide and I'll cover the glass and cover over that little metal edge that comes up which is very thin and then I'll have a region right above the gold foil where that metal makes a contact to the semiconductor and now I've got bronze thing the Brawn detector a metal against a semiconductor and what happens there is if I go in the direction where it doesn't carry much current there's a voltage between the metal and the semiconductor that voltage has to end up on charges and those charges are the things that are carrying a current in the semiconductor so I should be able to change the current carrying capability of the semiconductor by putting a voltage on the gold foil you absolutely clear reasoning it's the device we call a mesfet today it works beautifully and in fact each of you has one in your cell phone it's very good for being the final amplifier in a in a radio transmitter of which there's one in every cell phone of course Lilienfeld worked for two years to try to make his invention work couldn't make it work so he said well the problem is I can't really get this Junction to behave like Braun's idea where if you go in one direction you have a voltage but not much current I can't get that to happen I'm always get more current than I should I don't know why that is but I could put an insulator there I would still have an electric field between the control element which he had before as this gold foil now I'm gonna make a whole substrate made of aluminum and then of course it's easy to make a thin film of a little robach side-on aluminum and that was known and then I can deposit on top of that my semiconductor film and I'll make one spot in the middle you can see there on the figure which is narrow the reason you want to make it narrow is you don't want the electrons or whatever the things are they're carrying the current you have to go very far because it would take too long so if you want something that works fast you have to have the path that the current goes through be short and of course we've played that game for a long time now so this is the device which also dominates modern world it's called the MOSFET that stood for metal-oxide-semiconductor field-effect transistor and the field effect means there's an electric field that changes how many current carrying charges there are in the semiconductor this is by far the dominant type of device you have in your cell phones or computers today so both of lillienfeld inventions are not only workable but they're extremely important this one being by far the most numerous well meanwhile his colleagues wore thither theoretically inclined we're trying to figure out what's going on in solids what makes a mental conductive what makes an insulator an insulator and what makes these crazy things in between be in between they had this idea you could just take Schrodinger's equation which treats the electron as a wave and if it were out in a vacuum it would be a nice parabola so if you just looked at the left hand figured and you just sort of smeared out those dark curves it would be a nice parabola that's the energy of a propagating electron going as a square of the momentum well even classically an electron has a kinetic energy of you know the momentum squared over the mass so it should be a parabola but if you it's a wave and you put it in a crystal well the crystal has positive charges if you're going along it isn't just out in the vacuum that positive charges so there will be an interference between the wave of the electron and this periodic structure so act like a diffraction grating for electrons and what does that do well they found that it would create gaps in the energy that an electron could have and still propagate in the crystal and so if they blew up the the left-hand curve and look at it in the right-hand pictures they're the very bottom of the curve had an energy which is going like the square of the of the propagation constant which debroglie had said was the momentum okay that's right but then it got to the point where it's funny it come and the instead of being curving upward it would be curving downward and I got to a place where there weren't any propagating solutions and then you go up in energy a while on and then it starts another propagating solution and the curvature of these ways is like one over the mass in the equation for propagation well so that was interesting because if you had an electron at the top like there that electron is in a region where the energy is slowing down the propagation instead of continuing to increase it well that corresponds to something different and if you just think about the mass of these things being won over the curvature then down at the bottom the mass of those little electrons shown by those little hash marks is positive because the curvature is positive well when you get up to the the top one there the curvatures are the other way so that's something that has a negative mass which means that they push on it'll accelerate in the other direction that was the answer to why to some of these things that are carrying current look like they're positive things they're called holes and a lot of about half of the devices that you have in your cell phone work with those has the carriers of current and the other half work with ordinary electrons and then World War two started and everybody went to war and in the u.s. they didn't get into the war until 1941 but you could see it coming so there was a lab set up in MIT called the rad lab that building there they would just took over the whole building and they built a little hutch on the top which you can see and on the right-hand upper right hand picture you can see people working on electronics and antennas that was in that little building on the top there and lower right picture shows a glassblowing shop they had so they can make their own vacuum tubes and they could make the little packages for semiconductor devices and do all that stuff and they also contracted with industrial firms in particular Bell Labs and General Electric and they contracted with a bunch of materials labs that worked on getting better materials and guess what materials they picked well Picard had shown that the only reliable point-contact detector you could make was silicon well why did they need a semiconductor detector they had vacuum tubes well you're gonna put these radar sets in airplanes and they could only be that big and to get a beam it's just like optics to get up narrow beam you had to have what we get off mirror and that had to be many wavelengths across so for that that's only a fraction of a meter you needed waves that were like a centimeter wavelength and 30 gigahertz in today's nomenclature there are no vacuum tube would come close to that the very highest frequency vacuum tubes at that time were would go to a gigahertz so you had to go to higher frequencies and that meant you had to have something which was very small and the only thing they had that was very small was bronze detector so here's bronze detector on the right now what they make they made Picard's detector because it was the only thing it can be made reliably and the lower picture there shows how it was built it was built with a little piece of silicon you can see it's labeled there and a cat's whisker but just what Ron had done just what Picard had done just make it and make it reliable and fly it and that's what they did so the upper-left there is the current voltage curve for that device that was made in their ad lab and the right-hand one shows bronze device they're the same thing a cat's whisker on a semiconductor it just turned out that silicon was the most reliable according to Picard so they worked harder on it and they also worked on its its twin in column for germanium because some people thought that would work better and both things were developed and there was a lot of materials work done to make the whole thing work and it worked and they flew the radars halfway through the war a German scientist Schottky he was a dogged fellow he had been trying to understand the metal semiconductor contact every since Braun had shown it and he kept coming up with theories and they kept being wrong and being shown to be wrong by experiments many of which Braun had done and by 1942 it turns out that the German publication zeitschrift for physique was still being published even though the war was raging and Shakti published this paper which said okay I know how this works he took selenium which was a semiconductor and found that the impurities that were in selenium were atoms that had an extra positive charge that could come off we would say today they made a p-type semiconductor and so if you put a metal up against it there would be some relationship between the energy of electrons in the metal and the energy of these positive things over in them in the semiconductor and if I put a positive voltage on the metal that was a direction that not much current would flow because there's an energy barrier there and that's this five-plus thing and so what would happen well there's a positive voltage on the metal I have to have a negative charge on the semiconductor to absorb of electric field lines that come from the positive metal so what would that be well if I push the positive charges that are the mobile things that will move if I push those back then what's left is the atoms for those positive charges came from and they have a net negative charge left over because they were neutral atoms when when they came into the crystal and so the region that had the electric field in it would expand as I made a larger positive voltage on the metal and indeed he could measure the capacitance which is like 1 over the thickness and went just exactly the thing you would guess that the voltage is the electric field times the distance the electric field is the amount of charge which goes like the distance so the capacitance should go like 1 over the voltage squared and if you plot it it did that so that was the first theoretical understanding of what went on in these metal semiconductor contacts and that turned out to be a key thing one of the labs that was contracted by the rad lab to do work on these point contact bronze style Picard style detectors for radar was Bell Laboratories and they quite properly saw that here someplace there's a semiconductor device that will amplify so Bardeen who's a left hand figure in this in this publicity picture said hey when we have a reverse biased point contact there'll be a region around it which you push the carrier's out of just like Schottky said and if you put another little contact up there you should be able to measure the extent of that region so Bratton who's the right-hand figure on this on this slide was a good experimentalist and he made a thing with two little manipulators that she could make two point contacts and get them real close and as they were measuring what happened the metal contact on the Left they notice it affected the current in the metal contact on the right well that's interesting the current that you put into the metal on the Left a fair fraction of it came out of the metal on the right you'd say well there's no gain in that you actually lost some of the charge yes but the metal on the right is a reverse bias Junction it's the one where if you change the voltage the current doesn't change much so that's a high impedance device and on the left that's a forward bias device if you change the voltage a little it changes the current a lot so if the two currents are anywhere near you get a gain that's like the ratio of the impedances so to test that theory out you went to the bottom right diagram there you just make a transformer and you use a few turns on the left hand one and a lot of turns on the right hand one and so you tune them both to the same frequency and it oscillated like a banshee and that was the first understood active device it turned out there had been devices that would oscillate these Picard style devices Picard had found some that would oscillate and there was a Russian by the name of loss of that had gotten a little better at making things that would last late but they were never really understood are controllable but it was this was the first time that there was a thing that it looked like you could control well Bardeen than being a fantastic and like said I'm gonna understand this thing how could a metal contact work to put things that would carry current that were of the opposite signage you would expect from that picture of shot keys and it turned out that he figured out how it was there was this energy barrier for the the current carriers that were there in the in the semiconductor they would go over the barrier to the left from the right hand side of the picture but also there were the other sign of carrier which on above the Fermi level that would go to the right over the barrier and become opposite sign carriers in the valence band and that was what was being observed it was not the current that was coming the way you would expect and the way it worked for all the other things it was coming the opposite way and that was a property of germanium and that was why the first transistor was discovered in germanium well Shockley was not to be left out of this party he was not at all shy about getting into the party and he actually was the one that figured out really what was going on if bardeen's theory was right then what it really was like was you had one kind of semiconductor that had one kind of carrier let's say n-type had electrons and then you had the other one on the right-hand side that was n-type and then in the middle you put up a little sliver of p-type so if you forward bias the junction between the N and the P on the left side and reversed biased the NP Junction on the right-hand side then you had a thing that worked just like that point-contact transistor did except you could make this reliably and after we've worried about that though and that was the birth of the transistor as we know it so Shockley actually deserves his his position there in the center of the photograph this was the photograph from the publicity photograph for the nobel prize that was given to the three of them well then how do you actually make these things well they're a bunch of things tried and early transistors were Kluge's of various sorts and finally a sky pen cough whose real name you couldn't pronounce so he changed him invented this incredibly incredibly simple and reliable way of making transistors you took a piece of germanium and you put a little pellet of indium on both sides and you put in an oven until the indium melted and the indium dissolved a little bit of the germanium and it was very controllable because this solubility of germanium and indium goes up very slowly with temperature so you could get a temperature where it went in just just a nice controlled amount and then you cooled it down and the germanium and that had dissolved in the indium crystallized back out on the germanium and that was their bullet germanium that was there was n-type and indium is a p-type dopant it takes an electron away and what's left as a hole in the valence band this was a PNP transistor P from the indium on the left P from the indium on the right and n-type because that's what you started with incredibly easy to manufacture these things they would only operate up to about a megahertz but they were a transistor Raytheon brought out one called the CK 7:22 and that was the only thing you could afford to do in lab and so a lot of us grew up making these very low frequency circuits out of this includes e transistor but they were only a dollar which was a lot more in the 50s than it is now well then it was clear that this was a route to manufacture ability and Phil Cole who was big in electronics in those days came up with a fantastic technique for making transistors they were gonna make they were gonna tool up to have an automated factory which I was able to tour in the either late 50s early 60s and what they did is they had a little flake of germanium and they would they had a little spout on both sides and they would squirt electrolyte on the geranium and if you put a current into the electrolyte in one direction it would etch away the germanium and if you put the current in the opposite direction it would deposit indium so this was an automated way to make these PNP transistors and well how do you know when to stop well it turned out that germanium absorbs infrared light and of course the the absorb the amount you absorb goes exponentially with the thickness so you shined light this made a little light pipe this electrolyte you shined light infrared light down one and looked at how much came the other side it was very easy then to know when you get this much light you reverse the polarity and you deposit the indium and then and they had this factory running the thing and get indexed over to the next thing and their wires would be put on and the whole thing came out of the package transistor cut Chuck cut chunk cut um and I've I got to tour that that Factory in Bluebell Pennsylvania very impressive so here was an automated way to make economically feasible transistors and also you could reliably make the region in the middle of the base region very thin this way and that made good transistors and instead of a megahertz like the CK 722 these transistors will oscillate at 70 megahertz that's means you can make radio out of them so they thought they really had it made but meanwhile some Bell Labs guys were working away on silicon and they said I can put impurity atoms in silicon by diffusing them and if I want to make a junction I just diffusion and it was known how to diffuse impurities into silicon but I only want to diffuse them in a very narrow area well I just oxidized the silicon silicon oxide one of the best insulators in the world and then I use a photo lithographic process which are well known from the printing industry and I cut a hole in it with hydrofluoric acid you can dissolve silicon oxide with hydrofluoric acid and that won't attack the silicon underneath so it stops this right place and then I apply some other impurity atoms and I heat them up and they go diffuse in and I've just made myself a junction right where I want it and I didn't have to have Jets of stuff and deposit things hanging out on the air this was just on a surface of silicon and if you diffuse things in right you could diffuse one kind of dopant in a short distance and then diffuse another kind in a longer distance and if you had started with the right dopant in this case you started with a low concentration of n-type material and then you diffused n-type material in but you disused in a p-type material that diffused faster so what you got was the bottom left picture there the net doping the net concentration of impurities was n-type near the surface n-type far in and p-type in the middle that was an NPN transistor well electrons in silicon go faster than holes do so this was a fast transistor and they showed because you can really control the depth of the junctions by this diffusion process which is just a chemical process you raise the temperature a certain amount and the diffusion happens at a certain rate extraordinarily well controllable down to fractions of a micron so you could make a transistor that had sub micron dimensions vertically by this diffusion technique and all of a sudden the transistors instead of being 70 megahertz were 400 megahertz and silicon had by it had bypassed germanium as the speed record and you could make a whole bunch of these transistors on a single wafer of silicon so that's when it dawned on people that we weren't gonna make economical transistors by a machine that could chunk katana katana and it would do a Qing and it would do depositions and stuff we were gonna make economical transistors by making lots of them on a single piece of silicon that was a huge huge then started the migration of expertise from the East Coast centers to the west coast this was the first one pacific semiconductors inc just right down here in Culver City and they quite properly saw that there had been a big pure are over transistors because the vacuum tube people said yeah you can make these little transistors and they'll be good for hearing aids and then as they got faster well they'll be good for little transistor radios but we're gonna be making the things that make transmitters work our vacuum tubes we can miniaturize those and we can make them work at higher frequencies and we can beat you guys and there was a period where there was a lot of pulling and hauling over could you make military specification transistors that would compete with vacuum tubes and now the jury was out on that but the issue came down to Kent you can make these little transistors and they can work at high frequencies but that's because they're small and the people at Pacific semiconductors saw that you can make a big transistor by making a bunch of little ones in parallel so they started doing that very early on and by 1956 they had transistors like this on the market that would dissipate 50 watts they'd run it as good amplifiers at a hundred megahertz you could make radio transmitters that and so that settled the argument you don't need vacuum tubes to make all the technology and as this artform evolved that's what happened well the biggest single step in this evolution process toward the technology we have today was this one bill Shockley had come to Caltech for a year and he met Arnold Beckman and he convinced Arnold Beckman the bankroll him to start a little research lab up in Palo Alto and he did that and he went around and hand-picked a bunch of very bright young PhDs and within months they had record-breaking transistors that they were making and everybody was excited and bill said now transistor that's all known we're here to do something new and he quite properly saw that if you made a four layer device instead of a three layer device you could make something that would switch power really well and that's the kind of thing that's in all the light dimmers today so he was right but we didn't know what he didn't see was switching power wasn't gonna be the explosive market it was representing information electronically that was going to be the explosive market he couldn't see that these guys did see it so they got together and they said we're out of here they went back and met Sherman Fairchild got him the bankroll this little company that of course was called Fairchild because he put up the money and here we are there's Gordon Moore there's Shah her knee and there's Bob Noyce and those are the people who made the key steps in our modern technology well meanwhile tech technology was evolving everywhere and Texas Instruments was not to be left behind and they were the first ones to introduce a good high-frequency silicon transistor and Jack Kilby looked at this and he said you're making transistors on a wafer I could connect them up and make a circuit because if I have a diffusion in a wafer I can make a resistor out of that and if I have three diffusions vertically I can make a transistor out of that and I can hook them up and that would be a transistor all on the same wafer and that was right but what Jack didn't have was a way of hooking him up so he had these flying leads that you can see on this picture you could bond them up with bonding wires well that's not a integrated circuit that's a circuit that's all made on one piece of silicon but it had to be hooked up by individual external wires so he had a good vision but no way to get there what he didn't know was at that very moment John Ernie had invented this thing we call this planar process and what it was was you started from one side of the silicon and you diffused in the first layer and meanwhile while you diffused it in you let an oxide grow on the surface that was the insulator then you could cut another hole in that oxide and diffuse in another region and now you had a transistor and then you could cut holes in the oxides where you wanted them and you talked about any one of the elements of the transistor and they were all separate and in Jean her knees patent the figure 10 down at the bottom you can see you could get a wire to any part of the transistor to the emitter to the base and of the collector and that wire could be the aluminum that you use to make a contact and of course Bob Noyce signed Jean her knees patent disclosure and he looked at that and he said John we can make an integrate a real integrated circuit with this so Bob filed that patent and so the her knee and noise insights or what gave us our modern integrated circuit and Jack had a really good idea but couldn't make it well there been some important and mentioned since then I have to mention this one because Kong anatella John Italia at Bell Labs worked for many many years to try to figure out that silicon-silicon dioxide interface it's a complicated thing and everybody who's tried to make silicon devices runs into all of the things that can happen there that'll kill you but these guys that Bell finally kept at it until they were able to make really reliable silicon-silicon dioxide interface --is and that made possible Lilienfeld second invention the MOS transistor here you had a metal on top the oxide on the silicon and then the silicon was your Semiconductor so that was the first real commercial field-effect device and as you can see it's a very simple device you don't have to have so many layers vertically one more thing that was important if you're gonna make logic out of a circuit that means you're going to represent a 1 with some voltage and not 0 with a different voltage and that means that you have to have a transistor to pull the output up to the voltage that represents a 1 and you have to have a different transistor to pull it down to a voltage which represents a zero well Frank Wanless also at Fairchild quite properly saw that if you did things right you could hook up like figure 5 there which had it a top transistor which is a p-channel device so if you made the gate low it turned it on and that pull to the output up and if you made the bottom transistor an n-channel device you could hook the same input to it and when that input went up it would pull the output down but when the output was steady both devices were off let me say that again properly when the voltage on the input was steady one device would be off and the other one would be on so the output would be held either at the bottom potential or at the top potential but there wouldn't be any current flowing because always one device would be off and by hooking devices in series and in parallel you could do logic that's what figure 8 figure 7 or about and then the big one on figure 8 you could do it on an integrated circuit and he spelled out the way you would make that integrated circuit so that's the modern CMOS and that's what's in all your computers and that's what's in your cell phones and of course I have to mention a couple of other things Bob Denard at IBM came up with a way a manufacturable way of storing charge temporarily on a transistor gate and then coming back and reading it and in the process of reading it refreshed it so that even though it was decaying away you didn't have to have something pulling it up in something pulling it down because that took a lot of it took at least two transistors but this way you only had that one transistor and you just came back once in a while and refreshed the charge that was there that's the dynamic Ram which is the workhorse memory and your computers and I can't resist there was one more thing that lillienfeld did and that was his very first thing 40 years before he had come up with this mesfet device and nobody had ever been able to make one so this one was made right here at Cal Tech and that's how I got to know about lillienfeld and actually it it made me feel really good that I can finally make something that Lou Ian felt had completely clearly invented what was never able to make I just wish he could live to see it well from then on Gordon Moore 1965 he made this plot he gave me this this copy he also gave me one of the original ones but I couldn't find it the original ones are in the black and Gordon said you know we're getting more and more transistors on the same chip every year it turned out that what had been a quest for new and better electron devices became a quest for how do you do the ones we know work that are smaller and faster and dissipate less power and just put your head down and do it and the people that did that beat out all the people are trying to be fancy and what's happened is Moore's law and this is log to the base 2 so we started at 64 devices on a chip we thought that was pretty neat now we have billions and it's still going but a lot slower so you always wonder looking back does it teach us something well they're really three threads in this period and I've only really talked about one of them what just occasional references but you know you'd think looking at the history that this this vacuum tube stuff was a giant diversion from the straightforward evolution of you all have this in your handout so well but that wasn't true because they wouldn't have been a market for semiconductor devices if there hadn't been radio and there was nothing that was gonna make a reasonable radio broadcast technology except the vacuum tubes at the time so they made the market for semiconductor devices and without that market they wouldn't have been developed so it it's kind of odd people often ask me what's next we're on Moore's law and it's slowing down Steve here is a a big proponent of doing the same thing better and faster for the next generation and he's made a lot of these field programmable devices that are leading ads use of silicon and even Steve is getting a little queasy about well Moore's law is not going so fast as it used to is there gonna be a next generation of stuffin and what's it gonna be well you don't know about these things I can tell you from personal experience over here we are working hard to make new and better electronic devices and then Bob Noyce comes up with a way of hooking them together I never thought about that it's a good idea then Gordon comes law and says you don't need new and better kinds of devices just make them smaller and and keep going and now Steve probably did that more effectively than anyone I know didn't think of either of those things at the time well there's something that's just happened in the last 10 years and that is the way devices are used was really structures that we knew about things that would multiply things that would add things that would store and things that would you know sequence and you could make computers that way you can make memories that way you can do all that stuff and then here comes this new thing which we're all working on 30 years ago that we called neural networks back then and it turns out that now the technology for some reason will do those really well and so people are all making different interactions different structures on their silicon than they did and Steve things actually still in the running for that and that's a good thing that was not an expected thing when we were working on it we thought that would happen and nobody would give us the time of day that was 30 years ago and now everybody's all nuts about it the principal hasn't changed at all it's just that we got enough stuff and it runs fast enough that now it's feasible to do that in a commercially effective way I wouldn't have expected that I expected that 30 years ago I gave up so we don't know what's at least I don't know what's coming next there are lots of ideas maybe it's vacuum tubes you know you notice here in the middle column down by the bottom Ken shoulders in the 60s was making vacuum tubes that were a micron in size they worked really well and he figured he could make integrated circuits out of that at about the same time that Fairchild was off making integrated circuits never went anywhere well right now our very own axial shear is making little tiny vacuum tubes and now they're nano scale not micro scale and the kind of physics that goes with devices that size scales really well and electrons go really well through vacuum it's a good medium maybe vacuum tubes have come out who knows I'm not gonna poopoo anything but it's going to be very interesting in the next 50 years been 50 years since Moore's Law and one of the things that's happened as all the people that were good at figuring out how the devices work and inventing new ones and stuff they're all retired they're not and that lore has not been passed on to the educational system so right now we don't have a group of people who are good at inventing the next technology but there are a few out there and I wouldn't count them out so let's watch and let's let's root for the next generation you're welcome to come up and look at the at the exhibits afterward [Applause] oh boy carver has offered to answer a couple of questions so but please wait for us to give you the microphone because this is being recorded so questions so thank you very much cover that was a wonderful as always talk what was the biggest kind of like mistake that people made in your mind in terms of the way they did things that didn't pan out what was the thing that they shouldn't have done I mean in retrospect it these things are easier to see for me was that lillienfeld didn't discover selenium it's such a tragedy that he didn't get to see his devices back when he knew they would work and he was working on them and he was an incredibly talented experimentalist and and neither one of them would work with the stuff he chose so he stuck with copper sulphide fita just gone to selenium ada had it and then we'd had a whole different evolution curve would vacuum tubes have been obsolete I think not I think we still would have had a race between the vacuum tube and the transistor but it wouldn't have taken so long that to me is that is the ultimate tragedy in this story any other questions oh here's one over here what do you think about bottom-up assembling of devices autonomous self-assembly of devices do you think it's a future molecular devices and so you don't need to have like that huge good question salute ah what about self assembling devices that might be the next thing my guess is it's gonna be a little like our you know what's now called deep learning that it'll take a while for it to become but you know there may be an application area where it just fits and you get an early early start on it that's certainly what I'd be watching oh I should say something about that if you're gonna have a self-assembling technology it's not going to be a hundred percent reliable so the application is going to have to be fault tolerant and people are getting better at that but not on the scale that I think it's going to be required for for that to become commercially viable and neural systems are the only things we know that that tolerate massive kinds of of errors and learn around them and it could be that there's a some kind of interface between those two things how would you recommend that we go about in particular Caltech go about creating the generation of of the creative people needed to figure out what these next things are going to be well one of the advantage we have here at Caltech is that there are a lot of very unorthodox directions being explored and you don't expect all of them to be a home run at once but we get more than our share and I think as long as we don't get a big group thing going where everybody's pulling the same direction there'll be enough creative things that are being explored that will we'll have our share of the of the ones that push things forward what you don't want to do is just get a big group think where everybody has to be marching to the same drummer and Caltech is better at that than most places cover I know you've been heavily involved both in research here in Caltech industry so you're in a good position to answer I also wondered transistor was invented at Bell Labs but the 8080 system was the electric eighteen T which was the mother company of Bell Labs was one of the last companies to actually benefit when the transistor and adopted for its own natural system what is the economics politics put into action by maybe I touchy but not white well of course I wasn't at the Bell System but I can tell you that by the time we were doing this stuff the Bell System was heavily interested in being the big dog in the applied science area and that didn't help in terms of creativity but also there was the consent decree that they were not allowed to to benefit immediately from their invention so that kind of put a damper on on the commercialization and they were forced to license everybody on reasonable terms which they did and did very well so that put a damper on it but you know as late as as the MOS transistor they were leading edge in terms of the research that was going on at Bell so it's complicated i I don't you know I couldn't give you a one-liner your your beautiful recount of the history so far it seems the even though it went through the two world wars though it was mostly just put aside as a distraction to the technology development but but nowadays technology has a very strong political side and I wonder how you think about you know politics and you know what's going on outside of technology influence how technology would go ahead you know in these years were in the future well it's certainly true that technology today has a huge groupthink going so and it's a natural result of the fact that that's what's been rewarded they're just doing the same thing better and so any new thing has to be much much better not just better in order to to get any to break through any group think that's that's just doing what we know how to do I think Cal Tech's in a better position to do that than other places it's still not easy because you're going against the flow and every one of these inventions I mean these are incredibly clever people that understood perfectly well what was going on after you know a lot of this transistor stuff had settled and we had bipolar transistors the NP ins and the PNP s before the field effect devices had taken over there was a analysis that appeared in one of the magazines about Lillian fells invention and the guy that it that analyzed it thought it was a minority carrier device because he could only think one way now we had minority carrier devices so that was what it must have been so the groupthink thing is extremely strong and so if you if you're gonna go past it or against it it's not gonna be easy I think we're getting late yeah look at that and I'll be here if you have not to tell you can maybe oh yeah oh yeah yeah we can do that while everybody's getting something to eat you

Channel: caltech

Views: 6,586

Rating: 4.952569 out of 5

Keywords: Caltech, science, technology, research, Carver Mead, Semiconductors

Id: qhJaq3kl6Dc

Length: 93min 8sec (5588 seconds)

Published: Fri Jun 07 2019