We see a lot of sort of figures and numbers for kind of how small things are getting in computing, right? And I thought who better to come and talk to than Phil Moriarty, would you call yourself a nano-scientist? H Oh I guess I would it depends on which funding agency I'm applying to funding for, but yeah nano-scientist is a good one *laughs* Yeah it's great to be back on Computerphile it's been a long time I don't know a couple of years or something like that? Um.. So yeah we do, within the group here we manipulate individual atoms that's what we do so we work at the level of not just single atoms and single molecules but actually looking inside single molecules The state of the art is no longer, at least in the research and development community not in the commercial or industrial community but in the research and development, particularly in the universities, State of the art isn't seeing atoms we've been able to do that for years. In fact, we could see atoms as long ago as 1955. Um, with something called the field ion microscope.At the beginning of the 80s however, it all changed because there was in instrument called the scanning tunneling microscope developed, Which is basically a sharp probe, you bring it in close to a surface, you move it back and forth and you measure a force or an interaction and the important thing is if you can make this probe atomically sharp, then your resolution is at the single atom level. So we can see atoms we can move them, but as I said the state of the art is actually single bonds looking at single electron orbitals, single chemical bonds, and manipulating those. Commercially, that technology doesn't exist and its gonna be quite a while before we get down to devices which are truly mass produced at the single atom level. But if you look at the, you know, from the 70s I think we were at the sort of 10 micrometer level, so 10 micrometers, a micrometer is a thousandth of a millimeter, a millionth of a meter.
We are now at the 14 nano-meter level, so the really offhand way to try and get some vague handle on it, is, for many of you out there perhaps not for me but for many of you out there- your hair will grow roughly about a nano-meter a second, Particularly if you're in your, you know, teens, 20s Your hair's growing out at about that rate of about a nano-meter a second. In the context of single atoms, the diameter of an atom is about a few tenths of a nano-meter When we're working at 14 nano-meters and the feature sizes on semiconductor chips are around about 14 nano-meters We're talking, you know, tens of atoms, we're talking about about 50-60 atoms wide, the features So it's um, it's really quite remarkable you know? We went from a thousandth of a millimeter down to 50 atoms in the course of, you know, 50 years or something like that So you mentioned the features, is that things like the individual wires that go to make a transistor? Yeah, exactly, that's exactly, so it's the feature size really is to do with the all the different elements on the chip in terms, largely the transistors and the wiring and the different types of components you have in the chip that have been just scaled all the way down to that level. At the moment, the way it works and the way it worked for decades is that you control where the electrons are, you control the electron's charge and electrons will respond to electric fields so if you've got a battery, and you've got two metal plates and you put a battery across them What you have between those plates is whats called an electric field if you put an electron in there, then that electron will respond to the electric field A great deal of the electronics around us is based on silicon, there are other compounds, like Gallium arsenide for example But a great deal of the micro and the nano electronics industry is based on silicon. S: So does the silicon work a bit like the plastic, or maybe the PCB, if I said the green PCB, that's plastic with wires on it- That's a really good question, no the silicon's much more active than that, much more active, it's not like you pattern these features in the silicon as just a passive substrate, you're actually using the silicon and the electrons in the silicon and you're controlling where the electrons in the silicon go and you can take the raw silicon and you can dope it, you can add impurities deliberately to add, to introduce more electrons, or indeed, actually, to take electrons out.
So we got electrons and holes. And that means you can change dramatically how when you put a battery on this thing, or when you put a power source on it how those electrons flow, and then by, in turn, patterning little metal features on top of that, you can apply electric fields and you can control where the electrons go. You generally switch on or switch off the flow of the electrons, you trap the electrons in a region of space. Now, the problem with this technology is that it's starting to run out of steam.
It's been running out of steam for quite some time. And there's been lot's of nay-saying going on since the early 80s Saying it's going to stop -it's going to fail and the semiconductor industry is extremely clever and comes up with new ideas time and time again. But, you can't beat physics and were going to come very soon -we're now at tens of atoms when we get down to features that are just a few atoms, or say ten atoms across then, we've got to take into account that once you get down to that level it's quantum, which means that you can no longer just think of the electron as a little hard billiard ball, which is the picture that all of us have in our heads it's actually got a wave-like character which is not to say it would be easy if it were just that the electrons spread out in space that's not what happens it means that under certain circumstances it behaves as if it were a wave and if you find that confusing good, because so many of us find it confusing. This is raw quantum physics As a physicist, it's not that you understand it.
You just get used to it. Once you get down to this level you get this wave-like characteristic and just like waves will spread out so you might want to trap the electron in this region of space but it's starting now -due to this really small size- at the point where we've got to take that wave-like nature into consideration and the electron could spread out. We call that tunneling. And that means you want to trap it in space, but in fact you just can't.
It's so slippery, it's tunneling away. So I'm gonna find a really stupid question cause I am full of them today I've heard that quantum computing is a good thing, so I mean... Oh, So, oh, so, no, So there's a different, yeah right, *bleargh* So there's quantum computing which is where you absolutely exploit that. So there is classical computing which is what we have now.
And than we have quantum computing. Now those are two very, very different paradigms. So instead this is quantum effects on classic computing Perfect! Yes. Quantum effects on classical computing instead of just being pure quantum, The sort of mindset in the industry and the mindset in some areas of academia is How do we work around these effects? I think we are gonna have to start, particularly when we get down to these really small sizes We are gonna have to stop seeing them as something that's a, erm, you know to the detriment of device and exploit it instead Well we can do amazing things when we start thinking of that wave-like character And the move towards a much more quantum mindset Where instead of trying to work around these things, we exploit them So we work with silicon a great deal in the lab It's quite shiny...
It is very very shiny. So, it's polished on that side. So, when we do our experiments to manipulate atoms we take a chunk of this, a little bit less than a centimeter squared, something like that we put it in a ultrahigh vacuum and we heat it up to about 1200C. And that drives off the oxide that is on the surface and just after it cools down, we can see atoms.
So it's actually straight forward to see individual atoms on Silicon surfaces. That's not how the semiconductor industry works because to see individual atoms, particularly the way we do it and the way many other groups in the world do it it takes many hours to get good images To form the components and pattern the wafer, you use something called photolithography or something called electron beam lithography and photolithography has been the standard throughout the industry for very verry many years but it is, again, running out of steam. So you take the wafer, you coat it with a thin film of polymer or plastic and then you put what is called a mask in front of it or stencil basically. And then you shine light through that stencil and then you expose the plastic at the surface in particular regions according to the pattern and the important this is when this particular polymer is exposed to light, it becomes soluable The regions that aren't exposed aren't soluable so then you can put it in an etch and you can remove those regions where it's soluble and leave those regions where it's unsoluable That's the fundamental process, that's the fundamental process in a nutshell And do they do different layers of this as well? So, it's extremely clever because you're limited by the natural wavelength of light light has a wavelength and therefore that wavelength ultimately determines the feature size and with the current processing which is about 14 nano-meters, they do something really really clever which is they use one mask, expose, and then they put another mask and just offset it a little bit which is really clever and given this is 14nano-meters, when I say little bit, that's a tincy tiny bit and actually getting those in registry is quite tricky So, that's where the industry is at the moment To go beyond that, really to get down to these features which are 10 atoms or a few atoms across then you have to start thinking about lots of other issues and lots of other approaches One way of doing that is to move to a much shorter wavelength a much shorter wavelength means much higher energy and we have something called Extreme Ultraviolet so that's far off the end of the Ultraviolet range And you can get down to -or the standard they're aiming for is 13.5Nm wavelength light and that's quite high in energy.
It's really high in energy and that will enable you to get the feature size much much smaller You can really push the resolution, you can really push the feature size down by using instead of photons of light, using electrons and you can squeeze the wavelength of electrons down much much smaller the problem with that is it's a serial -with light you have your sample, you have your wafer and you just bathe it in light with electrons, it's a raster beam which means it is, instead of a parallel process, it's a serial process which means it's incredibly slow So electron beam gives you much better definition and much better resolution but for the semiconductor industry, it's a real pain because it's so unbelievably slow So... And for us, when we manipulate atoms it's exactly the same process we use a single probe, we bring it in close to a surface so we can see atoms and we can manipulate those atoms and move them around, but it's excruciatingly slow because it's a serial process This is the crystal structure of silicon. Just assume all the atoms are the same color but each sphere here, each ball, represents a single silicon atom and the silicon crystallizes in a form where you have these tetrahedrons so if you take one silicon atom here, this purple one. It is connected to 1, 2, 3, 4.
And that's because in chemical terms, silicon has four valence electrons It wants to form four valance bonds We have these, really silly at times, frustrating; this is chemistry, this is physics. It's all part of one integrated hole even computer science is part of that integrated hole, sometimes. ...But that's sort of off the point of what we were discussing It's fun, but just to show you in action that um....
No comments:
Post a Comment