 Hello everyone you hear me Yeah, good, right. So before I start on this any other physicists in the room hands up Okay, so those people I apologize that this is incredibly Not interesting and not in-depth as you might like it any complete non physicists in the room and To you people I apologize that it might be a little bit quick and and covering a lot of stuff so I am I My name is Dan whether all I am a detective physicist at the University of Oxford. I work on some of the things you see in the picture It Modern science is done using detectors of some kind basically our only window into the world is electromagnetic radiation Even for example a LIGO detector which detects gravitational waves does it buy the weak coupling of those gravitational waves into Interferometers which are electromagnetic and so electromagnetic detectors are very important And what I want to try and illustrate to you guys is some of the basic concepts that unify all these things in the picture so the top left there we have the LST telescope which is When it's built it's currently being built will be the Most powerful digital camera in the world. I mean 3.2 giga pixels So you need a fail of hard drives for it thousand images a night produces about 20 terabytes of imaging data every night for 10 years So that's not my problem. Luckily On the bottom left there. We have the the Atlas detector at the LHC and CERN And the bottom right as well and in the which you've probably heard about if you've been to any of the talk There's a couple of talks yesterday And there's a couple of other talks about the LHC and related things I think and in the center We have the humble Sony DSLR These devices to the interested party look very different and they look very different to the outsider as well There is a middle ground that connect all these things and that is that they are basically Digital cameras for light made with silicon and that's what I'm going to talk about how those work and so basically There are a lot of steps in this process a digital camera or an imaging sensor It's something that goes from that top left box photons to the bottom left box digital images and most people are Just interested in that That that left arrow there if you're in this room, I'm going to assume you're slightly interested in the other few arrows Where we go via nearly free electrons in semiconductors and we confine those electrons and then we turn them into a voltage Right, that's basically how imaging sensors work. We turn photons into electrons. We catch the electrons We count how many electrons there are simple There's a few details and we'll try and spend the next while doing so incidentally if anybody During the while I go through this once To ask me anything like I'd like to be asked because this is intended to be introductory talk for people who are interested so Yeah, so we'll start off with the em spectrum the em spectrum is Covers radio waves visible light to x-rays. They're all photons There's wave particle duality of course so photons are waves and particles for the purposes of Sensors that work in solid state like silicon you almost always want to consider things as photons and photons are little packets of energy and Visible light photons are the ones that are striking your eye So you can see this picture at the moment typically have a couple of electron volts of energy. That's Some small number times 10 to the minus 19 joules So it's a small amount of energy per photon, but those are actually quite high energy photons compared to to radio waves The thing to remember about photons and electromagnetic spectrum is longer wave lengths means lower energy So x-rays have very short wavelengths Very high energy photons, okay And the energy of a photon in the visible corresponds to what we perceive as color so red things Have lower energy photons and blue things The underlying thing behind how all sensors work is the photoelectric effect all solid state sensors This was not discovered by Einstein in 1905. It was discovered in the the 1870s But it was explained by Einstein in 1905 in terms of what was then new thing called quantum theory And he won the Nobel Prize for that in 1921 so interestingly although Einstein came up with general relativity Most people consider that to be his greatest achievement. He also came up with a lot of interesting results in statistical thermodynamics The one he actually won the Nobel Prize for was the photoelectric effect The photoelectric effect basically is the effect where you have electrons bound around an atom So an atom has a nucleus and it has electrons in orbital states around it and If you collide a photon Into an electron with sufficient energy it can be released from that atom Okay, so that is that is basically once you understand that you understand the mechanism by how all optical sensors work we have a Lump of silicon which has atoms in it Photons come in if they have sufficient energy we can knock electrons out the atoms and then we will measure those electrons with electronics, right? So any questions at that stage? Okay Of course, it's not as simple as that even Those who love to to mess around with quantum mechanics will have heard of the Pauli exclusion principle or the pep Basically when you have a solid Your electrons are a delocalized throughout that solid in most types of solid is crystal lattice structure and this is This is largely due to the translational symmetry of crystal crystal structures The power the exclusion principle says no two electrons can be in the same energy state There are spin as well, which adds a factor of two, but in this case, it's not much of a complication Um, so what that means is in a big lump of something like a silicon crystal No two electrons can occupy exactly the same energy state Which means there are trillions of possible energy states in that silicon crystal and each electron must occupy one And so what happens is those energy states are very close to each other and they form bands so so these lines at the bottom here this at the bottom right that is called a band structure and it's basically a diagram of Momentum along the bottom. So basically direction which direction electrons traveling versus energy at the top and the orange lines are Positions where an electron can have that energy quantum mechanics allows it to have that energy by the power the exclusion principle The blue shaded region is a thing called the band gap and that is a region where they simply do not exist Possible energy states for electrons. So in a solid if you have an electron in silicon That energy there in the blue region cannot be occupied by an electron quantum mechanics doesn't allow it, right? and this is useful for us because Electrons fill up from the bottom lower energy is at the bottom and so in normal conditions Those states those orange states at the bottom are largely full and the ones above the gap are largely empty And that means that we can shoot some energy in With enough energy we can pop electrons from underneath the gap to on top of the gap And then they'll be free to move around This is not a unique property of silicon there are many semiconductors the thing about silicon that makes it so ideal Is that we happen to live in a world where we have a room temperature of about 20 degrees? And at that temperature silicon's band gap The there's a thing called the Fermi energy, which is the energy that electrons want to have basically It's the thermal energy that every electron on average wants Silicon at temperatures that we live at that Fermi energy is right in the middle of the band gap And that means that silicon is a very convenient semiconductor for us to use at temperatures that we live at Because at room temperature There are very few electrons that have crossed over and that Fermi energy is exactly what we want it in the gap So our electrons are sitting just below the gap and we can fire energy in and get them above the gap very easily There are other semiconductors which only become semiconductors at minus a hundred pretty useless to make a camera that you want to carry around with Crucial to this is the concept of holes so To think like this you have to kind of think of two diagrams at once You have to think of energy a diagram with energy going up the scale like we had in the previous slide and then a separate diagram with space If I move an electron from underneath that gap to over the gap using energy I've left behind a hole and that is the technical term a hole And if these this electron moves around a bit in space, but at that high energy It's moved away from where the hole was and the hole now looks like a separate entity And it turns out that statistically you can treat these holes just as particles We call them quasi particles with positive charge So in a semiconductor you have charge which is carried by free electrons in the conduction band and charge Which is carried by free holes in the valence band, which are just the bits left behind when we promoted the electron This is all very mind-bending. I hope it'll become clear sooner But we have to get this this stuff out of the way if I can tell you how a camera works, right? Forget that The the really interesting thing about how electrons move in a semiconductor and this is something that most people won't tell you Introductory, but I think it's important is that We're probably a lot of us familiar with the f equals ma equation there newton's second law from from basic physics in free space or in air to some extent If I put a force on a particle it will accelerate that means it doesn't just stay the same velocity It gets faster and faster as long as I maintain the force, right? The same is not true if I get a ball bearing and try and push it through some honey I can push with a lot of force But what will happen is the ball bearing will reach terminal velocity through the honey And then if I push it harder it will actually cavitate the honey around it, right? It won't actually make the ball bearing faster and you can verify this in a very messy but easy to do experiment and the interesting thing about Electrons in silicon which you need to get your head around before you truly understand it is that that is how electrons move in silicon That is the difference between in silicon and just electrons fired through a vacuum tube Electrons fired through a vacuum tube. We put a force on them. They'll keep accelerating up to relativistic speeds in Silicon you put that force on they are moving through honey silicon is like honey for electrons, okay? They are moving through honey. They will reach their terminal velocity straight away And then they just drift and we call it drifting. That's why we call it drift and There's a parameter between how much force you put on and how fast it goes called mobility And so higher the mobility the faster your electrons are moving, right? But they're not moving like in free space where you just accelerate them So that's how electrons move around silicon and how we move electrons around silicon Once we want to start making devices Interesting useful devices. We need to think old doping Basically what doping does is it introduces either excess electrons or excess holes locally into the silicon So if I replace one silicon atom with a phosphorus atom I get because phosphorus has five atoms in its outer shell silicon has four we get an extra electron there And if I replace it with a boron we get one less electron there, but if you think about it one fewer electron is one extra hole, right? so The really interesting thing about silicon is not just that it's a convenient semiconductor that we can use at temperatures that we live at But also that you can introduce these handy other elements which happen to work very well and bond very well into the silicon lattice Which give us extra carriers either holes or electrons as you see fit basically now Technologically, it's very difficult to dope something below about 10 of the 12 percent of me a cubed So pure the purest silicon we can make has something like 10 to the 10 Impurity atoms per centimeter cubed which sounds like a lot, but that's one impurity atom per hundred billion silicon atoms So it's not that many, right? But but yeah, so so introducing these these Dope and atoms allows us to control the density of electrons locally in the silicon and that's very important The most important thing to understand about how some is about semiconductor physics is a pn junction I recognize I'm going through this very fast This is to give you guys if you want to understand this This was explained to me in another graduate physics class about 10 years ago. They're just over and I Thought I understood it when they first told me it and then about six years later. I suddenly realized they understood it so Don't worry if this doesn't make sense at first An n-type bit of silicon is just one that has extra phosphorous in and a p-type is extra boron So we have extra electrons on the left side an extra holes on the right side So imagine you have two blocks of stuff. This one has extra electrons in this one has extra holes in we're gonna put them together Now this is similar to what happens if I spray a can of air freshener in the corner of the room Really hard if I spray an entire can of air freshener into the corner of the room at first You can't smell the air freshener Okay after a short while these guys here can smell it and eventually everyone can smell it But the smell gets diluted. That's a process called diffusion And it's basically a statistical process if you have a high concentration of anything Like if you put a blob of dye into some flowing water, right? At first you can see the dye just just sort of the color Forming eventually it spreads throughout the entire volume of water So electrons in silicon move in a similar way and that's because they have this honey like motion They can't just be accelerated around they drift around So if I put a block with extra electrons next to a block with extra holes What happens is the electrons start going towards where there are fewer and the holes start going towards where there are fewer and it Collide in the middle and when the electrons and holes collide because one's just a gap for the other one they can recombine and and disappear essentially and The electrons don't disappear physically. What's happening is that electron is falling into that hole Okay, so the electrons aren't disappearing physically, but the effect of that If these electrons and holes diffuse into each other and then combine is that you leave behind unscreened Ions unscreened nuclei, which are positive, right? So what's happening is electrons on holes are diffusing in just because there's more of them over here and more of them over there they diffuse in towards each other they're recombining and The result of that is that in the center You have an unscreen some unscreened carrier atoms and those unscreened atoms because they're big fixed Charges cause an electric field and that electric field pushes back on the carriers, right? So the carriers naturally want to diffuse in But by diffusing in the annihilate that annihilation causes a field which pushes back It's very complicated, but that is the basic idea once you truly understand that you understand everything about semiconductor physics That's important. Right. I mean there's obviously people do career I mean I have a career in this right, but that's the basic idea if you can understand that give it five years Because there's a few complications, but if you can understand that you you have a grasp on everything that region in the middle Where and so so obviously the more extra electrons that were here and the more extra holes were here the faster They diffuse and the faster they diffuse the more field it takes to push them back So you can control the size of this region by how many electrons and holes there were and that's very important This region is called a depletion region and it's the most important part of how a camera works So this depletion region has a huge amount of fixed positive charge and That is creates a field especially if I bias it put a voltage across it So if I just drop an extra electron in there that electron is going to zoom off to one side depending on which way I put the voltage in right and That is basically how we collect electrons in an imaging sensor We deplete a big volume of semiconductor and then the photons come in kick extra electrons out of nowhere Into that depletion region and they get sucked up towards where we put the voltage Right, it's like a vacuum cleaner once you have a depletion region You can stick a voltage on it and you can hover those electrons up towards that voltage And once you can hover up electrons in a spatially resolved manner you have a camera and Hopefully that's kind of you can see that Forget the most capacitor the the first I'm getting on to how you actually build these things now It's very important to talk about the ccd for two reasons three reasons First reason is because it was invented. It wasn't invented first, but a successful one was built first in 1969 ccd chance for charge coupled device you see why in a second The second reason the ccd is important is because these guys won a Nobel Prize for it in 2009 Certainly the only Nobel Prize that will be one for building image sensors. I think along with a guy who Was one of the grandfathers of fibroptic communications The third reason it's important to talk about ccd's is because they are still widely used in scientific applications for various reasons But they are not used at all whatsoever in commercial products anymore almost none The only commercial ccd vendor left, which will sell you a ccd is Sony and they are shutting down their ccd fab in 2018 What are we going to do? Basically there are reasons for that the reason why ccd's are no longer really manufactured commercially is because if you look at that diagram On the top left that's how a ccd is constructed. You require overlapping bits on the top overlapping gates There is no other modern IC integrated circuit technology that requires those overlaps so modern CMOS fabs just cannot do the overlaps They could do if they threw a few billion dollars at it, but why if you're only doing ccd? So what we actually have nowadays is the thing called a CMOS active pixel sensor Which is it works similarly I'm gonna explain that but it's slightly different on the left You have how a ccd is constructed on the right you have how an aps is constructed in an aps You have a readout in every pixel in a ccd. You just have one readout and we shuffle electrons towards it. I'm gonna describe that ccd is important the way it works is We have buckets these capacitors like I said you put voltage across above a depletion region You can suck electrons up and if we have arrays of of gates We can turn this one on suck electrons to here and we can turn this one on and turn this one off And the electrons are now here So a ccd is very much like those games where you have board of wood and Magnet under it and like a ball bearing or a disc on the top and you can move the magnet under under the under the wood And it moves the ball bearing around on the top. That's exactly how a ccd works But with electrons and with voltage So what happens is we build a big array of these capacitors. They're called MOS capacitors The photons arrive we've biased it put voltage on it So the electrons get sucked up one of the capacitors and then we shuffle them out to the edge of the chip Where we have a handy transistor, which looks a bit like that It looks exactly like that in fact, which was why I was a bit worried about this talk being recorded, but never mind No one's all Yeah, no one tell anyone right so how the ccd output works basically is you shuffle this charge along to your output You drop it on a capacitor and a capacitor in silicon is just a bit of n-type on the bit of p-type Two details me to tell you how why that works, but and If you remember how a capacitor works Q equals c times v charge equals capacitance times voltage, right? So if we stick a voltage on that capacitor, this is a very delicate process It's not as easy as I'm making it sound, but you drop a bit of charge into it that voltage will change, okay? So it's as simple as that you drop some charge into the capacitor by shuffling it along The voltage on that capacitor changes and then we tie that voltage very ever so gently into the gate of a MOSFET and Once you've got the gate of a MOSFET at a correct voltage You can just hump a huge amount of current through the other side of the MOSFET and you've now got a solid signal That's amplified if you just try to put a hundred electrons through a wire Which is what comes out of a ccd low light conditions You're not gonna have a good time because a hundred electrons will just get instantly wiped out by thermal currents in the wire So the most difficult part of actually building these things engineering wise is to get those delicate few hundred electrons that you've collected ever So carefully in these buckets drop them in a capacitor and Couple it into a big meaty MOSFET or other power transistors are available, but I'm telling you now They're all MOSFET That that can get that can really drive some current through and then we just measure that voltage And so now we know the voltage we measure out the other side of those that MOSFET is Related to it's not generally exactly proportional to but related to how many electrons we dropped in the capacitor, right? So it's a capacitor into the gate of a MOSFET current comes through measure this voltage tells you how many electrons you had very simple a Few extra details to build a full image sensor might want to color one I don't know why you might do and To do that what you do is you just have neighboring pixels which have optical filters on top of them red green and blue in a pattern Which is the most common one is called a Bayer pattern and so There's all sorts of interesting algorithms for how you turn this into a proper color image because you actually only got one image It's red one image is green and one image is blue and but underneath that it works exactly the same way So far what I've said is is you know how these things work There are obviously difficulties in actually designing the thing If you're someone who wants to design a camera chip a sensor chip for for any kind of optical stuff You have to be concerned about getting those photons absorbed now Silicon is not that reflective As you can see there actually at 400 nanometers It's about half of the photons you fire on it bounce straight back So you've got to sort that out somehow you don't want to lose half the light before you start We use that with a thing called anti reflection coatings which are black magic and no one has a clue how they work The other thing you've got to be concerned about is if you want a camera that can detect very well in the reds or the infrared You can't go too far into the infrared because once you're below the bandgap You just won't be throwing any electrons across it at all, but The absorption depth in silicon that means how far a photon travels through silicon before it on average gets absorbed in infrared Starts to get out into the centimeters and near infrared So if you want something that can see infrared you want a big thick piece of silicon and to give you some context by really really thick I'm talking about a hundred microns But that's a lot thicker than most other types of chips Your phone camera is typically four or five microns of active silicon Which will get the blues nicely gets the reasonable reds nicely One trick that you can do this was started to be done in the 1990s Is basically build your chip as normal which is the left picture there front illuminated The problem with that is especially on a CMOS chip you have layers of metal on top of the silicon So you've got silicon and oxide layer six layers of metal sometimes and in between those are other oxide layers The problem with that is if you've ever looked at a sheet of aluminium you'll notice it's not very transparent so Even a very thin sheet of aluminium is not very transparent most photons don't get through it and we're talking about 20 or 30 nanometers thick typically So what you can do simply it's not simple But in principle is simple you make your chip as normal You you you flake it off the wafer of silicon flip it over put it in the package that way and hey presto Now your photons are just seeing silicon Very simple idea and they thought of it in the 70s, but it took 20 years to actually build one that worked at all But if you do that you instantly get a huge boost in in both the blue and the red on on on how much light you can capture Yes, okay Yeah, thanks. So interestingly there is a process Related to Poisson statistics, which is actually a very simple process to talk about it to anyone who's interested where you can Calibrate your camera in electrons You can always calibrate any camera in Units of electrons and that's because they're quantum mechanical particles and they're subject to Poisson statistics So you match the Poisson statistics with what the numbers you are seeing you can calibrate it in electrons So I can take a camera that camera with a bit of messing around providing get the raw pixel output and tell you how Many electrons and therefore how many photons arrived at it? That's like this is incredible. I mean this is how modern astronomy works It's kind of a dirty little secret, but that's how you do absolute light measurements nowadays So I thought that was interesting. I'll just go for a couple of bits this on the bottom left is a picture of where I work now I'm not in this picture. Thankfully because the hair netting is atrocious We're working on Atlas silicon strip detectors are built there some of them With without a big particle physics experiments, you're not generally concerned about light particles you're concerned about big hump and grit like neutrons and muons and stuff that have mass and They can just blast blast through a piece of silicon like a bullet through a bit of MDF and just dump a huge amount of charge Into the silicon instead of just one Electron for a visible photon But after you've done that the principle of collecting it Sucking it up from a depletion region reading it out by dumping it on a capacitor is Identical to the camera in your phones You've got to be a bit concerned about radiation damage because you're basically putting a thing with the energy of speeding train through these Hundreds of times a second, so it's a bit more difficult, but that's how that works medical imaging things like the MediPix sensor which which we also work on is Interesting that doesn't use silicon it use the thing called mercury cadmium telluride normally which is which is a different semiconductor It turns out if you bond a semiconductor to a bit of silicon So you do the readout on this bit of silicon you do the collection on that bit of silicon You can optimize both separately and get much nicer images and finally the bit I work closest on is the LST camera This is a 3.2 giga pixel camera consisting of 21 sets of 3 by 3 CCDs each of which is 4,000 by 4,000 pixels So it is quite large It's about 5 foot 5 tall That's the focal plane of the camera is about 5 foot 5 tall the block down there in the bottom right is Paul Conor the guy who designed it or at least designed the focal plane sensors part of it These are done with CCDs Because CCD we've been making them for 30 years. They're nice and linear you can make them very thick 300 microns if you want and they collect electrons beautifully and That will do and thank you very much if anyone wants to talk to you about any of this anytime I'm more than welcome to chat about it