 Professor Gaim gave us an excellent discussion of the basic physics of graphene. I'm going to focus on applications, applications of graphene in electronics and of the electronics. From the physics that we've heard, they're all important. But I wanted to stress one important point for electronic devices. So that involves the client handling that you've heard. Because it affects the most important property of today's electronic devices, namely the confinement of electrons by electrostatic means. And we start with graphene, let's say it's end-opt electron carriers. And then we apply a gate field. So we produce a p-type section on the graphene. And then on the other side, we have the electron, again, end-type graphene. So we have an end-p, end-type of junction. We heard about pseudo-spin, which, like real spin, is conserved. In this situation, we have 0-band gap graphene. So it's a case of tunneling without a gap. And it conserves pseudo-spin, which is, again, you don't actually need the pseudo-spin. It is based on the orthogonality of the pi and pi star-wave functions of graphene. So an electron from the conduction band on the end side can very nicely tunnel conserving pseudo-spin into the vellus band of the section and exit again in the conduction band on the other electrode. So we go back to this Gapless-Hindermann tunnel. It has been described a long time ago and used, actually, as semiconductors. Incident electrons on a barrier that's very sharp, as indicated here, would lead to full transmission. This is the basis also of the absence of backscattering in graphene. Other directions of electrons incident on the barrier will be reflected. And what the actual transmission coefficient would be depends very much on the shape of this barrier. The sharper the barrier, the easier the transmission up to 100% transmission. As a result, you cannot confine electrons by electrostatic means by a long-range potential. So if I look at the next slide, if we look at the current versus voltage, gate voltage of a field effect device, we get the characteristic v-shape by v. This is because graphene is an ampipolar device that is depending on what kind of bias we apply on the gate. We can either have hole or electron conduction. But the most important consequence because of the client tunneling, the gate can never turn off completely the current. There will always be a leakage. And of course, this is very essential because if you cannot turn off the device, you cannot have a digital switch. So you cannot do what, say, silicon does or gallium arsenide or other conventional semiconductors. Typically, the neutrality point is shifted from zero gate bias. And that's again because Professor Guy mentioned the existence of electron hole puddles, the fact that the insulator supporting the graphene has trapped charges. And that gives rise to the minimum conductance, which, again, is a function of the trapped charges in the substrate. So the substrate has a very profound effect on the gating characteristics of graphene. A different view that engineers have in describing graphene is actually that of a material with a band cap. And that is based on the fact that there is an angular dependence on the scattering. So you can describe graphene as having on the average a band gap. And this is a trick that electrical engineers and device physicists use to apply all the developed theories for conventional semiconductors to graphene. You consider that you have a valence and conductual band with an effective gap. I just bring it up just to show that there are other ways of thinking about graphene. So we already heard about the fantastic properties of graphene. And here is a set of properties that are of particular interest in electronics, high mobility, the fact that you can tune the current as we saw by a gate field, ultra-themed body of the device, the ability to carry very high current densities, excellent thermal conductivity that's very important, and a large optical absorption. But I should stress that when we are interested in the intrinsic properties of a material, we try to decouple as much as possible from the environment. Now graphene is a single atomic layer. It's all surface. So it interacts very strongly with the environment. When you do basic physics, you use low temperatures, you suspend the graphene, and so on. When you want to have applications, the application itself would determine what the environment should be. Of course, almost all devices work at room temperature or higher. They are encapsulated inside complex structures. So it is important to understand that interactions with the environment can drastically change most of these properties. Professor Gaim stressed that point. Here are just list types of scattering mechanisms, for example, that are important in transport. Scattering by impurities. And the impurities typically are not in graphene. They are on the substrate that you place graphene on. These are charged impurities. We use insulators, which are notorious for trapping charges. And they are long-range scatterers. They tend to dominate transport in graphene. Neutral defects, short-range scatterers, can be important. And of course, invalidate a lot of the things we said about abscess of scattering and so on, because they distort the structure, the local structure of graphene. Charge transfer between a droplet out of mode to the substrate and so on can induce doping. Change the chiroconcentration and scattering. Surface roughness can be very important. And then we have inelastic scattering mechanism. Of course, phonons of the graphene itself. Something that is usually neglected in discussions is the graphene has a very high optical phonon frequency, something like five times higher than that of Gali-Marsenite. So typically, the scattering in electronic devices as dominated by optical phonons is not very strong in graphene. However, we have new scattering mechanisms, such as coupling of the conduction electrons with the surface phonons of the polar substrate. They are always present, even in conventional devices. But the field is generated, the case exponentially away from the surface. And in a finite thickness electronic device, that field has decayed. But graphene is a one atomic layer, so it's right smack at that surface of the polar insulator. That leads to doping. So at the end, the mobility that you measure is determined by all by sum of all these scattering mechanisms. So if you want to have applications, especially in a complex structure, you have to evaluate and control all these different scattering mechanisms to get reproducible as you need for an application to resolve. The other very important aspect is bringing in the carriers and taking them out. Transporting graphene may be great, but how are we going to bring in those carriers and take them out? And that is the issue of contents. The traditional way of measuring the resistance of contents is shown here. You just make a channel of different lengths, and you measure the resistance. Then you plot the resistance as a function of channel length, and you extrapolate to zero channel length. This is called the transfer length method. And you find that for graphene, the zero length resistance depends very strongly on the gate voltage you apply. And you can see here contact resistance versus gate bias. Of course, the resistance is highest near the direct point. But on top of that, you also have a temperature dependence. Contact resistance in graphene is temperature dependent. The key conclusion is this is for a palladium metal contact. And as far as I know, actually, this is the smallest contact resistance observed so far, which is 200 ohms micron. That's a very large resistance. And if you go closer to the direct point, it can be kilo ohms. So the contact resistance is comparable to the resistance of the channel itself, and can dominate the performance. So let's see where this resistance comes from. We have graphene on an insulator. We deposit the metal contact. So the first thing that happens is graphene and the metal have different work functions. So it would be a change of charge. Charge transfer. So the graphene would be doped. As a result of that doping, there would be a dipole layer formation that the carriers have to cross. It's one barrier. They also have to propagate a finite distance under the contact that's the transfer length. And then exiting in the channel, they will find graphene that is differently doped because it hasn't been doped by the metal. So you will have a doping junction akin to a P-in junction in a semiconductor device. So at the end, these carriers encounter a number of blocks. And you can understand that looking at the Rolf-Andauer's resistance, contact resistance, which would be proportional to a transmission coefficient. This is the sum of these processes. And the number of modes, one of modes that carry the current, and that's determined by doping. And of course, the doping is here and different from there. So whichever is the smaller number of modes will dominate. The other important aspect is, depending on what this transfer length is, we'll determine how small you can make device. Because in electronics, you have a pitch for the device, it's a spacing. And if your contacts are required to be big, you don't gain much by making the active channel small. So you have to minimize both of them. And one can generate a simple model that will describe the propagation under the metal and the transmission in terms, describe it in terms of two lengths, the mean-free path under the graphene and a coupling length that depends on the interaction between graphene and the metal. And sometimes this is actually, you want to minimize that kind of ratio. Of course, in electronics, the game is scaling, making things smaller and smaller. And we talked about the binostructural graphene. We have a symmetry between electron and whole states. But if you take some graphene, in this case, the sample was from chemical depot deposition graphene with palladium contacts, you measure the resistance versus gate bias at a large channel length, 240 nanometers. You get an almost symmetric resistance voltage curve. And then you can start scaling it. And you see that it becomes more and more asymmetric. And eventually, you get at 40 nanometers a very asymmetric shape. And on top of it, you start seeing oscillations. As the length changes, of course, the transport mechanism changed. And at this point, actually, we are in the ballistic regime. So now electrons, actually, I shouldn't say ballistic. We are in the coherent regime. And now electrons can interfere and give us these Fabry-Perot-like oscillations. The behavior can be understood very easily, considering that palladium, which has high work function, p-dops graphene. So when we apply a positive bias, we create a p and p junction. And what I showed you before, that will lead to resistance. And the asymmetry comes from the contacts. So the electric characteristics are really dominated by the contact. And you can see that, visually, using near field source circuit, photo conductivity of a graphene palladium system, you measure the photo conductivity bringing in with a tip light source. And you scan it along the channel. You see the contacts here for negative gate bias. We have a biostructure that looks like these. So you have a p, p prime, p kind of biostructure. You apply a positive voltage. And you can see the formation of p and p junctions. Of course, depending on the detailed transport mechanism inside the graphene, the dependence of properties changes. For example, here I show dependence of mobility on carrier density. Ballistic, of course, we're talking about ballistic conductance proportional to square root. For columbic barrier scatterers, we have a mobility that's independent of the carrier density. But for impurity scattering, we have 1 over n. In general, the diffusive regime, the properties depend on what kind of impurities or scatterers you have. That brings to, again, the electrical engineer's point of view. Electrical engineers have been trying for a long time to understand shorter and shorter devices, particularly in 3-fives. And they have come very close to ballistic regime. There's always sun scattering. And we know about mobility. The normal mobility we all talk, which is proportional to the mean free path. But once you start having ballistic mobility, then they introduce a somewhat artificial term called ballistic mobility. And this is based on the fact that the mean free path can never be smaller than the channel length. Because you have the contacts, which are scattering. And they write the effective mobility in terms of Matheson's rule. So in that formulation of the problem, in scale devices, high mobility doesn't play a role. Everything saturates. So for example, if you start with a mobility, normal mobility of 10,000, the mean free path is 160 nanometers. So the ballistic mobility would be only 600. And the effective mobility would be about 570. If you go to 100,000, you still get 600 mobility. So this is something to consider. The effect of the contacts is paramount. Taking a graphene that's made by CVD and has poor properties, say a mobility above 1,000, and looking at the minimum conductivity as a function of the aspect ratio of the device with over length, you see that already at the aspect ratio of 25, which is about 50 nanometers, even earlier, we are very close to theoretically predicted ballistic limit. So if we are talking about scale devices, we don't have to worry that much about mobility. And there are some other reasons that we'll get to that. What we have to worry is, of course, getting good graphene. Large area wafer scale graphene. And there have been several approaches in the growth of graphene. One introduced back in 75 in the Phillips labs involves taking silicon carbide, heating it at high temperature 1,600 degrees, where silicon evaporates. And the carbon atoms that are left behind reorganize to form a graphene layer. And this is an AFM image of such a graphene type of sample. At the high temperatures used, typically, you get bunching of steps and flat terraces. And you can measure the whole mobility of such a sample. Here's mobility versus carrier density. And the whole mobility curve can be understood in terms of a combination of Coulomb and Sorter-Age scattering. The point is that we have a strong carrier density dependence. And you can get high mobilities at low carrier densities. But for electronic devices, you have to work at higher densities because you need to have at least milliamp per micron currents. Topography plays a very important role. And this is, again, the same image that I saw you. And the point is that if you, by making devices in different orientations within the steps, you find that a single step, this is about 10 nanometers high. This is a bunched step. Can it reduce resistances of the order of 10 kilomicron? So topography is important. Also, point effects in graphene play a role. They affect the temperature dependence of the carrier density I won't discuss is too complicated. But there are specific effects in silicon carbide with an energy of about 70 million electron volts. Another big technical problem is that graphene is inert. It's a polar inert. So if you want to deposit, as in other normal semiconductors, an insulator, it's very difficult. It doesn't nucleate properly. So if you want to put, for example, half-nume dioxide, here's 10 nanometers of half-nume dioxide. And you can see there are many, many gaps. Initially, nucleates at steps and defects. It's never very homogeneous unless you go to very high coverage. And of course, you want thin as possible films. So as a result of that, you need to do something to the surface, usually sort of prime it with something. This is one example. Early on, we used a polymer, a very thin seed layer of polymer, which covers uniformly the surface and then use atomic layer deposition to complete, and then you get a good surface. So the question now is, how do you use graphene in electronics? We know it's a zero-gap semiconductor. We cannot completely confine them. A typical on-off ratio of the current in graphene is of the order of 10, 20, depending on the quality of the graphene. But to make digital devices, you need at least 10 to the 4. So at this point in time, pristine graphene cannot be used for the kinds of things silicon can do. However, we have finite current ratio. We have high carrier mobilities and drive currents and trascoductance, which suggests that graphene can be ideal for analog applications, particularly fast applications, RF. And these applications are very widespread. They are growing these days because everything that involves wireless communications from cell phones to radars, sensors, biomedical imaging, security at the airport involves very high-frequency transistors and amplifiers. So there is a big demand for this. And high-end devices in that regime are not like silicon devices. They are very expensive. They cannot be produced massively except the lower frequencies. They are hand-selected and very, very pricey. So here's an example of a structure of a transistor for RF application based on graphene. We draw graphene on full wave first and make the devices. This was made out of silicon carbide, as I mentioned. And here are some typical results of graphene devices. At this point, we are operating at high frequencies, so we cannot use resistances and other things to describe the material because everything is mixed in capacitance, inductance, resistance. So we describe the operation of the device in terms of wave propagation, so-called S matrix approach. And we can define a number of metrics. For example, the current gain. And another important metric is the so-called cutoff frequency, F sub t, which is the frequency at which the current gain becomes one. That is, we stop having current gain. And if we measure the current gain versus frequency for here two devices, one is 550 nanometer channel length. The other is smaller, 240. You see first that there is scaling. That is, as you decrease the channel length, the cutoff frequency increases. This is a room temperature with a rather modest mobility of about 1,500. And already at 240, you can reach 100 gigahertz. That was significant because if you were trying to do that with silicon with the same channel length, you will probably get no more than 40 gigahertz. Now, the typical procedure, of course, is to start decreasing the channel length. And already silicon is at 22 nanometers, in order of mind to do smaller. But instead of doing that, I will go back. It's the same foil I had before. And I remind you this that there is a strong topography dependence. And try to make devices a little bit more carefully. So here we make the device to make sure that lies on the signal terrace, that no steps are crossed. And when you do that, then you look again at the scaling. And at the same length, 210 was 240 before, as close as possible, we see that now we have doubled the f sub t. So we have cutoff now at 210 gigahertz. And as I said, the currents are very good. We have drain currents of over 2 milliamps per micron. And looks promising. Another thing that needs to be done to increase the performance is to decrease these spaces, the ungated regions. They are the resistance. But we cannot overlap the gauge, the source, and drain because the capacitance, parasitic capacitance, will come in and slow the device down. So the fact that the graphene doesn't insulate or don't nucleate in graphene can be used to make something called self-aligned gating. And I want this to be technical. Another approach that can be used, that we started using, is graphene by CBD. So examples already. The advantage here is that, unlike silicon carbide, it saturates a monolayer. Only very small portions around the effects can become double layers. And the great thing about it is that you can now grow single graphene crystals, like this called graphene snowflake. It's inexpensive. But the most important thing, and you just use copper foils, you can get any copper foil, put it with a hydrocarbon, methane, ethylene, whatever, and make the graphene. The key advantage is that then you can peel that graphene and paste it on anywhere you want. For example, you can put it on a transparent polymer. Or in our case, you can put it on silicon wafers up to 8-inch wafers. And I think Professor Hong will discuss this technique probably very extensively. The next worry is heating, which is currently the major roadblock in silicon electronics. Power dissipation limits the growth. These days, you don't get more powerful chips because of the dissipation. As the density increases, you cannot cope with the power. And we have used Raman thermometry to look at these effects on graphene. And as you can see, as you increase the electrical power, you get very large temperature increases. And you can see it in the color mark here is we increase the drain bias how hot the graphene gets. And there's also some degree of anisotropy in the heating because, obviously, the middle gets hotter than the edges. And if you try to simulate the heat dissipation, you see, of course, near the contacts you have efficient dissipation. But for a long device, it is extreme heating. And for this particular geometry, we find that about 70% of the power is dissipated through the substrate. So that is a key question then. Where do you want to place graphene if you have the ability to peel it off and put it on anything you want? So the properties of the substrate, of course, if you want something for commercial applications to be readily available, compatible with wafer-scale coverage, you don't want to have charged traps, which are the main problem of silicon dioxide. You don't want to have something hydrophilic because water, oxygen, system on SIO2 and other polar insulators are the causes of doping. And you want, of course, high thermal conductivity. That made us conclude that maybe diamond type of films, SP3 carbon, which, like diamond itself, is known to have very high thermal conductivity being non-polar and having all these other properties is used extensively to cover memory, hard disks, and so on is possibility. So we try that. Here, again, results with CVD graphene. We are not very good at growing CVD graphene yet. So mobilities are over the order of 1,000. But here we saw the scaling for CVD graphene, 550 nanometers. We get 26 gigahertz, 140, 70 gigahertz, all, of course, at room temperature, and at 40 nanometers, 155 gigahertz. So you can see that we had no problems whatsoever with doping, or moisture, or traps, or anything. And proving the quality of the graphene, this looks extremely promising. The other thing is we looked at the temperature dependence of these devices. So we measured for the first time the frequency, the f sub t, from room temperature to liquid helium. And as you can see, there is hardly any difference in these devices for all lengths, which implies that graphene has another excellent property. There is no carrier freeze out. So you can use it, for example, even in space without having any deterioration of properties. So that looks very encouraging. Of course, cutoff frequency is not the only one, the only interesting thing. You have to have power. The applications that graphene has to displace generate power. There are amplifiers. For internet, we need stations that will be used to transmit movies, big files. And for that, you need power. And that is more difficult for reasons not interesting to graphene. But it's getting there. We're starting getting a unilateral power gain that is comparable to f sub t. How do we optimize the graphene devices and circuits? Basically, what you need is self-gain for graphene to be useful. And self-gain is the ratio of transconductance and output conductance. Unfortunately, these two quantities go in opposite directions. You want the output conductance to be 10 to 0. That is, you want this curve to be flat. You need the transconductance to go to infinity. So you want maximum slope. It turns out that this goes faster. So as in other semiconductor devices, you need current saturation. That is, the current has to reach a certain volume and remain flat as a function of drain bias. In all discusses. So what we need is good mobility, good gating to increase the transconductance. We need current saturation to increase the output conductance. We need good high-K dialectics to affect the current saturation. And as I mentioned, self-aligned juicers to minimize the so-called axis resistance. For circuits, you also require to minimize the contact resistance that enters in the gain, power gain. If you optimize the gating, then you can start getting current saturation that you need. And you also increase a lot the transconductance. This is just some simulations that show how critical the contact resistance is. If you have F sub t is a function of channel lens for different contact resistance, you see how fast the performance drops as a function of the contact resistance. Really dominates the performance. Then, of course, the next step is to try to integrate devices. Graphene is planar material, so it's easy to fabricate easier than some other material to fabricate individual devices. But then, in circuits, you have to fabricate not only the transistors, but all the passive elements, inductors, capacitors, and so on, on graphene. So there, you have to develop patterning techniques that can operate on graphene. And there are adhesion problems and other things. So just one example of making a unipolar frequency mixer that involves inductors and other components on graphene. And the frequency mixer is, of course, you put two frequencies in, and you generate some indifference frequencies in every radio and every place. Because you want to have a high frequency, but to process it at a lower frequency. So this is our graphene transistors, our symbol for graphene transistors. So you bring in one frequency and another. Here they are, the local oscillator and the RF. And you generate the sum and the difference. And now you have taken a 4 gigahertz frequency and made it 200 megahertz. You can easily process. The advantage of graphene in this case, this is not optimized by far. But the advantage of this particular design is that, first, it works even in heavily doped samples that don't so direct point. And it has superior thermal stability. It operates even at higher temperature without loss, unlike conventional semiconductors. So probably, I'm running late, I should jump to a little bit of the optical properties and how we can use it. Graphene, the many-body effects are not very strong. So we can imagine the optical absorption, a single-party transition, an interbund transition. We already heard about the universal optical absorption which is, we'll see to some extent, frequency-dependent. It's about 2.3% for normal incidence for free graphene. For multilayers, for energies above about half a volt, it's additive. So by looking at an absorption spectrum, you can get how many layers. And then we have also intramund transition, which is the drudet type of transition in the very far IR, that can directly provide you the carrier density of graphene. And then you have an additional property that first described by Professor Kim, the polyblocking, which is simply the fact that if you don't have states, and your thermal level is there, there is no absorption. So by tuning the thermal level by a gate, you can tune what wavelengths will be absorbed. The absorption spectrum of graphene over a very wide energy range provided by Tony Hines of Columbia is shown here. You see, of course, it says semi-metals that it absorbs over the entire region. Indivisible and near IR, the absorption is relatively constant, about 2.3%, as we said. Up here, you have a plasmonic effect to complicate the picture. But in general, it's high. For us, the most interesting part is the far IR, where we have drudet absorption that can get very high. If we take graphene, we see the characteristic drudet behavior with frequency. We can chemically dope it and increase it up, in this case, up to 40% of the incident light. And you can use it diagnostically if you want. Here is over in the mid IR, the absorption. From this type of spectrum, you can get the Fermi energy by the polyblocking, as I mentioned. You can dope it chemically. And you see that the polyblocking moves far out of the region that we can probe. So that leads to an increased absorption in the druder regime and a decrease in the mid IR visible because we have an oscillators trend. Some rule. For us, the first application we explore in photonics is that of photodetectors. Photodetectors, of course, look a little bit unusual for graphene given that it doesn't have a backup. But when we did photoconductivity experiments, where we put graphene, source and drain electrodes, but didn't apply a bias between them, brought in the light with an optical tip in a near field microscopy, and scanned the light over the surface of graphene and detected the photocardin. We saw that the photocardin was localized around the contacts. And if you think about it, we mentioned at the very beginning that metal in contact with graphene will create charge transfer. And that charge transfer, of course, will lead to band bending. And the band bending will create a local electric field. So near the contacts, if we radiate near the contact, there will be a band bending, which we can manipulate with a gate without applying a drain bias. So if we excite one contact, the electrical hole pairs that are produced will be separated by the intrinsic field, built-in field, and we get a photocard. And if we excite the entire device, we get nothing. Because the field acts in opposite directions at the two contacts, so there is no net field. But if you excite near one contact, then you get a photocard. So we decided to use this as a photo detector. Advantage, of course, of course, is because we don't need to apply a drain bias, we have no dark current and, therefore, no short noise. So we did that in measure photo conductivity. These are results of the photo response and photo responsivity for 1.5 micro light, which is used in optical communications as a function of the modulation frequency of the light. And the highest frequency we can measure now allowed with equipment we have is 40 gigahertz. And as you see, the photo response is essentially flat. A little decrease that you see comes actually from the cable's note graphene. And if we measure the photo responsivity, DC, and high frequency, essentially, they coincide. This tells us that the graphene photo detectors are very fast. My post-doc went back to Austria and tells me that through an optical technique, now he can measure up to 270 gigahertz with no problem. So the advantage here, of course, is that you have a universal type of photo detector that is responsive to just about every frequency. The disadvantage is that because of the single atomic layer, the current you get, the photo response, is not very high, cannot compete, say, with 3.5s. But also, we have the problem that we have to radiate near the contact. So the first thing we wanted to do is correct that problem. So typically then, if we have two electrodes, identical metal, the potential will look like this, the red, symmetric. And as I said, left and right will cancel out. But if you use two metals with different work functions, they will lead to different bond bending. And so we made structures like this in the digitated electrodes, one with a high and low work function metal, like titanium, another with high work function, like palladium. And then we have the gate, no applied bias. For an arbitrary gate, we have cancellation, again, of the photo currents. But tuning the gate, we can get to a regime where we have positive contributions only. And as a result, you can first radiate the whole area. So you have a large detection area. And you get a 15 times enhancement in the photo resistivity. This is not the end, of course, because one could first make more layers. But also, you can enhance the absorption of graphene coupling to plasmas, for example. There are many ways of doing that with silicon wave guides, noble metals, even to plasmas of graphene itself. But you can get very high absorption and performance. And to test this, we applied it to optical communications. So we used the graphene photodetector to detect an optical signal, again, at the communications frequencies and use eye diagrams to check how faithfully the interpretation of the optical data is done. And we, again, limited only by our ability to measure high frequencies, we could measure up to 10 gigabits per second internet rate with the graphene photodetector. So these are some of the things we are doing. I wanted to add some conclusions. I really feel that graphene can have many important applications in both electronics and photonics. The key advances, in my opinion, are the excellent transport and optical properties, the thinness. Something that is not stressed enough, in my opinion, is the possibility of low price and ease of fabrication. People always compare with Re5s and so on, which are established technologies but expensive, require MBE or other difficult systems, expensive systems. I think, particularly, CVD graphene can bring a tremendous abundance there. Another great advantage, which, again, conventional semiconductors don't share, is integration with silicon technology. Graphene cannot be used in isolation because you don't have the digital part. But it can be integrated with silicon technology. The critical issues is material. We need high quality, large area, homogeneous, and low price graphene. We need control toppings. We need low resistance contacts and gates and good insulators. And also, I believe that we need more science, particularly science of graphene and interaction under what I would call real-life conditions. Anyway, I would like to thank all of my collaborators to remain unnamed. And also, IBM and DARPA for the support of the project. Thank you very much. You mentioned that mobility of electrons in graphene can be measured at small distances, or circa 15 nanometers. Is it isotropic or may depend on the direction? I think it would be isotropic. I mean, first, I didn't say that, really, it would be measured at small. It doesn't have physical meaning. Because in very small dimensions, basically, it's ballistic. The concept of ballistic mobility is sort of artificial. But I don't think where you can measure mobility, I would be amisotropic, unless there is one report, at least. But it's a very structured surface. And silicon carbide, people have done measurements with tips, metallic tips, to put down the tip at one point and circle over the other. And define anisotropy. But that is because the impurities in the silicon carbide are anisotropically distributed. They tend to collect near steps. But intrinsically, graphene should not be anisotropic. I'm Norbentland Imperial College London. The FT values you were reporting for the transistors are very impressive. However, they're still well below the values you can achieve in gallium, arsenide, indium, phosphide. What is the physical limitation? I mean, you said that the mobility doesn't play a role. It's limited by the geometrical dimension anyway. It's a matter of time. You're comparing a technology that developed over 40 years versus one that's less than a year old. Nothing is optimized. And there are still issues with the substrate, the gates. I think the insulator is a major problem still. And the quality of graphene is not reproducible and up to par. So individual devices, I don't see any intrinsic limitation. But that's why I stress, in my opinion, the goal is not to go to 400, 500 gigahertz. There are reports of germanium devices going to 700. The applications, they are very minimal. May I add one more question? Is it a realistic vision to think about, let's say, RFID, plastic electronic graphene or is that? Yes, yes. My feeling is that the major advantage of graphene is simplicity and price. And the sweet spot, if you want, for applications is in the range of 50 to 60 gigahertz. There is a market for very high frequencies for military applications. Don't forget that the propagation of electromagnetic waves is limited by the atmosphere. For point-to-point secret communications, you can use very high frequencies. But for long distance, you can't. And the frequencies are controlled. So I wouldn't say that graphene has a future in cell phones because there you cannot improve anything really. Everything is controlled. The frequency, the power you can use because of health effects. So it is primarily internet, diagnostics, terahertz for imaging. I can't tell you all our plans, but I don't think the search for higher and higher frequencies is the ultimate goal. Because companies have to make money. And if the market is small, although needs develop as potential disappear. What is special about plating as a contact? Is it just high-density states? Well, you need a metal that is stable. You don't want it to be oxidizable. You cannot use copper or anything like that. So it's noble. It has a high work function. It sticks well to graphene. I mentioned about charge transfer and so on. But most metals will not stop at charge transfer. We'll actually hybridize with graphene and make it an insulator. So by trial and error, we came up with as the best metal for us. People have also used titanium. We have used titanium, too. There are other possibilities. But I think for us, the optimal metal, also price. Gold is not good. And it's had these days extremely expensive. So palladium is the best compromise. Nathaniel Riem from University of Surrey. You mentioned a lot of potential applications in your talk so far. Which one excites you personally the most at the moment? Which one are you most excited about at the moment? I don't know. I came from the optics field. So maybe interest of more in the optics intrinsically, but I work for IBM. And IBM is not into optics. And I'm supported by DARPA, which interests at only high frequencies. So what I'm interested in and what the reality is, working in the industry is different than in university. More focus, University of Sheffield. I'm trying to understand the optical absorption spectrum you showed from Tony Hines. So if I remember the one which, as mentioned in Manchester, it was more or less completely flat across the visible spectrum. If that's, I don't know if Andrew can confirm that. And yet the one you were showing from Tony Hines is clearly as a big absorption in the blue. So what's the difference? I hope you have a very large energy rates. The flatness appears indivisible in the RIR. The spectrum you showed was clearly in the blue. So it's already the absorption is going up, or what? You don't have a single part of the transitions over there. It has been described by some theorists as a result of excitonic interactions. There may be there. But traditionally, that part of the spectrum has been considered as plasmonic. Yeah, we did measure before Tony Hines. It's in T-Zero-R-B or a white range. And we see the same, essentially. Some samples show flat and visible frequencies. Other, we're all the peak, but the peak is there. Sometimes narrow, then it gives flat. Sometimes we're all the look. Now that's a tall T-Zero-R-B. It was just before this many of our Tony Hines. That peak has been traditionally used as a probe of how doped graphite is. It shifts very strongly. And there are sensors based on that absorption. And as I mentioned, also the intensity by sound rule observation is a measure of the density, carrier density in both graphite and graphene. For the electron spectroscopy, it's used that peak all the time to test graphinization and so on. In the old days, in surface science, always you see the heels and other techniques. It's always there from carbon separated from transition metals and so on.