WEBVTT 1 00:00:05.110 --> 00:00:06.110 Mark Kushner: Your tenure 2 00:00:12.910 --> 00:00:13.749 Mark Kushner: can't do that. 3 00:00:14.190 --> 00:00:21.690 Mark Kushner: It gives me great pleasure to introduce our seminar speaker for today. Professor David B. Graves, at Princeton University. 4 00:00:21.930 --> 00:00:39.400 Mark Kushner: David received his Phd. In 10 points from University of Minnesota, working with Jensen, and he joined the faculty at the University of California, Berkeley, where he was for many years before recently 2,020 5 00:00:39.440 --> 00:00:47.610 Mark Kushner: joining class person, class of physics, research associate director and joining the faculty at Princeton University. 6 00:00:47.970 --> 00:00:54.529 Mark Kushner: David is an international leader in several areas of low temperature platform science and engineering. 7 00:00:54.620 --> 00:01:05.260 Mark Kushner: He's been pioneering contributions to our understanding of planet surface interactions for semiconductor fabrication through measurements and do modeling 8 00:01:05.630 --> 00:01:34.889 Mark Kushner: David in a totally different field and pretty much defined science base for class and medicine. The treatment of living tissue with atmosphere plasmas along the way. David has served in several leadership roles that are determining research policy and logical plasmus with service, with service. Our National Academy of Committees, and most recently as chair of the DOE, basic research needs project 9 00:01:34.910 --> 00:01:38.270 Mark Kushner: on platform science remind electronics fabrication. 10 00:01:38.650 --> 00:01:49.940 Mark Kushner: David has received several awards for his contributions, including the Abs. Alice Prize and the Plaza prize of the plasma Science and Technology division of the Abs. 11 00:01:50.130 --> 00:01:58.169 Mark Kushner: The title of David's talk today is modeling and simulation and plasma surface interactions can handle fabrication. 12 00:01:58.180 --> 00:02:05.190 Mark Kushner: But before they would begins we need to present them with the august. Let's see log 13 00:02:05.230 --> 00:02:18.160 Mark Kushner: that has the distinguishing characteristic of still fewer Lipsy Mug recipients and placid than Nobel Prize winners. And David. 14 00:02:21.580 --> 00:02:22.390 Mark Kushner: hey, Jensen. 15 00:02:23.300 --> 00:02:30.229 Mark Kushner: thank you, Mark. Thank you. Yes, this is this is very nice. My wife would be very happy for a note with another bug. She loves bugs. 16 00:02:30.370 --> 00:02:34.880 Mark Kushner: Okay, so thanks for the introduction mark and thanks for the invitation. 17 00:02:35.760 --> 00:02:44.399 Mark Kushner: yeah. So today I will talk about not the plasma, medicine stuff or plaza biology stuff, but the Nana fabrication stuff. 18 00:02:44.860 --> 00:02:48.689 Mark Kushner: as Mark mentioned. So my talk today is. 19 00:02:50.170 --> 00:03:10.969 Mark Kushner: oh, before I start. Yeah, I wanted to mention the just acknowledging mostly what I'm gonna talk about today are Md simulations from our own results. So Joey Bell is a postdoc with me. Former student, Dave Humbers, now been a consultant for many years, and it really helped us get back into this, because for many years we were kind of out of it. We kind of came back to the sending interest stuff more recently. 20 00:03:11.010 --> 00:03:20.390 Mark Kushner: As you'll see one of the things I'll talk about today are some comparisons to experiments that we did with Vince Donnelly and Shen Howe at the University of Houston. 21 00:03:20.450 --> 00:03:31.049 Mark Kushner: We have been working with people at lamp research for the last few years, Karen Cannaric and her team. We've had many discussions about Ali atomic layer etching with them, and I wanted to acknowledge them. And then and then the funding. 22 00:03:31.530 --> 00:03:48.360 Mark Kushner: Okay, so the summary of my talk, my talk is actually gonna be fairly simple. I'm gonna start with a discussion of plasma technology that's relevant to the semiconductor industry and how it's evolved from the the yesterday today and hopefully for tomorrow things have changed. 23 00:03:48.440 --> 00:03:52.560 Mark Kushner: and this has implications for what we study and using plasmas. 24 00:03:52.920 --> 00:04:00.220 Mark Kushner: So. And then I'll talk a lot about the second thing. How can we model or simulate and simulate plasma surface interactions, the atomic scale 25 00:04:00.770 --> 00:04:12.709 Mark Kushner: in a way that's that's convincing and believable. And we understand what some of the limitations of this of these approaches are. and actually a good bit of my talk today is to to show you that 26 00:04:13.030 --> 00:04:23.279 Mark Kushner: kind of relatively simple and straightforward assumptions and approximations can in some cases lead to predictions that are in pretty good agreement with experiment. 27 00:04:23.310 --> 00:04:45.739 Mark Kushner: And it's kind of a proof of principle to show this approach can work, and then towards the end, I'll talk more a little bit more about the challenges and opportunities as we, as we move forward, have actually had a number of really nice discussions today. There are a number of people here at Michigan who are working on related things. A lot of people doing animistic simulations are are grappling with some of the same issues. So actually, it's been quite fruitful for me. 28 00:04:45.770 --> 00:04:47.079 Mark Kushner: Okay. 29 00:04:47.300 --> 00:04:57.939 Mark Kushner: so, Mark, you should recognize this. I borrowed it from you, and it's kind of a busy slide. So let me go through a little bit. This we were using this for our workshop last year. 30 00:04:58.160 --> 00:05:27.129 Mark Kushner: So the the and the aster here is is to this report, that Mike Mark referred to that we put together. This, this was this workshop, and this report was sponsored by the office of Fusion Energy Sciences at the DOE, and we came up with a series. The group came up with a series of rec recommendations on priority research opportunities. And they're in this report which you can get through this URL, if you're interested in it, plasma surface interactions is part of that's an important part. But it's not the only part, as you can see. 31 00:05:27.580 --> 00:05:33.379 Mark Kushner: So the this this plot is this sort of Moore's law. 32 00:05:33.400 --> 00:05:48.430 Mark Kushner: plot! Something you've probably seen before on the vertical axis is a log scale of the number of transistors per chip, and then the the horizontal axis is just linear in year going from, if you can just read it, 1,970 out to about 2020. 33 00:05:48.540 --> 00:06:18.419 Mark Kushner: And the fact that this is showing kind of a straight slope is the Moore's law idea every couple of years that double or so the number of transistors per chip. And we're up to, you know, maybe tens of of of 1,000,000,000,000,000 transistors on the ship. So everybody knows this story, and everybody knows that this is the reason why transistors got really cheap, and there's more transistors on the plant than their cells and human bodies, or something. Once you're approximated. 34 00:06:18.460 --> 00:06:39.100 Mark Kushner: and the the line that that Mark drew here so below this line without plasma processing. This is starting right in the late seventies, early eighties but before that you could make chips like the one on the right here. This Intel 80386 chip, because the characteristic dimensions were on the order of microns, and you didn't need plasma to get that really precise etching. 35 00:06:39.580 --> 00:06:53.800 Mark Kushner: But after that time, in order to increase the number of transistors per chip, of course you had to etch these features much closer together, and the only way to do that is, with the directional and isotropic etching associated with plasma. Of course, coupled with lithography. 36 00:06:54.150 --> 00:07:13.030 Mark Kushner: So that's great. And from the eighties early eighties to today this progression has been enabled by plasmas without plasmus we couldn't have done it. There's no question about that. It's an enabling technology. There were other enabling technologies too obviously, lithography had to evolve as well among many others. 37 00:07:13.750 --> 00:07:35.339 Mark Kushner: but it sort of begs the question. Well, that's great plasma, you know, you did this great job in the past. Well, what's what's gonna what's important in the future. How should we think about the challenges as we move forward? And I borrowed from another colleague. This was Eric Joseph from Ibm research. In fact, I stole this from the presentation he made at our workshop. 38 00:07:35.460 --> 00:07:45.619 Mark Kushner: basically, he's making the point that scaling has now been replaced by innovation and materials and structures scaling being the idea that you just simply make 39 00:07:45.930 --> 00:08:03.729 Mark Kushner: the transistor's smaller and smaller every couple of years, and you can pack more of them onto the chip. So starting really with dimensions below, about 19 nanometers or so traditional gayness is this tan color. And this game by innovation is showing with the blue color. 40 00:08:03.740 --> 00:08:16.240 Mark Kushner: And so a lot of the the innovation that's occurred certainly below the 45 or 32 nanometer nodes has been in new materials and new types of devices 41 00:08:16.700 --> 00:08:34.869 Mark Kushner: and new device architectures. And and this is where the industry is going. This is how devices are getting better and better, cheaper and cheaper, more and more powerful rather than simply just scaling to to smaller and smaller sizes. And so we'll talk a bit about what the implications are for plasma in this transition. 42 00:08:35.419 --> 00:08:58.810 Mark Kushner: So just to remind you, what these field effect transistors look like, I should have noted my post, Doc. Or share with some folks at Ppl. Henry Wu, who used to be at Tsmc. Actually made this slide. So if if you're familiar with the field effect, transistor this a source, a gate, and a drain, and you you can, of course, the channel between just below the gate dielectric is where the electrons flow. 43 00:08:59.070 --> 00:09:20.439 Mark Kushner: You could sort of see how Planar processing is is was used historically to make these devices where you you deposit a film and then you pattern it. You etch it. You. Maybe you dope it or something. Then you can deposit another film and you pattern that. And you can see these structures are there. You know you can make this player by layer by layer. Fabrication. 44 00:09:20.690 --> 00:09:26.609 Mark Kushner: Now, what's happened is that this is on the left to this plant Planner futt where the gate links. 45 00:09:26.710 --> 00:09:45.489 Mark Kushner: The distance that the electrons or or holes would travel on the order! A few tenths of nanometers, and then so called, fin fets, fin F. Ets with 3 gates. The idea is, the gate is now wrapping around the channel on 3 sides, as opposed to to one surface. Those were shrunk to maybe 10 to 20 nanometer gate lengths. 46 00:09:45.490 --> 00:09:56.880 Mark Kushner: And now the industry is going to. So we're called gate all around transistors. And so the gate is all the way around this this channel that you can see here 47 00:09:57.040 --> 00:10:07.049 Mark Kushner: and here the characteristic lengths are on the 10 nanometer zone. So you see, for sure this is this transition in sizes that we're talking about. 48 00:10:07.330 --> 00:10:33.499 Mark Kushner: And I also, I wanted to use this picture as well that Eric provided for us from Ibm, the so-called Nanowire transistor. This is a false color image of the different materials. So there's there's there's silicon, there's crystalline silicon in the middle, there's sio 2 and hafnium based electric and tanal nitride out here and you can see the colors. But if you look at the link scale here, you know, that's 5 nanometers. So we're talking about devices that are on the order of a few nanometers now. 49 00:10:33.560 --> 00:10:41.110 Mark Kushner: So in addition to making things smaller, there, there are new materials being introduced as well. What are the implications for plasma? 50 00:10:41.640 --> 00:10:57.579 Mark Kushner: Well, the main implication is the increasingly important role of atomic scale processing. So when you're dealing with, with features like the size, interfaces and surfaces really start to dominate, and that has implications for plasma and a few more examples from Eric. 51 00:10:58.140 --> 00:11:05.830 Mark Kushner: I won't go into details on spacer etch, but going from a 40 nanometer half pitch to a 5 nanometer half pitch 52 00:11:05.880 --> 00:11:17.809 Mark Kushner: with double quadruple and octuple patterning. Yeah patterning. You could see that the the features are just getting much much, much smaller. This is from a Tokyo electron presentation in 2,014. 53 00:11:18.010 --> 00:11:35.570 Mark Kushner: Another example is this so-called self aligned contact etch, which I'm not gonna go into detail. But what they're pointing out here is the convention lets you get this etching on the shoulder, which is undesirable, but with what they're referring to as quasi atomically or etching they can improve the selectivity in that device process. 54 00:11:35.770 --> 00:12:01.979 Mark Kushner: And there's another one here. This is from Lam, where they showed that Ali atomic layer etching was giving them a much better flat edge front as opposed to the micro trenched features shown to the left, and then a similar effect with with this, with these materials and quasi atomic layer, etching not to go into great detail about the devices, but simply to say, you know, controlling things at the atomic level as as become much, much more important. 55 00:12:02.280 --> 00:12:07.329 Mark Kushner: And of course, as we introduce new materials that will also continue to be even more important. 56 00:12:08.130 --> 00:12:26.169 Mark Kushner: So just to remind you, this is a slide that I adapted from Sungbo, my son Samsung visitor. What we're talking about here, if you remember, maybe most of you already know this. You start with exposure and lithography. In this case extreme UV is going through a mask and exposing a photoresist 57 00:12:26.210 --> 00:12:39.100 Mark Kushner: and then this is developed you, and and you just have the the photoresist that was not exposed. And then the part that we're particularly interested in is a plasma. Etch you you you transfer that pattern to the underlying film. 58 00:12:39.320 --> 00:12:54.310 Mark Kushner: and then finally the strip and a clean. But the point is that what we're talking about here is one part of this of this process a very critical part during the etched step, when the plasma is in contact with these materials, what is happening at these surfaces? 59 00:12:54.920 --> 00:12:57.619 Mark Kushner: This is kind of the topic of of today. 60 00:12:58.980 --> 00:13:09.799 Mark Kushner: and it's, of course, true that plasmas are also used for deposition and cleaning. Just a couple of little cartoons here. This was kind of an ald atomic layer deposition conformal process. 61 00:13:09.890 --> 00:13:15.510 but but cleaning. So removing contaminants on the surface is actually becoming a really big deal 62 00:13:15.540 --> 00:13:26.179 Mark Kushner: both in the semiconductor industry as well as I've started collaborating with some folks at Princeton who are making quantum devices at least the quantum devices that are made within film processing. 63 00:13:26.460 --> 00:13:47.409 Mark Kushner: They're trying to leverage the technology in silicon cmos in these quantum devices. And there, even a single atom being out of place, I'm told, can act as a parasitic qubit, and so any roughness at the surface, any contaminants to the surface are really a problem they they control. They affect the the coherence times. 64 00:13:47.900 --> 00:13:51.760 Mark Kushner: And so plasma can be can be used for this as well. 65 00:13:51.960 --> 00:13:57.000 Mark Kushner: Yeah. And another little cartoon that I borrowed from Mark here, recognize this. 66 00:13:58.530 --> 00:14:02.750 Mark Kushner: I wanted just to point out to you that if we kind of condense down 67 00:14:02.990 --> 00:14:10.450 Mark Kushner: by the way, this is this is a sketch of a typical plasma etch reactor. Excuse me with with inductively coupled power at the top. 68 00:14:10.940 --> 00:14:25.829 Mark Kushner: and a couple of capacitive sources at the bottom, and the the basic idea, really, that I wanted to to point out here is that at the simplest level, what these low temperature plasmas are doing is creating free radicals by electron impact association 69 00:14:25.900 --> 00:14:28.189 Mark Kushner: at relatively low gas temperatures 70 00:14:28.340 --> 00:14:49.780 Mark Kushner: and ionizing the gas shown here in the Argonine. And then both of these species come down to the surface react in this case with silicon, and then an etch product. SICO, N is is produced. And so what what I'm really going to talk about today, and especially in the context of atomic layer etching is, can we simulate using 71 00:14:49.810 --> 00:15:00.500 Mark Kushner: molecular dynamic simulations, atomistic scale simulations? Can we simulate the details of what's happening at that etch point and understand what's what's going on. This is this is the task. 72 00:15:01.460 --> 00:15:02.130 Mark Kushner: Thank you. 73 00:15:02.850 --> 00:15:07.679 Mark Kushner: Okay, so let me talk a little bit about the indie simulation procedure. 74 00:15:07.860 --> 00:15:13.089 Mark Kushner: This slide, or what's sort of related to it, we could probably have multiple lectures on. 75 00:15:13.100 --> 00:15:26.819 Mark Kushner: and I don't want to go into a lot of detail on how we do the Md but the basic idea is the following. we are approximating a semi-infinite surface with a three-dimensional slab. 76 00:15:26.880 --> 00:15:30.759 Mark Kushner: and we're seeing the side view of it here. The little yellow balls are silicon. 77 00:15:30.830 --> 00:15:35.040 Mark Kushner: So the lateral boundaries, which are about a little over 3 nanometers across. 78 00:15:35.240 --> 00:15:42.760 Mark Kushner: They're periodic in the simulation. So anything in the simulation that goes out one side comes in the other side. Anything that goes out the back comes in the front. 79 00:15:43.200 --> 00:15:47.419 Mark Kushner: Now, if your layer's not large enough. that will give you artifacts. 80 00:15:47.560 --> 00:15:54.589 Mark Kushner: so you just have to make sure it's it's big enough laterally that that doesn't happen and you're simulating. Then a semi-infinite surface. 81 00:15:54.920 --> 00:16:04.979 Mark Kushner: There's no patterning here. By the way, this is what the industry would call blanket processing first, and then it's about 5 nanometers deep. The bottom 2 layers are fixed. They're not moving 82 00:16:05.600 --> 00:16:16.040 Mark Kushner: and if ions or neutrals coming in at the top penetrated down to those 2 fixed layers, you would have artifacts as well. So the layer has to be deep enough 83 00:16:16.370 --> 00:16:19.160 Mark Kushner: that that doesn't happen. 84 00:16:19.440 --> 00:16:26.640 Mark Kushner: But so if you do that, then this will simulate the semi infinite surface. 85 00:16:27.890 --> 00:16:41.689 Mark Kushner: Now, the way Md. Works some of you probably know this very well is you need to have a force field or an inner atomic potential. So every atom is feeling every other atom in the, in the, in its immediate vicinity. 86 00:16:41.900 --> 00:16:47.800 Mark Kushner: and by taking the negative gradient of that potential. You get the net force on every atom. 87 00:16:47.930 --> 00:16:55.510 Mark Kushner: and you simply solve in these classical calculations you simply solve Newton's equations of motion. F equals ma. and you move the atoms around 88 00:16:55.860 --> 00:17:21.249 Mark Kushner: so they're they're vibrating because of of thermal motions, and then if they're hit at the top, of course, they will move. The potentials have to allow for the breaking of bonds and the reforming of bonds. Okay, so we use a class of anatomic potentials called reboot reactive, empirical bond order potentials. These are algebraic expressions that basically say, if I'm a silicon atom sitting here. And I look at my neighbors. 89 00:17:21.640 --> 00:17:24.480 Mark Kushner: how far away they are from. I mean, what others around. 90 00:17:24.569 --> 00:17:36.310 Mark Kushner: I can. I can tell you what the the potential energy is for those interactions. So it has to be smart enough to know that silicon wants to form 4 bonds at a particular angle, and the bronze strength is a certain value, etc. 91 00:17:36.640 --> 00:17:44.869 Mark Kushner: So there's a bunch of parameters in these algebraic forms which you have to fit. You can fit it to thermodynamic data. You can fit it to dft calculations, etc. 92 00:17:46.140 --> 00:17:56.120 Mark Kushner: It assumes all the atoms are uncharged and in their ground electronic state it is not hard to imagine situations where these approximations are not correct. Okay? 93 00:17:56.600 --> 00:17:57.960 Mark Kushner: And so 94 00:17:57.980 --> 00:18:01.619 Mark Kushner: the more realism you put in the chemistry of the physics. 95 00:18:02.130 --> 00:18:06.569 Mark Kushner: the harder it is. Do the calculation. So you could do density functions here in every one of those atoms. 96 00:18:06.710 --> 00:18:14.330 Mark Kushner: and do so-called Ab initio. Ind. Right? You could do that. And today you can, but not very many atoms, and not for very long. 97 00:18:14.490 --> 00:18:20.270 Mark Kushner: and we have to simulate thousands of impacts on a layer this size to get realistic changes to the surface. 98 00:18:20.570 --> 00:18:26.059 Mark Kushner: So it's on us to show that these approximate approaches are legitimate. 99 00:18:26.180 --> 00:18:30.830 Mark Kushner: And that's a lot of what I'm going to tell you today is that I think in some cases I could, I could show that to you. 100 00:18:31.990 --> 00:18:33.420 Mark Kushner: There's another problem 101 00:18:33.590 --> 00:18:36.999 Mark Kushner: when you put the wafer in the chamber, and you etch 102 00:18:37.150 --> 00:18:40.700 Mark Kushner: you etch for minutes hundreds of seconds. 103 00:18:41.350 --> 00:18:53.830 Mark Kushner: When you do the nd simulation, and you advance the atoms. Your delta T. In your finite difference. Calculation is at least a fifth a second, although maybe even smaller. I was not here. 104 00:18:53.860 --> 00:19:04.279 Mark Kushner: in order to conserve energy, you have to take extremely small time steps. Now you're not taking 10 to the power. 15 steps to go. A second isn't going to happen. 105 00:19:04.760 --> 00:19:10.070 Mark Kushner: so you have to be smarter than that. So what we do is the following, and it's up to us to justify it. 106 00:19:10.570 --> 00:19:12.600 Mark Kushner: We say, well. 107 00:19:12.650 --> 00:19:15.600 Mark Kushner: when an ion or a radical come into the surface. 108 00:19:15.710 --> 00:19:22.310 Mark Kushner: their interaction time, the important stuff is happening within a picosecond or a few picoseconds. 109 00:19:23.570 --> 00:19:29.109 Mark Kushner: And then what we'll do then is we'll just basically say I'm going to jump a millisecond 110 00:19:29.770 --> 00:19:37.539 Mark Kushner: and assume nothing else has happened. Maybe something absorbs. But I'm going to pick another point at random on my surface and hit it with another neutral or ion. 111 00:19:37.710 --> 00:19:42.300 Mark Kushner: And I do this thousands to tens of thousands of times until we have a steady state layer. 112 00:19:42.430 --> 00:19:47.419 Mark Kushner: Now, you might say, wait a minute. In that millisecond or microsecond, or whatever 113 00:19:47.470 --> 00:19:53.740 Mark Kushner: things can diffuse things can react, etc. Things can. All sorts of things could happen. And it's true. 114 00:19:54.500 --> 00:19:58.189 Mark Kushner: What we have done is assumed initially that that's not important. 115 00:19:58.270 --> 00:20:01.659 Mark Kushner: And we're going to make some predictions and compare to experiment 116 00:20:02.450 --> 00:20:09.090 Mark Kushner: in some cases, for sure, it's important, those those processes. But this is what we've done so far. 117 00:20:09.140 --> 00:20:14.280 Mark Kushner: So the little schematic shows. Here we have the radical or ion impact. By the way, the impacts are all 118 00:20:14.390 --> 00:20:26.589 Mark Kushner: independent. If you take this 10 square nanometer surface, and you have a few milliamp per square centimeter to 10 milliamp per square centimeter. The ions impact that surface once every 119 00:20:26.970 --> 00:20:29.190 Mark Kushner: 10 or 100 microseconds 120 00:20:29.350 --> 00:20:34.579 Mark Kushner: that the iron currents are low enough, that the impacts are all independent. And the same is true for the neutrals. 121 00:20:34.740 --> 00:20:41.729 Mark Kushner: So each in practice, each one of those impacts is essentially independent of every other one. So you just have to accumulate enough impacts. 122 00:20:42.110 --> 00:20:50.760 Mark Kushner: And we we typically do a product sweep. So if there's weekly bound products, we'll remove them, and then we thermostat. So this is constant energy 123 00:20:51.170 --> 00:21:04.470 Mark Kushner: with with periodic boundaries laterally, and a fixed boundary at the bottom. So if I bring a 200 ev argon ion into that surface, that 200 ev. Is pretty soon within a few 150 s is shared with the rest of the atoms in that cell. 124 00:21:04.500 --> 00:21:10.799 Mark Kushner: But that's not physical. In reality that energy would go down to the backside of the wafer, etc. 125 00:21:11.470 --> 00:21:21.189 Mark Kushner: So as long as a cell is large enough and the atoms don't heat up too much in a few picoseconds, it doesn't affect the results, and we test it simply by running it longer with a larger cell. 126 00:21:21.620 --> 00:21:24.520 Mark Kushner: So for 200 ev ions. And below, this is about right. 127 00:21:25.620 --> 00:21:37.000 Mark Kushner: So we thermostat it back to 300 kelvin, or whatever we assume the wafer is, and we go it and do it again. And we do this thousands to tens of thousands of times. and then we get the statistics and we compare them to experiment. That's what we're doing. 128 00:21:37.680 --> 00:21:51.660 Mark Kushner: is it? Okay? But there's some papers here where we describe some details. Okay, so Joey Vela, last year this this paper came out this year, said, why don't we take our in atomic potentials once we fit them 129 00:21:51.910 --> 00:21:54.020 Mark Kushner: no more, adjusting, no cheating. 130 00:21:54.450 --> 00:22:10.189 Mark Kushner: and compare to some beam experiments that were done in the late 90 S. Jane Chang, when she was at at Mit with herps on, did some really nice work in a beam system. She shot radicals and ions at a silicon surface and then measured 131 00:22:10.340 --> 00:22:14.369 Mark Kushner: the etch yield. That is, the silicon removed per argon ion. 132 00:22:15.010 --> 00:22:27.990 Mark Kushner: And it turns out, if you plot it as a function of the ratio of the fluxes of the Radicals to the ions. You can do a Langmur Henschelwood type calculation, and you get this. You get this kind of plot that she's showing with the dotted lines 133 00:22:29.300 --> 00:22:39.899 Mark Kushner: the dark. The squares were at 100 ev. The triangles were 60 Ev. And the circles were at 35 ev. So Joey did the same energies. 134 00:22:40.290 --> 00:22:49.570 Mark Kushner: 35, 16100. And you can see the error bars in his open symbols. That's what the Md. Is predicting for the same energies and flux ratios and the and the agreement isn't perfect. 135 00:22:49.940 --> 00:22:53.349 Mark Kushner: But in our business it's pretty good. I'm fairly happy with it. 136 00:22:53.470 --> 00:23:04.429 Mark Kushner: plus I like to point out, although the experimentalists get mad when I do this is that their measurements of the fluxes and of the energies isn't perfect either. So who knows who's right? Who's wrong. 137 00:23:04.820 --> 00:23:15.519 Mark Kushner: And then and there's another one where Jane was looking at silicon and chlorine radicals, and organized at 100 ev. And a flux ratio of 600. So it was this point, like right here. 138 00:23:16.530 --> 00:23:21.149 Mark Kushner: And then she changed the angle of incidence of the of the of the ions. 139 00:23:21.510 --> 00:23:27.310 Mark Kushner: and you could see what Joe Joey's predictions were reasonably in agreement with her measurements. 140 00:23:28.840 --> 00:23:48.839 Mark Kushner: Here she did experiments, and also Tinichi, Tachi, and Japan also did these chlorine, atomic chlorine ions on Silicon, and you can see we plotted this now as a square root of ion energy. It's often linear. With that. And again Joey's Joey's results are shown here. And again, I want to remind you there was no adjusting parameters here 141 00:23:50.250 --> 00:23:55.980 Mark Kushner: and then finally, again, using these 100 ev results at 142 00:23:56.240 --> 00:24:01.730 Mark Kushner: for for the, for the chlorine to Argonine ratio. Jane also made measurements of chlorine molecules. 143 00:24:02.330 --> 00:24:08.970 Mark Kushner: and you see that the etch yields are lower, and Joey's predictions are in pretty good agreement with that as well. 144 00:24:09.820 --> 00:24:11.980 So I assert, based on this. 145 00:24:12.040 --> 00:24:18.939 Mark Kushner: that all of these assumptions and approximations that we made in doing these Md. Simulations, at least, for these beam experiments are not bad. 146 00:24:19.360 --> 00:24:21.720 Mark Kushner: and the results should be pretty reliable. 147 00:24:21.900 --> 00:24:33.719 Mark Kushner: I think that it's not completely clear in our community that, you know, we aren't cheating when we do this right? It really does work for for these conditions. Now, the plasmas are different. 148 00:24:34.260 --> 00:24:36.079 Mark Kushner: For example, there's photons. 149 00:24:36.430 --> 00:24:46.859 Mark Kushner: There's lots of other species present as well. But I like to use this comparison, to say at least in principle. This approach can give us results that are experimentally valid. 150 00:24:46.930 --> 00:24:50.350 Mark Kushner: and that follow the trends of the experiments fairly well. 151 00:24:50.870 --> 00:25:03.460 Mark Kushner: Excuse me, question this. This is maybe a really stupid question, but what produces the error bars on the simulate. Is it statistical noise? Yeah. So what Joey does was he? He starts at different points. 152 00:25:03.620 --> 00:25:08.980 Mark Kushner: So so when you when you, when you do the simulations, you just pick points at random on the surface. 153 00:25:09.010 --> 00:25:12.680 Mark Kushner: so he'll do like 3 independent simulations. 154 00:25:12.890 --> 00:25:20.650 Mark Kushner: and the numbers will be a little bit different, because there's a finite number of atoms. You get some fluctuations. And so he's just using the approximate standard deviation. 155 00:25:20.850 --> 00:25:23.240 Mark Kushner: That's what it is. Now, it's not a dumb question at all. 156 00:25:24.430 --> 00:25:29.279 Mark Kushner: Be, be, yeah, be free to ask a question. If anybody wants to. Happy to answer 157 00:25:29.960 --> 00:25:38.199 Mark Kushner: video. Okay, okay. all right. So let's talk about Ali. So we wanted to simulate atomic layer etching. 158 00:25:38.250 --> 00:25:49.889 Mark Kushner: and this little cartoon is intended to show you what people nominally think is going on with atomic layer etching. Now atomic layer etching like atomic layer deposition is broken up into 2 parts. 159 00:25:50.050 --> 00:25:55.509 Mark Kushner: The idea is the first chemical modification step is done with in this case, without a plasma. 160 00:25:55.560 --> 00:25:59.849 Mark Kushner: molecular chlorine comes in dissociatively, can absorbs and forms like a monol layer. 161 00:26:00.880 --> 00:26:14.719 Mark Kushner: You could do that with a chlorine plasma as well. but we're starting with the simplest possible case of just molecular chlorine. and then what happens is you get rid of the chlorine flow of the chamber, and you turn on an Argon plasma. 162 00:26:15.020 --> 00:26:22.170 Mark Kushner: and you impact that surface with organized at different energies. And if everything's perfect, you simply remove that chlorinated silicon layer. 163 00:26:22.360 --> 00:26:25.020 Mark Kushner: the idea being the silicon silicon bond 164 00:26:25.060 --> 00:26:33.550 Mark Kushner: between the first and the second layer of silicon has been weakened because the presence of the chlorine, and so in principle, you could remove simply the silicon chloride 165 00:26:33.760 --> 00:26:43.140 Mark Kushner: and not damage the underlying film. If the ions energy is below the sputtering physical, sputtering threshold of the silicon. 166 00:26:43.440 --> 00:26:44.759 Mark Kushner: Then this is possible. 167 00:26:45.070 --> 00:26:57.849 Mark Kushner: This is kind of the party line picture that people have now. It gets a little more complicated. People realize that it's never a simple model. But if you, if you look, they they'll send. Well, okay, there's a finite multi-model 168 00:26:58.360 --> 00:27:05.539 Mark Kushner: region. But the ions just remove that uniformly, and then they stop when the silicon's exposed. 169 00:27:05.910 --> 00:27:12.550 Mark Kushner: And as what I'm going to tell you today is that the Md. Simulations tell us that's not what happens at all. It's quite different 170 00:27:13.940 --> 00:27:14.790 Mark Kushner: now. 171 00:27:15.030 --> 00:27:20.579 Mark Kushner: we started working with our colleagues at Lamb research. Wonderful group. 172 00:27:21.060 --> 00:27:36.859 Mark Kushner: But they don't have diagnostics on their plasmids. These are commercial tools, you can't really measure what's going on. Fortunately, my old friend and colleague, Vince Donnelly, at the University of Houston, was doing precisely these experiments with Samsung support. And we said, Hey, let's get together. 173 00:27:37.110 --> 00:27:46.669 Mark Kushner: And so here's a sketch of his experiment. This is Kinchen. How is the grad student? So there's an inductive coil and dielectric plate. Here 174 00:27:46.830 --> 00:27:57.669 Mark Kushner: feed gases come in the top, and the silicon wafer. That's a piece of a silicon wafer on the on the substrate, which is biased Rf. Bias. You control the ion energy. 175 00:27:57.790 --> 00:28:04.000 Mark Kushner: and Vince has 2 major measurements. With a laser interferometer they can measure in situ the etch rate. 176 00:28:04.150 --> 00:28:14.769 Mark Kushner: they can see how fast the film is is going down. and then, by looking at the optical emission just above the substrate surface. you can make some conclusions about what are the species? 177 00:28:14.870 --> 00:28:18.140 Mark Kushner: What species are leaving during that Argonine bombardment? 178 00:28:18.510 --> 00:28:25.100 Mark Kushner: Both of these can be compared directly to the Md simulation results. And so it looked like a good way for us to further test 179 00:28:25.380 --> 00:28:26.350 Mark Kushner: our approach. 180 00:28:28.010 --> 00:28:34.390 Mark Kushner: So here's a little schematic of what they do. So there's the first 2 s 181 00:28:34.700 --> 00:28:44.750 Mark Kushner: the chlorine flow. The argon is always on at like 80 sccm. The chlorine's at 20. So for a couple of seconds they just flow CL. 2 into the chamber. There's no plasma 182 00:28:46.140 --> 00:28:59.570 Mark Kushner: The vince is measuring the the. I didn't show it here, I guess, but these these green squares are the pressure. You can see there's a millitore and a half or so variation, because he's changing the total flow rate. 183 00:28:59.910 --> 00:29:02.870 Mark Kushner: And so the pressure fluctuates a little bit, but not too much. 184 00:29:03.600 --> 00:29:09.009 Mark Kushner: and then, after 2 s, they turn on the plasma that the chlorine flow goes to 0. 185 00:29:09.240 --> 00:29:13.260 Mark Kushner: They turn on the plasma with a self-bias here of about minus 60 volts. 186 00:29:14.030 --> 00:29:28.050 Mark Kushner: So this 2 from seconds 2 to 5 is when the Argon ions are coming in and knocking off the chlorinated silicon, and then his step 3. They actually just turned off the bias power. But they still have a plasma and then step 4. Everything's off. 187 00:29:29.490 --> 00:29:34.930 Mark Kushner: And here are some sketches of his measured optical emission. 188 00:29:35.300 --> 00:29:39.320 Mark Kushner: So he's measuring 3 different products. Silicon atoms. 189 00:29:39.610 --> 00:29:42.519 Mark Kushner: SICL. Radicals, and SICL. 2. 190 00:29:43.180 --> 00:29:47.820 Mark Kushner: Now I will note that the silicon optical emission in the blue here. 191 00:29:49.110 --> 00:29:52.030 Mark Kushner: and this is the period where the argon ions are hitting. The surface 192 00:29:52.050 --> 00:29:58.700 Mark Kushner: has been corrected for dissociative excitation. It turns out, silicon. You'll get silicon emission from sicl. 193 00:29:59.560 --> 00:30:02.109 Mark Kushner: So when the Scl. Comes into the plasma 194 00:30:02.340 --> 00:30:15.909 Mark Kushner: plasma electrons hit it, knock off silicon and chlorine. Some of that silicon's in an excited electronic state when it decays back down. If it's a photon, it just looks like atomic silicon. So you have to correct you. You want just the the original silicon. So we did that. 195 00:30:16.600 --> 00:30:23.840 Mark Kushner: And what you can see is, there's there's this big peak of emission except for the silicon. and then it drops down, and then kind of plateau's off. 196 00:30:24.400 --> 00:30:38.650 Mark Kushner: This is step 3 where? The biases off with the plasmos on. I'm not going to talk about that section actually, Vince, that I don't understand what's going on here. But what we will do is we model this first part of the the chlorination 197 00:30:39.690 --> 00:30:43.480 Mark Kushner: of the silicon and the and D, and we model the Argonine bombardment. 198 00:30:45.310 --> 00:30:47.820 Mark Kushner: How do we do alright? 199 00:30:48.330 --> 00:30:58.220 Mark Kushner: So some details on how we start. I won't go into this in great detail. You've got to be a little careful about how many species hit the surface to relate that back to what's happening experimentally, etc. 200 00:30:58.480 --> 00:31:03.200 Mark Kushner: And we just actually, this, this paper just got accepted. Jvst. 201 00:31:04.100 --> 00:31:12.559 Mark Kushner: all right. So here's kind of kind of the punchline of much of the talk for the Ali part. So these are side views of the silicon layer. 202 00:31:12.600 --> 00:31:17.889 Mark Kushner: It was exposed to to chlorine. This is with no 203 00:31:18.660 --> 00:31:25.380 Mark Kushner: no Argonine bomb barbit. After several cycles. This is after one modal layer is 10 to the fifteenth 204 00:31:25.410 --> 00:31:27.250 Mark Kushner: ions per square centimeter. 205 00:31:27.520 --> 00:31:46.419 Mark Kushner: one monolayer on the cell is like 72 atoms. So a monolayer by influence is 72 Ergon ions that they did the service. and you can see what happens is, the chlorine from the previous sealed exposure is is segregated to the top. There's this amorphous region where there's some mixing. By the way, what joy has done here. He's made the silicon translucent. 206 00:31:46.440 --> 00:31:50.609 Mark Kushner: so you could see the chlorine deeper into the layer. The chlorine is the kind of the greenish pulse. 207 00:31:50.710 --> 00:31:55.929 Mark Kushner: and you can see as as the argon imabarbit increases, you're removing. The chlorine 208 00:31:55.970 --> 00:32:01.919 Mark Kushner: is not so much on the surface. And this yeah, this is a hundred ev, this is the top view. 209 00:32:02.130 --> 00:32:16.670 Mark Kushner: So at at 0 monolayers, basically, it's it's fully saturated. And by this 50 monolays of Argon bombardment most most of the chlorine has been removed from the top surface. But recognize how different this is from the from the from the cartoon. The original cartoon. 210 00:32:17.610 --> 00:32:19.220 Mark Kushner: which I'll show you again 211 00:32:20.600 --> 00:32:32.899 Mark Kushner: this one where you know. Yes, you can get something approaching this model layer adsorption of chlorine. But the Argonne bombardment doesn't just do this. It does a lot of other things, too. 212 00:32:35.250 --> 00:32:43.550 Mark Kushner: and in particular it mixes. It creates this amorphous region about a nanometer deep. and it mixes the chlorine into that layer. 213 00:32:45.140 --> 00:32:47.599 Mark Kushner: and this has implications, of course. 214 00:32:48.350 --> 00:33:02.280 Mark Kushner: So if you look at the silicon etch rate and the chlorine uptake, this is a series of cycles. There's 5 cycles. The red is the chlorine. Uptake the first cycle, so the chlorine comes up, and it kind of saturates. 215 00:33:02.410 --> 00:33:04.979 Mark Kushner: And then at this point the Argonians come on. 216 00:33:05.080 --> 00:33:09.830 Mark Kushner: and the chlorine coverage of the layer. By the way, this is the total chlorine in the layer 217 00:33:09.940 --> 00:33:17.989 Mark Kushner: expressed as a number per area. and this is the blue. The blue is a silicon etched. You can see that we lost about 2 angstroms of silicon 218 00:33:18.210 --> 00:33:22.449 Mark Kushner: and that period. Then you turn on the chlorine again. Plaza is off. 219 00:33:22.490 --> 00:33:24.099 Mark Kushner: goes up a bit more. 220 00:33:24.510 --> 00:33:38.130 Mark Kushner: Again it falls with the Argonine bombardment. We lose more silicon. more chlorine, and then the silicon is lost, etc., and after about 3 or 4 cycles, it's come to a steady, roughly, a harmonic, steady state, a periodic, cyclic, steady state. 221 00:33:39.080 --> 00:33:42.520 Mark Kushner: So this is our prediction of what's happening. 222 00:33:42.620 --> 00:33:45.470 Mark Kushner: How do we compare to the measurements? 223 00:33:45.710 --> 00:34:02.369 Mark Kushner: Actually, before I go there, I just want to point out that again. Kind of at the cartoon level. What's really happening, we think from the Md. Is that? Yes, you're getting silicon chloride sputtered off. But you're also getting atomic. Silicon sputtered. Off and I'll show you an atomic chlorine. This surprised our industrial colleagues. 224 00:34:02.380 --> 00:34:04.419 Mark Kushner: They were. They were not expecting this 225 00:34:04.850 --> 00:34:17.439 Mark Kushner: and you get this mixed layer where the chlorine mixes almost uniformly in this region, and then, of course, the whole thing is is going down steadily, as it's etching, so you can kind of think of silicon is kind of coming up from below. 226 00:34:17.540 --> 00:34:21.479 Mark Kushner: This, I think, is the is the picture that is emerging from the simulations. 227 00:34:22.840 --> 00:34:25.599 Mark Kushner: What about comparison to edge per cycle? 228 00:34:25.780 --> 00:34:41.359 Mark Kushner: So Vince and and Kinchen measured how much Argonne they lost, how many nanometers per cycle as a function of our Argonine energy for the fluances they were using, and you can see that's the the black symbols. The red triangles are Joey's predictions. 229 00:34:41.670 --> 00:34:45.210 Mark Kushner: So at the higher energies there's some disagreement. 230 00:34:45.260 --> 00:34:50.670 Mark Kushner: But I would note that we just assume a mono energetic ion beam. 231 00:34:50.980 --> 00:35:01.520 Mark Kushner: Mark can can tell you that with the Rf. Modulation of the iron energy in the sheath and some neutral scattering things aren't aren't so simple, so we don't expect it to be too quantitative. 232 00:35:02.790 --> 00:35:09.410 Mark Kushner: And here's a prediction of the edge products that come off during the Argonne bombardment. This is 80 ev. 233 00:35:09.950 --> 00:35:20.590 Mark Kushner: so these blue symbols are atomic chlorine. That was the primary thing that came off initially, which makes sense actually, because the chlorine's right at the surface, or comes in, sputters it off. 234 00:35:20.870 --> 00:35:22.880 Mark Kushner: but that falls quite rapidly. 235 00:35:23.530 --> 00:35:36.660 Mark Kushner: And this would be a a fraction of a second in the, in, the, in the experiments. The sicl 2 is the are these triangles? They fall off even faster. Sicl also drops quickly. 236 00:35:37.080 --> 00:35:47.590 Mark Kushner: And but what's interesting is, we predict the silicon atom sputtering to be basically flat across the whole organizable Mormon. just physical sputtering of atoms. 237 00:35:50.480 --> 00:35:55.719 Mark Kushner: So let's compare to Vincent Kensen's measurements. So here's the experiments. This is S. Icl. 238 00:35:55.950 --> 00:36:13.390 Mark Kushner: As a function of their argon bind dose at 3 energies, 45, 80 ev. And 215. You can see that initially. Here are the simulations. You get more coming off initially with the 200 Ev. Of course, but it falls rapidly. So the initial drop in emission 239 00:36:13.570 --> 00:36:15.179 Mark Kushner: matches really well 240 00:36:15.250 --> 00:36:17.770 Mark Kushner: with experiments. But there's this plateau. 241 00:36:18.170 --> 00:36:19.499 Mark Kushner: There's this plateau. 242 00:36:19.540 --> 00:36:25.710 Mark Kushner: We are not predicting this long tail of the oes of the products coming off the surface. 243 00:36:27.280 --> 00:36:29.579 Mark Kushner: we see the same thing with SICO. 2. 244 00:36:30.380 --> 00:36:34.960 Mark Kushner: The the initial rapid drop is is is well approximated. 245 00:36:34.990 --> 00:36:38.309 Mark Kushner: but then this this long tail of oes 246 00:36:38.640 --> 00:36:44.100 Mark Kushner: is not, is not is not seen in the simulations, and Vince and I went many, many months 247 00:36:44.560 --> 00:36:46.519 Mark Kushner: trying to figure out what the heck is going on. 248 00:36:48.160 --> 00:36:59.549 Mark Kushner: Here's silicon atoms. Actually, this was really quite interesting. Notice. It does not show the same thing because it's not surprising. The other products were chlorine related. So as the chlorine got spread away from the surface, the mission dropped. 249 00:36:59.720 --> 00:37:05.150 Mark Kushner: So what Joey did was he used the 80 ev. The green symbols 250 00:37:05.620 --> 00:37:23.850 Mark Kushner: the average. This plateau here as just arbitrarily, is one, and then the other measurements at 2, 15 and 45 were quantitative. With respect to that, that normalization. We did the same thing with the simulations, and you can. So you can see that going from one to 5 or 6 251 00:37:24.330 --> 00:37:27.190 Mark Kushner: we went and experimentally went from one to maybe 5. 252 00:37:27.410 --> 00:37:31.319 Mark Kushner: So that factor of 5 is real. It's a quantitative prediction. 253 00:37:31.380 --> 00:37:34.440 Mark Kushner: and the drop from 80 to 45 is also real. 254 00:37:34.890 --> 00:37:44.970 Mark Kushner: So the simulation's getting things really well for the silicon atoms, you might say. Well, big Deal, you're but you're you're simulating physical sputtering. You'd better get that right? 255 00:37:45.290 --> 00:37:46.390 Mark Kushner: Okay, we did. 256 00:37:47.800 --> 00:37:59.189 Mark Kushner: So what is the cause of this this difference in in the Md with the Emd and the experiment after 10 model areas of eye influence. Well, one of the things we thought about first was, Oh, Mom. 257 00:38:00.060 --> 00:38:09.030 Mark Kushner: we're only simulating. you know, a very small time of a few picoseconds of the molecular chlorine on the silicon. 258 00:38:10.090 --> 00:38:16.989 Mark Kushner: Yes, you get to dissociate chemies eruption very quickly, but maybe over a couple of seconds. That chlorine's diffusing deeper 259 00:38:17.150 --> 00:38:18.670 Mark Kushner: even at room temperature 260 00:38:19.530 --> 00:38:24.710 Mark Kushner: if you get it wrong. And there's actually more chlorine in the layer than we predicted. And that's the reason that that 261 00:38:25.020 --> 00:38:28.040 Mark Kushner: Vince and Kinchen see this tale of emissions. 262 00:38:29.350 --> 00:38:37.440 Mark Kushner: So that was one possibility. The other possibility is, it's well known in atomic layer etching. And you can get products coming off the wall during the Argon exposure 263 00:38:38.250 --> 00:38:50.090 Mark Kushner: and that's not included in the simulation. So of course, we wouldn't expect to see those extra chlorines or silicons coming in. And so what Vince and Kinchen are doing like right now, in fact. 264 00:38:50.260 --> 00:39:01.560 Mark Kushner: actually was literally yesterday we had our our weekly meeting, and what they're doing is changing the residence time. So they're keeping the chlorine to argon ratio the same, but just changing the total flow. 265 00:39:01.930 --> 00:39:07.979 Mark Kushner: and at higher flows. The residence time is shorter, should be sweeping those products out if they're coming on the walls from the walls. 266 00:39:08.670 --> 00:39:18.400 Mark Kushner: And what he did. The the first experiment was, they lowered the flow rate. So the resin time was longer, and indeed the peak was higher. 267 00:39:18.450 --> 00:39:28.450 Mark Kushner: and and the the exponential decline in the sicl and SICL. 2 were longer, which is exactly what you'd expect if that stuff was coming off the walls. 268 00:39:28.910 --> 00:39:32.600 Mark Kushner: So I keep my fingers crossed. I hope that's true. 269 00:39:34.290 --> 00:39:38.919 Mark Kushner: But let's we say, well, has it to joy. What if you just remove that plateau 270 00:39:39.280 --> 00:39:46.910 Mark Kushner: from the optical optical emission measurements and assuming it's just from the walls that we can get rid of with higher flow, and we'll compare that to the Md. 271 00:39:47.550 --> 00:40:06.160 Mark Kushner: And here's what we found. So I've I'm plotting this differently. Here's sic on the left, sicl 2 on the right. This is at 45 ev, this is an 80 ev. This is a 250 Ev. And I saw that I was almost like, Okay, this is almost too good. I would not expect the agreement to be quite this good again. It was just normalized the 80 ev results 272 00:40:06.270 --> 00:40:08.640 Mark Kushner: at one you can see here. 273 00:40:09.610 --> 00:40:16.980 Mark Kushner: So if we remove those extra species coming in, it seems this Md. Approach is is working pretty well. 274 00:40:18.020 --> 00:40:20.069 Mark Kushner: At least that's our tentative conclusion now. 275 00:40:21.220 --> 00:40:23.070 Mark Kushner: Oh, and actually. 276 00:40:23.210 --> 00:40:37.530 Mark Kushner: it turns out, chlorine emission is much harder to see, because the excitation to the excited state for the chlorine is higher, you need more energetic electrons, so they only did this one experiment at 70, Ev Agony's and W. When we compared to Joey's 80 Ev. Again, the 277 00:40:37.600 --> 00:40:42.140 Mark Kushner: solid symbols being the measurements. The agreement is almost surprisingly good. 278 00:40:42.860 --> 00:40:45.120 Mark Kushner: So it appears that it's it's getting this right? 279 00:40:46.730 --> 00:40:55.680 Mark Kushner: Okay? So I think what I'd like to do is just sort of kind of summarize the the picture here again to get it in your head. 280 00:40:55.690 --> 00:41:05.819 Mark Kushner: and II showed something similar to you before. This is chlorine. This is this is the chlorine for one cycle, the chlorine saturating a little below 1.5 bottle layers total. 281 00:41:06.530 --> 00:41:08.570 Mark Kushner: or it might look something like like this. 282 00:41:08.990 --> 00:41:11.800 Mark Kushner: And then when you turn the Argon ions on. 283 00:41:12.030 --> 00:41:17.230 Mark Kushner: you get the chlorine composition layer dropping. You see it from one to 50. It drops down. 284 00:41:17.620 --> 00:41:25.070 Mark Kushner: and we plotted this now with silicon units of model layers and impact number monolers. So the slope of that line is actually the yield. 285 00:41:26.230 --> 00:41:33.159 Mark Kushner: At the beginning, when the chlorine concentration is high, which is up here. the yield is about point 1 1 5. 286 00:41:34.130 --> 00:41:40.360 Mark Kushner: And then later on, when the chlorine concentration is low. When we're down here, we're doing predominantly physical sputtering 287 00:41:41.340 --> 00:41:45.890 Mark Kushner: point. 4 is about the physical sputtering yield for argon ions at 100 you need on silicon. 288 00:41:46.590 --> 00:41:50.440 Mark Kushner: So, the point being that the process is dynamic. 289 00:41:50.790 --> 00:41:52.559 Mark Kushner: the layers changing with time. 290 00:41:53.930 --> 00:41:59.889 Mark Kushner: And this is something I think, that's not fully appreciated in the industry. Again, that was the kind of the party line picture 291 00:42:00.050 --> 00:42:05.610 Mark Kushner: and the Md, simulation is giving us this, these significant subsurface mix. 292 00:42:07.260 --> 00:42:13.170 Mark Kushner: Okay, so I wanted to show this picture to you. This is actually what we call a snapshot movie 293 00:42:13.370 --> 00:42:21.640 Mark Kushner: where Joey took the side view of the layer after each argon ion impact and then just strung them together like a thousand of them or something like that. 294 00:42:22.110 --> 00:42:25.840 Mark Kushner: You've got to be a little careful. If you look at this, you might think that 295 00:42:26.030 --> 00:42:36.050 Mark Kushner: the Md. Is predicting that the layer is liquid like that's not true, because each and each impact is independent. So it's not really like a liquid. But it sort of is 296 00:42:36.150 --> 00:42:39.450 Mark Kushner: is your average over seconds of impact. But you can just see 297 00:42:39.530 --> 00:42:47.679 Mark Kushner: how the Argonine bombardment is is mixing this region of the silicon. There's quite a bit of motion to the silicon. The chlorine, of course, is eventually leaving. 298 00:42:48.210 --> 00:42:56.220 Mark Kushner: and I think this is this is a better picture of of what I am. Bombardment is doing, and in this this 299 00:42:56.700 --> 00:43:07.209 Mark Kushner: in this system. Okay, so let me give you some concluding remarks. So plasma assisted. Aol is intrinsically dynamic. The layer changes continuously during the cycle. 300 00:43:07.990 --> 00:43:14.870 Mark Kushner: The Argonne bombardment efficiently mixes the top nanometer. So this is something, I think people in the industry really weren't aware of 301 00:43:16.060 --> 00:43:29.309 Mark Kushner: this mixed layer concentration changes from chlorine, rich at the surface to well mixed with declining chlorine concentrations during the ergonomical apartment, set the edge shield starts relatively high, then gradually reduces to physical sputtering. 302 00:43:29.440 --> 00:43:44.059 Mark Kushner: The experiments seem to support these major conclusions about the edge per cycle and the edge products are in pretty good agreement, at least in the early part of the cycle. We think in the later part, it's just due to these these reactor effects we've we've not included in the Md. 303 00:43:45.360 --> 00:43:51.480 Mark Kushner: So let me just say a few things about some of the challenges and opportunities here, as I see it, at least. 304 00:43:51.590 --> 00:44:09.720 Mark Kushner: so if you're doing classical MDF surfaces which are on the order of these 10 square nanometers or so, a few, 5 to 10 nanometers deep. And you're following these picosecond timescale processes. You can capture many important processes in in Ltp surface interactions. 305 00:44:10.090 --> 00:44:13.259 Mark Kushner: What we call prompt processes can often dominate. 306 00:44:14.550 --> 00:44:18.019 Mark Kushner: however, longer timescale activated 307 00:44:18.160 --> 00:44:19.710 Mark Kushner: thermal processes 308 00:44:20.080 --> 00:44:22.640 Mark Kushner: can be important, including diffusion 309 00:44:23.040 --> 00:44:30.979 Mark Kushner: as can processes that require explicit treatment of chargers and electric fields and photons and electronic excitation. 310 00:44:31.240 --> 00:44:36.129 Mark Kushner: Those things are not included at all in these simulations, and we know that they can be important. 311 00:44:36.530 --> 00:44:39.799 Mark Kushner: So that's one of the frontiers in this field. 312 00:44:40.830 --> 00:44:50.409 Mark Kushner: I think this has come up several times today with my discussions with people interested in animistic simulations. The most important immediate need is a systematic and reliable way 313 00:44:50.510 --> 00:44:59.869 Mark Kushner: to develop suitable in atomic potentials or equivalently force fields for any combination of atoms. There are many new materials being introduced in industry that need to be simulated. 314 00:45:00.560 --> 00:45:10.280 Mark Kushner: And there's a whole bunch of different machine learning methods. People are in all different applications. They're really moving fast in this direction, and it could be a game changer for us. 315 00:45:10.520 --> 00:45:12.010 Mark Kushner: Course, we're hoping it will be. 316 00:45:14.410 --> 00:45:21.489 Mark Kushner: And finally, one of the things I've been dreaming about for a while now is to be able to do atomistic simulations of future profile evolution. 317 00:45:21.740 --> 00:45:31.339 Mark Kushner: So using this picture that Eric gave me gave us of the nanoscale transistor, the 5 nanometers across, you know, and maybe a little bigger than that. 318 00:45:31.500 --> 00:45:39.349 Mark Kushner: We're not that far from doing those calculations today, if we had in atomic potentials for the tantalum and the nitrogen. oxygen, so forth, but have to. 319 00:45:39.510 --> 00:45:50.260 Mark Kushner: And I think that's a frontier. It seems to me that the industry really could use these kinds of calculations and and understand a lot better. What's what's going on? I think it's a real opportunity for the field. 320 00:45:51.370 --> 00:45:57.370 Mark Kushner: With that I'm gonna I'm gonna leave it. Thank you very much for your attention. 321 00:46:01.100 --> 00:46:03.909 Mark Kushner: And thank you, David, are there questions. 322 00:46:04.920 --> 00:46:07.250 Mark Kushner: Got that? Okay? 323 00:46:07.860 --> 00:46:09.300 Mark Kushner: Sure? So 324 00:46:09.620 --> 00:46:22.200 Mark Kushner: the energetic argon Ion is coming in with enough energy to etch so just butter off does it not also have enough energy to potentially ionize the atom 325 00:46:22.580 --> 00:46:35.600 Mark Kushner: it actually has enough energy. But that's at these energies. It won't do it. Yeah, if you had an electron coming in at. you know, above the ionization potential. Yes, for sure. 326 00:46:35.740 --> 00:46:39.689 Mark Kushner: but but not not an energetic ion, I think. 327 00:46:39.960 --> 00:46:49.190 Mark Kushner: for like, and these are actually fast neutrals coming, but even if it were an iron it it would take it would take more energy. It's it's at least it's 328 00:46:49.250 --> 00:46:51.030 Mark Kushner: it's a good question. Well. 329 00:46:51.670 --> 00:46:59.969 Mark Kushner: so if the if the ion energy is high enough, you will get secondary ions, and of course you get secondary electrons, as you know, you can get that you can get emission 330 00:46:59.980 --> 00:47:06.419 Mark Kushner: from surfaces. This that's none of that's included in these kinds of simulations. I wouldn't call it ionization. 331 00:47:06.860 --> 00:47:16.059 Mark Kushner: Exactly. I mean, it's like emission of of electrons from the bands, or something like that. But yes, that can happen. Of course it happens all the time. 332 00:47:16.760 --> 00:47:19.719 Mark Kushner: Okay. can I see second one? 333 00:47:20.010 --> 00:47:30.440 Mark Kushner: I was surprised. 334 00:47:31.080 --> 00:47:39.369 Mark Kushner: this amorphous layer, so experimentally can they like probe just first nanometer on the surface and see like? Is it amorphous, and 335 00:47:39.390 --> 00:47:56.549 Mark Kushner: does it have mixing with chlorine like subsurface? They could do this, but as far as I know, they haven't. So I know that when Humber did similar calculations back in 2,005 or something. We're still doing this. We published a paper on how the Argonauts amorphize the layers. 336 00:47:56.890 --> 00:48:06.129 Mark Kushner: and we worked for a bit with Erwin Kessels and the gang and Einhoven, and they they were doing similar exposures, and then they were using ellipsometry 337 00:48:07.500 --> 00:48:19.840 Mark Kushner: and something else. Now, maybe, Ramon, to look at this, and we have a paper from, like, you know, 15 years ago, where there was reasonable agreement between this amorphous layer thickness and the energy 338 00:48:20.320 --> 00:48:23.550 Mark Kushner: in industry. I don't think they've made that connection. 339 00:48:23.770 --> 00:48:31.780 Mark Kushner: For one thing, the atomic layer etches that they do are usually more complicated than just argon and silicon but 340 00:48:32.080 --> 00:48:35.949 Mark Kushner: they typically don't do. I think I'm not speaking out of turn here. 341 00:48:36.010 --> 00:48:42.890 Mark Kushner: The industrial folks typically don't do much surface analysis of that plus also, after the processes they always clean them. 342 00:48:43.370 --> 00:48:50.619 Mark Kushner: So there's like a, you know, a a wet clean. And so some of the debris that might be left behind is is lost. 343 00:48:50.940 --> 00:49:02.840 Mark Kushner: Actually, that's kind of a something that we don't talk much about. But I think it's really important. It's just an opportunity for automatic simulations. What happens when those wet cleans are occurring which really going on during that period? 344 00:49:02.980 --> 00:49:06.440 Mark Kushner: You would think that would be important. Yeah, thank you. 345 00:49:07.110 --> 00:49:32.370 Mark Kushner: Okay? Hi, thanks for your nice talk. I have a question about your starting since a situation where you did not expose your surface yet to the organ. Yeah. And so then the surface has reacted to the chlorine. But then it seems already amorphous. How? How would I? Oh, so what I was showing there were were snapshots after several cycles. 346 00:49:32.400 --> 00:49:47.040 Mark Kushner: Okay. But there was also a picture of 0. Ml, yeah. But that was 0. Ml, of the new argon cycle. Okay? So so that yeah. So that's not your stop with the crystal layer. But thing is the ions amorphise so quickly 347 00:49:47.230 --> 00:49:54.010 Mark Kushner: that it doesn't take long before. What you're dealing with is anymore. 5 layer. So yeah, we always start with a 100 and crystal service 348 00:49:54.420 --> 00:49:55.120 click. 349 00:49:55.930 --> 00:50:14.599 Mark Kushner: I'm also interested in that amorphous layer. So I was wondering. It seems like something that could be measured, what the relative probabilities are of the silicon atom being removed from, say, the top of the same orbits layer, or halfway down, or at the bottom, I mean, since it is 350 00:50:14.630 --> 00:50:25.930 Mark Kushner: seemingly fairly well mixed, I mean, but the sputtering is occurring almost completely from the top. Okay, yeah. The the ones in the middle will move around because of knock on collisions, but they won't get out 351 00:50:26.160 --> 00:50:30.139 Mark Kushner: right. It's only the ones that are at the surface that happen to get a bump off the company. 352 00:50:30.340 --> 00:50:40.359 Mark Kushner: So I think I think the ones that leave, I mean, we haven't looked at this for many, many years. But yeah, they're they're they're the ones very close to surface, which is why, when you have 353 00:50:40.580 --> 00:50:46.229 Mark Kushner: a monolayer, at least of chlorine. It's not surprising that you're getting physical spotting or chlorine atoms 354 00:50:46.250 --> 00:50:57.390 Mark Kushner: initially, because they're the ones that are right at the surface. Okay, do you expect that this adds, like a large amount of thickness, to say the death that a single daily cycle would run 355 00:50:57.810 --> 00:51:11.730 Mark Kushner: well, you mean with daily goes there. What they're doing is they're measuring the total silicon loss over like 10 or 20 or 100 cycles. And then just then they then they take the total delta and divide by the number of cycles, and they say that's their etch per cycle. 356 00:51:12.290 --> 00:51:19.589 Mark Kushner: But you know you're still, even though you're modifying that near surface region. The question is, what's your net loss of silicon from the surface? 357 00:51:19.700 --> 00:51:27.380 Mark Kushner: So in some sense it's independent of what modifications you're making to the top surface at the end of the day. How many silicons have I lost? 358 00:51:27.470 --> 00:51:32.120 Mark Kushner: Right? And if you say I've got a 10 to the fifteenth silicons per square centimeter. 359 00:51:32.480 --> 00:51:44.560 Mark Kushner: That's what we call one monolayer, that if you, if you've lost that amount, then we would say, one monolre has been removed, even though, in that top region where there's this modification to the layer, it's not quite a monol layer. 360 00:51:45.810 --> 00:51:47.250 Mark Kushner: if that makes sense. 361 00:51:49.260 --> 00:52:04.700 Mark Kushner: Yes. So you mentioned that the time step has to be very small in the simulation. So what but can what physical but can be services dominant that we need to stabilize? So it's it's basically the thermal motion you're you're resolving the vibrational timescales of the atoms 362 00:52:04.760 --> 00:52:09.890 Mark Kushner: right as they as they. Just. You know they're sitting there 300 Kelvin. They're going back and forth in their potential wells. 363 00:52:09.960 --> 00:52:15.689 Mark Kushner: And the time step's got to be small enough that you resolve that vibrational motion for us. 364 00:52:16.350 --> 00:52:18.500 Mark Kushner: typically, it's on the order of a femtosecond. 365 00:52:18.590 --> 00:52:27.980 Mark Kushner: So if it depends on the species hydrogen, you've got to take smaller time steps because it's so light. But that's that's the order of magnitude and the the big way you check this is if it's energy conservation. 366 00:52:28.600 --> 00:52:32.420 Mark Kushner: So as long as you conserve energy to some degree. 367 00:52:32.510 --> 00:52:38.329 Mark Kushner: And there was. There was this discussion earlier today about this very subject. Then we say, Okay, the timestep's small enough. 368 00:52:40.240 --> 00:52:42.099 Mark Kushner: So do you need a longer. 369 00:52:42.150 --> 00:53:00.519 Mark Kushner: I don't double to the simulation, because your time step is kind of hitting the round of ever of the double. I'm sorry I didn't quite catch the question. so, because I think the times of 10 to minus 15 is kind of close to the round of ever of the commuters. So do you need a longer 370 00:53:00.830 --> 00:53:01.910 Mark Kushner: teachings. 371 00:53:01.950 --> 00:53:07.130 Mark Kushner: Well, the real problem is the fact that if you want to try to simulate experiments. 372 00:53:07.320 --> 00:53:16.279 Mark Kushner: I mean you can. You can make the time step as small as you like, right? But then you have to integrate longer to have something happen that you want to compare to an experiment. 373 00:53:16.580 --> 00:53:24.130 Mark Kushner: and that's why we came up with this idea. So other people who do Md. They they do the same thing. But they say, Oh, it's I have a higher flux. 374 00:53:24.640 --> 00:53:27.740 Mark Kushner: So I'm saying, I'm jumping a millisecond between impacts. 375 00:53:27.840 --> 00:53:34.010 Mark Kushner: And they're saying, No, I'm just bringing the next one in right away. I'm just calling it a higher flux. I think a better way to say that 376 00:53:34.170 --> 00:53:36.780 Mark Kushner: is, you're assuming nothing's happening between impacts. 377 00:53:37.880 --> 00:53:42.479 Mark Kushner: So if it's a kind of a so a first order or pseudo. First order process. 378 00:53:42.530 --> 00:53:45.060 Mark Kushner: then then it's not flux dependent. 379 00:53:46.070 --> 00:53:57.339 Mark Kushner: at least not for these. For other things it can be, but it's really mainly. Can I integrate long enough so that I can look at my results and sensibly compare them 380 00:53:57.380 --> 00:54:00.300 Mark Kushner: to experiment, which takes seconds to minutes 381 00:54:02.020 --> 00:54:03.400 Mark Kushner: back to Sky 382 00:54:03.410 --> 00:54:30.200 Mark Kushner: Silicon. 383 00:54:30.330 --> 00:54:43.770 Mark Kushner: Well, what you what you can do is and this this is how you you can. The way you fit these intertime potentials is you do dft calculations. or you look at like known quantities, bond strengths, bond ankles. 384 00:54:44.050 --> 00:55:06.069 Mark Kushner: You know. So the parameters in those force fields are selected. So they're matching reasonably well. So you know. Humbert did this years ago. Others do it, of course, as well. Where you do. You know whole bunch of dft calculations and clusters, and then you adjust the parameters in your classical force field until you get about the same bond strength, bond angles. 385 00:55:06.340 --> 00:55:09.549 Mark Kushner: maybe even barriers in some cases for reactions. 386 00:55:10.530 --> 00:55:27.379 Mark Kushner: It it's not perfect right? And you're using like an average atom kind of yeah, well, it depends. It depends on the orientation of the other atoms. Right? Cause. These are many body potentials. So the silicon chlorine bond depends on whether or not that silicon has another chlorine. And if there's chlorine behind that 387 00:55:27.610 --> 00:55:32.779 Mark Kushner: right, so these many body potentials can take some of this into account. But it's imperfect. 388 00:55:33.100 --> 00:55:43.420 Mark Kushner: The one of the reasons that we do these kinds of calculations to compare to experiment is to say, yes, of course we're averaging. There's a lot of assumptions and approximations here as you criticize. 389 00:55:43.530 --> 00:55:52.450 Mark Kushner: But at the end of the day, when you do things like average et shields. Get it about right? You get coverage is about right. You get modifications of surface steps about right 390 00:55:53.090 --> 00:56:02.030 Mark Kushner: then. Okay, we're making approximations. But we're getting results that can be interpreted and compared to experiment. And it's in all modeling right? Every model is an approximation. 391 00:56:02.670 --> 00:56:06.169 Mark Kushner: And the question is what's appropriate for what you're trying to do. 392 00:56:06.460 --> 00:56:12.940 Mark Kushner: And it wasn't obvious when we started this in the Mid ninetys, that this made sense for these problems. It really wasn't 393 00:56:13.170 --> 00:56:18.700 Mark Kushner: is rubbish. Is it going to be like, you know, terrible comparison to experiment? Or is it going to be okay? 394 00:56:19.160 --> 00:56:25.430 Mark Kushner: And it turned out to be okay. But there's still plenty of problems and plenty of challenges. 395 00:56:29.360 --> 00:56:30.990 Mark Kushner: Are there any other questions 396 00:56:32.070 --> 00:56:39.679 Mark Kushner: popped up in the chat? Oh, chat! Okay, I should stop sharing 397 00:56:40.910 --> 00:56:43.330 Mark Kushner: photo hurts. Sorry? 398 00:56:44.240 --> 00:56:46.350 Mark Kushner: okay, thank you, David 399 00:56:48.140 --> 00:57:00.550 Mark Kushner: cracks. You observe some cracks and fissures on sicl or sif surface. Do you still see those cracks in your simulation experiments now are those cracks real physics, you know. Those cracks came from the stilling of rubber potential. 400 00:57:00.910 --> 00:57:03.490 Mark Kushner: I think it wasn't. It might have been Humber. 401 00:57:03.510 --> 00:57:09.360 Mark Kushner: II don't think we saw those with the Turso potential, II think. Oh, should I type this. 402 00:57:09.890 --> 00:57:20.859 Mark Kushner: and the person can hear me. we haven't seen those kind of cracks without still in your Webber. And I think the 403 00:57:22.330 --> 00:57:25.119 Mark Kushner: yeah, it's a good question. The answer is, no. 404 00:57:25.770 --> 00:57:26.850 Mark Kushner: we're not seeing them. 405 00:57:27.440 --> 00:57:38.579 Mark Kushner: Yeah, the details of the potential matters right? If you change potentials, if you go from sales or whatever to terse off through xf or do whatever. You will get quantitatively different numbers 406 00:57:38.970 --> 00:57:39.810 Mark Kushner: for sure. 407 00:57:40.300 --> 00:57:45.619 Mark Kushner: Qualitatively, they all give about the same results, but quantitatively give different results, which is what you'd expect. 408 00:57:46.210 --> 00:57:50.320 Mark Kushner: So don't necessarily believe these calculations too quantitatively. 409 00:57:51.820 --> 00:57:53.970 Mark Kushner: Yeah, that's good memory. 410 00:57:55.330 --> 00:57:58.560 Mark Kushner: Okay? Any other questions. 411 00:57:59.880 --> 00:58:02.280 Mark Kushner: No, David, thank you very much.