WEBVTT 1 00:00:00.329 --> 00:00:00.870 Richard Gottscho: Okay. 2 00:00:05.910 --> 00:00:19.980 Mark Kushner: hey Good afternoon, my name is mark pusher and i'm director of mitzi, it is my pleasure to introduce Dr Richard God true executive Vice President and chief technology officer and then research. 3 00:00:20.610 --> 00:00:37.080 Mark Kushner: rick is one of the world's foremost leading innovators in plans of processing and leaves lamb research, one of the world's leading innovator innovating companies and developing Plaza processing equipment for semiconductor fabrication. 4 00:00:38.100 --> 00:00:45.300 Mark Kushner: received his bachelor of science from Pennsylvania State University and his PhD from MIT and physical chemistry. 5 00:00:46.110 --> 00:00:54.390 Mark Kushner: He then joined bell laboratories, where he headed research departments and electronic materials electronic packaging and flat panel displays. 6 00:00:55.170 --> 00:01:01.110 Mark Kushner: That work established many of the principles that we now follow for Plaza processing of semiconductors. 7 00:01:01.980 --> 00:01:14.040 Mark Kushner: rick join their research in 1996 where he served in many roles from director to executive Vice President across the deposition and clean business units. 8 00:01:14.820 --> 00:01:30.120 Mark Kushner: He has authored landmark papers in Plaza processing sciences has countless pad patents had delivered countless invited lectures and it served on journal editorial boards and program committees for the major conferences in plasma science and engineering. 9 00:01:31.350 --> 00:01:40.710 Mark Kushner: is also provided to extremely valuable surface to the national academies and national research council and writing and managing major policy reports. 10 00:01:41.820 --> 00:01:56.130 Mark Kushner: records, we see several awards, including the ABS Peter mark memorial award eds plasma science and technology prize they tried processing pores and nisha was a award and the to think reward. 11 00:01:57.120 --> 00:02:05.250 Mark Kushner: rick is fellow of the American physical society and the American vacuum society and a member of the National Academy of engineering. 12 00:02:06.210 --> 00:02:19.500 Mark Kushner: By any measure Dr gotcha oh it's an intellectual leader of the Plaza processing community and a day to day example of how advances and fundamental science can be rapidly translated to technology. 13 00:02:20.880 --> 00:02:25.230 Mark Kushner: rick's talk today is particularly relevant given. 14 00:02:26.490 --> 00:02:38.700 Mark Kushner: The prospect of a new national program in semiconductor manufacturing this program is creating helpful incentives to produce semiconductors for America to act, the chips act. 15 00:02:39.390 --> 00:02:52.170 Mark Kushner: In fact, the Department of Energy today announced a new program for national laboratories in semiconductor manufacturing and we look forward to collaboration with industry under this program. 16 00:02:53.490 --> 00:03:04.980 Mark Kushner: The titles rick's presentation today is rethinking he art of plasma edge, but before risky ken's, we need to thank him for taking time out of his busy schedule to virtual visit us. 17 00:03:05.520 --> 00:03:16.380 Mark Kushner: And as part of our things we virtually present the mitzi mug among the most prized possessions, of the plasma scientists can have sorry there is. 18 00:03:17.580 --> 00:03:23.310 Mark Kushner: Thank you, thank you very much for taking the time and the podium is yours. 19 00:03:24.420 --> 00:03:34.920 Richard Gottscho: Well, thank you mark it's great pleasure to be able to spend this time with you all, I wish, as I was just saying that it could be done live by it's. 20 00:03:35.520 --> 00:03:51.840 Richard Gottscho: i'd love to come and visit the campus and and and meet you all in person and to see the facilities you've got spend more time with you, alas, the pandemic prevents us from doing that, today, but I do appreciate the opportunity and the time with you all. 21 00:03:53.100 --> 00:03:53.610 Richard Gottscho: i've. 22 00:03:55.530 --> 00:03:59.220 Richard Gottscho: Let me just make sure you can see my screen again. 23 00:04:01.110 --> 00:04:02.070 Mark Kushner: Yes, we can. 24 00:04:05.490 --> 00:04:08.130 Richard Gottscho: Okay i'm hold on in here i've got a problem. 25 00:04:10.200 --> 00:04:11.700 Richard Gottscho: All right, and still okay. 26 00:04:14.940 --> 00:04:15.900 Richard Gottscho: Mark okay. 27 00:04:17.310 --> 00:04:19.830 Nerissa Draeger: We see presenter view now so both slides. 28 00:04:21.690 --> 00:04:25.410 Richard Gottscho: So I was afraid of okay so we'll just get rid of the presenter view that okay. 29 00:04:26.490 --> 00:04:27.600 Nerissa Draeger: Yes, single slide now. 30 00:04:27.810 --> 00:04:32.670 Richard Gottscho: Alright, so i've entitled this talk rethinking the art of plasma. 31 00:04:33.870 --> 00:04:34.470 Richard Gottscho: and 32 00:04:35.580 --> 00:04:42.600 Richard Gottscho: I don't use the word art likely I i'm a firm believer that for innovation to occur. 33 00:04:44.160 --> 00:04:51.390 Richard Gottscho: For the creativity process to thrive, you really need to balance the right side of your brain with the left side of your brain. 34 00:04:52.740 --> 00:05:04.770 Richard Gottscho: And the artistic or the left side of the brain really can combine with rigorous science and engineering, I think, is the foundation of most. 35 00:05:05.790 --> 00:05:19.470 Richard Gottscho: great innovations, and so I want to start off with a little bit of perspective from the art world with respect to etching so let's start with Rembrandt in the 1600s. 36 00:05:20.760 --> 00:05:33.330 Richard Gottscho: Who was a pioneer of an etching technique, where he would cover up a plate with resin and then he would meticulously scratch his designs through that resin with a soft needle. 37 00:05:33.900 --> 00:05:47.430 Richard Gottscho: Then immersive and dilute hydrochloric acid to bring out that relief pattern, and he would do this over and over and over again to construct the kind of self image that you see here. 38 00:05:48.240 --> 00:06:09.510 Richard Gottscho: So he was a pioneer of multiple patterning if you will back in the 1600s, but of course etching is actually much older than that and it's not just the result of manmade operation, so there are natural etching events or processes that occur in nature, such as the. 39 00:06:11.550 --> 00:06:25.020 Richard Gottscho: acid water erosion of limestone, for example, there was the ancient Egyptians etched in stone woodblock printing and choking engravings in Asia. 40 00:06:25.740 --> 00:06:42.450 Richard Gottscho: back and 500 ad and so forth, until the modern age of the modern information age and from beginning in the really with the invention of the transistor in 47 the invention of the integrated circuit about 10 years later. 41 00:06:43.500 --> 00:06:49.530 Richard Gottscho: And the semiconductor the age of silicon and semiconductors really. 42 00:06:50.700 --> 00:07:03.990 Richard Gottscho: started after that and has flourished ever since and here's an image from Intel, and so I really want to talk about this kind of etching, of course, today, but recognize that there is a lot of artistic. 43 00:07:06.330 --> 00:07:16.020 Richard Gottscho: elements to the development of edge processes, even today, the work that's done by process engineers around the world. 44 00:07:17.190 --> 00:07:22.830 Richard Gottscho: Is a very creative process driven as much by intuition as science and engineering. 45 00:07:23.970 --> 00:07:30.360 Richard Gottscho: And it's truly remarkable in my mind to see what those engineers can accomplish. 46 00:07:30.810 --> 00:07:41.040 Richard Gottscho: So that picture is a segment of an integrated circuit, and so I just want to by way of introducing the topic talk or remind you and, in most cases i'm sure. 47 00:07:41.370 --> 00:07:53.100 Richard Gottscho: How chips are made and it starts with a wafer on which you, you do a deposition and maybe a thermal process, it may be a plasma assisted process to put down a thin film. 48 00:07:53.940 --> 00:08:04.950 Richard Gottscho: Then, in order to pattern that film that could be, for example, a conducting film and you want to create wires that will end up being interconnect that will in their connect one device or another. 49 00:08:06.090 --> 00:08:21.900 Richard Gottscho: You create that pattern by creating a mask projecting light through that mask exposing a photo resist on top of the film I see actually the the film is kind of missing in here at this point so imagine that's the film you're going to attach. 50 00:08:23.160 --> 00:08:23.760 Richard Gottscho: and 51 00:08:25.590 --> 00:08:33.150 Richard Gottscho: So you have a photo sensitive film you you expose the pattern you develop the pattern, and then you go into an editor and you transfer that. 52 00:08:34.080 --> 00:08:45.090 Richard Gottscho: photo generated pattern into the underlying film and again the edge processes that are used primarily in the industry today not exclusively, but primarily in the most critical edges. 53 00:08:45.390 --> 00:08:57.810 Richard Gottscho: Are all done using plasma processing, and this is where i'm going to focus my attention for the rest of the talk following that there's some strip processes removed the photo resist and then some clean processes. 54 00:08:58.500 --> 00:09:19.200 Richard Gottscho: which today are mostly wet but there's a trend to to replace those with dry processes to clean up any remaining residues and particles, so why plasma etching and plasma etching was introduced into semiconductor manufacturing integrated circuit manufacturing in the late 1970s. 55 00:09:20.640 --> 00:09:24.930 Richard Gottscho: And the reason the primary reason at that time was it gave. 56 00:09:26.160 --> 00:09:29.430 Richard Gottscho: It provided the ability to do an ice a Tropic etching. 57 00:09:30.750 --> 00:09:42.930 Richard Gottscho: If you do wet etching, it is naturally ice a Tropic, although it can sometimes proceed along Crystal and planes and be claws I and isotopic in that sense, and the whole point of. 58 00:09:43.380 --> 00:09:51.180 Richard Gottscho: Pursuing and and isotopic edge solution was simply that you could cram more and more patterns or structures. 59 00:09:51.900 --> 00:09:57.840 Richard Gottscho: closer together, so you had a higher density of patterning if you could do and isotopic etching. 60 00:09:58.260 --> 00:10:09.990 Richard Gottscho: And this results from the acceleration of ions across a plasma sheath which is rigorously analogous to a depletion layer in a semiconductor device so it's a region. 61 00:10:10.950 --> 00:10:19.650 Richard Gottscho: That the boundary between the plasma, which is caused by neutral consists of a soup of electrons and ions and free radicals and the surface. 62 00:10:20.550 --> 00:10:28.680 Richard Gottscho: Which is controlled at some potential and because there's a potential drop between the plasma and the surface, a sheath forms. 63 00:10:29.340 --> 00:10:44.700 Richard Gottscho: That accelerates ions in one direction and that accelerates etching in that direction that's The primary reason plasma etching was introduced into semiconductor manufacturing now since that time, there have been many other benefits. 64 00:10:45.930 --> 00:10:54.720 Richard Gottscho: uncovered and developed in plasma etching so, for example, lithography tends to suffer from line edge roughness, particularly on the leading edge. 65 00:10:54.990 --> 00:11:05.520 Richard Gottscho: most advanced lithography it's one of the biggest problems that line edge roughness ultimately translates into variability and devices and can affect overall performance and yield. 66 00:11:06.240 --> 00:11:23.700 Richard Gottscho: That line edge roughness can be smoothed or reduce not completely eliminated using edge processes or combination of cyclical deposition and edge processes simple minded way to think about that, as you can chop off ruff ruff corners protrusions and you can fill in crevices. 67 00:11:24.840 --> 00:11:39.660 Richard Gottscho: Another benefit is to go beyond the optical limits of with all griffey in terms of the resolution of that's that's dependent on the wavelength, of the light and the optics of the with the graphic system and today. 68 00:11:40.380 --> 00:11:58.020 Richard Gottscho: Those those limits we've we've been going beyond those limits for more than 20 years by using techniques during at such as shown here, where you would have a deposition step to conform we coat. 69 00:11:59.370 --> 00:12:06.720 Richard Gottscho: A structure that was defined lithograph Basically, this is showing actually coating of other resist, for example, but it might be in a subsequent step. 70 00:12:07.320 --> 00:12:13.500 Richard Gottscho: and effectively shrink a whole is shown here and then do your directional ash, so you can go beyond. 71 00:12:13.980 --> 00:12:27.150 Richard Gottscho: The the the the the physical limits of lithography by using deposition and etching techniques and the the last benefit is in the area of area selective etching. 72 00:12:28.020 --> 00:12:33.330 Richard Gottscho: This comes out of an Intel paper where they talk about what they really want the. 73 00:12:33.630 --> 00:12:44.160 Richard Gottscho: whole world once is to have four different materials, each of which could be epsilon actively to the other that opens up all kinds of possibilities and the fabrication of integrated circuits. 74 00:12:44.550 --> 00:12:53.070 Richard Gottscho: And one of the ways in which that's done in in the plasma etch world is we actually again deposit some material, while we're etching. 75 00:12:53.520 --> 00:13:08.430 Richard Gottscho: And the material that's deposited say on an oxide surface versus a nitride surface has a different reactivity with the underlying material such that when I NS hit it, you have either a high probability of forming some volatile edge products in the case of the oxide. 76 00:13:09.540 --> 00:13:15.480 Richard Gottscho: And the opposite, in the case of the nitride and that way you can generate almost an artificial so activity. 77 00:13:16.680 --> 00:13:21.690 Richard Gottscho: That facilitates the fabrication of most advanced devices, so how does. 78 00:13:23.190 --> 00:13:34.680 Richard Gottscho: How does that work i'll come back the plasma etching in a second, but I wanted to just kind of generalize what what in fact is etching it's obviously the removal of material somehow from a surface. 79 00:13:35.160 --> 00:13:45.510 Richard Gottscho: And the best way, I think, to think about that as you got to overcome the binding energy you not here, which is basically that energy required to pull an atom. 80 00:13:46.470 --> 00:13:56.010 Richard Gottscho: From from the solid state and and and take it out either into solution and carry it away or into the gas phase and pump it away. 81 00:13:56.370 --> 00:14:07.200 Richard Gottscho: And you can think of that is there's kind of four different ways in which you can put sufficient energy in to overcome a not and therefore remove material there's mechanical and. 82 00:14:07.770 --> 00:14:14.520 Richard Gottscho: You can imagine very small tweezers think of that in terms of perhaps atomic force microscope moving Adams around not. 83 00:14:14.910 --> 00:14:22.800 Richard Gottscho: Really mainstream or practical there's chemical watching, which is mainstream and practical there's thermal etching which. 84 00:14:23.610 --> 00:14:36.930 Richard Gottscho: is also practical not widely used today but we'll see as more advanced structures are being fabricated with complicated three dimensional architectures thermal based. 85 00:14:37.380 --> 00:14:49.710 Richard Gottscho: etching is going to grow in importance and then there's physical etching or sputtering where you simply bombard the surface, with an inert species, as opposed to a chemically active species over here. 86 00:14:50.070 --> 00:14:56.940 Richard Gottscho: And that energy is imparted into the lattice and ultimately results in the inject a rejection of an atom now. 87 00:14:57.600 --> 00:15:14.910 Richard Gottscho: How does plasma etching work well effectively plasma etching is a combination of chemical watching and physical sputtering but what's quite interesting and as first highlighted as far as I know, in 1979 by coburn and winners, who are at IBM at the time. 88 00:15:16.410 --> 00:15:25.020 Richard Gottscho: There is a dramatic synergy when you bring both the chemical reactions and the inert gas ions together. 89 00:15:25.680 --> 00:15:33.870 Richard Gottscho: And what coburn winners did was very simple and elegant experiment beautiful experiment that I think everybody quotes today still. 90 00:15:34.350 --> 00:15:48.840 Richard Gottscho: And that is the first irradiated a silicon surface with xenon by fluoride, which was a proxy for fluorine atoms and they got some very low level of ice spontaneous and isotopic etching. 91 00:15:49.710 --> 00:16:00.090 Richard Gottscho: And they they also had an iron beam in the system and they turn the neutrals off and just run the ions they get some a lower rate that corresponds to sputtering. 92 00:16:01.200 --> 00:16:07.770 Richard Gottscho: But Lo and behold, when they turn both the radicals in the ions on or the xenon die for I, in this case and the ions on together. 93 00:16:08.220 --> 00:16:14.490 Richard Gottscho: They get more than the sum of the two X rates, so there is a that's what we mean by the Ai n neutral synergy. 94 00:16:15.120 --> 00:16:19.800 Richard Gottscho: And it's really interesting to note also that they get this transient right at the beginning. 95 00:16:20.340 --> 00:16:26.910 Richard Gottscho: and come back and talk about the origins of that shortly, but fundamentally it's because the surface was covered. 96 00:16:27.270 --> 00:16:39.120 Richard Gottscho: With xenon di fluoride and saturated now you bring in the ions and you get an accelerated rate, but as you read the ions on you get a lower steady state surface concentration of xenon die for. 97 00:16:39.780 --> 00:16:49.410 Richard Gottscho: Now you can model this effect, quite simply, and it's something that we did in a paper I published in 1992 it's kind of frightening to think how long ago that was. 98 00:16:50.940 --> 00:17:02.370 Richard Gottscho: And this actually was a review Article so we we didn't per se derive this ourselves, but we we we looked at what was in the literature and brought it all together. 99 00:17:03.180 --> 00:17:15.270 Richard Gottscho: And it's a pretty simple formulation and very, very powerful now it assumes there's no sputtering and spontaneous APP so those would be additional terms that you could add on here, but this is the synergy term. 100 00:17:15.900 --> 00:17:30.000 Richard Gottscho: That depends on both iron flux and neutral flux, it is derived from from two basic considerations, one is it assumes that there are reactive, that there are sites on the surface, to which. 101 00:17:30.480 --> 00:17:40.980 Richard Gottscho: An action can bind and then depending on the surface concentration of those react and bound sites when an iron comes in. 102 00:17:41.730 --> 00:17:52.620 Richard Gottscho: They are more easily sputtered than the substrate material you've lowered the E naught, if you will, by modifying the surface and that's why that's how you get this synergistic. 103 00:17:53.070 --> 00:18:00.030 Richard Gottscho: effect the energy dependence is the other consideration and that's derived was derived by a guy named steinbrueck whole. 104 00:18:01.140 --> 00:18:15.540 Richard Gottscho: I think, also in the 70s, or maybe in the 80s and he showed that that should scale as the square root of the iron energy above a certain threshold energy So if you just take that equation, and you look at a couple women in cases for. 105 00:18:18.600 --> 00:18:28.500 Richard Gottscho: Low values of the neutral to iron flux ratio or low values of the neutral flux your etching is limited by how many neutrals you have available. 106 00:18:28.920 --> 00:18:39.600 Richard Gottscho: And it'll scale linearly with neutral flux, on the other hand, if you have lots of neutrals around but compared to the the ions so a large neutral iron ratio. 107 00:18:40.200 --> 00:18:53.940 Richard Gottscho: Then you're the rate is limited by the iron flux, but that I am that that X Ray which is limited by i'm fox is a function of iron energy is given by this formula here, and that was verified experimentally. 108 00:18:54.990 --> 00:19:06.870 Richard Gottscho: Not first time bye bye Jane Chang UCLA actually this is when she was a graduate student at MIT and 97 there was some earlier work, but this is one of the most exhaustive studies that that. 109 00:19:07.590 --> 00:19:20.310 Richard Gottscho: At that point in time, and so that's why I use this example so and every etching system i'm i'll spend a lot of time talking about silicon in Korean because it's been studied the most silicon Korean argon. 110 00:19:21.600 --> 00:19:38.610 Richard Gottscho: But every etching system plasma edge material system exhibits this behavior it is, it is universally applicable now conventional plasma etching then proceeds by. 111 00:19:39.210 --> 00:19:47.700 Richard Gottscho: Having a reactor, and this is one one configuration that's widely used throughout the world, there are two basic. 112 00:19:48.510 --> 00:19:58.950 Richard Gottscho: configurations that are widely used one is inductive Lee coupled plasma, which is what shown here you apply an rf frequency you inductive we couple energy into and form a plasma. 113 00:19:59.370 --> 00:20:09.150 Richard Gottscho: And then you have a separate rf energy source to bias, the way for an extract those ions to get that NIC Tropic etching effect the other basic configuration. 114 00:20:09.480 --> 00:20:17.970 Richard Gottscho: Is a capacitive we coupled reactor where this top point is just it would be a conducting plate, as opposed to a dielectric barrier. 115 00:20:18.360 --> 00:20:29.940 Richard Gottscho: With a coil on top, and it would just be a plane plane or electrode parallel play kind of geometry, and each reactor has different applications for different reasons beyond the scope. 116 00:20:30.420 --> 00:20:42.090 Richard Gottscho: of my talk here today and not really relevant to the the art of plasma etching message that I want to try and get across now this this is today still. 117 00:20:42.510 --> 00:20:54.690 Richard Gottscho: The primary means by which plasma etching is performed to make integrated circuits in volume production and there's a huge problem with it and it simply put. 118 00:20:55.770 --> 00:21:07.470 Richard Gottscho: The dissociation of your reactants let's say it's corrine CO2 or hdr a B is is occurs by electron impact bombardment. 119 00:21:08.130 --> 00:21:22.290 Richard Gottscho: And I n ization whether it's argon or it could be the Korean itself to make Korean plus or Korean molecular plus is also done by electron impact and there's one self consistent electron energy distribution function. 120 00:21:23.370 --> 00:21:36.060 Richard Gottscho: That is determined by how much power you put in what your pressure is what your gases are, and given that recipe condition it's it's just a fixed function, and so the. 121 00:21:36.930 --> 00:21:48.660 Richard Gottscho: probability for dissociation the probability for ionization is given by the overlapped area between that electron energy distribution function and the Cross sections for dissociation and ionization. 122 00:21:49.200 --> 00:22:01.380 Richard Gottscho: And therefore, the generation of neutrals and the generation of ions and remember I just got done explaining that the the plasma edge mechanism is governed by the ratio of ions the neutrals. 123 00:22:02.580 --> 00:22:12.360 Richard Gottscho: But that's kind of fixed in this reactor geometry it's very difficult to vary that ratio, because their generation rates are coupled by. 124 00:22:13.560 --> 00:22:16.140 Richard Gottscho: The same electron energy distribution function. 125 00:22:17.190 --> 00:22:25.440 Richard Gottscho: The loss rates might be different, but that's a function of reactor conditions and forgiven set of reactors that's also a fixed that's going to be a fixed ratio. 126 00:22:26.490 --> 00:22:36.510 Richard Gottscho: There are other problems in conventional etching Now this is this is rapidly changing but it used to be that we just turn the plasma on we touch it for a period of time. 127 00:22:36.870 --> 00:22:48.630 Richard Gottscho: Maybe we changed some conditions as we went from one layer to the other, but basically the plasma was on all the time and and the problem with that is that you accumulate damage. 128 00:22:49.590 --> 00:22:58.710 Richard Gottscho: In the surface, that your etching by having this complicated soup constantly bombarding the surface, as in, as in the coburn winters experiment. 129 00:22:59.370 --> 00:23:07.290 Richard Gottscho: You saw that gee if, when you first turn on those if you got a much higher rate than when you just left it on which. 130 00:23:07.860 --> 00:23:19.260 Richard Gottscho: tells us, in fact, the clues were there back in 1979 but it took us probably 20 to 30 years to really understand the significance of that result it's kind of embarrassing to admit it, but it's true. 131 00:23:20.430 --> 00:23:31.680 Richard Gottscho: So that's that's Another drawback of of this approach and the the other so a couple of ionization dissociation leads to a very small process window very challenging to develop. 132 00:23:32.280 --> 00:23:41.310 Richard Gottscho: The right process really takes an artist to tune it in, and the other problem is is transport limitations, because in this situation. 133 00:23:43.620 --> 00:23:56.340 Richard Gottscho: As when I say simultaneous and continuous etching it's not a self limiting kind of reaction, the bombardment of the ions and the neutrals is occurring all the time and the synergistic effect is happening all the time. 134 00:23:56.730 --> 00:24:08.610 Richard Gottscho: So I say it's not self limiting is it, but it is limiting in the sense that you get this interface interface will damage and roughness just accumulates with time you never get you never have a chance to reset the surface. 135 00:24:10.020 --> 00:24:16.920 Richard Gottscho: Because the reactions tend to be either neutral limited or I unlimited or somewhere in between it's transport limited it's how. 136 00:24:17.190 --> 00:24:28.860 Richard Gottscho: The rate occurs, the rate is dependent on how fast you bring those ions and neutrals to the surface, and this leads to a very strong aspect ratio dependence, to the edge which the industry has been. 137 00:24:29.400 --> 00:24:46.170 Richard Gottscho: Fighting forever, to this day, and you can see, the example of it here, and you can get it to go either way, generally speaking, smaller aspect ratios by aspect ratio, I mean the the the width of a feature the diameter of the feature. 138 00:24:47.370 --> 00:24:58.140 Richard Gottscho: Divided into its its depth so depth over with, I should say or depth over diameter and the smaller, that is, the larger the solid angle. 139 00:24:58.590 --> 00:25:08.490 Richard Gottscho: Of the features of 10s, and so the higher the neutral flux, now the ions generally don't vary too much from as a function of aspect ratio because they're highly directional. 140 00:25:09.270 --> 00:25:18.510 Richard Gottscho: Although, to the extent that the surface can charge up differentially it can deflect the ions and that effect, also scales with aspect ratio. 141 00:25:18.960 --> 00:25:28.350 Richard Gottscho: The transport of neutrals can either be a shadowing mechanism that is a neutral hits aside was stuck in lost or you could have. 142 00:25:29.280 --> 00:25:35.130 Richard Gottscho: A low sticking coefficient on the sidewall and the neutrals come in and they bounce around till they hit the bottom it. 143 00:25:35.520 --> 00:25:45.240 Richard Gottscho: From a scaling perspective, it turns out, it doesn't matter all those mechanisms scale with aspect ratio and that's bad news for making devices, where you have. 144 00:25:45.660 --> 00:25:54.300 Richard Gottscho: All kinds of different aspect ratios you're trying to etch simultaneously on the wafer and you're trying to do it, simultaneously, for economic reasons. 145 00:25:56.220 --> 00:26:02.070 Richard Gottscho: And then another problem with this approach is you're susceptible to wait for scale non uniformity, this is. 146 00:26:02.970 --> 00:26:14.100 Richard Gottscho: device scale non uniformity, if you will, but you also have way for scaling on uniformity, there are very natural gradients that are effectively impossible to eliminate. 147 00:26:15.030 --> 00:26:20.550 Richard Gottscho: In a plasma reactor, and the reason the basic reason is the wafers finite size. 148 00:26:21.000 --> 00:26:26.160 Richard Gottscho: I keep telling our customers if you give us infinite size wafers we can make, we can give you a perfect uniformity. 149 00:26:26.520 --> 00:26:38.100 Richard Gottscho: Or if you're willing to throw away the last two inches of your way, for we can give you a perfect uniformity and the other 10 inches of the wafer and strangely enough, they just kind of look at me with crossed eyes. 150 00:26:39.720 --> 00:26:54.030 Richard Gottscho: But because of that discontinuity in the material properties, where the wafer ends and something else begins you you create a gradient a chemical potential gradient which results in a gradient in. 151 00:26:55.170 --> 00:27:08.370 Richard Gottscho: Every neutral species in the plasma you actually have an electrical discontinuity that produces a potential gradient in the plasma and therefore a bending of the sheath which means. 152 00:27:08.820 --> 00:27:28.080 Richard Gottscho: Not only does the flux change radio way, but the ions tend to come in at an angle, which creates a lot of issues again that problem exists when the reaction is transport limited as opposed to being limited by how fast things happen on the surface. 153 00:27:29.700 --> 00:27:38.610 Richard Gottscho: And those are those are fundamental limitations of simultaneous and continuous plasma etching that is not self limiting in any any way. 154 00:27:39.690 --> 00:27:40.770 Richard Gottscho: So enter. 155 00:27:42.240 --> 00:27:52.650 Richard Gottscho: The new art of etching which I can't emphasize enough, is based on academic research and simulation, and here I want to digress a little bit and just mention. 156 00:27:53.280 --> 00:28:05.100 Richard Gottscho: That I met mark kushner actually I don't know if I met you in 1980 mark, but I remember reaching out to you, I think, in about 1980, which is when I join bell laboratories, or maybe it was at one. 157 00:28:06.600 --> 00:28:20.820 Richard Gottscho: Fresh out of doing a postdoc at MIT and the physics department and I, and I was hired at bell labs to study plasmas because they were in production and nobody actually understood how they work and what was going on. 158 00:28:22.770 --> 00:28:35.130 Richard Gottscho: The Cobra and winters paper just come out actually by the time I started to work at at bell labs and that was the first seminal paper that started to point the direct point us in the right direction mechanistic way. 159 00:28:36.150 --> 00:28:41.730 Richard Gottscho: But I, so I started I set up some diagnostic experiment, I was making all kinds of measurements. 160 00:28:42.930 --> 00:28:44.520 Richard Gottscho: And we used to call those. 161 00:28:45.660 --> 00:28:55.890 Richard Gottscho: g lat papers papers, just like jeez look at that, but we didn't understand what we were looking at, and then I ran across a wonderful paper by by Mark kushner. 162 00:28:57.180 --> 00:29:13.440 Richard Gottscho: That where he had he was modeling plasma processes with all the complicated chemistry, and I was like wow I got to work with this guy he's he's this is, this is what we need to marry with the experiments and that was the beginning of a very long and fruitful relationship. 163 00:29:15.060 --> 00:29:29.010 Richard Gottscho: Because mark mark was very receptive to my to my initial overture but in 2006 mark and his team, started to look at how atomic layer etching and i'll talk a little bit more about what we mean by that. 164 00:29:30.270 --> 00:29:40.380 Richard Gottscho: But basically it's designed to get around these fundamental limitations of simultaneous and continuous plasma etching and so in simulation they modeled the atomic. 165 00:29:41.190 --> 00:29:54.120 Richard Gottscho: layer etching where they alternately expose the system to some reactive gases and then and then bombarded and effective we create a two step reaction that each have one, each of which is self limiting. 166 00:29:55.650 --> 00:30:09.660 Richard Gottscho: This was verified in the laboratory this this particular dielectric a daily process was verified in the laboratory about seven years later at the University of Maryland by gottlieb airline and his group. 167 00:30:11.010 --> 00:30:24.150 Richard Gottscho: and lamb leverage this learning and it's it almost feels like a law of nature, but it's about 10 years for any new idea to actually reach commercial fruition and in 2016. 168 00:30:24.630 --> 00:30:33.270 Richard Gottscho: About 10 years after marks seminal paper we introduced an atomic way retching process and high volume manufacturing for logic devices. 169 00:30:34.650 --> 00:30:44.550 Richard Gottscho: So what is atomic way or etching i've already alluded to it if we go back to the the theory of plasma etching and we talked about the neutral limited regime and the I unlimited regime. 170 00:30:44.970 --> 00:30:54.900 Richard Gottscho: Basically, instead of doing everything, at the same time, and leaving the plasma on it all the time we break it up into two steps, the first step is a surface modification step. 171 00:30:55.470 --> 00:31:12.330 Richard Gottscho: it's very much analogous to atomic layer deposition or your first functional eyes the surface and then you bring in a converting agent and grow a film atomic layer by atomic layer in this case we're modifying the surface, imagine coring ads or being on the silicon. 172 00:31:13.440 --> 00:31:23.790 Richard Gottscho: there's no I am bombardment around, ideally, and you, you reach a limit where it saturates so once you saturate the surface of silicon with Korean you make silicon. 173 00:31:24.240 --> 00:31:34.770 Richard Gottscho: chloride loyalties si si si si si si si l three sal for is volatile so typically if you make that it's going to come off that would be spontaneous setting. 174 00:31:35.760 --> 00:31:43.110 Richard Gottscho: But it reaches a limited saturates it's self limiting critically important, because then the reaction can't proceed. 175 00:31:43.950 --> 00:31:49.260 Richard Gottscho: And then you come in with a source of energy, energy typically I mean implies matching. 176 00:31:49.890 --> 00:32:01.860 Richard Gottscho: I n bombardment and you remove that modified layer, and if you tune your energy just right you only modify you only remove that modified layer so again, you have a self limiting. 177 00:32:02.460 --> 00:32:15.660 Richard Gottscho: Removal step self limiting modification step self move self limiting removal step and that has huge advantages it gets it gets so now you're no longer transport limited. 178 00:32:16.230 --> 00:32:21.480 Richard Gottscho: That you're limited by how fast, you can absorb and modify the surface, on the one hand. 179 00:32:21.930 --> 00:32:31.290 Richard Gottscho: And then, how fast, you can remove that modified layer and if they are both self limiting reactions it doesn't matter how fast the stuff comes down that affects throughput. 180 00:32:31.860 --> 00:32:48.030 Richard Gottscho: But does not affect the properties of the edge, so you get atomic we smooth touching, as you can see here, this is, this is an Al di analog, this is a an Ai le analog atomic we smooth surfaces you get aspect ratio independent etching. 181 00:32:49.320 --> 00:33:00.330 Richard Gottscho: Because it doesn't matter how long it takes for things to go down a high aspect ratio feature versus a low aspect ratio if you're willing to wait long enough to it reaches itself limit. 182 00:33:00.870 --> 00:33:19.290 Richard Gottscho: There is no aspect ratio dependence because you're not transport limited and same thing for macroscopic or way for scale non uniformity same mechanism not transport limited surface rate limited and you get atomic we uniform results over the entire way for. 183 00:33:22.860 --> 00:33:32.160 Richard Gottscho: Now I said a lie, is it produces smoother surfaces, I want to delve a little bit more into the mechanism for how that occurs. 184 00:33:34.200 --> 00:33:46.290 Richard Gottscho: And this was again go back to two marks groups papers, this one in 2007 and and mark was doing simulation showing that, in an ri ve. 185 00:33:46.650 --> 00:33:54.210 Richard Gottscho: that's conventional plasma acting or basically everything's on at the same time, so little bit of a misnomer stands for reacted by an etching. 186 00:33:54.750 --> 00:33:59.670 Richard Gottscho: In the days when people use beams but it but it's referred more generic way to plasma matching now. 187 00:34:00.240 --> 00:34:08.100 Richard Gottscho: And he predicted you get a very atomic way rough surface because you're you're nothing is self limiting so you you, you will you. 188 00:34:08.430 --> 00:34:22.230 Richard Gottscho: modify the surface, but at the same time you're modifying it you're deserving things creating more reactive sites and so that leads to a very complicated mix structure that is very, very rough on an atomic scale, by contrast with. 189 00:34:22.980 --> 00:34:29.250 Richard Gottscho: The self limiting reaction approach of a le and here's marks prediction here's our experimental results. 190 00:34:29.880 --> 00:34:45.060 Richard Gottscho: very, very, very good agreement qualitatively, and you can see that the the profile has a much flatter edge front than what you got over here as a result of doing taking an elite approach so. 191 00:34:45.690 --> 00:34:52.440 Richard Gottscho: One of the questions we ask is you know between these two experiments here's here's an ra Korean argon at 50 volts that's the. 192 00:34:53.190 --> 00:34:58.110 Richard Gottscho: The mean energy of the ions and here's Korean and argon at 50 volts and, of course. 193 00:34:58.650 --> 00:35:12.000 Richard Gottscho: These aren't being done at the same time they're their stage the sequential and time it's a cyclical process, this is continuously on but but there's there's another big difference between the two, and that is when you have. 194 00:35:13.080 --> 00:35:18.120 Richard Gottscho: Everything going on at the same time you not only have our gun I am bombardment you have Korean I am bombardment. 195 00:35:19.170 --> 00:35:31.110 Richard Gottscho: And there's a huge difference, depending on whether that the the surface is bombarded with a rare gas and enter an inner species versus a reactive species and it's pretty easy to understand if it's reactive. 196 00:35:31.500 --> 00:35:46.560 Richard Gottscho: it's going to break bonds and restructure the surface and that that induces this mixing and roughness whereas if you just imagine a an unmodified surface has shown here or a very thin layer on top. 197 00:35:47.100 --> 00:35:57.300 Richard Gottscho: And then, an argon comes in it's it's just sputtering it's not doing any chemistry and that actually tends to to create smoothing smoothing. 198 00:35:58.320 --> 00:36:07.620 Richard Gottscho: And there are a couple different ways to think about that one is you just have this injection of ballistic energy or thermal induced. 199 00:36:08.040 --> 00:36:19.920 Richard Gottscho: which produces a kind of a heat wave or phone on wave that can smooth the surface, gives the atoms enough energy to move around and overcome because the the surface, would like to be smooth. 200 00:36:20.670 --> 00:36:26.100 Richard Gottscho: So any of these protrusions or divots tend to get filled in if you if you kind of heat things up, if you will. 201 00:36:27.180 --> 00:36:35.520 Richard Gottscho: Also, if you have a protrusion and now you think about a lie, again, where your first modify the surface, you have a much higher concentration. 202 00:36:36.480 --> 00:36:48.270 Richard Gottscho: In a given volume for for something that's sticking out versus a flat surface there's more local availability of reacting, so this will tend to react faster now when you hit it with the ions. 203 00:36:49.380 --> 00:37:05.850 Richard Gottscho: Also, the the spotter yield for both chemically modified surfaces, as well as the bear surface it's a very strong angular dependence, and that will tend to smooth out protrude or reduce protrusions and, lastly, you can imagine. 204 00:37:07.470 --> 00:37:19.410 Richard Gottscho: an iron coming in and sputtering something off and it comes off at some angle and doesn't escape before hits something else, and then it comes back and then maybe that edge product read deposits and it can feel things in that way. 205 00:37:20.070 --> 00:37:24.870 Richard Gottscho: So we took all these mechanisms together and we did some Monte Carlo simulations. 206 00:37:25.950 --> 00:37:36.810 Richard Gottscho: To try and understand the smoothing benefits of a we and and, in particular, we were curious what's the limit, I mean how rough can you start and how smoothly, can you make it. 207 00:37:37.410 --> 00:37:48.030 Richard Gottscho: So we took a we took a surface with a to nanometre rms pre edge, this is a tungsten film which we're matching with atomic were matching. 208 00:37:49.110 --> 00:38:07.530 Richard Gottscho: And those are the dots here, and then the Green is the result of the Monte Carlo simulation and you can see that, with the more haley cycles, we expose the way for to we get more and more smoothing down to essentially an atomic limit. 209 00:38:08.580 --> 00:38:13.980 Richard Gottscho: And it doesn't matter I mean you can start with a rough surface or you can start with an atomic we flat surface. 210 00:38:14.400 --> 00:38:21.930 Richard Gottscho: And what will happen as you as you expose it is it reaches a surface roughness that's equivalent to about plus or minus one atom. 211 00:38:22.800 --> 00:38:30.180 Richard Gottscho: So we think we understand that pretty well and it's an inherent benefit of a week that you cannot get out of normal plasma etching. 212 00:38:30.900 --> 00:38:42.570 Richard Gottscho: So that's all great stuff where's the bad news, well, first and foremost because of the cyclical nature of the process and the need to go to a self limiting. 213 00:38:43.290 --> 00:38:49.260 Richard Gottscho: State and there's a trade off there, but that actually takes a long time it takes a long time to. 214 00:38:49.770 --> 00:38:57.960 Richard Gottscho: First, I got to expose it to chlorine and I got to pump out the Korean because I don't want Korean ions around when I do my energetic I am bombardment otherwise I get all that mixing. 215 00:38:58.410 --> 00:39:10.170 Richard Gottscho: Then I introduce my argon Ionized my art on it, but it removes the modified layer, then I gotta gotta go back and forth, so that kind of kills throughput and rough order about by a factor of 10. 216 00:39:10.830 --> 00:39:21.300 Richard Gottscho: Today, relative to conventional plasma etching has all those benefits but it comes at a very big price factor of 10 reduction in throughput is a significant. 217 00:39:21.750 --> 00:39:29.460 Richard Gottscho: detriment, the other problem is there are limits to the self limiting behavior so, for example. 218 00:39:30.060 --> 00:39:37.770 Richard Gottscho: You can have photons in a plasma we coordinate the surface, we actually run a Plaza and we just don't accelerate ions to the surface. 219 00:39:38.130 --> 00:39:43.020 Richard Gottscho: And they're photons very energetic photons floating around that can induce photo chemistry. 220 00:39:43.650 --> 00:39:56.070 Richard Gottscho: And that's going to give you some spontaneous etching during the, the first step, where all you're trying to do is modify the surface, you still have sputtering going on, so even once you've removed all the coring silicon Korean Maltese. 221 00:39:57.210 --> 00:40:00.210 Richard Gottscho: You still have energetic I am bombardment and. 222 00:40:01.980 --> 00:40:06.390 Richard Gottscho: That can that that leads to not ideal behavior from a ui perspective. 223 00:40:07.950 --> 00:40:16.590 Richard Gottscho: i'm going to talk more about about these but there's also these are to certain extent manageable and and. 224 00:40:17.490 --> 00:40:32.310 Richard Gottscho: deal with them, this one is really a really challenging problem and is in the domain of reactor design the kind of things that lamb research does, but after you pump out that coring it's naive to think there's still no Korean and the system Korean. 225 00:40:33.360 --> 00:40:38.790 Richard Gottscho: molecules and atoms will absorb on the walls of the reactor and during your argon sputter step. 226 00:40:39.060 --> 00:40:48.690 Richard Gottscho: can come off those walls and contribute to re modification of the surface, and that leads to non ideal haley behavior and, finally, you could have a. 227 00:40:49.140 --> 00:40:58.770 Richard Gottscho: rough surface is coming in surface imperfections and and the defects and the surface, will match at different rates and so you may lose some of the benefits of a lie, for that reason. 228 00:40:59.790 --> 00:41:08.940 Richard Gottscho: For all these reasons, we do we define something we call a lie synergy which is very simple measurement that you do you look at the. 229 00:41:09.300 --> 00:41:22.230 Richard Gottscho: The X amount per cycle under aly conditions and you subtract the etching you get just from the Korean step and you subtract the sputtering you get just from the sputter step or step B. 230 00:41:23.550 --> 00:41:36.600 Richard Gottscho: And and convert that to a percentage, and so, if neither one of these reactions is taking place your synergy is 100% and we consider that ideal a le behavior. 231 00:41:37.020 --> 00:41:44.850 Richard Gottscho: Anything less than that of course is less than ideal, so we use that as a metric to characterize the degree of a le ideology. 232 00:41:45.510 --> 00:41:58.110 Richard Gottscho: And that and that's an important concept to just see in a second now, how do you control how much sputtering goes on, or how much spontaneous etching goes on, we want to suppress both We only want the synergistic effect. 233 00:41:59.520 --> 00:42:03.720 Richard Gottscho: It helps to look at the energetics of what's going on So the first thing is. 234 00:42:04.050 --> 00:42:13.890 Richard Gottscho: to absorb the Korean or modify the surface, you have an energy of absorption way to look at think of it as shown here, and you actually want that to be a very small number. 235 00:42:14.550 --> 00:42:17.400 Richard Gottscho: So that the absorption is very rapid. 236 00:42:18.390 --> 00:42:28.590 Richard Gottscho: Now the subsequent steps with Korean absorbing on top of the Korean you'd actually like that to be a high number, so that you only get one monolayer of react and sticking to the surface. 237 00:42:28.890 --> 00:42:34.980 Richard Gottscho: Although if you have Korean on top of Korean it's not the end of the world, it just gives you when the iron comes in, you have that much more opportunity. 238 00:42:35.400 --> 00:42:39.810 Richard Gottscho: To remove material, but you want this to be a low number you want it to be fast. 239 00:42:40.530 --> 00:42:47.340 Richard Gottscho: Then you've got a disruption barrier you got to get over so this is where you got to think about what I in energy to I want. 240 00:42:48.000 --> 00:42:54.540 Richard Gottscho: To get over this barrier because i've got a stable species sitting on the surface and and I know out here. 241 00:42:55.200 --> 00:43:07.050 Richard Gottscho: it's actually higher energy so i've got a kinetic barrier, I have to overcome that's the role of the energetic I am bombardment and the amount of iron energy, I need is is determined by this barrier height. 242 00:43:08.400 --> 00:43:22.950 Richard Gottscho: And then the last thing is I don't want my energy to be so high that I actually do a lot of sputtering, which is what this binding energy number is all about and i'll point out you'll notice right away, I got. 243 00:43:26.820 --> 00:43:30.540 Richard Gottscho: And i'm hitting this thing with iron energies of typically 50 volts. 244 00:43:32.220 --> 00:43:41.490 Richard Gottscho: And the reason those numbers are so widely different, and yet this methodology still works is when an iron hits a surface that energy is dissipated. 245 00:43:42.000 --> 00:43:51.180 Richard Gottscho: primarily in the phone ons vibrations of the lattice, if you will, and most of that energy is not effective, we doing. 246 00:43:51.660 --> 00:44:07.230 Richard Gottscho: Chemistry in terms of overcoming this barrier here or overcoming this barrier here so and a rough rule of thumb is to think of 10 to one you're going to lose you know roughly 10 about 10% of the energy is actually efficient in overcoming the barrier. 247 00:44:08.550 --> 00:44:11.850 Richard Gottscho: So it's it's useful to think in terms of an elite window. 248 00:44:13.680 --> 00:44:25.500 Richard Gottscho: That is, if your iron energies to low you're not going to you can struggle to get over this barrier gonna have incomplete removal a le ideology will be less than one, and if the iron energy is too big. 249 00:44:26.610 --> 00:44:30.660 Richard Gottscho: you're going to sputter and you're going to lose a lie benefit so there's a. 250 00:44:31.800 --> 00:44:40.980 Richard Gottscho: A window here of where the you know it's like goldilocks it's you know the this energy is just right. 251 00:44:43.230 --> 00:44:56.220 Richard Gottscho: So it turns out that the size of that window how wide, it is, and from a manufacturing point of view volume manufacturing point of view you want that window to be as large as, as you can possibly get it. 252 00:44:56.760 --> 00:45:03.870 Richard Gottscho: And it's a very nice qualitative correlation between the binding energy of the species. 253 00:45:04.230 --> 00:45:15.630 Richard Gottscho: And the size of that window and it just you can you can understand it in terms of of what you see here if you make this really big it's going to open up the window between the energy needed to get over this barrier. 254 00:45:15.960 --> 00:45:21.660 Richard Gottscho: And yet stay under this barrier, so you you suppress the sputtering and so. 255 00:45:22.380 --> 00:45:31.740 Richard Gottscho: We looked at quite a few elements in the periodic table and and one thing i'll say is that atomic way or etching can basically be done with any. 256 00:45:32.280 --> 00:45:51.000 Richard Gottscho: Material system it's it's a universally it's a universal phenomenon, but the size of the window is not the size of the window scales with with binding energy so here's carbon binding energy 7.7 vs windows roughly 40 volts germanium really hard to a le. 257 00:45:52.920 --> 00:46:04.950 Richard Gottscho: Because the binding energies fairly low compared to the disruption, energy and you kind of have to use your artistic imagination to see that there's a window at all, there you go over here to tan on. 258 00:46:05.340 --> 00:46:21.180 Richard Gottscho: On it's it's huge window because of the high binding energy and silicon the prototypical system everybody likes to study is actually not the best system to study a early on, because it's got a fairly well binding energy not as bad as germanium. 259 00:46:22.500 --> 00:46:34.320 Richard Gottscho: The other thing that that kind of scales, with the binding energy everything's tends to scale with the binding energy, so the sputtering threshold we just talked about that scales, with a binding energy that's straightforward to understand. 260 00:46:35.070 --> 00:46:54.450 Richard Gottscho: The synergy parameter, therefore, because it's connected to this scales, with a binding energy and a euphemism for man it's really etching slowly is higher resolution that is it's easier to control how much material you remove and that it gets a lot easier at higher binding energies. 261 00:46:56.730 --> 00:46:59.250 Richard Gottscho: Because it's it's just more difficult to remove material. 262 00:47:00.780 --> 00:47:02.160 Richard Gottscho: For a given given energy. 263 00:47:03.330 --> 00:47:04.020 Richard Gottscho: So. 264 00:47:05.070 --> 00:47:09.390 Richard Gottscho: Now I want to, I want to take a closer look at this concept of an Ai le window. 265 00:47:10.740 --> 00:47:21.000 Richard Gottscho: And, and point out that the concept actually comes again from the atomic layer deposition world which inspired the atomic layer etching community. 266 00:47:22.110 --> 00:47:35.640 Richard Gottscho: And it was first pointed out by i'm not sure how to say the name per in in in 2005 where he said the temperature range so Al di is largely the energy sources, typically temperature, as opposed to high end bombardment. 267 00:47:36.420 --> 00:47:47.940 Richard Gottscho: Where an amd processor fulfills the requirement of self terminating reactions are really fulfills the promise the benefits of Al di that is the same thing the atomic we smooth. 268 00:47:48.870 --> 00:47:59.520 Richard Gottscho: Know aspect ratio perfectly uniform but there's a finite temperature range and it's it can be explained in similar fashion to what I just did for a lie. 269 00:48:00.060 --> 00:48:08.970 Richard Gottscho: that the primary difference in a lie is we use iron energy instead of temperature as as a dish, and this is the nature of the windows that I just showed. 270 00:48:09.450 --> 00:48:14.700 Richard Gottscho: Now, the thing is it's a little misleading because this window. 271 00:48:15.390 --> 00:48:25.020 Richard Gottscho: This measurement and all the measurements that I just showed, you are done for a fixed iron dose, that is, we we dial up a set of conditions that determines the flux of ions the wafer. 272 00:48:25.500 --> 00:48:35.280 Richard Gottscho: And we leave that flux on for a certain period of time, at a certain energy and we map out this curve and what we discovered very recently. 273 00:48:35.790 --> 00:48:48.630 Richard Gottscho: was actually there isn't one window there's a whole family of windows and the family of windows is a function of the iron dose or forgiven energy. 274 00:48:49.500 --> 00:49:00.090 Richard Gottscho: How long you leave the ions on forgiving energy and flux, how long you even on so here's what i've been showing you mostly are very long time, relatively speaking, five seconds. 275 00:49:01.110 --> 00:49:09.960 Richard Gottscho: at low iron energies and you know the window, this is silicon Korean again, so the windows kind of small and you can see, relative to the sputtering. 276 00:49:11.220 --> 00:49:24.450 Richard Gottscho: If you don't have 100% synergy in fact for silicon Korean argon it's about 80% under these conditions, but if I reduced the time I leave the ions on to two tenths of a second. 277 00:49:25.650 --> 00:49:33.540 Richard Gottscho: holy cow now, I find that first of all my window shifts to very high energy compared to what I had before. 278 00:49:34.770 --> 00:49:36.270 Richard Gottscho: And it's much wider. 279 00:49:37.800 --> 00:49:40.530 Richard Gottscho: it's it's much less dependent on the iron energy. 280 00:49:41.700 --> 00:49:53.580 Richard Gottscho: And these these are normalized scales if I told you, the absolute numbers I i'd have to kill you all but they're they're relative to each other, that the same so basically we're getting the same amount removed. 281 00:49:54.090 --> 00:50:02.580 Richard Gottscho: In a fraction of the time, so if you think about that that means i'm removing stuff much faster if I integrated over a long period of if I do many pulses. 282 00:50:03.420 --> 00:50:17.970 Richard Gottscho: So this is, we believe, one potential a path to the throughput problem of a le is to actually operate under very short pulse conditions iron pulse conditions. 283 00:50:19.170 --> 00:50:33.210 Richard Gottscho: And then we can go to higher energy remember the yield scales is the square root of the energy and so you get a kind of a double benefit here you much shorter cycle times, if you can switch fast enough plus inherently higher X rates. 284 00:50:35.730 --> 00:50:47.940 Richard Gottscho: But there's there's some interesting science here as well, because, as I said, there's a whole family of windows here if I change this time parameter i'm going to get different. 285 00:50:49.350 --> 00:50:56.190 Richard Gottscho: Energy dependencies for the self limiting behavior of the removal step, so we we looked at that and. 286 00:50:57.000 --> 00:51:01.800 Richard Gottscho: Go back to the same theory that I started this talk with, and you can show that the. 287 00:51:02.430 --> 00:51:18.090 Richard Gottscho: The time dependence of the removal of the modified layer is a very simple function one minus exponential and the exponential coefficient is is tap a ti where Kappa is the yield, that is, how many. 288 00:51:18.690 --> 00:51:30.570 Richard Gottscho: surface silicon carbide species, for example, are removed per incident iron times the flux, of the iron and the yield again is energy dependent depends on the square root of the energy. 289 00:51:31.680 --> 00:51:37.920 Richard Gottscho: divided by the surface density of the material and, if you look at a fairly long times were you're asking. 290 00:51:38.640 --> 00:51:55.230 Richard Gottscho: haven't removed 97% of of the modified layer you can kind of forget about the threshold, energy and that that time scales with one over the square root of the eye and energy, which is this curve here, and these are three of our data points. 291 00:51:56.400 --> 00:52:05.580 Richard Gottscho: I haven't I mean we should do more work to fill in the hole curve, but that's where we are, as of today, this was just published earlier this year. 292 00:52:06.150 --> 00:52:16.800 Richard Gottscho: This observation and then, if you take the two families of curves that I showed here, and you plot them instead of as a function of iron energy Kappa te. 293 00:52:17.640 --> 00:52:28.650 Richard Gottscho: Lo and behold, they fall in the same curve, so this is a very powerful concept, it says there's really is only one a le window providing us the right parameter to plot the normalized edge. 294 00:52:29.010 --> 00:52:37.620 Richard Gottscho: amount per cycle against and that's not I an energy but a slightly more complicated function of iron energy times time. 295 00:52:39.540 --> 00:52:41.010 Richard Gottscho: So in conclusion. 296 00:52:42.360 --> 00:52:45.690 Richard Gottscho: That saying really is an art and it's really is a lot of fun. 297 00:52:47.040 --> 00:52:56.130 Richard Gottscho: I mean I started in this business in 1980 and it's 2021 and i'm i'm still discovering things learning things and having fun. 298 00:52:57.210 --> 00:53:05.280 Richard Gottscho: And the kinds of things that we're watching and the way in which we're watching today I couldn't even imagine 10 years ago, let alone 40 years ago. 299 00:53:07.230 --> 00:53:18.420 Richard Gottscho: Atomic way or etching is what we're referring to as the new art form for etching it's much simpler simpler to design and control because the coupling between neutrals and ions is eliminated. 300 00:53:19.140 --> 00:53:24.780 Richard Gottscho: In fact there's a whole new world here that that people haven't even begun to explore, and that is. 301 00:53:25.380 --> 00:53:32.070 Richard Gottscho: When you do everything at the same time, you get the soup that I referred to it's very difficult to control the concentrations of species. 302 00:53:32.580 --> 00:53:45.090 Richard Gottscho: When you break things up into self limiting steps you can actually do chemistry on the surface and know what you've done and design molecules the way you want them to bind to the surface. 303 00:53:45.690 --> 00:54:03.900 Richard Gottscho: Like you can't do in normal plasma processing that's a whole new field that actually hasn't hasn't even gotten off the ground, yet people are still using all the same old suspects as as far as actions are concerned, typically you know bromine chlorine flooring oxygen hydrogen. 304 00:54:05.190 --> 00:54:11.880 Richard Gottscho: Much more complicated molecular species can be envision that would give you an unprecedented level of control. 305 00:54:13.020 --> 00:54:16.110 Richard Gottscho: The process window scales, with a surface binding energy. 306 00:54:17.460 --> 00:54:31.920 Richard Gottscho: critically important use inert gas ions and step B or you'll get all that mixing and you lose all the benefits of atomic way retching benefits or smoothing aspect ratio independence macroscopic uniformity really do atomic dimensions. 307 00:54:32.940 --> 00:54:45.240 Richard Gottscho: throughput and management of non synergistic effects remain as the big challenges and very active areas of research that's where I think designer molecules could play a much bigger role in the future of etching. 308 00:54:46.830 --> 00:54:52.380 Richard Gottscho: smoothing as an inherent benefit of a we for all the reasons that that I showed. 309 00:54:54.000 --> 00:55:08.070 Richard Gottscho: And we've uncovered a universal scaling relationship that unifies all the various windows for a le and collapses it into one functional curve so with that i'll just say that. 310 00:55:09.510 --> 00:55:21.870 Richard Gottscho: we've been rethinking the art of batch going from weathering a sandstorm etching a metal phases latest is we're doing a le on extreme ultraviolet resist or. 311 00:55:22.950 --> 00:55:36.750 Richard Gottscho: And, and in so doing, we can we can it's really critical to make very rounds smooth holes, to make the most advanced integrated circuits that go into the most advanced systems. 312 00:55:37.560 --> 00:55:47.940 Richard Gottscho: on earth today, and with that i'll close acknowledging my fellow artists who worked with me on all the work just presented, and in particular I want to acknowledge. 313 00:55:50.310 --> 00:55:53.550 Richard Gottscho: Okay there it is Karen can Eric who. 314 00:55:54.690 --> 00:56:02.820 Richard Gottscho: Was the lead investigator on all this work put these slides together and has just been a fantastic colleague for me to work with. 315 00:56:04.170 --> 00:56:12.540 Richard Gottscho: And and really is all these people, but especially Karen make it make it so much more fun Thank you and i'm open. 316 00:56:13.770 --> 00:56:15.480 Richard Gottscho: happy to take questions or comments. 317 00:56:19.740 --> 00:56:34.410 Mark Kushner: So rick Thank you very much for that presentation it's great to see that the progress and the the advances have been made over the last many years, so are there questions from the audience. 318 00:56:36.120 --> 00:56:50.670 Mark Kushner: And you can either put it in the chat or just unmute yourself and ask the question we have one question in the chat from Andre Andre do you want to unmute yourself and ask the question yourself. 319 00:56:51.450 --> 00:57:10.680 Andre Fils Antoine: Sure, I guess um yeah thanks for the talk and yeah my name is Roger so I just want to ask on I heard you mentioned the photons I guess causing like a spontaneous sputtering and I guess that sounds like an unwanted effect so if that's true like, how do you mitigate that. 320 00:57:13.380 --> 00:57:24.900 Richard Gottscho: Yes, it is an unwanted effect in general, although photon assisted etching is a whole technology in and of itself it's never become mainstream people have worked on it for many years. 321 00:57:25.950 --> 00:57:35.490 Richard Gottscho: There are some applications where it's where it's done there's also photo induced electro chemistry, but in general in the for the applications, I was referring to its unwanted. 322 00:57:36.030 --> 00:57:48.960 Richard Gottscho: And I don't know if we I don't think we are really controlling it, the one thing that you can do is control surface temperature and and and you know if you get the temperature down even a photons absorb. 323 00:57:51.390 --> 00:57:51.990 Richard Gottscho: You can. 324 00:57:53.100 --> 00:57:56.400 Richard Gottscho: minimize those effects be one strategy. 325 00:57:58.560 --> 00:58:05.490 Richard Gottscho: Whereas, even if you over the temperature the ions still have enough energy to overcome the the the disruption barrier. 326 00:58:06.690 --> 00:58:19.470 Richard Gottscho: But my belief is that these effects are going on in our reactors, they are producing some low level of spontaneous setting, and so it and they'll they'll detract from the ideology of the elite process. 327 00:58:20.790 --> 00:58:25.800 Richard Gottscho: And it's it's you can, as you twiddle the knobs on a reactor. 328 00:58:27.330 --> 00:58:36.990 Richard Gottscho: And there are a lot of different combinations of knobs like more than 10 to the 14 that you can create you are undoubtedly changing the spectrum of radiation that's hitting the wafer. 329 00:58:37.470 --> 00:58:54.930 Richard Gottscho: And so I think the artists are finding solutions around that problem without actually realizing it, so I think it's a really big question about how you can control or minimize photon induced reactions in such an environment. 330 00:58:56.070 --> 00:59:01.050 Richard Gottscho: Very then there's a bunch of people working on trying to solve that problem that's a really big problem. 331 00:59:02.400 --> 00:59:12.420 Richard Gottscho: I don't think it's really big problem from a practical point of view we actually don't understand how big it is because it's inherently there and it's very hard to isolate the effects. 332 00:59:15.450 --> 00:59:16.170 Andre Fils Antoine: Okay, thank you. 333 00:59:19.260 --> 00:59:21.150 Mark Kushner: Any other questions. 334 00:59:24.660 --> 00:59:25.470 Mark Kushner: Okay i'll see. 335 00:59:25.530 --> 00:59:27.240 Richard Gottscho: This I see one here about. 336 00:59:27.300 --> 00:59:31.410 Mark Kushner: hope it does disappear, please go ahead, ask the question unmute yourself. 337 00:59:37.200 --> 00:59:37.620 Mark Kushner: Oh. 338 00:59:37.800 --> 00:59:46.440 Richard Gottscho: So I got what I see one here do you foresee the use of Ai le and higher aspect ratio etch applications such as channel whole action in the near future. 339 00:59:49.620 --> 01:00:05.730 Richard Gottscho: I hesitate to say never about anything, because every time I say something like that I get proven wrong, like the next day, but that is a very challenging applications, one that we're actually trying to tackle but it's it's difficult because. 340 01:00:07.410 --> 01:00:17.730 Richard Gottscho: In highest ratio feature you still in an area we you have this dosing step or modification step step, a Korean modifying silicon, for example. 341 01:00:18.300 --> 01:00:30.570 Richard Gottscho: And you got to get those reactants down to the bottom of the highest pick ratio feature you you don't want to do anything to the sidewalls you just want to etch the bottom, and if you start going through the math and there's experimental work. 342 01:00:31.680 --> 01:00:37.050 Richard Gottscho: For an aspect ratio of 50 to one it is many, many minutes. 343 01:00:37.740 --> 01:00:47.490 Richard Gottscho: So it's not practical in fact what's really happening, even in the conventional plasma reactor, we believe, is the transport of the reactive species. 344 01:00:47.850 --> 01:00:56.280 Richard Gottscho: occurs from starts with an iron in the gap in the plasma state as it crosses the sheath and approaches the surface it's o'shea neutralized. 345 01:00:56.670 --> 01:01:07.440 Richard Gottscho: And comes down as a fast neutral, otherwise we wouldn't even have practical etching processes today for for channel whole action in 3D Nan. 346 01:01:07.920 --> 01:01:20.520 Richard Gottscho: So, if you think about that as the mechanism for the dosing and you start with a high energy reactive species it's what I told you, you don't want to do to get atomic. 347 01:01:23.250 --> 01:01:36.000 Richard Gottscho: smoothness on the bottom surface or a daily behavior so there's some really big challenges and highest ratio effing i'm not saying they want to be overcome, but they haven't been overcome, as of today. 348 01:01:36.270 --> 01:01:37.980 Richard Gottscho: it's big opportunity for breakthrough. 349 01:01:39.060 --> 01:01:46.530 Fred Lewis Terry: So I think that that gets partly at what I was frightening about is whether or not you still have to fret about. 350 01:01:48.000 --> 01:01:59.280 Fred Lewis Terry: Doing sidewall passive ation tracks and and also worrying about charging of the sidewalls and and focusing and deflection of ions and all of that, even in the world. 351 01:02:02.610 --> 01:02:13.410 Richard Gottscho: yeah I think in general Fred the answer to that is yes, you do need to worry about that, and in a lot of our modeling work for Ai we we haven't really thought deeply. 352 01:02:13.740 --> 01:02:27.840 Richard Gottscho: Yet about the sidewall passive ation, but if you imagine you've you've somehow passive aided the sidewall and the inert gas ions you're coming in with our in the right energy regime. 353 01:02:29.520 --> 01:02:34.200 Richard Gottscho: I think you're you're not as likely to attack that sidewall, as you will, on the bottom. 354 01:02:35.280 --> 01:02:41.100 Richard Gottscho: But it's I think that's also still an open question we I don't think we fully understand. 355 01:02:42.540 --> 01:02:46.950 Richard Gottscho: The how elite how well it will work or won't work. 356 01:02:48.030 --> 01:02:52.140 Richard Gottscho: as a function of how your past evading the sidewall as a function of the aspect ratios. 357 01:02:55.530 --> 01:03:15.150 Mark Kushner: another couple of questions from the chat from Tom nail Horn, do you foresee adopting new hybrid a le Al di techniques such as he been created plasmas, as demonstrated by nrl large area plasma processing source. 358 01:03:16.710 --> 01:03:17.250 Richard Gottscho: well. 359 01:03:18.750 --> 01:03:20.040 Richard Gottscho: two parts of that question. 360 01:03:22.110 --> 01:03:34.500 Richard Gottscho: Al di processes have already been incorporated into commercial plasma edge solutions i'm not necessarily in concert with ellie. 361 01:03:35.040 --> 01:03:43.590 Richard Gottscho: But just as a way of doing conformal passive Asian followed by not a conventional Plaza process because nobody's really. 362 01:03:44.190 --> 01:03:52.230 Richard Gottscho: Processing anymore with things on all the time it's a Pole sequence but it's not in the daily limit where things are self limiting. 363 01:03:52.830 --> 01:04:12.180 Richard Gottscho: But the out so Al di and combined combined with with pulsed plasma etching is actually in volume production today's commercial practice now electron beam induced plasmas I confess I don't know too much about that particular project or about those plasmas in general. 364 01:04:13.500 --> 01:04:16.410 Richard Gottscho: And I so maybe i'll just leave it at that. 365 01:04:19.050 --> 01:04:25.140 Richard Gottscho: yeah I think i'll just leave it at that I I i'm not quite I mean, I know that there's been a lot of work in that area. 366 01:04:25.920 --> 01:04:44.970 Richard Gottscho: But it has yet to reach commercial implementation, and I think that's probably because the benefits of driving a plasma electron beam injection haven't been substantially better than a conventional inductive we coupled plasma that's post or capacitive we couple plasmid let's pulse. 367 01:04:46.050 --> 01:04:50.580 Richard Gottscho: But again, my caveat is I don't know that much about those experiments. 368 01:04:52.440 --> 01:04:55.080 Mark Kushner: Another question from the chat from Nick. 369 01:04:56.130 --> 01:05:11.040 Mark Kushner: Thank you very interesting talk, you mentioned that plasmas are used primarily for an n isotopic etching but also that it Tropic application to required in the future, do you think plasmids can also be utilized for isotopic etching oh. 370 01:05:11.160 --> 01:05:13.710 Richard Gottscho: Absolutely we're doing that now it's. 371 01:05:14.820 --> 01:05:22.230 Richard Gottscho: Basically, if you turn off the bias, or you or you work or use the plasma as a source of neutrals. 372 01:05:23.940 --> 01:05:44.610 Richard Gottscho: But eliminate the the ions by passing those neutrals through gritted plate, for example, then you can do isotopic etching with no iron bombardment and know photon bombardment and there are products that we enter a competition sell today and semiconductor FABs that do that. 373 01:05:46.230 --> 01:05:47.730 Richard Gottscho: You can also imagine. 374 01:05:49.080 --> 01:05:58.470 Richard Gottscho: thermal we driven processes with reagents that are not generated in a plasma, so a plasma was isotopic edge. 375 01:05:59.460 --> 01:06:06.540 Richard Gottscho: reaction Steve George at university of Colorado who we've been working with a lot on this has done a lot of work in this area. 376 01:06:07.110 --> 01:06:24.870 Richard Gottscho: And you can do actually atomic way or etching under those conditions where you modify the surface again you, you can change your gas environment bring in a another reacting and use temperature to control the rate of that reaction those tend to be highly selective. 377 01:06:27.870 --> 01:06:31.050 Richard Gottscho: isotopic etches with atomic scale precision. 378 01:06:32.190 --> 01:06:33.720 Richard Gottscho: And those kinds of. 379 01:06:34.920 --> 01:06:47.010 Richard Gottscho: processes are absolutely essential for making what so called the so called gate all around or nano sheet devices, where you have to basically. 380 01:06:48.210 --> 01:06:49.770 Richard Gottscho: You know you create a. 381 01:06:51.540 --> 01:07:02.220 Richard Gottscho: Free standing being by removing the material all around and anchoring it on both ends that's that's a nano sheet or nanowire and that's done by highly selective isotopic etching. 382 01:07:03.960 --> 01:07:08.970 Mark Kushner: A potential take one last question urban you have your hand up. 383 01:07:10.440 --> 01:07:17.820 Erwin: Yes, thank you mark this is our castles and over university, so I was wondering, I very much. 384 01:07:18.420 --> 01:07:33.060 Erwin: need this nice results with these very high iron iron energies and these balls plus master that's pretty exciting, do you think that the future of alias, especially in that direction, or do you think there was also a lot of work to be done at the lower energies. 385 01:07:34.950 --> 01:07:44.910 Richard Gottscho: i'm pretty biased pardon the pun towards the high energy regime for the reasons that I said it has inherently higher X Ray and therefore throughput. 386 01:07:46.620 --> 01:07:55.170 Richard Gottscho: And that's probably the biggest reason it also because the windows so much bigger from a manufacturing control perspective it's much more forgiving. 387 01:07:56.610 --> 01:08:06.540 Richard Gottscho: So that that's the way I think this will shape up now Eo, even though I showed you in 2016 lamb introduced an elite process based on. 388 01:08:06.990 --> 01:08:19.080 Richard Gottscho: marks simulation work 10 years earlier, is in volume production, there are very few elite processes actually in production that's an exception to the rule and the reason is low throughput primarily. 389 01:08:19.800 --> 01:08:30.720 Richard Gottscho: And so I think the high energy regime is one of the things necessary to implement to get around that throughput limitation it's not enough. 390 01:08:31.800 --> 01:08:36.210 Richard Gottscho: But it's one of the it's one of the one of the approaches that that we're pursuing. 391 01:08:38.040 --> 01:08:39.360 Erwin: Thank you yeah. 392 01:08:40.230 --> 01:08:40.560 and 393 01:08:41.610 --> 01:08:45.210 Mark Kushner: Maybe i'll reserve the very last question for me. 394 01:08:46.560 --> 01:08:59.100 Mark Kushner: Now as moore's law from a dimensional perspective is are grinding to a halt, more advances and Mike electronics are coming from more complex use of materials more complex structures. 395 01:08:59.760 --> 01:09:10.860 Mark Kushner: You think that that Al way is going to become even more important because it can be more selective with these new large variety of new materials. 396 01:09:11.460 --> 01:09:21.570 Richard Gottscho: Yes, absolutely I take issue with your original premise, but besides that scaling is continuing as well, but, but your anyway. 397 01:09:22.200 --> 01:09:37.080 Richard Gottscho: Yes, that's what it's trying to say in the in the gate all around structures are Nana sheets, they are highly reliant on very selective isotopic edges and I believe a lot of those solutions will end up being done by a le. 398 01:09:38.370 --> 01:09:40.380 Richard Gottscho: The primary challenge again. 399 01:09:41.400 --> 01:09:49.920 Richard Gottscho: In a situation like that is getting the throughput up and in a lot of those reactions you don't have any I am bombardments so you don't have to worry about sputtering. 400 01:09:50.370 --> 01:09:57.150 Richard Gottscho: But you still have to control any spontaneous chemistry, but that's a lot easier to do in that environment than in a plasma environment. 401 01:09:59.250 --> 01:10:12.690 Mark Kushner: Eric, thank you very, very much for the seminar and for taking this long period for for questions, we hope that we'll be able to get you out here in person at at some point where we're working on it. 402 01:10:12.960 --> 01:10:20.040 Richard Gottscho: make me too, and thank you for the opportunity to present our work to you and and for all the questions. 403 01:10:21.450 --> 01:10:23.610 Richard Gottscho: very, very provocative Thank you yeah. 404 01:10:23.700 --> 01:10:29.100 Mark Kushner: Thank you and thank you all view attendees we hope to see you at our next mitzi seminar. 405 01:10:32.010 --> 01:10:32.310 Richard Gottscho: bye.