WEBVTT 00:00:00.000 --> 00:00:01.550 align:middle line:84% JUDY HOYT: The curvature of the surface, 00:00:01.550 --> 00:00:03.342 align:middle line:84% and they want to look at the oxidation rate 00:00:03.342 --> 00:00:06.920 align:middle line:84% as a function of how curved the surface was. 00:00:06.920 --> 00:00:09.620 align:middle line:84% So this is a top view or a bird's eye 00:00:09.620 --> 00:00:16.070 align:middle line:84% view here of his structure in part A. But look on slide 3. 00:00:16.070 --> 00:00:18.870 align:middle line:84% Part B you're looking at a side view here. 00:00:18.870 --> 00:00:22.100 align:middle line:84% So you can see he had etched this cylindrical structure 00:00:22.100 --> 00:00:23.410 align:middle line:90% in the top. 00:00:23.410 --> 00:00:26.300 align:middle line:90% It's got a circular pattern. 00:00:26.300 --> 00:00:29.810 align:middle line:84% After preparing this surface, he then put it in the furnace 00:00:29.810 --> 00:00:33.650 align:middle line:84% and grew this thermal oxide on all those free surfaces, 00:00:33.650 --> 00:00:35.030 align:middle line:90% which is shown here in blue. 00:00:35.030 --> 00:00:37.650 align:middle line:90% That's the silicon dioxide. 00:00:37.650 --> 00:00:39.860 align:middle line:90% So that was a real experiment. 00:00:39.860 --> 00:00:42.620 align:middle line:84% And then, just like we saw in the previous photo 00:00:42.620 --> 00:00:44.300 align:middle line:84% from Marcus and Shang, he put down-- 00:00:44.300 --> 00:00:46.730 align:middle line:84% he deposited polycrystalline silicon, 00:00:46.730 --> 00:00:50.090 align:middle line:84% and that's simply used for contrast during the SEM 00:00:50.090 --> 00:00:52.070 align:middle line:84% analysis and to protect the oxide. 00:00:52.070 --> 00:00:54.290 align:middle line:84% So he has got this polysilicon over everything, 00:00:54.290 --> 00:00:56.900 align:middle line:84% so from a top view all you would see is the poly. 00:00:56.900 --> 00:01:00.420 align:middle line:84% And then to reveal this oxide and what 00:01:00.420 --> 00:01:02.790 align:middle line:84% it looks like looking down, he then 00:01:02.790 --> 00:01:06.720 align:middle line:84% lapped or polished off the top, the poly. 00:01:06.720 --> 00:01:09.810 align:middle line:84% So he's about halfway down the pillar, partway 00:01:09.810 --> 00:01:10.500 align:middle line:90% down the pillar. 00:01:10.500 --> 00:01:12.850 align:middle line:84% And then a top view is shown here in the lower right. 00:01:12.850 --> 00:01:14.587 align:middle line:90% You can what he would see. 00:01:14.587 --> 00:01:16.420 align:middle line:84% You'd see this polysilicon, which is, again, 00:01:16.420 --> 00:01:17.520 align:middle line:90% just a contrast medium. 00:01:17.520 --> 00:01:21.150 align:middle line:84% But you would see this white ring here 00:01:21.150 --> 00:01:23.340 align:middle line:84% is the silicon that was initially oxidized, 00:01:23.340 --> 00:01:26.070 align:middle line:84% and the blue ring all around it is the oxide that 00:01:26.070 --> 00:01:27.780 align:middle line:84% grew on the inner part of the cylinder 00:01:27.780 --> 00:01:29.460 align:middle line:84% and the outer part of the cylinder. 00:01:29.460 --> 00:01:31.320 align:middle line:84% And then you can vary the cylinder size. 00:01:31.320 --> 00:01:33.180 align:middle line:84% He can create a whole lot of these cylinders 00:01:33.180 --> 00:01:34.380 align:middle line:90% on the same wafer. 00:01:34.380 --> 00:01:37.410 align:middle line:84% And look at how the oxide grows, what 00:01:37.410 --> 00:01:39.090 align:middle line:84% its oxidation rate is as a function 00:01:39.090 --> 00:01:40.690 align:middle line:90% of the radius of curvature. 00:01:40.690 --> 00:01:44.700 align:middle line:84% So it's a very nice scientific experiment 00:01:44.700 --> 00:01:48.820 align:middle line:90% to look at these dependencies. 00:01:48.820 --> 00:01:53.290 align:middle line:84% If you go to slide number 4, these are some actual-- 00:01:53.290 --> 00:01:56.590 align:middle line:84% the upper left is a scanning electron micrograph. 00:01:56.590 --> 00:01:58.930 align:middle line:84% Not transmission, but it's an SEM, 00:01:58.930 --> 00:02:01.930 align:middle line:84% so we're looking at backscattered electrons. 00:02:01.930 --> 00:02:04.690 align:middle line:84% And there is some contrast associated 00:02:04.690 --> 00:02:06.860 align:middle line:90% with the different materials. 00:02:06.860 --> 00:02:10.479 align:middle line:84% And this is sort of a typical result of a plan view SEM. 00:02:10.479 --> 00:02:15.580 align:middle line:84% And an artist's drawing of what is seen in the SEM 00:02:15.580 --> 00:02:17.420 align:middle line:90% is shown down here. 00:02:17.420 --> 00:02:19.510 align:middle line:84% So just looking first at the SEM, 00:02:19.510 --> 00:02:22.690 align:middle line:84% this wall right here-- so now we're looking down on the chip 00:02:22.690 --> 00:02:25.610 align:middle line:84% or on the surface of the wafer, and he's etched a wall. 00:02:25.610 --> 00:02:26.110 align:middle line:90% Right? 00:02:26.110 --> 00:02:28.580 align:middle line:84% Looking down along certain directions. 00:02:28.580 --> 00:02:36.280 align:middle line:84% This is probably along the 110, the 110 equivalent, 110 planes. 00:02:36.280 --> 00:02:37.975 align:middle line:84% He's also etched these regions here. 00:02:37.975 --> 00:02:39.100 align:middle line:90% Look at on the upper right. 00:02:39.100 --> 00:02:41.600 align:middle line:90% This is a cylinder of silicon. 00:02:41.600 --> 00:02:45.580 align:middle line:84% This darker region is the SiO2, and surrounded by that 00:02:45.580 --> 00:02:46.377 align:middle line:90% is polysilicon. 00:02:46.377 --> 00:02:48.710 align:middle line:84% If you want to look down here, it's a little bit easier. 00:02:48.710 --> 00:02:51.920 align:middle line:90% Everything's labeled. 00:02:51.920 --> 00:02:56.210 align:middle line:84% The thing you see right away is that the oxide grown 00:02:56.210 --> 00:03:00.120 align:middle line:84% is thinner for both a concave and a convex corner. 00:03:00.120 --> 00:03:03.530 align:middle line:84% So here's an example right here where the silicon was rounded, 00:03:03.530 --> 00:03:05.000 align:middle line:90% but it's in a concave corner. 00:03:05.000 --> 00:03:07.850 align:middle line:84% And the thickness of the oxide, which is this dark region, 00:03:07.850 --> 00:03:12.020 align:middle line:84% is much thinner here than it is on a flat surface here or here. 00:03:12.020 --> 00:03:15.560 align:middle line:84% So it's thinning up when you're oxidizing a corner, 00:03:15.560 --> 00:03:17.750 align:middle line:84% and even a convex corner as well, 00:03:17.750 --> 00:03:20.600 align:middle line:84% the oxide growing is a little thinner. 00:03:20.600 --> 00:03:25.310 align:middle line:90% 00:03:25.310 --> 00:03:28.517 align:middle line:84% In fact, the effect seems to be a little more pronounced 00:03:28.517 --> 00:03:31.100 align:middle line:84% for this concave corner where it really looks like there's not 00:03:31.100 --> 00:03:34.560 align:middle line:90% much oxide grown here at all. 00:03:34.560 --> 00:03:37.400 align:middle line:84% So let's say we're just talking about the oxidation 00:03:37.400 --> 00:03:38.490 align:middle line:90% of the cylinders. 00:03:38.490 --> 00:03:41.510 align:middle line:84% Let's say just the silicon cylinder, this pillar right 00:03:41.510 --> 00:03:44.660 align:middle line:84% here, and there are differences in the oxidation rate 00:03:44.660 --> 00:03:47.600 align:middle line:90% along the different directions. 00:03:47.600 --> 00:03:49.310 align:middle line:90% What would you think? 00:03:49.310 --> 00:03:50.720 align:middle line:90% People have any suggestions? 00:03:50.720 --> 00:03:52.130 align:middle line:90% Anybody have any ideas? 00:03:52.130 --> 00:03:55.810 align:middle line:84% If I take a cylinder of silicon that was etched-- 00:03:55.810 --> 00:04:01.180 align:middle line:84% the wafer was originally 100 and I have a cylindrical surface. 00:04:01.180 --> 00:04:03.820 align:middle line:84% Let's say I'm growing in the thin oxide regime. 00:04:03.820 --> 00:04:06.358 align:middle line:84% Why do you think you might get different thicknesses going 00:04:06.358 --> 00:04:07.900 align:middle line:84% around the perimeter of the cylinder? 00:04:07.900 --> 00:04:08.650 align:middle line:90% What's one reason? 00:04:08.650 --> 00:04:12.740 align:middle line:90% 00:04:12.740 --> 00:04:14.550 align:middle line:90% Any ideas? 00:04:14.550 --> 00:04:18.750 align:middle line:84% How about thinking about what's the orientation of those planes 00:04:18.750 --> 00:04:19.890 align:middle line:90% as I make a cylinder. 00:04:19.890 --> 00:04:24.360 align:middle line:84% If I etch a cylinder, I'm exposing all different planar 00:04:24.360 --> 00:04:25.120 align:middle line:90% orientations. 00:04:25.120 --> 00:04:27.520 align:middle line:84% So locally, at each point in the cylinder, 00:04:27.520 --> 00:04:31.260 align:middle line:84% there's sort of a different crystal plane that's exposed. 00:04:31.260 --> 00:04:33.510 align:middle line:84% So you might expect, if you're in the thin oxide 00:04:33.510 --> 00:04:35.510 align:middle line:84% regime, at least there might be some differences 00:04:35.510 --> 00:04:38.620 align:middle line:84% due to the fact that I'm looking at different orientations. 00:04:38.620 --> 00:04:41.670 align:middle line:90% So that's one effect. 00:04:41.670 --> 00:04:44.760 align:middle line:84% Any other ideas on any other potential 00:04:44.760 --> 00:04:47.045 align:middle line:84% effects that you might think about? 00:04:47.045 --> 00:04:48.420 align:middle line:84% Etching a small cylinder compared 00:04:48.420 --> 00:04:51.270 align:middle line:90% to etching a flat wafer. 00:04:51.270 --> 00:04:53.680 align:middle line:90% Orientation is one. 00:04:53.680 --> 00:04:54.180 align:middle line:90% All right. 00:04:54.180 --> 00:04:55.890 align:middle line:84% Well, let's go ahead and see, then we'll 00:04:55.890 --> 00:04:59.860 align:middle line:84% give the answer away in the next couple slides. 00:04:59.860 --> 00:05:02.285 align:middle line:84% In fact, you go on to slide 5, and it doesn't completely 00:05:02.285 --> 00:05:04.660 align:middle line:84% give you the answer but it shows you at least the results 00:05:04.660 --> 00:05:07.470 align:middle line:90% that Gao got. 00:05:07.470 --> 00:05:11.760 align:middle line:84% This is a plot from his work, and the y-axis 00:05:11.760 --> 00:05:13.840 align:middle line:84% is the normalized oxide thickness. 00:05:13.840 --> 00:05:16.930 align:middle line:84% So this is normalized thickness to the thickness 00:05:16.930 --> 00:05:18.710 align:middle line:84% he grew on the flat surface of the wafer. 00:05:18.710 --> 00:05:20.590 align:middle line:84% So he measured the thickness on the flat part 00:05:20.590 --> 00:05:23.050 align:middle line:84% and then he measured the thickness on these cylinder 00:05:23.050 --> 00:05:26.740 align:middle line:90% sides and he just divided. 00:05:26.740 --> 00:05:32.200 align:middle line:84% On the flat part he grew about 500 nanometers or half a micron 00:05:32.200 --> 00:05:35.020 align:middle line:90% of oxide on all surfaces. 00:05:35.020 --> 00:05:36.850 align:middle line:84% And by the way, a flat surface here-- 00:05:36.850 --> 00:05:39.760 align:middle line:84% he's plotting this normalized thickness as a function of 1 00:05:39.760 --> 00:05:43.730 align:middle line:84% over r, where r is the radius of the cylinder. 00:05:43.730 --> 00:05:47.710 align:middle line:84% So if you're on a flat surface, the radius of curvature, 00:05:47.710 --> 00:05:48.860 align:middle line:90% the radius is infinity. 00:05:48.860 --> 00:05:50.170 align:middle line:90% So 1 over r is zero. 00:05:50.170 --> 00:05:52.697 align:middle line:84% So right at this point zero, this corresponds to-- and all 00:05:52.697 --> 00:05:54.280 align:middle line:84% the curves converge, that means you're 00:05:54.280 --> 00:05:55.970 align:middle line:90% oxidizing a flat surface. 00:05:55.970 --> 00:05:59.050 align:middle line:84% As you go out to higher and higher, 1 over r, 00:05:59.050 --> 00:06:01.730 align:middle line:84% that means that r itself is getting smaller 00:06:01.730 --> 00:06:05.230 align:middle line:84% so you're getting a tighter and tighter-- a smaller diameter 00:06:05.230 --> 00:06:09.310 align:middle line:84% cylinder just so you get yourself calibrated. 00:06:09.310 --> 00:06:11.440 align:middle line:84% And he's got two different types of curves 00:06:11.440 --> 00:06:16.330 align:middle line:90% here, convex radii and concave. 00:06:16.330 --> 00:06:18.730 align:middle line:90% If we just go back one-- 00:06:18.730 --> 00:06:20.860 align:middle line:90% whoops. 00:06:20.860 --> 00:06:24.250 align:middle line:84% One way, one slide to slide number 4, 00:06:24.250 --> 00:06:26.550 align:middle line:90% here's a concave cylinder. 00:06:26.550 --> 00:06:29.270 align:middle line:84% So you're oxidizing on the inside of a surface, 00:06:29.270 --> 00:06:31.700 align:middle line:84% like you're in the cave here so you're oxidizing. 00:06:31.700 --> 00:06:34.280 align:middle line:90% And here's a convex surface. 00:06:34.280 --> 00:06:36.020 align:middle line:84% You have silica and you're oxidizing 00:06:36.020 --> 00:06:37.670 align:middle line:90% on the outside of the cylinder. 00:06:37.670 --> 00:06:40.610 align:middle line:90% So you can do it either way. 00:06:40.610 --> 00:06:42.170 align:middle line:84% You can imagine just geometrically, 00:06:42.170 --> 00:06:43.790 align:middle line:84% knowing that oxide has to expand, 00:06:43.790 --> 00:06:46.970 align:middle line:84% it would be different if you're oxidizing inside a cave, 00:06:46.970 --> 00:06:50.690 align:middle line:84% in a concave area, versus outside on the outer surface. 00:06:50.690 --> 00:06:52.730 align:middle line:90% And in fact, you see this. 00:06:52.730 --> 00:06:56.030 align:middle line:84% Look at the concave radii, all these 00:06:56.030 --> 00:06:57.950 align:middle line:84% dashed lines compared to the solid lines. 00:06:57.950 --> 00:07:01.520 align:middle line:84% The normalized oxide thickness at a given temperature, 00:07:01.520 --> 00:07:04.550 align:middle line:84% say here at 900 degrees, is a lot thinner 00:07:04.550 --> 00:07:07.020 align:middle line:84% for the dashed compared to the solid. 00:07:07.020 --> 00:07:12.440 align:middle line:84% So a concave cylinder, he found a lot less lower oxidation 00:07:12.440 --> 00:07:15.270 align:middle line:84% rate, oxidizing the inside of the cylinder as opposed 00:07:15.270 --> 00:07:16.145 align:middle line:90% to the outer surface. 00:07:16.145 --> 00:07:19.400 align:middle line:90% 00:07:19.400 --> 00:07:22.670 align:middle line:84% And in addition to you see that difference right 00:07:22.670 --> 00:07:24.800 align:middle line:84% off the solid versus the dashed, the other thing 00:07:24.800 --> 00:07:27.230 align:middle line:84% is look at it as a function of temperature. 00:07:27.230 --> 00:07:31.220 align:middle line:84% 1,200 degrees, it's pretty darn flat. 00:07:31.220 --> 00:07:32.880 align:middle line:90% 1,100, almost flat. 00:07:32.880 --> 00:07:35.090 align:middle line:84% So at those temperatures, the oxidation rate 00:07:35.090 --> 00:07:38.210 align:middle line:84% doesn't seem to depend very much on the curvature 00:07:38.210 --> 00:07:39.800 align:middle line:90% of the cylinder at all. 00:07:39.800 --> 00:07:41.450 align:middle line:90% It's oxidizing equally rapidly. 00:07:41.450 --> 00:07:43.220 align:middle line:84% But at low temperatures, so should I 00:07:43.220 --> 00:07:47.360 align:middle line:84% go down to right between 1,000 and 900, 00:07:47.360 --> 00:07:50.960 align:middle line:84% you get a big retardation in the oxidation rate. 00:07:50.960 --> 00:07:55.160 align:middle line:84% So there's a much more pronounced effect when 00:07:55.160 --> 00:07:56.943 align:middle line:90% you're at low temperatures. 00:07:56.943 --> 00:07:59.360 align:middle line:84% That's kind of consistent what we saw in Marcus and Shang, 00:07:59.360 --> 00:08:02.750 align:middle line:84% the 950 degrees C pillar oxide had a much more 00:08:02.750 --> 00:08:03.990 align:middle line:90% of a corner effect. 00:08:03.990 --> 00:08:07.160 align:middle line:84% So this is scientifically in agreement 00:08:07.160 --> 00:08:11.030 align:middle line:84% with that qualitative result. OK. 00:08:11.030 --> 00:08:13.340 align:middle line:84% So this was the actual data that he got. 00:08:13.340 --> 00:08:15.620 align:middle line:84% And given this data, then he had to come up 00:08:15.620 --> 00:08:17.420 align:middle line:90% with some kind of explanation. 00:08:17.420 --> 00:08:20.090 align:middle line:84% That's what we're going to talk about. 00:08:20.090 --> 00:08:23.430 align:middle line:84% What are the physical mechanisms to explain his results? 00:08:23.430 --> 00:08:25.910 align:middle line:84% The first one we already said because we know it 00:08:25.910 --> 00:08:28.160 align:middle line:84% from last time, relatively simple, 00:08:28.160 --> 00:08:32.120 align:middle line:84% is the fact that the crystal orientation of the silicon 00:08:32.120 --> 00:08:36.530 align:middle line:84% is changing along the surface of a cylinder or any surface 00:08:36.530 --> 00:08:38.150 align:middle line:90% that's not flat. 00:08:38.150 --> 00:08:40.340 align:middle line:84% And we know in the thin oxide regime, that 00:08:40.340 --> 00:08:41.780 align:middle line:90% impacts the oxidation rate. 00:08:41.780 --> 00:08:43.610 align:middle line:84% So at sufficiently low temperatures 00:08:43.610 --> 00:08:46.820 align:middle line:84% or when he was growing the thinner oxides, 00:08:46.820 --> 00:08:48.270 align:middle line:90% that's going to have an effect. 00:08:48.270 --> 00:08:51.380 align:middle line:84% Now, of course, it's said that on the flat regions 00:08:51.380 --> 00:08:54.160 align:middle line:90% he was growing 5,000 angstroms. 00:08:54.160 --> 00:08:55.498 align:middle line:90% That's not particularly thin. 00:08:55.498 --> 00:08:58.040 align:middle line:84% So depending on the temperature which he was growing that at, 00:08:58.040 --> 00:09:02.728 align:middle line:84% you probably don't expect a huge effect in his results 00:09:02.728 --> 00:09:03.770 align:middle line:90% from crystal orientation. 00:09:03.770 --> 00:09:06.440 align:middle line:84% But nevertheless, you have to take into account. 00:09:06.440 --> 00:09:08.450 align:middle line:84% If you're oxidizing a trench and you're 00:09:08.450 --> 00:09:10.610 align:middle line:84% growing a very thin oxide, the orientation 00:09:10.610 --> 00:09:11.520 align:middle line:90% will impact the rate. 00:09:11.520 --> 00:09:13.640 align:middle line:90% We saw that last time. 00:09:13.640 --> 00:09:17.240 align:middle line:84% These other things, though, are more telling 00:09:17.240 --> 00:09:19.700 align:middle line:84% and more having to do with what he actually saw. 00:09:19.700 --> 00:09:25.240 align:middle line:84% The two dimensional diffusion of the oxidant, 00:09:25.240 --> 00:09:27.710 align:middle line:84% you need some kind of numerical technique 00:09:27.710 --> 00:09:30.180 align:middle line:84% to solve the diffusion equation in multiple dimensions. 00:09:30.180 --> 00:09:34.230 align:middle line:84% So if I just go back a couple of slides-- 00:09:34.230 --> 00:09:38.950 align:middle line:84% let's go back to slide now, say, number 4 00:09:38.950 --> 00:09:46.048 align:middle line:84% and look at oxidizing this concave corner. 00:09:46.048 --> 00:09:48.820 align:middle line:84% As the oxide grows, the oxygen or the water, 00:09:48.820 --> 00:09:50.650 align:middle line:84% whichever is he's using as the oxidant, 00:09:50.650 --> 00:09:54.380 align:middle line:84% has to diffuse in two dimensions through this structure. 00:09:54.380 --> 00:09:57.790 align:middle line:84% So any changes due to orientation or whatever 00:09:57.790 --> 00:10:00.160 align:middle line:84% in that thickness in which it has to diffuse through 00:10:00.160 --> 00:10:01.390 align:middle line:90% are then going to affect-- 00:10:01.390 --> 00:10:03.730 align:middle line:84% eventually will affect the diffusion through that. 00:10:03.730 --> 00:10:05.620 align:middle line:84% So we have a two dimensional problem now. 00:10:05.620 --> 00:10:07.245 align:middle line:84% Before we had a simple one dimensional. 00:10:07.245 --> 00:10:10.060 align:middle line:84% Deal-Grove we just had to diffuse from the top surface 00:10:10.060 --> 00:10:11.230 align:middle line:90% to the interface. 00:10:11.230 --> 00:10:13.420 align:middle line:84% Well, we have to do the same thing but now 00:10:13.420 --> 00:10:16.160 align:middle line:90% in a two dimensional problem. 00:10:16.160 --> 00:10:22.390 align:middle line:84% So as the shapes change, we'll see diffusion of the oxidant 00:10:22.390 --> 00:10:26.270 align:middle line:90% will take effect. 00:10:26.270 --> 00:10:30.120 align:middle line:84% Let's go back then to slide number 6. 00:10:30.120 --> 00:10:32.850 align:middle line:84% The third one and probably one of the most important 00:10:32.850 --> 00:10:36.450 align:middle line:84% in explaining Gao's results is the stress 00:10:36.450 --> 00:10:39.480 align:middle line:90% due to the volume expansion. 00:10:39.480 --> 00:10:42.180 align:middle line:84% We know that oxide layers that are formed on silicon 00:10:42.180 --> 00:10:44.410 align:middle line:90% are under compressive stress. 00:10:44.410 --> 00:10:44.910 align:middle line:90% OK? 00:10:44.910 --> 00:10:46.650 align:middle line:84% So they're kind of being compressed 00:10:46.650 --> 00:10:48.387 align:middle line:90% to get onto the wafer. 00:10:48.387 --> 00:10:50.220 align:middle line:84% Even when they're in the planar case, right? 00:10:50.220 --> 00:10:56.640 align:middle line:84% Just on a planar silicon wafer surface without any curvature. 00:10:56.640 --> 00:10:58.650 align:middle line:84% These stresses can be increased quite a bit 00:10:58.650 --> 00:11:03.120 align:middle line:84% on a curved surface, because the volume expansion is going 00:11:03.120 --> 00:11:04.530 align:middle line:90% to be confined dimensionally. 00:11:04.530 --> 00:11:08.750 align:middle line:84% So here's a way of picturing this on a curved surface. 00:11:08.750 --> 00:11:11.250 align:middle line:84% So the central region that's colored in gray, 00:11:11.250 --> 00:11:14.070 align:middle line:84% let's say that's my silicon pillar. 00:11:14.070 --> 00:11:16.290 align:middle line:84% Let's say I've grown a certain amount of oxide 00:11:16.290 --> 00:11:18.360 align:middle line:84% already that's on this outer region. 00:11:18.360 --> 00:11:20.010 align:middle line:84% That's the oxide that I've grown, 00:11:20.010 --> 00:11:22.860 align:middle line:84% and now I'm just growing a little bit of new oxide. 00:11:22.860 --> 00:11:25.290 align:middle line:84% Remember, in oxidation the reaction 00:11:25.290 --> 00:11:27.400 align:middle line:84% takes place at the interface, not at the surface. 00:11:27.400 --> 00:11:31.590 align:middle line:84% So to grow this little band here of new oxide, 00:11:31.590 --> 00:11:35.520 align:middle line:84% let's say a few angstroms, 10 angstroms that I'm growing, 00:11:35.520 --> 00:11:38.280 align:middle line:84% the oxidant has to fuse into the silicon surface, 00:11:38.280 --> 00:11:41.190 align:middle line:84% has to react at the surface, and then the new oxide 00:11:41.190 --> 00:11:44.200 align:middle line:84% has to push out on the old oxide to make room for itself. 00:11:44.200 --> 00:11:46.860 align:middle line:84% It's got to push up and push down and consume 00:11:46.860 --> 00:11:48.840 align:middle line:90% some of the silicon. 00:11:48.840 --> 00:11:51.120 align:middle line:84% When I was on a flat silicon wafer, 00:11:51.120 --> 00:11:54.180 align:middle line:84% the oxide could just push up and push out 00:11:54.180 --> 00:11:56.740 align:middle line:84% a little bit at the edge of the wafer if it had to. 00:11:56.740 --> 00:11:57.910 align:middle line:90% But I'm not. 00:11:57.910 --> 00:11:59.980 align:middle line:84% I'm on a curved surface here where 00:11:59.980 --> 00:12:05.480 align:middle line:84% the oxide is completely attached to itself, so to speak. 00:12:05.480 --> 00:12:09.760 align:middle line:84% So it has to actually expand along 00:12:09.760 --> 00:12:11.350 align:middle line:84% the direction of the circumference 00:12:11.350 --> 00:12:12.230 align:middle line:90% in this direction. 00:12:12.230 --> 00:12:16.090 align:middle line:84% So it has to do work, more work on the oxide above it 00:12:16.090 --> 00:12:19.000 align:middle line:84% than it would have to do on a planar surface, 00:12:19.000 --> 00:12:22.600 align:middle line:84% because it's really confined all the way around that surface. 00:12:22.600 --> 00:12:26.503 align:middle line:90% 00:12:26.503 --> 00:12:27.920 align:middle line:84% Really, the curved surface has got 00:12:27.920 --> 00:12:29.420 align:middle line:84% to flow in order to get this to happen 00:12:29.420 --> 00:12:31.160 align:middle line:84% So you have the solid material this glass 00:12:31.160 --> 00:12:33.083 align:middle line:90% that has to essentially flow. 00:12:33.083 --> 00:12:35.000 align:middle line:84% So there's a certain amount of stress built up 00:12:35.000 --> 00:12:37.580 align:middle line:90% as a result of that. 00:12:37.580 --> 00:12:40.940 align:middle line:84% The deformation of the surface can be quite large. 00:12:40.940 --> 00:12:42.437 align:middle line:84% Forget about the cylinder for now. 00:12:42.437 --> 00:12:43.520 align:middle line:90% That was Gao's experiment. 00:12:43.520 --> 00:12:47.690 align:middle line:84% Let's do something a little more realistic, like LOCOS. 00:12:47.690 --> 00:12:48.470 align:middle line:90% What is LOCOS? 00:12:48.470 --> 00:12:51.650 align:middle line:84% Remember, LOCOS is you start with a silicon wafer. 00:12:51.650 --> 00:12:55.100 align:middle line:84% You pattern part of it with silicon nitride 00:12:55.100 --> 00:12:56.660 align:middle line:90% under which the oxidant-- 00:12:56.660 --> 00:12:58.760 align:middle line:84% through which the oxidant cannot diffuse. 00:12:58.760 --> 00:13:00.140 align:middle line:90% So you're locally oxidizing. 00:13:00.140 --> 00:13:03.230 align:middle line:84% You're not oxidizing over here on the right in this lower 00:13:03.230 --> 00:13:03.920 align:middle line:90% right picture. 00:13:03.920 --> 00:13:06.810 align:middle line:84% You're oxidizing over here, only over here on the left. 00:13:06.810 --> 00:13:10.340 align:middle line:84% So you're locally growing a thick oxide. 00:13:10.340 --> 00:13:11.990 align:middle line:90% But look at the surface. 00:13:11.990 --> 00:13:15.020 align:middle line:84% The top surface of this oxide here. 00:13:15.020 --> 00:13:17.360 align:middle line:84% Remember, at one point when it first starts out, 00:13:17.360 --> 00:13:19.220 align:middle line:90% it's relatively flat. 00:13:19.220 --> 00:13:21.710 align:middle line:84% This top surface has to stretch, in order 00:13:21.710 --> 00:13:25.530 align:middle line:84% to maintain this shape, by as much as 15%, 20%. 00:13:25.530 --> 00:13:31.430 align:middle line:84% So the oxide is itself having to stretch and deform. 00:13:31.430 --> 00:13:36.080 align:middle line:84% All that involves doing work and stress buildup. 00:13:36.080 --> 00:13:37.970 align:middle line:84% Even, for example, also the nitride 00:13:37.970 --> 00:13:42.170 align:middle line:84% itself is pushing down on the oxide. 00:13:42.170 --> 00:13:44.280 align:middle line:84% In order to lift this nitride layer, 00:13:44.280 --> 00:13:46.620 align:middle line:90% the oxide has to do work. 00:13:46.620 --> 00:13:49.920 align:middle line:84% So there's definitely stress buildup in these regions 00:13:49.920 --> 00:13:53.100 align:middle line:84% near corners or on curved surfaces, 00:13:53.100 --> 00:13:58.110 align:middle line:84% and that slows down the oxidation rate. 00:13:58.110 --> 00:14:00.460 align:middle line:84% So how does it slow down the oxidation rate? 00:14:00.460 --> 00:14:04.620 align:middle line:84% Well, let's look, for a moment, at slide number 7. 00:14:04.620 --> 00:14:08.040 align:middle line:84% These stresses can impact a couple of things, 00:14:08.040 --> 00:14:11.040 align:middle line:84% both transport of the oxygen through the oxide, 00:14:11.040 --> 00:14:14.800 align:middle line:84% like a diffusion process, and the interface reaction rate. 00:14:14.800 --> 00:14:16.110 align:middle line:90% And here's an example again. 00:14:16.110 --> 00:14:19.200 align:middle line:84% This is showing a small portion, a quarter 00:14:19.200 --> 00:14:21.780 align:middle line:84% or a section of a silicon surface 00:14:21.780 --> 00:14:24.960 align:middle line:84% that's curved because it's been etched or whatever, 00:14:24.960 --> 00:14:27.540 align:middle line:84% and the oxide that's already been grown here 00:14:27.540 --> 00:14:32.160 align:middle line:84% is on the outside, this annulus shown in the medium gray. 00:14:32.160 --> 00:14:35.400 align:middle line:84% This little white annulus, this little white region, 00:14:35.400 --> 00:14:37.560 align:middle line:84% the nascent oxide-- nascent meaning new. 00:14:37.560 --> 00:14:39.900 align:middle line:84% It's the new little piece of oxide that's grown. 00:14:39.900 --> 00:14:43.560 align:middle line:84% This little growing oxide has to do mechanical work 00:14:43.560 --> 00:14:48.030 align:middle line:84% against the oxide above it, because it has to push it out, 00:14:48.030 --> 00:14:50.040 align:middle line:90% and it's constrained. 00:14:50.040 --> 00:14:53.700 align:middle line:84% So this mechanical work actually can 00:14:53.700 --> 00:14:56.820 align:middle line:84% modify the energy, the activation energy 00:14:56.820 --> 00:14:59.610 align:middle line:84% of the process, the amount of energy 00:14:59.610 --> 00:15:03.280 align:middle line:84% that it takes to do that process at a given temperature. 00:15:03.280 --> 00:15:07.440 align:middle line:84% So what we write is EA, the activation energy 00:15:07.440 --> 00:15:11.670 align:middle line:84% for the surface reaction rate under the stress case. 00:15:11.670 --> 00:15:14.430 align:middle line:84% When we're oxidizing in a curved surface, 00:15:14.430 --> 00:15:17.280 align:middle line:84% it's going to be equal to EA on a flat surface, 00:15:17.280 --> 00:15:20.040 align:middle line:84% whatever that might be-- say two electron volts-- 00:15:20.040 --> 00:15:22.710 align:middle line:90% Plus some extra terms. 00:15:22.710 --> 00:15:27.270 align:middle line:84% And there's a term that depends on the normal stress. 00:15:27.270 --> 00:15:29.010 align:middle line:90% Sigma n is the normal stress. 00:15:29.010 --> 00:15:30.450 align:middle line:84% Normal, so it's in this direction, 00:15:30.450 --> 00:15:33.010 align:middle line:90% in the radial direction. 00:15:33.010 --> 00:15:38.070 align:middle line:84% The sigma n times some fitting parameter VR plus sigma t 00:15:38.070 --> 00:15:40.710 align:middle line:84% is the stress in the tangential to the growing surface, 00:15:40.710 --> 00:15:44.670 align:middle line:84% so along the circumference, times some other fitting 00:15:44.670 --> 00:15:48.040 align:middle line:90% parameter volume Vt. 00:15:48.040 --> 00:15:52.720 align:middle line:84% So then k sub s in the curved or stress case is just 00:15:52.720 --> 00:15:55.990 align:middle line:84% the ordinary k sub s times these two exponentials-- again, 00:15:55.990 --> 00:15:58.930 align:middle line:90% this is an activation energy-- 00:15:58.930 --> 00:16:02.600 align:middle line:84% minus sigma n VR over kt and a sigma t term. 00:16:02.600 --> 00:16:04.870 align:middle line:84% So basically the surface reaction rate 00:16:04.870 --> 00:16:06.490 align:middle line:90% is going to go down. 00:16:06.490 --> 00:16:09.130 align:middle line:84% The amount it goes down will depend exponentially 00:16:09.130 --> 00:16:13.030 align:middle line:84% on these two stress terms, and these are fitting parameters 00:16:13.030 --> 00:16:17.292 align:middle line:84% that we put in-- essentially you can put in the model. 00:16:17.292 --> 00:16:19.750 align:middle line:84% And so you can see then how you can introduce a temperature 00:16:19.750 --> 00:16:21.070 align:middle line:90% dependence to this effect. 00:16:21.070 --> 00:16:24.218 align:middle line:84% Depending on how strongly temperature dependent 00:16:24.218 --> 00:16:26.510 align:middle line:84% is will depend on the numbers you put in for VR and Vt. 00:16:26.510 --> 00:16:29.030 align:middle line:90% 00:16:29.030 --> 00:16:29.530 align:middle line:90% OK. 00:16:29.530 --> 00:16:32.050 align:middle line:84% So let's go on to slide number 8, and that 00:16:32.050 --> 00:16:33.790 align:middle line:90% was the surface reaction rate. 00:16:33.790 --> 00:16:37.690 align:middle line:84% You can intuitively understand how that surface reaction 00:16:37.690 --> 00:16:40.450 align:middle line:84% might be affected by the fact that it takes more energy. 00:16:40.450 --> 00:16:42.640 align:middle line:84% You have to do work in order for that reaction 00:16:42.640 --> 00:16:45.940 align:middle line:84% to take place that you didn't have to do on a flat surface. 00:16:45.940 --> 00:16:49.840 align:middle line:84% How about the diffusivity and the solubility? 00:16:49.840 --> 00:16:52.690 align:middle line:84% Well, how could that be affected by stress? 00:16:52.690 --> 00:16:58.300 align:middle line:84% Well, this is meant to be a highly schematic picture 00:16:58.300 --> 00:17:01.840 align:middle line:84% of a ring structure that would exist in silicon dioxide 00:17:01.840 --> 00:17:07.764 align:middle line:84% where these black regions of black circles are the silicon. 00:17:07.764 --> 00:17:11.589 align:middle line:90% 00:17:11.589 --> 00:17:16.270 align:middle line:84% We have oxygen here as well taking up 00:17:16.270 --> 00:17:19.420 align:middle line:84% a certain amount of space, say, in the bonds. 00:17:19.420 --> 00:17:22.569 align:middle line:84% There is some sort of region in here 00:17:22.569 --> 00:17:26.440 align:middle line:84% where there is sort of free space or interstitial 00:17:26.440 --> 00:17:30.380 align:middle line:84% space where the oxidant can diffuse through. 00:17:30.380 --> 00:17:34.040 align:middle line:84% And let's say p is a hydrostatic pressure in the growing oxide, 00:17:34.040 --> 00:17:36.980 align:middle line:84% so there's a certain amount of pressure in the solid. 00:17:36.980 --> 00:17:40.400 align:middle line:84% Well, as you put the solid under pressure, it reduces that. 00:17:40.400 --> 00:17:42.935 align:middle line:84% You can imagine it reducing the interstitial space 00:17:42.935 --> 00:17:43.560 align:middle line:90% in the network. 00:17:43.560 --> 00:17:44.720 align:middle line:90% So you have this network. 00:17:44.720 --> 00:17:47.190 align:middle line:84% Although it's somewhat random, it has this structure, 00:17:47.190 --> 00:17:49.220 align:middle line:84% and inside the network there's this open space. 00:17:49.220 --> 00:17:50.970 align:middle line:84% As you put the whole solid under pressure, 00:17:50.970 --> 00:17:53.480 align:middle line:84% but that bond's compressed, the amount 00:17:53.480 --> 00:17:54.980 align:middle line:84% of interstitial space in the network 00:17:54.980 --> 00:17:56.870 align:middle line:84% is going to be reduced to a certain extent. 00:17:56.870 --> 00:17:58.970 align:middle line:84% And if you're oxidant, your water molecule, 00:17:58.970 --> 00:18:00.800 align:middle line:84% your oxygen molecule has to diffuse 00:18:00.800 --> 00:18:02.210 align:middle line:90% through that interstitial space. 00:18:02.210 --> 00:18:04.550 align:middle line:84% You can imagine the diffusivity or the diffusion rate 00:18:04.550 --> 00:18:06.170 align:middle line:90% is going to go down. 00:18:06.170 --> 00:18:08.030 align:middle line:84% As I compress on the oxide, it makes 00:18:08.030 --> 00:18:10.970 align:middle line:84% it harder for the molecules to diffuse through. 00:18:10.970 --> 00:18:13.550 align:middle line:84% It might even affect the solubility, 00:18:13.550 --> 00:18:16.140 align:middle line:84% as you might imagine, hypothetically, 00:18:16.140 --> 00:18:19.310 align:middle line:84% that the amount of water vapor molecules 00:18:19.310 --> 00:18:21.380 align:middle line:84% or oxygen you can stuff in there maybe 00:18:21.380 --> 00:18:25.010 align:middle line:84% would be affected by how much pressure this solid is under. 00:18:25.010 --> 00:18:27.740 align:middle line:84% So again, mathematically we can do what we did before. 00:18:27.740 --> 00:18:30.790 align:middle line:84% We write the stress diffusivity as the unstressed-- 00:18:30.790 --> 00:18:33.340 align:middle line:84% or not unstressed, but the normal planar diffusivity 00:18:33.340 --> 00:18:35.080 align:middle line:84% through the oxide times something that 00:18:35.080 --> 00:18:39.010 align:middle line:84% goes exponentially it depends on the pressure, which 00:18:39.010 --> 00:18:41.830 align:middle line:84% is the hydrostatic pressure is due to the oxide stress, 00:18:41.830 --> 00:18:46.900 align:middle line:84% times a constant, which is a fitting parameter VD over kt. 00:18:46.900 --> 00:18:50.740 align:middle line:84% So again, D will go down as you go under stress, diffusivity. 00:18:50.740 --> 00:18:52.930 align:middle line:84% And you might hypothesize the same thing 00:18:52.930 --> 00:18:55.930 align:middle line:84% for the solubility, a similar dependence. 00:18:55.930 --> 00:18:56.650 align:middle line:90% OK. 00:18:56.650 --> 00:18:59.140 align:middle line:90% Let's go on to slide number 9. 00:18:59.140 --> 00:19:01.750 align:middle line:84% So just as shown in your text, I'm 00:19:01.750 --> 00:19:03.910 align:middle line:84% just summarizing then the impact of stress 00:19:03.910 --> 00:19:05.920 align:middle line:84% on our basic oxidation parameters 00:19:05.920 --> 00:19:09.040 align:middle line:84% that we know from our Deal-Grove model. 00:19:09.040 --> 00:19:11.340 align:middle line:84% The parameters from Deal-Grove, k sub 00:19:11.340 --> 00:19:14.140 align:middle line:84% s, the reaction rate D and C star all 00:19:14.140 --> 00:19:16.180 align:middle line:84% get modified in a way that depends 00:19:16.180 --> 00:19:20.510 align:middle line:84% on the stress in an exponential fashion. 00:19:20.510 --> 00:19:22.960 align:middle line:90% So it's very sensitive over kt. 00:19:22.960 --> 00:19:25.780 align:middle line:84% So the stress effect will be different 00:19:25.780 --> 00:19:28.640 align:middle line:90% at different temperatures. 00:19:28.640 --> 00:19:32.420 align:middle line:84% So these parameters, these VR, Vt, VD, and Vs, 00:19:32.420 --> 00:19:33.750 align:middle line:90% they're reaction volumes. 00:19:33.750 --> 00:19:35.780 align:middle line:90% They are designed in SUPREM-IV. 00:19:35.780 --> 00:19:37.970 align:middle line:84% They are put in as fitting parameters. 00:19:37.970 --> 00:19:39.800 align:middle line:84% So they have some default values, 00:19:39.800 --> 00:19:42.320 align:middle line:84% but people typically modify them as needed 00:19:42.320 --> 00:19:44.570 align:middle line:84% to fit the shape that's observed experimentally 00:19:44.570 --> 00:19:49.430 align:middle line:84% from LOCOS or from a pillar or from your structure. 00:19:49.430 --> 00:19:52.460 align:middle line:84% In practice, this number V sub s ends up 00:19:52.460 --> 00:19:55.760 align:middle line:84% being zero, which essentially people believe 00:19:55.760 --> 00:19:57.950 align:middle line:84% that the solubility is not dramatically 00:19:57.950 --> 00:20:00.240 align:middle line:90% itself impacted by the stress. 00:20:00.240 --> 00:20:05.180 align:middle line:84% And so this extra term here, e to the PVs 00:20:05.180 --> 00:20:06.620 align:middle line:90% over kt equals to one. 00:20:06.620 --> 00:20:08.780 align:middle line:84% And so people generally believe it's 00:20:08.780 --> 00:20:14.360 align:middle line:84% the k sub s and the D that are being modified by the stress. 00:20:14.360 --> 00:20:16.950 align:middle line:90% 00:20:16.950 --> 00:20:17.450 align:middle line:90% OK. 00:20:17.450 --> 00:20:20.040 align:middle line:84% So let's go on to page or slide 10. 00:20:20.040 --> 00:20:21.590 align:middle line:84% There's one more parameter that we 00:20:21.590 --> 00:20:25.220 align:middle line:84% need because we said the oxide is a glass and it's flowing, 00:20:25.220 --> 00:20:30.560 align:middle line:84% and the glass flows differently or its ability to flow 00:20:30.560 --> 00:20:33.450 align:middle line:90% depends on temperature. 00:20:33.450 --> 00:20:39.650 align:middle line:84% And that is expressed through a viscosity parameter, eta. 00:20:39.650 --> 00:20:42.170 align:middle line:84% And in fact, the stresses that are in the glass 00:20:42.170 --> 00:20:45.320 align:middle line:84% or in the oxide are high enough that the viscosity 00:20:45.320 --> 00:20:47.780 align:middle line:84% itself, its rate of flow actually 00:20:47.780 --> 00:20:51.330 align:middle line:84% is a function of the stress in order to model it accurately. 00:20:51.330 --> 00:20:55.370 align:middle line:84% So this is an equation that's been derived 00:20:55.370 --> 00:20:58.190 align:middle line:84% and that is used in SUPREM that gives 00:20:58.190 --> 00:21:02.640 align:middle line:84% reasonably good agreement with the experimental observations. 00:21:02.640 --> 00:21:06.800 align:middle line:84% So the viscosity as a function of stress 00:21:06.800 --> 00:21:10.460 align:middle line:84% is equal to the stress independent viscosity eta, 00:21:10.460 --> 00:21:12.950 align:middle line:84% but it's a function of t, temperature, 00:21:12.950 --> 00:21:15.710 align:middle line:90% times these sort of parameters. 00:21:15.710 --> 00:21:18.260 align:middle line:84% A shear stress in the oxide, so it's actually 00:21:18.260 --> 00:21:21.300 align:middle line:84% linearly dependent on the stress in the oxide. 00:21:21.300 --> 00:21:24.380 align:middle line:84% So if this goes up, the viscosity goes up, 00:21:24.380 --> 00:21:28.490 align:middle line:84% and something that depends on the hyperbolic sine. 00:21:28.490 --> 00:21:31.700 align:middle line:84% So that's just an equation that people have derived 00:21:31.700 --> 00:21:33.093 align:middle line:90% more or less empirically. 00:21:33.093 --> 00:21:35.510 align:middle line:84% So we have these parameters that are changing with stress. 00:21:35.510 --> 00:21:38.640 align:middle line:90% 00:21:38.640 --> 00:21:40.940 align:middle line:90% So let's go on to slide 11. 00:21:40.940 --> 00:21:42.690 align:middle line:84% You can probably get the impression by now 00:21:42.690 --> 00:21:46.050 align:middle line:84% there's no way you can easily do these calculations by hand. 00:21:46.050 --> 00:21:47.280 align:middle line:90% Deal-Groves, piece of cake. 00:21:47.280 --> 00:21:48.870 align:middle line:90% You sit down. 00:21:48.870 --> 00:21:51.475 align:middle line:84% You just simply solve or integrate the equations. 00:21:51.475 --> 00:21:52.350 align:middle line:90% They're all analytic. 00:21:52.350 --> 00:21:53.490 align:middle line:90% These are non-analytic. 00:21:53.490 --> 00:21:55.860 align:middle line:84% The shape of the growing oxide changes with time. 00:21:55.860 --> 00:21:59.373 align:middle line:84% k sub s, D, and eta all change with stress, 00:21:59.373 --> 00:22:01.290 align:middle line:84% and the amount of stress is changing with time 00:22:01.290 --> 00:22:03.930 align:middle line:84% as the oxide has to push on a thicker and thicker thing 00:22:03.930 --> 00:22:05.650 align:middle line:90% above it. 00:22:05.650 --> 00:22:07.590 align:middle line:84% So with all this time dependence, 00:22:07.590 --> 00:22:09.480 align:middle line:84% you really need a numerical simulator. 00:22:09.480 --> 00:22:12.480 align:middle line:84% You need a computer to do these calculations for you 00:22:12.480 --> 00:22:14.670 align:middle line:90% and to integrate them in time. 00:22:14.670 --> 00:22:16.140 align:middle line:84% This is exactly what has been done 00:22:16.140 --> 00:22:20.160 align:middle line:84% in SUPREM-IV, which can model the shapes of oxide 00:22:20.160 --> 00:22:21.030 align:middle line:90% on curved surfaces. 00:22:21.030 --> 00:22:23.322 align:middle line:84% You'll have a chance to do this on one of your homework 00:22:23.322 --> 00:22:24.630 align:middle line:90% problems, hopefully. 00:22:24.630 --> 00:22:27.840 align:middle line:84% And this is a reasonably common thing in IC technology, 00:22:27.840 --> 00:22:29.970 align:middle line:84% that you don't typically always grow 00:22:29.970 --> 00:22:31.795 align:middle line:84% an oxide over the entire surface. 00:22:31.795 --> 00:22:33.670 align:middle line:84% Very often there are patterns on the surface. 00:22:33.670 --> 00:22:37.680 align:middle line:84% In fact, the very first step is to isolate the active area 00:22:37.680 --> 00:22:41.520 align:middle line:84% from the non-active region, and it's often 00:22:41.520 --> 00:22:43.350 align:middle line:84% done by the process called LOCOS we 00:22:43.350 --> 00:22:44.910 align:middle line:90% talked about the first lecture. 00:22:44.910 --> 00:22:48.420 align:middle line:84% Poly buffered LOCOS, slightly fancier version, 00:22:48.420 --> 00:22:50.100 align:middle line:90% and shallow trench isolation. 00:22:50.100 --> 00:22:51.420 align:middle line:90% So here's an example of LOCOS. 00:22:51.420 --> 00:22:54.030 align:middle line:84% Remember, you start with a silicon substrate. 00:22:54.030 --> 00:22:57.540 align:middle line:84% You put down a thin oxide, which is a pad oxide, which 00:22:57.540 --> 00:22:59.300 align:middle line:84% reduces the stress on the silicon 00:22:59.300 --> 00:23:01.800 align:middle line:84% due to the silicon nitride, and you put down silicon nitride 00:23:01.800 --> 00:23:02.550 align:middle line:90% and pattern it. 00:23:02.550 --> 00:23:07.140 align:middle line:84% Silicon nitride is a perfect or reasonably good mask, 00:23:07.140 --> 00:23:09.270 align:middle line:84% as oxygen cannot diffuse through it. 00:23:09.270 --> 00:23:13.500 align:middle line:84% So the oxidant diffuses in here and only oxidizes 00:23:13.500 --> 00:23:15.150 align:middle line:90% on the left side. 00:23:15.150 --> 00:23:16.820 align:middle line:84% So this is what the local structure just 00:23:16.820 --> 00:23:19.340 align:middle line:84% in your mind, as a starting structure before you 00:23:19.340 --> 00:23:22.810 align:middle line:84% do oxidation, that's what it looks like. 00:23:22.810 --> 00:23:26.110 align:middle line:84% Let's go on to slide 12 and see what that LOCOS structure now 00:23:26.110 --> 00:23:30.850 align:middle line:84% looks like in a simulation, the SUPREM simulation, 00:23:30.850 --> 00:23:32.330 align:middle line:90% illustrating the stress effects. 00:23:32.330 --> 00:23:35.710 align:middle line:84% So there are two simulations that are shown here 00:23:35.710 --> 00:23:37.880 align:middle line:90% and let's start on the left. 00:23:37.880 --> 00:23:41.310 align:middle line:84% And what this is, the y-axis is in microns 00:23:41.310 --> 00:23:45.440 align:middle line:84% so that's vertical in depth from the silicon surface, 00:23:45.440 --> 00:23:50.450 align:middle line:84% and the x-axis is in microns as well. 00:23:50.450 --> 00:23:52.550 align:middle line:84% This point right here corresponds 00:23:52.550 --> 00:23:55.400 align:middle line:90% to where the mask edge is. 00:23:55.400 --> 00:23:58.350 align:middle line:84% So we're doing a two dimensional simulation. 00:23:58.350 --> 00:24:00.770 align:middle line:84% And what you see is this upper region 00:24:00.770 --> 00:24:03.830 align:middle line:84% is the silicon nitride, as you can see, 00:24:03.830 --> 00:24:07.220 align:middle line:84% and the yellow region underneath is the silicon surface. 00:24:07.220 --> 00:24:11.330 align:middle line:84% Look how the silicon surface is shaped in this case. 00:24:11.330 --> 00:24:15.530 align:middle line:90% And the field oxide is out here. 00:24:15.530 --> 00:24:17.630 align:middle line:84% Now an important parameter that people 00:24:17.630 --> 00:24:20.383 align:middle line:84% care about, if you're scaling silicon circuits, is you 00:24:20.383 --> 00:24:22.550 align:middle line:84% want to be able to put the devices closer and closer 00:24:22.550 --> 00:24:25.470 align:middle line:84% together, as close as you can get them. 00:24:25.470 --> 00:24:28.010 align:middle line:84% So in order to be able to do that, 00:24:28.010 --> 00:24:30.380 align:middle line:84% you need to grow the field oxide where you want it, 00:24:30.380 --> 00:24:33.560 align:middle line:84% and you don't want field oxide in the active region. 00:24:33.560 --> 00:24:36.410 align:middle line:84% Any part of the field oxide that encroaches or pushes 00:24:36.410 --> 00:24:38.150 align:middle line:84% its way into the active region means 00:24:38.150 --> 00:24:40.640 align:middle line:84% that it's scaling the active region in which you 00:24:40.640 --> 00:24:43.340 align:middle line:84% can make your device so you don't have as much control. 00:24:43.340 --> 00:24:45.900 align:middle line:84% You don't get a nice, well defined active region. 00:24:45.900 --> 00:24:48.920 align:middle line:84% So in this model, you can see what's 00:24:48.920 --> 00:24:52.640 align:middle line:84% happened is the field oxide has encroached in. 00:24:52.640 --> 00:24:55.370 align:middle line:84% The mask edge start here at x equals zero, 00:24:55.370 --> 00:24:59.270 align:middle line:84% it's encroached in from x equals zero by about-- 00:24:59.270 --> 00:25:02.300 align:middle line:90% oh, about, say, half a micron. 00:25:02.300 --> 00:25:06.470 align:middle line:84% The oxidant has diffused in here and has oxidized 00:25:06.470 --> 00:25:08.520 align:middle line:90% underneath the nitride. 00:25:08.520 --> 00:25:11.330 align:middle line:84% So you have this region where the silicon is sloped 00:25:11.330 --> 00:25:13.700 align:middle line:84% and the encroachment distance, I would quote here 00:25:13.700 --> 00:25:14.750 align:middle line:90% as being half micron. 00:25:14.750 --> 00:25:20.650 align:middle line:84% So your active device really starts here in this region. 00:25:20.650 --> 00:25:23.340 align:middle line:84% So this is the model run, though, without stress. 00:25:23.340 --> 00:25:26.813 align:middle line:84% And people did these models and they found, oh, it 00:25:26.813 --> 00:25:28.980 align:middle line:84% doesn't really agree with what we get experimentally 00:25:28.980 --> 00:25:30.420 align:middle line:90% when we grow LOCOS. 00:25:30.420 --> 00:25:32.280 align:middle line:84% We do a cross-section and we look at it. 00:25:32.280 --> 00:25:33.547 align:middle line:90% It's not what it looks like. 00:25:33.547 --> 00:25:35.130 align:middle line:84% In fact, what they found it looks more 00:25:35.130 --> 00:25:37.170 align:middle line:90% like what's on the right. 00:25:37.170 --> 00:25:39.350 align:middle line:84% And what's on the right is the SUPREM-IV model, 00:25:39.350 --> 00:25:42.830 align:middle line:84% but this time turning on the stress dependence. 00:25:42.830 --> 00:25:43.490 align:middle line:90% OK? 00:25:43.490 --> 00:25:48.620 align:middle line:84% So you have the diffusion rate through the oxide, 00:25:48.620 --> 00:25:51.320 align:middle line:84% the surface reaction rate are now all going down 00:25:51.320 --> 00:25:53.000 align:middle line:90% as a function of stress. 00:25:53.000 --> 00:25:55.310 align:middle line:90% And where is their stress? 00:25:55.310 --> 00:25:57.170 align:middle line:84% Well, there's a lot of stress in this region 00:25:57.170 --> 00:25:59.600 align:middle line:84% here, in this corner region, because the oxide has to push 00:25:59.600 --> 00:26:02.370 align:middle line:90% up against this nitride film. 00:26:02.370 --> 00:26:05.000 align:middle line:84% So what you see here is look at the amount of oxide 00:26:05.000 --> 00:26:07.760 align:middle line:84% that was grown here in this bird's beak area. 00:26:07.760 --> 00:26:10.940 align:middle line:84% There's quite a bit of it has grown, the thickness here. 00:26:10.940 --> 00:26:12.770 align:middle line:84% Compared to, in the bird's beak area, 00:26:12.770 --> 00:26:16.910 align:middle line:84% when you turn the stress model on not so much oxide has grown. 00:26:16.910 --> 00:26:17.600 align:middle line:90% OK? 00:26:17.600 --> 00:26:21.800 align:middle line:84% And as a result, with stress, the amount of encroachment, 00:26:21.800 --> 00:26:24.590 align:middle line:84% which I'm marking between these two lines now here, 00:26:24.590 --> 00:26:25.910 align:middle line:90% is quite a bit less. 00:26:25.910 --> 00:26:27.860 align:middle line:90% It's about a factor of 2, maybe. 00:26:27.860 --> 00:26:29.360 align:middle line:90% Here it encroached-- I'm sorry. 00:26:29.360 --> 00:26:30.530 align:middle line:90% I said 0.5. 00:26:30.530 --> 00:26:32.660 align:middle line:84% It actually encroached by 0.6 microns. 00:26:32.660 --> 00:26:35.700 align:middle line:84% The encroachment here, according to my eye, is about 0.3. 00:26:35.700 --> 00:26:37.400 align:middle line:90% So it's about half. 00:26:37.400 --> 00:26:39.920 align:middle line:84% That's because stress was included in the model. 00:26:39.920 --> 00:26:42.650 align:middle line:84% This turns out to be a lot closer to what people actually 00:26:42.650 --> 00:26:49.220 align:middle line:84% measured when they do LOCOS oxidation. 00:26:49.220 --> 00:26:50.300 align:middle line:90% That's for LOCOS. 00:26:50.300 --> 00:26:51.830 align:middle line:84% You say, well, not too many people 00:26:51.830 --> 00:26:53.600 align:middle line:90% do LOCOS in manufacturing. 00:26:53.600 --> 00:26:56.750 align:middle line:84% Well, there is still some LOCOS or LOCOS-like structures 00:26:56.750 --> 00:26:57.710 align:middle line:90% being done. 00:26:57.710 --> 00:27:02.420 align:middle line:84% People nowadays do shallow trench isolation or STI. 00:27:02.420 --> 00:27:05.150 align:middle line:84% Well, we still get corner effects and stress effects 00:27:05.150 --> 00:27:06.140 align:middle line:90% at STI. 00:27:06.140 --> 00:27:09.810 align:middle line:84% And I've taken this from an article or a short course 00:27:09.810 --> 00:27:12.050 align:middle line:90% given at IBM back in 1998. 00:27:12.050 --> 00:27:13.670 align:middle line:90% And what is shallow trench? 00:27:13.670 --> 00:27:14.930 align:middle line:90% Remember what we have. 00:27:14.930 --> 00:27:17.330 align:middle line:84% We have an active region here shown in the center 00:27:17.330 --> 00:27:20.120 align:middle line:84% under the gate, and isolating the devices 00:27:20.120 --> 00:27:22.662 align:middle line:84% we have field oxide, which is this thick oxide on either side 00:27:22.662 --> 00:27:23.578 align:middle line:90% all the way around it. 00:27:23.578 --> 00:27:25.010 align:middle line:84% In fact, goes all the way around. 00:27:25.010 --> 00:27:26.630 align:middle line:90% We've cut a cross-section here. 00:27:26.630 --> 00:27:31.250 align:middle line:84% And we do this by first etching a trench in the silicon 00:27:31.250 --> 00:27:34.130 align:middle line:90% and then doing some oxidation. 00:27:34.130 --> 00:27:37.370 align:middle line:84% We do a little thermal oxidation of that sidewall. 00:27:37.370 --> 00:27:40.040 align:middle line:84% And the engineers, the electrical folks, 00:27:40.040 --> 00:27:41.930 align:middle line:84% are very interested in the exact shape 00:27:41.930 --> 00:27:44.960 align:middle line:84% of this corner, because it turns out, 00:27:44.960 --> 00:27:46.880 align:middle line:84% depending on the exact shape of that corner 00:27:46.880 --> 00:27:51.140 align:middle line:84% and how thick the thermal oxide is, you'll get electric field 00:27:51.140 --> 00:27:54.680 align:middle line:84% spikes or build ups right there, or you'll get a region 00:27:54.680 --> 00:27:56.390 align:middle line:90% where the oxide is weak. 00:27:56.390 --> 00:27:59.900 align:middle line:84% And tends to break down and when the gate goes over that, 00:27:59.900 --> 00:28:02.580 align:middle line:84% the device can have breakdown effects, 00:28:02.580 --> 00:28:07.710 align:middle line:84% or you can actually cause unusual transistor effects just 00:28:07.710 --> 00:28:12.150 align:middle line:84% based on the shape of this corner and this oxidation. 00:28:12.150 --> 00:28:14.610 align:middle line:84% So there are two different simulations shown here. 00:28:14.610 --> 00:28:17.640 align:middle line:84% Up on the upper right is without any stress. 00:28:17.640 --> 00:28:19.380 align:middle line:84% And you see the thickness of the oxide 00:28:19.380 --> 00:28:23.290 align:middle line:84% as you go around the corner is reasonably uniform. 00:28:23.290 --> 00:28:24.450 align:middle line:90% It doesn't change too much. 00:28:24.450 --> 00:28:26.580 align:middle line:84% With stress effects, look at the corner. 00:28:26.580 --> 00:28:29.100 align:middle line:84% The oxide thickness is quite a bit thinner here. 00:28:29.100 --> 00:28:30.690 align:middle line:84% It's thinned up in the corner region, 00:28:30.690 --> 00:28:35.250 align:middle line:84% compared to on the flat surface or on the vertical surface. 00:28:35.250 --> 00:28:38.283 align:middle line:84% And this is a problem because that oxide gets too thin, 00:28:38.283 --> 00:28:40.200 align:middle line:84% you're going to tend to have breakdown effects 00:28:40.200 --> 00:28:42.900 align:middle line:84% or it's going to affect the electrical properties 00:28:42.900 --> 00:28:43.870 align:middle line:90% of this device. 00:28:43.870 --> 00:28:46.257 align:middle line:84% So it seems maybe to your eye initially, 00:28:46.257 --> 00:28:47.590 align:middle line:90% oh, what's the big deal of that? 00:28:47.590 --> 00:28:50.730 align:middle line:84% But when you go to simulate how these devices electrically 00:28:50.730 --> 00:28:52.390 align:middle line:84% behave, it makes a big difference. 00:28:52.390 --> 00:28:54.870 align:middle line:84% So the stress effect here comes in 00:28:54.870 --> 00:28:57.240 align:middle line:84% because, again, you have this curved surface that 00:28:57.240 --> 00:29:00.300 align:middle line:84% needs to stretch, and there's stress buildup whenever 00:29:00.300 --> 00:29:01.230 align:middle line:90% you have a corner. 00:29:01.230 --> 00:29:03.780 align:middle line:84% That tends to lower the oxidation rate. 00:29:03.780 --> 00:29:06.330 align:middle line:84% And that's also been observed experimentally. 00:29:06.330 --> 00:29:09.880 align:middle line:90% 00:29:09.880 --> 00:29:14.850 align:middle line:84% So let's go on to slide number 14. 00:29:14.850 --> 00:29:17.820 align:middle line:84% Stress also has another interesting effect. 00:29:17.820 --> 00:29:19.290 align:middle line:84% Let's say you're not doing shapes. 00:29:19.290 --> 00:29:21.420 align:middle line:84% So you don't do LOCOS, you don't do shallow trench. 00:29:21.420 --> 00:29:23.100 align:middle line:84% You're a very simple process person. 00:29:23.100 --> 00:29:25.110 align:middle line:90% You only oxidize flat wafers. 00:29:25.110 --> 00:29:25.800 align:middle line:90% OK. 00:29:25.800 --> 00:29:27.160 align:middle line:90% Fine. 00:29:27.160 --> 00:29:30.540 align:middle line:84% You can still see stress effects in the form of the history, 00:29:30.540 --> 00:29:33.630 align:middle line:84% of what the history of the oxide itself has gone through, 00:29:33.630 --> 00:29:35.550 align:middle line:90% what thermal history. 00:29:35.550 --> 00:29:37.470 align:middle line:84% Because we know even planar growth 00:29:37.470 --> 00:29:41.160 align:middle line:84% has some intrinsic stress during the oxide growth. 00:29:41.160 --> 00:29:43.080 align:middle line:84% The oxide is under a certain amount of stress 00:29:43.080 --> 00:29:44.080 align:middle line:90% even when it's planar. 00:29:44.080 --> 00:29:47.570 align:middle line:84% It's just under more stress when it's curved. 00:29:47.570 --> 00:29:50.270 align:middle line:84% But this stress can relax upon annealing. 00:29:50.270 --> 00:29:53.180 align:middle line:84% At a high enough temperature, above 900, 950, the glass 00:29:53.180 --> 00:29:56.780 align:middle line:84% can flow a little bit and the stress will relax. 00:29:56.780 --> 00:30:00.410 align:middle line:84% So when people measure the Deal-Grove linear parabolic 00:30:00.410 --> 00:30:03.230 align:middle line:84% rate constants that describe oxidation 00:30:03.230 --> 00:30:06.830 align:middle line:84% in this intrinsic stress state, the 1D stresses 00:30:06.830 --> 00:30:08.480 align:middle line:90% are already accounted for. 00:30:08.480 --> 00:30:09.560 align:middle line:90% OK? 00:30:09.560 --> 00:30:14.670 align:middle line:84% But let's look at the experiment that's shown up on this slide. 00:30:14.670 --> 00:30:18.410 align:middle line:84% Let's say we do a two step oxidation. 00:30:18.410 --> 00:30:20.840 align:middle line:84% And you decide, OK, I take two wafers. 00:30:20.840 --> 00:30:23.450 align:middle line:84% I put one in the furnace at 1,100 degrees 00:30:23.450 --> 00:30:25.730 align:middle line:84% and I grow a certain thickness of oxide. 00:30:25.730 --> 00:30:29.090 align:middle line:84% Say, 1,000 angstroms or whatever. 00:30:29.090 --> 00:30:33.390 align:middle line:84% And I put another wafer identical in a furnace at 800. 00:30:33.390 --> 00:30:35.610 align:middle line:84% I grow the same thickness, 1,000 angstroms. 00:30:35.610 --> 00:30:39.110 align:middle line:84% So I have two oxides, both the same thickness, 00:30:39.110 --> 00:30:41.220 align:middle line:90% grown at different temperatures. 00:30:41.220 --> 00:30:44.600 align:middle line:84% However, they look the same but they're not exactly the same. 00:30:44.600 --> 00:30:47.960 align:middle line:84% The intrinsic stress in this 1,100 degree C oxide 00:30:47.960 --> 00:30:51.590 align:middle line:84% is going to be less than the intrinsic stress in this 800 00:30:51.590 --> 00:30:53.450 align:middle line:84% degree oxide because the higher temperature, 00:30:53.450 --> 00:30:54.860 align:middle line:90% the oxide can flow more. 00:30:54.860 --> 00:30:57.740 align:middle line:84% There isn't as much built in stress. 00:30:57.740 --> 00:30:58.250 align:middle line:90% OK? 00:30:58.250 --> 00:31:00.500 align:middle line:84% So you say, well, this is under more stress than that. 00:31:00.500 --> 00:31:01.417 align:middle line:90% What's the difference? 00:31:01.417 --> 00:31:04.010 align:middle line:84% OK, maybe you'd see a little less, little difference 00:31:04.010 --> 00:31:04.970 align:middle line:90% in wafer curvature. 00:31:04.970 --> 00:31:06.440 align:middle line:84% That's a way you could measure it. 00:31:06.440 --> 00:31:07.100 align:middle line:90% All right. 00:31:07.100 --> 00:31:07.940 align:middle line:90% That's fine. 00:31:07.940 --> 00:31:10.350 align:middle line:84% Now you go to the next step in your process, says, 00:31:10.350 --> 00:31:15.200 align:middle line:84% I want to grow another oxide underneath these two at 800 00:31:15.200 --> 00:31:17.070 align:middle line:90% degrees. 00:31:17.070 --> 00:31:18.930 align:middle line:84% Now, if I hadn't said anything about stress, 00:31:18.930 --> 00:31:22.230 align:middle line:84% you'd say, well, Deal-Grove would say I have 1,000 Angstrom 00:31:22.230 --> 00:31:23.890 align:middle line:90% oxides here, 1,000 here. 00:31:23.890 --> 00:31:25.193 align:middle line:90% I put them both in the furnace. 00:31:25.193 --> 00:31:26.610 align:middle line:84% You just use Deal-Grove, you would 00:31:26.610 --> 00:31:28.920 align:middle line:84% get the exact same thickness on both wafers. 00:31:28.920 --> 00:31:30.480 align:middle line:84% Deal-Grove doesn't tell you anything 00:31:30.480 --> 00:31:33.090 align:middle line:84% having to do with the history of how you grew the prior oxide 00:31:33.090 --> 00:31:33.660 align:middle line:90% layer. 00:31:33.660 --> 00:31:35.580 align:middle line:84% It just depends on-- remember, just 00:31:35.580 --> 00:31:37.890 align:middle line:84% the t sub i, the thickness, the initial thickness. 00:31:37.890 --> 00:31:40.410 align:middle line:84% It cares nothing about what temperature 00:31:40.410 --> 00:31:42.180 align:middle line:90% that oxide was grown at. 00:31:42.180 --> 00:31:43.980 align:middle line:84% In reality, though, if you go to look 00:31:43.980 --> 00:31:46.560 align:middle line:84% at the wafer that comes out of the furnace after this 800 00:31:46.560 --> 00:31:50.792 align:middle line:84% degree C, these two wafers will have different thicknesses. 00:31:50.792 --> 00:31:51.750 align:middle line:90% And so how can that be? 00:31:51.750 --> 00:31:53.040 align:middle line:84% They're both in the same furnace. 00:31:53.040 --> 00:31:54.748 align:middle line:84% They're sitting right next to each other. 00:31:54.748 --> 00:31:56.910 align:middle line:84% They had the exact same thickness to start with. 00:31:56.910 --> 00:31:59.460 align:middle line:90% Why would it matter? 00:31:59.460 --> 00:32:01.820 align:middle line:84% Well, actually the wafer oxidized at high temperature 00:32:01.820 --> 00:32:05.870 align:middle line:84% is going to grow faster compared to the 800 degrees, because it 00:32:05.870 --> 00:32:07.620 align:middle line:90% has lower stress levels. 00:32:07.620 --> 00:32:10.790 align:middle line:84% So it has a little bit more open network. 00:32:10.790 --> 00:32:13.850 align:middle line:84% So you can diffuse through that oxide more rapidly, 00:32:13.850 --> 00:32:15.043 align:middle line:90% and it's a k sub s. 00:32:15.043 --> 00:32:16.460 align:middle line:84% Its surface reaction rate is going 00:32:16.460 --> 00:32:18.020 align:middle line:84% to be a little bit faster because it 00:32:18.020 --> 00:32:19.890 align:middle line:90% doesn't have the stress in it. 00:32:19.890 --> 00:32:21.560 align:middle line:84% So you can see these differences even 00:32:21.560 --> 00:32:24.170 align:middle line:84% if you're not doing shape surfaces, just 00:32:24.170 --> 00:32:25.170 align:middle line:90% in planar surfaces. 00:32:25.170 --> 00:32:27.590 align:middle line:84% Now we chose a very extreme example. 00:32:27.590 --> 00:32:32.150 align:middle line:84% 1,100 degrees, which is quite hot, above the flow point 00:32:32.150 --> 00:32:37.130 align:middle line:84% of the glass at 800, which is quite a bit different. 00:32:37.130 --> 00:32:39.777 align:middle line:84% Maybe you typically wouldn't do such a wide range, 00:32:39.777 --> 00:32:42.110 align:middle line:84% but it really does tell you that the history effects can 00:32:42.110 --> 00:32:45.710 align:middle line:84% be important when you're trying to do accurate modeling. 00:32:45.710 --> 00:32:50.310 align:middle line:84% And in fact, most simulators don't take this into account, 00:32:50.310 --> 00:32:52.160 align:middle line:84% so it's something you'd have to account for, 00:32:52.160 --> 00:32:54.020 align:middle line:84% either empirically in your process 00:32:54.020 --> 00:32:57.680 align:middle line:90% or calculate it yourself. 00:32:57.680 --> 00:32:59.550 align:middle line:84% That's sort of an interesting effect 00:32:59.550 --> 00:33:01.215 align:middle line:90% that goes beyond shape effects. 00:33:01.215 --> 00:33:04.090 align:middle line:90% 00:33:04.090 --> 00:33:06.360 align:middle line:84% So let's go on to slide 15, and we're 00:33:06.360 --> 00:33:08.700 align:middle line:84% going to talk a little bit about the evolution of what's 00:33:08.700 --> 00:33:10.680 align:middle line:90% called recess LOCOS. 00:33:10.680 --> 00:33:12.990 align:middle line:84% Before, earlier today I talked about LOCOS 00:33:12.990 --> 00:33:14.640 align:middle line:84% and we just said we put down a mask, 00:33:14.640 --> 00:33:16.470 align:middle line:84% we have a flat surface to begin with, 00:33:16.470 --> 00:33:19.200 align:middle line:84% and it becomes a little strangely shaped 00:33:19.200 --> 00:33:21.180 align:middle line:90% as we oxidize it. 00:33:21.180 --> 00:33:24.060 align:middle line:84% Remember, we had quite a bit of encroachment. 00:33:24.060 --> 00:33:27.660 align:middle line:84% Even 0.2 or 0.3 microns of encroachment is too much. 00:33:27.660 --> 00:33:29.040 align:middle line:90% And not only that. 00:33:29.040 --> 00:33:30.770 align:middle line:90% If you look back-- let's see. 00:33:30.770 --> 00:33:35.580 align:middle line:84% I'm going go back a couple slides to, say, slide 12. 00:33:35.580 --> 00:33:38.040 align:middle line:84% When I take the nitride off and I 00:33:38.040 --> 00:33:40.500 align:middle line:84% go to use this as a processing my chip, 00:33:40.500 --> 00:33:43.720 align:middle line:90% the surface is not very planar. 00:33:43.720 --> 00:33:46.020 align:middle line:84% So I have to do photolithography now on this surface. 00:33:46.020 --> 00:33:46.895 align:middle line:90% It's not very planar. 00:33:46.895 --> 00:33:50.200 align:middle line:84% I have this big oxide hump sticking up here in the field, 00:33:50.200 --> 00:33:51.870 align:middle line:84% and the field is much higher here. 00:33:51.870 --> 00:33:52.620 align:middle line:90% Look at the field. 00:33:52.620 --> 00:33:57.510 align:middle line:84% It's up here at 0.3 microns above the original surface. 00:33:57.510 --> 00:34:01.440 align:middle line:84% So I got this big mountain sticking up over in the field. 00:34:01.440 --> 00:34:03.150 align:middle line:84% Photolithography doesn't like to be 00:34:03.150 --> 00:34:05.470 align:middle line:84% done on surfaces that are rough or surfaces 00:34:05.470 --> 00:34:06.700 align:middle line:90% that have big steps. 00:34:06.700 --> 00:34:09.790 align:middle line:84% The depth of focus from your camera 00:34:09.790 --> 00:34:13.132 align:middle line:84% or whatever, when you're trying to focus down, 00:34:13.132 --> 00:34:14.590 align:middle line:84% the smaller the feature size you're 00:34:14.590 --> 00:34:17.860 align:middle line:84% trying to focus on, the depth of focus 00:34:17.860 --> 00:34:22.630 align:middle line:84% becomes more and more difficult. And so you the lithography tool 00:34:22.630 --> 00:34:24.550 align:middle line:84% might not be able to focus on this valley 00:34:24.550 --> 00:34:27.580 align:middle line:84% when it's well focused on the highest part of the mountain. 00:34:27.580 --> 00:34:30.800 align:middle line:84% So there's a big requirement with litho today, 00:34:30.800 --> 00:34:33.100 align:middle line:84% forcing us to make all our chips so we 00:34:33.100 --> 00:34:37.820 align:middle line:84% can keep the surfaces as flat as possible in every litho step. 00:34:37.820 --> 00:34:40.270 align:middle line:84% So a step like this that might have been acceptable, 00:34:40.270 --> 00:34:44.980 align:middle line:84% 0.3 micron step years ago when people were making large 00:34:44.980 --> 00:34:48.460 align:middle line:84% devices and the lithography tools didn't use really high 00:34:48.460 --> 00:34:51.250 align:middle line:84% power lenses, now we're using a high power lens, 00:34:51.250 --> 00:34:53.270 align:middle line:84% we cannot tolerate such a big step. 00:34:53.270 --> 00:34:55.960 align:middle line:84% So there's always a move in silicon ice industry 00:34:55.960 --> 00:34:57.760 align:middle line:84% to keep the surface as flat as you can. 00:34:57.760 --> 00:35:00.560 align:middle line:90% LOCOS doesn't do that. 00:35:00.560 --> 00:35:03.670 align:middle line:84% So before people invented shallow trench, STI-- 00:35:03.670 --> 00:35:05.860 align:middle line:84% which is, again, a very planar process 00:35:05.860 --> 00:35:08.680 align:middle line:84% because you use a polishing process 00:35:08.680 --> 00:35:10.570 align:middle line:84% to polish things down at the end-- 00:35:10.570 --> 00:35:12.310 align:middle line:84% they did something called recess LOCOS. 00:35:12.310 --> 00:35:13.313 align:middle line:90% So this was an attempt-- 00:35:13.313 --> 00:35:15.730 align:middle line:84% well, you say, well, you have this big mountain growing up 00:35:15.730 --> 00:35:17.080 align:middle line:90% on the side. 00:35:17.080 --> 00:35:18.670 align:middle line:90% Let's give it a handicap. 00:35:18.670 --> 00:35:21.460 align:middle line:84% Let's etch the silicon in the recess 00:35:21.460 --> 00:35:24.070 align:middle line:84% to begin with so that as the oxide grows, 00:35:24.070 --> 00:35:27.590 align:middle line:84% when it's finished, it'll end up being pretty flat. 00:35:27.590 --> 00:35:30.410 align:middle line:84% And so you try to compensate for that. 00:35:30.410 --> 00:35:32.158 align:middle line:84% And this is an example of how that works, 00:35:32.158 --> 00:35:33.700 align:middle line:84% and I'll show you it doesn't come out 00:35:33.700 --> 00:35:35.090 align:middle line:90% exactly as flat as you want. 00:35:35.090 --> 00:35:37.870 align:middle line:84% But let's say this is a starting structure. 00:35:37.870 --> 00:35:41.610 align:middle line:84% So I have nitride over here, little pad ox underneath, 00:35:41.610 --> 00:35:44.080 align:middle line:84% and I've etched this recess over on the left. 00:35:44.080 --> 00:35:46.380 align:middle line:84% The recess is maybe 0.3 microns deep, 00:35:46.380 --> 00:35:49.708 align:middle line:84% and I'm doing that so that the field oxide will be recessed 00:35:49.708 --> 00:35:51.000 align:middle line:90% and I'll get a flatter surface. 00:35:51.000 --> 00:35:54.230 align:middle line:90% 00:35:54.230 --> 00:35:56.238 align:middle line:84% And then I go after certain timesteps. 00:35:56.238 --> 00:35:58.280 align:middle line:84% So this is after I've grown a little bit of oxide 00:35:58.280 --> 00:36:00.380 align:middle line:84% in the center here, and you can already 00:36:00.380 --> 00:36:03.290 align:middle line:84% see, even for a thin oxide, immediately 00:36:03.290 --> 00:36:04.370 align:middle line:90% some corner effects. 00:36:04.370 --> 00:36:07.460 align:middle line:84% The oxidation rate is a little slower here and here 00:36:07.460 --> 00:36:08.870 align:middle line:90% compared to on the flat. 00:36:08.870 --> 00:36:12.380 align:middle line:84% And on the far right, you can see 00:36:12.380 --> 00:36:15.150 align:middle line:84% I'm starting to-- the corner effects are still evident 00:36:15.150 --> 00:36:17.780 align:middle line:84% and I'm starting to develop this bird's beak thing. 00:36:17.780 --> 00:36:20.883 align:middle line:84% The stress of a nitride is causing the oxidation rate here 00:36:20.883 --> 00:36:22.550 align:middle line:84% to be also a little slower, and the fact 00:36:22.550 --> 00:36:25.620 align:middle line:90% that it has to diffuse through. 00:36:25.620 --> 00:36:27.495 align:middle line:84% So let's go to the next time step. 00:36:27.495 --> 00:36:29.120 align:middle line:84% This is sort of a time evolution, which 00:36:29.120 --> 00:36:30.530 align:middle line:90% is easy to do in the computer. 00:36:30.530 --> 00:36:33.650 align:middle line:84% Here's the oxide again after a certain period. 00:36:33.650 --> 00:36:34.500 align:middle line:90% Once more. 00:36:34.500 --> 00:36:38.430 align:middle line:84% And finally, after 90 minutes, the structure is almost flat. 00:36:38.430 --> 00:36:42.530 align:middle line:84% So I achieved what I wanted in that in the flat regions 00:36:42.530 --> 00:36:44.600 align:middle line:84% out far away, in the field region here, 00:36:44.600 --> 00:36:48.530 align:middle line:84% the surface of the oxide is at zero or close to. 00:36:48.530 --> 00:36:50.180 align:middle line:84% Not zero, but it's at a height that's 00:36:50.180 --> 00:36:53.830 align:middle line:84% just about equal to the surface of the silicon. 00:36:53.830 --> 00:36:55.700 align:middle line:90% So that's flat. 00:36:55.700 --> 00:36:58.080 align:middle line:84% Once I strip the nitride off and etch it, 00:36:58.080 --> 00:37:01.030 align:middle line:84% I still have this little bump except for the bird's head. 00:37:01.030 --> 00:37:02.530 align:middle line:84% And if you look at this, you can see 00:37:02.530 --> 00:37:03.960 align:middle line:84% it looks like a bird's head and a bird's beak 00:37:03.960 --> 00:37:06.040 align:middle line:84% and the bird's eye would be right about there. 00:37:06.040 --> 00:37:08.160 align:middle line:84% So that we really can't get away from, 00:37:08.160 --> 00:37:10.620 align:middle line:84% and that's because the oxidant diffuses 00:37:10.620 --> 00:37:13.680 align:middle line:84% underneath that nitride, causes some oxidation, 00:37:13.680 --> 00:37:17.080 align:middle line:84% and the exact shape of the bird's head depends on, 00:37:17.080 --> 00:37:20.310 align:middle line:84% of course, what kind of stress model you use. 00:37:20.310 --> 00:37:21.960 align:middle line:84% But it's better than it was before. 00:37:21.960 --> 00:37:24.780 align:middle line:84% The lithography has a flat surface here, a flat here, 00:37:24.780 --> 00:37:26.530 align:middle line:84% but it's got this little bump here, 00:37:26.530 --> 00:37:30.390 align:middle line:84% which is not so easy to get rid of. 00:37:30.390 --> 00:37:35.420 align:middle line:84% So the next technology people develop is shown on slide 17. 00:37:35.420 --> 00:37:37.490 align:middle line:84% People thought of this other effect, 00:37:37.490 --> 00:37:39.750 align:middle line:84% other idea called the Swami process. 00:37:39.750 --> 00:37:41.730 align:middle line:84% This was done at Hewlett Packard, 00:37:41.730 --> 00:37:44.730 align:middle line:84% and this was even a little more sophisticated. 00:37:44.730 --> 00:37:48.830 align:middle line:84% They were trying to keep the occident from getting around. 00:37:48.830 --> 00:37:51.440 align:middle line:84% Remember, if we go back here-- let me go back for a second. 00:37:51.440 --> 00:37:55.580 align:middle line:84% The starting structure here, the problem is that at this corner, 00:37:55.580 --> 00:37:58.670 align:middle line:84% the oxygen can diffuse straight in and, right off the bat, 00:37:58.670 --> 00:38:00.080 align:middle line:84% start oxidizing, and that's going 00:38:00.080 --> 00:38:02.780 align:middle line:84% to form where the bird's eye or the bird's head ends up. 00:38:02.780 --> 00:38:06.500 align:middle line:84% So they thought, all right, let me put this nitride mask all 00:38:06.500 --> 00:38:08.900 align:middle line:84% the way down this the edge, and that'll 00:38:08.900 --> 00:38:11.060 align:middle line:84% keep the oxygen from going up there 00:38:11.060 --> 00:38:13.730 align:middle line:84% and reduce that amount of oxidation. 00:38:13.730 --> 00:38:15.600 align:middle line:84% So this was the Swami structures. 00:38:15.600 --> 00:38:17.400 align:middle line:84% They're very similar to recessed LOCOS. 00:38:17.400 --> 00:38:20.870 align:middle line:84% You create the recess, but they create an oxide mask here, 00:38:20.870 --> 00:38:23.250 align:middle line:90% and they also put a paddocks-- 00:38:23.250 --> 00:38:23.750 align:middle line:90% I'm sorry. 00:38:23.750 --> 00:38:26.720 align:middle line:84% A nitride mask along the edge of the recess, 00:38:26.720 --> 00:38:29.860 align:middle line:84% and even at the very bottom, at the foot. 00:38:29.860 --> 00:38:32.560 align:middle line:84% So trying to minimize the oxidation in the active device 00:38:32.560 --> 00:38:37.990 align:middle line:84% region by masking the etched sidewall. 00:38:37.990 --> 00:38:39.702 align:middle line:84% But they want to do this in a way 00:38:39.702 --> 00:38:42.160 align:middle line:84% that you don't build up too much stress, because after all, 00:38:42.160 --> 00:38:44.202 align:middle line:84% the oxidant will find its way over here and start 00:38:44.202 --> 00:38:45.970 align:middle line:90% growing in oxide. 00:38:45.970 --> 00:38:48.070 align:middle line:84% If the nitride is too thick here, 00:38:48.070 --> 00:38:49.878 align:middle line:90% it can build up too much stress. 00:38:49.878 --> 00:38:51.670 align:middle line:84% If you build up enough stress in the oxide, 00:38:51.670 --> 00:38:53.530 align:middle line:84% you can actually crack the silicon 00:38:53.530 --> 00:38:55.340 align:middle line:90% or introduce dislocations. 00:38:55.340 --> 00:38:57.340 align:middle line:84% They're really not cracks, they're dislocations, 00:38:57.340 --> 00:38:58.860 align:middle line:90% but that can be a problem. 00:38:58.860 --> 00:39:00.770 align:middle line:84% So you notice they did this in a way 00:39:00.770 --> 00:39:03.620 align:middle line:84% that the nitride is a little thinner along the sidewall 00:39:03.620 --> 00:39:05.780 align:middle line:84% and down here so that the flap can 00:39:05.780 --> 00:39:07.910 align:middle line:84% be lifted without too much stress being induced. 00:39:07.910 --> 00:39:11.900 align:middle line:90% 00:39:11.900 --> 00:39:14.180 align:middle line:84% So we're going to take this structure now, put it 00:39:14.180 --> 00:39:17.750 align:middle line:84% in the furnace at 1,000 degrees, in moisture or water 00:39:17.750 --> 00:39:20.240 align:middle line:84% vapor for quite a few minutes and see 00:39:20.240 --> 00:39:24.730 align:middle line:90% what happens as we oxidize it. 00:39:24.730 --> 00:39:29.530 align:middle line:84% So here on the left on slide 18 is the starting structure. 00:39:29.530 --> 00:39:32.400 align:middle line:84% And now look after 450 minutes of oxidation. 00:39:32.400 --> 00:39:33.980 align:middle line:90% This is the simulation. 00:39:33.980 --> 00:39:35.730 align:middle line:84% Unfortunately, I'm sorry about the colors, 00:39:35.730 --> 00:39:37.950 align:middle line:84% but I think you can see this top layer 00:39:37.950 --> 00:39:40.530 align:middle line:84% is the original nitride, silicon nitride. 00:39:40.530 --> 00:39:41.550 align:middle line:90% Look at the flap. 00:39:41.550 --> 00:39:44.970 align:middle line:84% The flap that was along this sidewall 00:39:44.970 --> 00:39:47.400 align:middle line:84% has been lifted up because the oxidant has moved 00:39:47.400 --> 00:39:50.040 align:middle line:84% its way under there and pushed it up, 00:39:50.040 --> 00:39:53.490 align:middle line:84% and then it's been-- it's quasi vertical over here. 00:39:53.490 --> 00:39:56.340 align:middle line:84% The flap has been pushed out by the oxidant that 00:39:56.340 --> 00:39:59.890 align:middle line:84% was coming under here from the left and pushing on it. 00:39:59.890 --> 00:40:02.230 align:middle line:84% So the thin flap lifts up during oxidation, 00:40:02.230 --> 00:40:04.300 align:middle line:84% but it's still reasonably flat surface. 00:40:04.300 --> 00:40:06.520 align:middle line:84% Not perfect, but they're trying to reduce 00:40:06.520 --> 00:40:11.620 align:middle line:84% the amount of bird's beaking and bird's head that formed. 00:40:11.620 --> 00:40:13.720 align:middle line:90% That's the simulation. 00:40:13.720 --> 00:40:19.370 align:middle line:84% And in fact, on the lower left, on slide 19, 00:40:19.370 --> 00:40:22.230 align:middle line:84% the upper diagrams are the same as what we just saw. 00:40:22.230 --> 00:40:24.590 align:middle line:84% But if we turn to slide 19 in the lower left, 00:40:24.590 --> 00:40:25.610 align:middle line:90% they have actually-- 00:40:25.610 --> 00:40:27.440 align:middle line:84% SUPREM-IV form can actually be used 00:40:27.440 --> 00:40:30.170 align:middle line:84% to calculate stress contours, so it's actually 00:40:30.170 --> 00:40:33.080 align:middle line:84% calculating the stress in the oxide of course at each point, 00:40:33.080 --> 00:40:36.710 align:middle line:84% and at each point, assigning an oxidation rate. 00:40:36.710 --> 00:40:41.750 align:middle line:84% That's how it knows how to calculate the shape. 00:40:41.750 --> 00:40:43.970 align:middle line:84% There is some stress buildup or concentration 00:40:43.970 --> 00:40:47.390 align:middle line:84% here here in this yellow region, so these different-- 00:40:47.390 --> 00:40:49.520 align:middle line:84% I don't have stress numbers associated with this, 00:40:49.520 --> 00:40:50.310 align:middle line:90% to be honest. 00:40:50.310 --> 00:40:52.940 align:middle line:84% But if you run SUPREM-IV, it'll tell you 00:40:52.940 --> 00:40:54.830 align:middle line:84% in each color what that corresponds to what 00:40:54.830 --> 00:40:57.620 align:middle line:90% stress level in the oxide. 00:40:57.620 --> 00:40:59.670 align:middle line:84% On the lower right, which is kind of interesting, 00:40:59.670 --> 00:41:00.770 align:middle line:90% this is an actual-- 00:41:00.770 --> 00:41:04.440 align:middle line:84% from the HP process, this is the actual experimental data. 00:41:04.440 --> 00:41:06.140 align:middle line:90% So they grew this. 00:41:06.140 --> 00:41:08.480 align:middle line:84% They made this Swami process, and then they 00:41:08.480 --> 00:41:11.900 align:middle line:84% did a polished cross-section, scanning an electron 00:41:11.900 --> 00:41:12.650 align:middle line:90% micrograph. 00:41:12.650 --> 00:41:15.680 align:middle line:84% And this is the act of silicon region afterwards. 00:41:15.680 --> 00:41:20.980 align:middle line:84% This region right here is the oxide and the nitride. 00:41:20.980 --> 00:41:25.050 align:middle line:84% And you can see if you compare the simulation upper right 00:41:25.050 --> 00:41:28.290 align:middle line:84% to the lower right, the actual data, it's not that far off. 00:41:28.290 --> 00:41:30.510 align:middle line:84% It reproduces some of the key features. 00:41:30.510 --> 00:41:32.400 align:middle line:84% You can see this little divot here, 00:41:32.400 --> 00:41:35.430 align:middle line:90% this little V shaped well. 00:41:35.430 --> 00:41:38.350 align:middle line:84% Well, that's also seen in the simulation. 00:41:38.350 --> 00:41:41.580 align:middle line:84% So the stress effects did a reasonably good job of getting 00:41:41.580 --> 00:41:44.070 align:middle line:84% the shape right here at the corner is not perfect, 00:41:44.070 --> 00:41:45.600 align:middle line:90% but it's close. 00:41:45.600 --> 00:41:48.840 align:middle line:84% The actual device shown here, the actual data 00:41:48.840 --> 00:41:51.780 align:middle line:84% seems to have a little bit more of a bird's head, then. 00:41:51.780 --> 00:41:53.910 align:middle line:84% Maybe the stress effects weren't perfectly modeled, 00:41:53.910 --> 00:41:58.200 align:middle line:90% but it's pretty darn close. 00:41:58.200 --> 00:42:01.620 align:middle line:84% People spend a lot of time calibrating their stress models 00:42:01.620 --> 00:42:03.240 align:middle line:84% and their stress coefficients to try 00:42:03.240 --> 00:42:07.570 align:middle line:84% to get close to what's measured experimentally. 00:42:07.570 --> 00:42:08.500 align:middle line:90% OK. 00:42:08.500 --> 00:42:12.300 align:middle line:90% Let's go on to slide 20. 00:42:12.300 --> 00:42:14.190 align:middle line:84% What I just talked about is specifically 00:42:14.190 --> 00:42:17.190 align:middle line:84% related to seamless processing devices, 00:42:17.190 --> 00:42:19.095 align:middle line:90% trying to make the active area. 00:42:19.095 --> 00:42:20.910 align:middle line:84% By definition, when you're doing that, 00:42:20.910 --> 00:42:23.940 align:middle line:84% you're doing a patterned oxidation, 00:42:23.940 --> 00:42:26.110 align:middle line:90% and you have stress effects. 00:42:26.110 --> 00:42:28.470 align:middle line:84% Here's a neat application of stress effects 00:42:28.470 --> 00:42:30.442 align:middle line:84% that has nothing to do with CMOS. 00:42:30.442 --> 00:42:32.400 align:middle line:84% For those of you who are interested in nanotech 00:42:32.400 --> 00:42:35.670 align:middle line:84% or nanostructures, on page 20, this 00:42:35.670 --> 00:42:38.520 align:middle line:84% is an application of stress oxidation effects 00:42:38.520 --> 00:42:40.290 align:middle line:90% to form nanowires. 00:42:40.290 --> 00:42:43.980 align:middle line:84% I took this from a PhD thesis of Harvey Lu at Stanford. 00:42:43.980 --> 00:42:47.652 align:middle line:90% He graduated in 1995. 00:42:47.652 --> 00:42:49.860 align:middle line:84% And what he was doing, and a lot of people since then 00:42:49.860 --> 00:42:51.810 align:middle line:84% have done the same thing, he was first 00:42:51.810 --> 00:42:55.320 align:middle line:84% etching a very small, tiny pillar in silicon. 00:42:55.320 --> 00:42:58.050 align:middle line:90% So a tall pillar. 00:42:58.050 --> 00:43:00.060 align:middle line:90% Say, a couple microns tall. 00:43:00.060 --> 00:43:02.280 align:middle line:84% But where the starting diameter of the pillar 00:43:02.280 --> 00:43:04.500 align:middle line:84% would maybe only be a couple hundred angstroms. 00:43:04.500 --> 00:43:06.100 align:middle line:90% Say 200 angstroms. 00:43:06.100 --> 00:43:08.580 align:middle line:84% So he starts with this little needle of silicon. 00:43:08.580 --> 00:43:12.130 align:middle line:84% And you can do a 200 Angstrom pillar or a 300 Angstrom 00:43:12.130 --> 00:43:15.520 align:middle line:84% diameter pillar with a very good electron beam lithography 00:43:15.520 --> 00:43:16.555 align:middle line:90% machine. 00:43:16.555 --> 00:43:18.430 align:middle line:84% At least in those days, that's about as small 00:43:18.430 --> 00:43:21.550 align:middle line:84% as people could get-- a couple hundred angstroms. 00:43:21.550 --> 00:43:23.770 align:middle line:84% You patterned the dot and then you etched the silicon 00:43:23.770 --> 00:43:25.780 align:middle line:90% into a nice pillar. 00:43:25.780 --> 00:43:28.120 align:middle line:84% And then he took that pillar, those pillars, 00:43:28.120 --> 00:43:30.860 align:middle line:84% and he put it in the furnace and oxidized it. 00:43:30.860 --> 00:43:33.520 align:middle line:84% And the idea was he wanted to get the pillar diameter down. 00:43:33.520 --> 00:43:36.620 align:middle line:84% He could not lithographically make the pillar any smaller, 00:43:36.620 --> 00:43:39.450 align:middle line:84% but he wanted to look at quantum effects, current conduction, 00:43:39.450 --> 00:43:41.110 align:middle line:90% in this very thin wire. 00:43:41.110 --> 00:43:42.610 align:middle line:84% He wanted to get it thinner somehow. 00:43:42.610 --> 00:43:45.550 align:middle line:84% And an obvious way to make it thinner is oxidize it, right? 00:43:45.550 --> 00:43:48.228 align:middle line:84% As you oxidize, you will consume from the top 00:43:48.228 --> 00:43:50.770 align:middle line:84% and from all the way around the sides of the little cylinder, 00:43:50.770 --> 00:43:53.658 align:middle line:84% you will consume the silicon by the SiO2 production. 00:43:53.658 --> 00:43:55.450 align:middle line:84% And then you can strip the oxide at the end 00:43:55.450 --> 00:43:58.613 align:middle line:84% if you want and make electrical measurement which turns out not 00:43:58.613 --> 00:44:00.280 align:middle line:84% to be so easy to contact a pillar that's 00:44:00.280 --> 00:44:01.810 align:middle line:90% only 20 angstroms. 00:44:01.810 --> 00:44:04.000 align:middle line:84% A little minor technical problem, 00:44:04.000 --> 00:44:05.390 align:middle line:90% but he was able to do it. 00:44:05.390 --> 00:44:08.032 align:middle line:84% And in fact, this is a cross-section TEM, 00:44:08.032 --> 00:44:09.490 align:middle line:84% a transmission electron micrograph, 00:44:09.490 --> 00:44:11.290 align:middle line:90% of one of Harvey's pillars. 00:44:11.290 --> 00:44:14.560 align:middle line:84% And what you can see here is two nanometers in diameter. 00:44:14.560 --> 00:44:15.690 align:middle line:90% So it's 20 angstroms. 00:44:15.690 --> 00:44:18.310 align:middle line:84% So it's a really tiny wire of silicon. 00:44:18.310 --> 00:44:21.490 align:middle line:84% And this amorphous looking material all the way around it 00:44:21.490 --> 00:44:26.660 align:middle line:84% on the sides is the silicon dioxide that he grew. 00:44:26.660 --> 00:44:29.180 align:middle line:84% This particular one was oxidized in dry ox 00:44:29.180 --> 00:44:32.195 align:middle line:90% at 875 for about 10 hours. 00:44:32.195 --> 00:44:35.357 align:middle line:90% 00:44:35.357 --> 00:44:37.190 align:middle line:84% And you can actually see the lattice fringes 00:44:37.190 --> 00:44:39.050 align:middle line:84% in the single crystal silicon that remained. 00:44:39.050 --> 00:44:40.508 align:middle line:84% What was really neat is what Harvey 00:44:40.508 --> 00:44:43.040 align:middle line:84% found is, depending on the temperature that he did this, 00:44:43.040 --> 00:44:47.310 align:middle line:84% he could oxidize that pillar and it would stop. 00:44:47.310 --> 00:44:49.620 align:middle line:84% The oxidation rate would go way down when he got down 00:44:49.620 --> 00:44:52.380 align:middle line:90% to a certain diameter pillar. 00:44:52.380 --> 00:44:55.440 align:middle line:84% And so that's quite convenient because then you 00:44:55.440 --> 00:44:57.390 align:middle line:84% could use this to form reasonably-- 00:44:57.390 --> 00:44:59.460 align:middle line:84% let's say lithography is not perfect, right? 00:44:59.460 --> 00:45:01.170 align:middle line:84% Some of the dots came out when he 00:45:01.170 --> 00:45:04.897 align:middle line:84% patterned at 200 angstroms, 250, maybe, 210. 00:45:04.897 --> 00:45:05.730 align:middle line:90% Something like that. 00:45:05.730 --> 00:45:06.630 align:middle line:90% Litho is not perfect. 00:45:06.630 --> 00:45:09.000 align:middle line:84% But he could get a reasonably wide range 00:45:09.000 --> 00:45:11.790 align:middle line:90% in the starting pillar diameter. 00:45:11.790 --> 00:45:14.880 align:middle line:84% The final pillar diameter, if he oxidized it long enough, 00:45:14.880 --> 00:45:16.900 align:middle line:90% was pretty darn uniform. 00:45:16.900 --> 00:45:19.830 align:middle line:84% Which was kind of nice, because you have a process now 00:45:19.830 --> 00:45:24.553 align:middle line:84% that's physically slowing down as it gets smaller. 00:45:24.553 --> 00:45:26.220 align:middle line:84% And so they all end up-- even though you 00:45:26.220 --> 00:45:27.570 align:middle line:84% have a large distribution to begin with, 00:45:27.570 --> 00:45:29.400 align:middle line:84% they all end up very close to a single diameter. 00:45:29.400 --> 00:45:30.570 align:middle line:84% If you're trying to make some device 00:45:30.570 --> 00:45:32.130 align:middle line:84% or some interesting quantum thing, 00:45:32.130 --> 00:45:35.430 align:middle line:84% that was a very nice side effect. 00:45:35.430 --> 00:45:38.910 align:middle line:84% It's a side effect of, he believed, the stress effect. 00:45:38.910 --> 00:45:41.460 align:middle line:84% That there was stress built up in the silicon 00:45:41.460 --> 00:45:44.280 align:middle line:84% and in the oxide in such a way that the oxidation 00:45:44.280 --> 00:45:47.160 align:middle line:84% rate, as the pillar got thinner and thinner, 00:45:47.160 --> 00:45:49.440 align:middle line:84% the oxidation rate would go down, 00:45:49.440 --> 00:45:51.120 align:middle line:84% and it ended up being somewhat self 00:45:51.120 --> 00:45:53.610 align:middle line:84% limiting at a certain diameter pillar. 00:45:53.610 --> 00:45:55.620 align:middle line:84% It's not exactly the same kind of effect 00:45:55.620 --> 00:45:59.327 align:middle line:84% that's going on perfectly as LOCOS, but it's related. 00:45:59.327 --> 00:46:01.410 align:middle line:84% It gives you an idea that you can use these stress 00:46:01.410 --> 00:46:04.320 align:middle line:84% effects in processes to create structures that you wouldn't 00:46:04.320 --> 00:46:06.580 align:middle line:90% be able to otherwise create. 00:46:06.580 --> 00:46:08.370 align:middle line:84% If he didn't have this effect, he'd 00:46:08.370 --> 00:46:10.830 align:middle line:84% have a very good chance of really ending up 00:46:10.830 --> 00:46:15.000 align:middle line:84% with very few pillars that were of similar diameter. 00:46:15.000 --> 00:46:17.900 align:middle line:90% 00:46:17.900 --> 00:46:18.400 align:middle line:90% OK. 00:46:18.400 --> 00:46:19.870 align:middle line:90% So just as a point to make. 00:46:19.870 --> 00:46:21.640 align:middle line:84% And if you're working on quantum devices, 00:46:21.640 --> 00:46:25.180 align:middle line:84% you'll see a lot of work where people do shaped oxidation. 00:46:25.180 --> 00:46:26.440 align:middle line:90% They don't just make pillars. 00:46:26.440 --> 00:46:29.110 align:middle line:84% They make small structures with electron beam. 00:46:29.110 --> 00:46:32.770 align:middle line:84% They pattern a silicon on insulator layer, 00:46:32.770 --> 00:46:35.350 align:middle line:84% and they take advantage of corner effects 00:46:35.350 --> 00:46:37.330 align:middle line:84% to create funny looking shaped structures that 00:46:37.330 --> 00:46:38.538 align:middle line:90% are used for quantum devices. 00:46:38.538 --> 00:46:40.660 align:middle line:84% So shaped oxidation stress effects 00:46:40.660 --> 00:46:43.780 align:middle line:84% are used all the time today in research other than just people 00:46:43.780 --> 00:46:47.090 align:middle line:90% doing CMOS devices. 00:46:47.090 --> 00:46:50.080 align:middle line:84% So what I want to go on to beyond the stress 00:46:50.080 --> 00:46:54.190 align:middle line:84% is to talk about now a more atomistic or microscopic 00:46:54.190 --> 00:46:56.990 align:middle line:90% or point defect based model. 00:46:56.990 --> 00:47:01.000 align:middle line:84% So far we've been using macroscopic models. 00:47:01.000 --> 00:47:03.010 align:middle line:90% We write chemical equations. 00:47:03.010 --> 00:47:04.870 align:middle line:84% We talk about diffusion coefficients. 00:47:04.870 --> 00:47:06.700 align:middle line:84% We're not really looking at an atomic scale 00:47:06.700 --> 00:47:10.525 align:middle line:84% of what's going on at the interface or in the oxide. 00:47:10.525 --> 00:47:12.700 align:middle line:84% There's also a very atomistic picture 00:47:12.700 --> 00:47:15.890 align:middle line:84% of the process that turns out to be very important. 00:47:15.890 --> 00:47:17.770 align:middle line:84% It's important because it explains 00:47:17.770 --> 00:47:19.270 align:middle line:84% certain physical phenomena that are 00:47:19.270 --> 00:47:22.240 align:middle line:84% critical to fabrication such as oxidation 00:47:22.240 --> 00:47:27.340 align:middle line:84% enhanced diffusion, oxidation retarded diffusion, 00:47:27.340 --> 00:47:31.240 align:middle line:84% and these things that tell us that the oxidation process 00:47:31.240 --> 00:47:34.570 align:middle line:84% itself dramatically changes the point defect distribution 00:47:34.570 --> 00:47:37.510 align:middle line:90% in the near surface region. 00:47:37.510 --> 00:47:39.580 align:middle line:84% Most of the ideas are related to the fact 00:47:39.580 --> 00:47:41.830 align:middle line:84% that there is volume expansion going on 00:47:41.830 --> 00:47:44.150 align:middle line:84% and the need for free volume at that interface. 00:47:44.150 --> 00:47:48.830 align:middle line:84% And so point defects play a role in that volume expansion. 00:47:48.830 --> 00:47:51.640 align:middle line:90% 00:47:51.640 --> 00:47:55.200 align:middle line:84% So let's go on to slide number 22. 00:47:55.200 --> 00:47:56.958 align:middle line:84% This is a kind of ugly looking equation. 00:47:56.958 --> 00:47:58.500 align:middle line:84% I don't really like to use it, so you 00:47:58.500 --> 00:48:00.393 align:middle line:84% don't have to look at every single term 00:48:00.393 --> 00:48:01.560 align:middle line:90% and understand it perfectly. 00:48:01.560 --> 00:48:04.447 align:middle line:84% But it sort of shows qualitatively what 00:48:04.447 --> 00:48:05.280 align:middle line:90% we're talking about. 00:48:05.280 --> 00:48:07.530 align:middle line:84% This is an equation we talked about in chapter 3 00:48:07.530 --> 00:48:10.440 align:middle line:84% when we were discussing SiO2 precipitation 00:48:10.440 --> 00:48:11.880 align:middle line:90% and using gettering. 00:48:11.880 --> 00:48:15.120 align:middle line:84% But the same equation essentially 00:48:15.120 --> 00:48:19.020 align:middle line:84% applies here when we're oxidizing a silicon surface. 00:48:19.020 --> 00:48:21.120 align:middle line:84% And what it tells us is that we have 00:48:21.120 --> 00:48:23.550 align:middle line:84% a certain number of silicon lattice sites, 00:48:23.550 --> 00:48:29.280 align:middle line:84% the silicon sub SI, combining with a certain number of oxygen 00:48:29.280 --> 00:48:34.470 align:middle line:84% species and combining with a certain number of vacancies. 00:48:34.470 --> 00:48:38.190 align:middle line:84% Notice that vacancies are being consumed in this process, 00:48:38.190 --> 00:48:40.170 align:middle line:90% forming SiO2. 00:48:40.170 --> 00:48:43.530 align:middle line:84% Vacancies at the surface are being consumed, presumably, 00:48:43.530 --> 00:48:46.230 align:middle line:84% because we need space for that oxide to move, 00:48:46.230 --> 00:48:47.870 align:middle line:90% to push around a little bit. 00:48:47.870 --> 00:48:51.220 align:middle line:84% So consuming a vacancy will help you find extra space, 00:48:51.220 --> 00:48:52.790 align:middle line:90% essentially, at the interface. 00:48:52.790 --> 00:48:54.790 align:middle line:84% So I consume a vacancy on the left. 00:48:54.790 --> 00:48:58.390 align:middle line:84% I form SiO2, and I actually form-- 00:48:58.390 --> 00:49:01.570 align:middle line:84% on the right side I also create these interstitials, 00:49:01.570 --> 00:49:03.280 align:middle line:84% a certain number of interstitials. 00:49:03.280 --> 00:49:04.370 align:middle line:90% Plus stress, of course. 00:49:04.370 --> 00:49:06.620 align:middle line:90% We talked about stress. 00:49:06.620 --> 00:49:09.130 align:middle line:84% So this oxidation reaction happening right here 00:49:09.130 --> 00:49:12.460 align:middle line:84% at the borderline between the pink and the white region here 00:49:12.460 --> 00:49:14.872 align:middle line:84% at star, they can consume vacancies 00:49:14.872 --> 00:49:16.330 align:middle line:84% or they can generate interstitials. 00:49:16.330 --> 00:49:19.060 align:middle line:84% So I've got a vacancy going in, interstitial going out 00:49:19.060 --> 00:49:22.780 align:middle line:84% either way in order to provide the volume needed 00:49:22.780 --> 00:49:23.750 align:middle line:90% for the reaction. 00:49:23.750 --> 00:49:29.530 align:middle line:84% So we need room to put these oxygen atoms at that interface. 00:49:29.530 --> 00:49:33.220 align:middle line:84% And that room comes from vacancies and also picking out 00:49:33.220 --> 00:49:38.040 align:middle line:84% silicon atoms from the lattice, creating interstitials. 00:49:38.040 --> 00:49:40.950 align:middle line:90% So let's go on to slide 23. 00:49:40.950 --> 00:49:43.830 align:middle line:84% This generation of interstitials is extremely important 00:49:43.830 --> 00:49:46.830 align:middle line:84% because it is used to explain the non-local effects 00:49:46.830 --> 00:49:50.010 align:middle line:84% of oxidation, such as oxidation enhanced diffusion, which 00:49:50.010 --> 00:49:54.420 align:middle line:84% we'll show, or oxidation retarded diffusion. 00:49:54.420 --> 00:49:55.890 align:middle line:84% That's because the interstitial can 00:49:55.890 --> 00:49:57.900 align:middle line:84% diffuse very far from the interface 00:49:57.900 --> 00:50:00.480 align:middle line:84% and change the diffusivity of dopants. 00:50:00.480 --> 00:50:04.260 align:middle line:84% So this shows how these models can actually be used. 00:50:04.260 --> 00:50:08.670 align:middle line:84% At this interface, about 1 in 1,000 of the oxidized silicon 00:50:08.670 --> 00:50:10.770 align:middle line:90% atoms is injected into the bulk. 00:50:10.770 --> 00:50:14.250 align:middle line:84% So for every thousand atoms that I oxidize, 00:50:14.250 --> 00:50:17.520 align:middle line:84% one atom goes into the bulk that wouldn't be there ordinarily 00:50:17.520 --> 00:50:20.670 align:middle line:90% if I were not oxidizing. 00:50:20.670 --> 00:50:22.140 align:middle line:90% So it doesn't sound like a lot. 00:50:22.140 --> 00:50:24.420 align:middle line:84% Only one atom out of 1,000 goes into the bulk 00:50:24.420 --> 00:50:25.830 align:middle line:90% of the ones out of the surface. 00:50:25.830 --> 00:50:27.247 align:middle line:84% OK, that doesn't sound like a lot. 00:50:27.247 --> 00:50:29.850 align:middle line:84% But their impact is huge in oxidation 00:50:29.850 --> 00:50:32.280 align:middle line:84% enhanced diffusion, oxidation retarded diffusion, 00:50:32.280 --> 00:50:34.260 align:middle line:84% and the growth of stacking faults. 00:50:34.260 --> 00:50:37.370 align:middle line:90% So it's a very important effect. 00:50:37.370 --> 00:50:40.520 align:middle line:84% And on slide 24, we have a picture 00:50:40.520 --> 00:50:44.810 align:middle line:84% from your text, a very schematic picture of what's 00:50:44.810 --> 00:50:46.820 align:middle line:90% going on when we do oxidation. 00:50:46.820 --> 00:50:49.070 align:middle line:84% So this is a LOCOS structure you recognize by now. 00:50:49.070 --> 00:50:50.840 align:middle line:90% Here's my nitride on the left. 00:50:50.840 --> 00:50:53.510 align:middle line:84% There's no oxidation taking place on the left. 00:50:53.510 --> 00:50:55.850 align:middle line:84% I am actively oxidizing on the right. 00:50:55.850 --> 00:50:59.097 align:middle line:84% You can see this thick layer of oxide that I've grown. 00:50:59.097 --> 00:51:00.680 align:middle line:84% At the surface here there are a number 00:51:00.680 --> 00:51:02.390 align:middle line:90% of processes taking place. 00:51:02.390 --> 00:51:07.020 align:middle line:84% This G is meant to represent the generation of interstitials 00:51:07.020 --> 00:51:07.745 align:middle line:90% or the injection. 00:51:07.745 --> 00:51:09.120 align:middle line:84% It's a certain rate of generation 00:51:09.120 --> 00:51:11.517 align:middle line:84% of interstitials that get injected and go into the bulk. 00:51:11.517 --> 00:51:12.100 align:middle line:90% Here they are. 00:51:12.100 --> 00:51:14.190 align:middle line:90% Here's an interstitial. 00:51:14.190 --> 00:51:16.680 align:middle line:84% There's also a flux of interstitials 00:51:16.680 --> 00:51:19.000 align:middle line:84% from the bulk to the surface, and they can recombine. 00:51:19.000 --> 00:51:21.420 align:middle line:84% So interstitials can be generated during oxidation. 00:51:21.420 --> 00:51:23.290 align:middle line:84% They can also find their way to the surface 00:51:23.290 --> 00:51:25.140 align:middle line:90% and they can recombine there. 00:51:25.140 --> 00:51:27.600 align:middle line:84% The interstitials can recombine in the bulk with vacancy. 00:51:27.600 --> 00:51:30.490 align:middle line:84% They can find a vacancy and go away. 00:51:30.490 --> 00:51:32.760 align:middle line:84% They can diffuse over to this surface, which 00:51:32.760 --> 00:51:36.720 align:middle line:84% is a non-oxidizing or inert place in recombine, 00:51:36.720 --> 00:51:38.160 align:middle line:90% or they can go into the bulk. 00:51:38.160 --> 00:51:41.230 align:middle line:84% And here, here's a buried dopant marker layer, 00:51:41.230 --> 00:51:43.950 align:middle line:84% which started out looking like this dark green. 00:51:43.950 --> 00:51:45.240 align:middle line:90% It was very narrow. 00:51:45.240 --> 00:51:47.670 align:middle line:84% And lo and behold, underneath where 00:51:47.670 --> 00:51:50.610 align:middle line:84% you're doing the oxidation, the boron diffuses a lot, 00:51:50.610 --> 00:51:52.260 align:middle line:90% so it becomes very wide. 00:51:52.260 --> 00:51:55.080 align:middle line:84% So they come in here and they disturb the boron diffusion. 00:51:55.080 --> 00:51:56.020 align:middle line:90% They enhance it. 00:51:56.020 --> 00:51:59.400 align:middle line:84% And so the junction here, the width of this boron profile 00:51:59.400 --> 00:52:02.700 align:middle line:84% is now quite wide underneath the oxide where it was growing. 00:52:02.700 --> 00:52:07.280 align:middle line:84% Under the nitride where there was no oxidation taking place 00:52:07.280 --> 00:52:08.350 align:middle line:90% I have inert diffusion. 00:52:08.350 --> 00:52:10.870 align:middle line:84% There's very little broad motion. 00:52:10.870 --> 00:52:13.250 align:middle line:84% And these stacking faults, which are in the substrate. 00:52:13.250 --> 00:52:16.000 align:middle line:84% Let's say we have a way of introducing stacking faults. 00:52:16.000 --> 00:52:17.950 align:middle line:90% They are found to grow. 00:52:17.950 --> 00:52:20.412 align:middle line:84% They actually get longer under the region 00:52:20.412 --> 00:52:21.370 align:middle line:90% where the oxidation is. 00:52:21.370 --> 00:52:24.580 align:middle line:84% So these are all indirect means that people 00:52:24.580 --> 00:52:26.530 align:middle line:84% use to understand what's going on in the point 00:52:26.530 --> 00:52:29.300 align:middle line:90% defects in this model. 00:52:29.300 --> 00:52:31.030 align:middle line:90% So let's go on to page 25. 00:52:31.030 --> 00:52:33.010 align:middle line:84% I mentioned about the stacking fault. 00:52:33.010 --> 00:52:35.320 align:middle line:84% I don't want to go through detailed crystallography, 00:52:35.320 --> 00:52:37.480 align:middle line:84% but stacking faults occur when you 00:52:37.480 --> 00:52:39.200 align:middle line:90% have an extra layer of atoms. 00:52:39.200 --> 00:52:42.335 align:middle line:84% So if I look here on the lower left, 00:52:42.335 --> 00:52:44.570 align:middle line:84% it has a certain stacking sequence. 00:52:44.570 --> 00:52:47.060 align:middle line:84% I insert this extra layer of atoms. 00:52:47.060 --> 00:52:50.090 align:middle line:84% It has a certain layer and a certain length in the crystal 00:52:50.090 --> 00:52:52.020 align:middle line:90% right at this point. 00:52:52.020 --> 00:52:54.110 align:middle line:84% So it's not a complete layer that goes all the way 00:52:54.110 --> 00:52:56.700 align:middle line:84% throughout the crystal, but it doesn't-- so it doesn't belong 00:52:56.700 --> 00:52:57.200 align:middle line:90% there. 00:52:57.200 --> 00:52:59.870 align:middle line:84% But I can grow this fault. I can make it 00:52:59.870 --> 00:53:03.020 align:middle line:84% longer within the crystal by adding extra silicon 00:53:03.020 --> 00:53:06.480 align:middle line:84% interstitial atoms, people believe, on either edge of it. 00:53:06.480 --> 00:53:10.160 align:middle line:84% So stacking faults, people use their growth rate 00:53:10.160 --> 00:53:12.230 align:middle line:84% as a measure of the injection rate 00:53:12.230 --> 00:53:15.960 align:middle line:84% or the injection of silicon interstitials 00:53:15.960 --> 00:53:18.300 align:middle line:90% if they grow or shrink. 00:53:18.300 --> 00:53:20.790 align:middle line:84% On slide 26 here's some actual photographs 00:53:20.790 --> 00:53:22.350 align:middle line:90% or optical micrographs. 00:53:22.350 --> 00:53:23.612 align:middle line:90% The surfaces were etched. 00:53:23.612 --> 00:53:25.320 align:middle line:84% Remember, we talked about defect etching. 00:53:25.320 --> 00:53:27.330 align:middle line:84% You can put it in a solution of acid 00:53:27.330 --> 00:53:32.520 align:middle line:84% that will preferentially etch where the crystal is imperfect. 00:53:32.520 --> 00:53:35.160 align:middle line:84% And in fact, these are stacking faults. 00:53:35.160 --> 00:53:36.330 align:middle line:90% There are pictures of them. 00:53:36.330 --> 00:53:38.250 align:middle line:90% They have a certain size here. 00:53:38.250 --> 00:53:39.960 align:middle line:84% These are all-- scale bar up here 00:53:39.960 --> 00:53:42.300 align:middle line:90% is 10 microns is shown up here. 00:53:42.300 --> 00:53:47.170 align:middle line:84% After 10, 20, 30, 40, 50 all the way down to 60 minutes, 00:53:47.170 --> 00:53:48.990 align:middle line:84% these are all at the same magnification. 00:53:48.990 --> 00:53:51.060 align:middle line:84% These stacking faults, these crystal defects 00:53:51.060 --> 00:53:54.330 align:middle line:84% have actually grown quite a bit at 1,200 degrees. 00:53:54.330 --> 00:53:58.050 align:middle line:84% This is during oxidation, during high temperature oxidation 00:53:58.050 --> 00:53:58.740 align:middle line:90% at the surface. 00:53:58.740 --> 00:54:00.870 align:middle line:84% These were down under in the bulk, 00:54:00.870 --> 00:54:02.910 align:middle line:84% so people believe that that oxidation 00:54:02.910 --> 00:54:06.285 align:middle line:84% was injecting interstitials and causing these defects to grow. 00:54:06.285 --> 00:54:09.920 align:middle line:90% 00:54:09.920 --> 00:54:12.670 align:middle line:84% In fact, we go on to slide 27, this 00:54:12.670 --> 00:54:17.080 align:middle line:84% is some data from the literature on the stacking fault length 00:54:17.080 --> 00:54:20.080 align:middle line:84% here on the y-axis as a function of oxidation 00:54:20.080 --> 00:54:23.020 align:middle line:90% time at different temperatures. 00:54:23.020 --> 00:54:25.220 align:middle line:90% Here 1,200 is up at the top. 00:54:25.220 --> 00:54:27.160 align:middle line:90% This is 1,050. 00:54:27.160 --> 00:54:30.550 align:middle line:84% And so you can see the stacking faults increases, 00:54:30.550 --> 00:54:33.130 align:middle line:84% this length increases with time of oxidation, 00:54:33.130 --> 00:54:36.290 align:middle line:84% and the growth rate is faster for higher temperatures. 00:54:36.290 --> 00:54:39.580 align:middle line:84% So if we go here, 1,050, 1,100, 1,150, 00:54:39.580 --> 00:54:41.380 align:middle line:90% the growth rate is faster. 00:54:41.380 --> 00:54:43.850 align:middle line:84% Of course, the oxidation rate is going up as well. 00:54:43.850 --> 00:54:47.200 align:middle line:84% So this gave people maybe a hint that the faster you 00:54:47.200 --> 00:54:52.120 align:middle line:84% oxidize, the more interstitials that you're injecting. 00:54:52.120 --> 00:54:55.210 align:middle line:84% The oxidation induced stacking fault growth rate 00:54:55.210 --> 00:54:59.020 align:middle line:84% is smaller for lower partial pressures of oxygen. So again, 00:54:59.020 --> 00:55:00.580 align:middle line:90% slower oxidation rates. 00:55:00.580 --> 00:55:03.790 align:middle line:84% This is an idea that the rate of oxidation 00:55:03.790 --> 00:55:07.270 align:middle line:84% determines the rate of injection of interstitials into the bulk. 00:55:07.270 --> 00:55:09.860 align:middle line:90% 00:55:09.860 --> 00:55:11.240 align:middle line:90% So let's go back. 00:55:11.240 --> 00:55:14.000 align:middle line:90% Let's go on to slide 28. 00:55:14.000 --> 00:55:16.645 align:middle line:84% This is the diagram I just went through. 00:55:16.645 --> 00:55:18.020 align:middle line:84% These are the different processes 00:55:18.020 --> 00:55:22.310 align:middle line:84% we talked about that contribute to this oxidation enhanced 00:55:22.310 --> 00:55:23.850 align:middle line:90% diffusion. 00:55:23.850 --> 00:55:28.340 align:middle line:84% So it's this generation at the interface where I'm oxidizing 00:55:28.340 --> 00:55:31.790 align:middle line:84% and the recombination, the balance of G and R, generation 00:55:31.790 --> 00:55:34.370 align:middle line:84% and recombination, determine what 00:55:34.370 --> 00:55:37.430 align:middle line:84% your net flux of interstitials is into the bulk 00:55:37.430 --> 00:55:39.500 align:middle line:90% away from the surface. 00:55:39.500 --> 00:55:41.120 align:middle line:84% It's this generation rate G that's 00:55:41.120 --> 00:55:43.980 align:middle line:84% directly proportional to the oxidation rate. 00:55:43.980 --> 00:55:45.950 align:middle line:84% So if I oxidize faster at a given temperature 00:55:45.950 --> 00:55:47.780 align:middle line:84% by upping the partial pressure, I 00:55:47.780 --> 00:55:52.340 align:middle line:84% will generate a larger rate of injection of interstitials 00:55:52.340 --> 00:55:55.980 align:middle line:84% and I'll get more boron diffusion in the Berry layer. 00:55:55.980 --> 00:55:59.363 align:middle line:84% And by the way, this Berry layer can be 2, 3, 4 microns away. 00:55:59.363 --> 00:56:00.530 align:middle line:90% It doesn't have to be close. 00:56:00.530 --> 00:56:03.570 align:middle line:84% The interstitials diffuse very quickly and very rapidly, 00:56:03.570 --> 00:56:05.000 align:middle line:90% so this is the problem. 00:56:05.000 --> 00:56:08.150 align:middle line:84% Oxidation can have effects at a distance. 00:56:08.150 --> 00:56:10.980 align:middle line:84% A micron or so away can greatly impact 00:56:10.980 --> 00:56:14.623 align:middle line:90% what's going on in the chip. 00:56:14.623 --> 00:56:16.290 align:middle line:84% So they diffuse away from the interface, 00:56:16.290 --> 00:56:18.930 align:middle line:84% and they enhance the boron diffusion rate. 00:56:18.930 --> 00:56:21.500 align:middle line:90% 00:56:21.500 --> 00:56:25.550 align:middle line:84% Here's an actual simulation of oxidation enhanced diffusion. 00:56:25.550 --> 00:56:28.850 align:middle line:84% What happened was the surface started out 00:56:28.850 --> 00:56:32.730 align:middle line:84% by being implanted uniformly with a certain dose of boron. 00:56:32.730 --> 00:56:38.060 align:middle line:84% So I have a very thin layer of boron up near the surface, 00:56:38.060 --> 00:56:40.280 align:middle line:84% and then we grow a paddocks and we grow 00:56:40.280 --> 00:56:41.820 align:middle line:90% a nitride layer on the left. 00:56:41.820 --> 00:56:45.950 align:middle line:84% So on the left there is going to be no oxidation taking place. 00:56:45.950 --> 00:56:49.590 align:middle line:84% We put it in the furnace at 850 for an hour, 00:56:49.590 --> 00:56:51.872 align:middle line:84% and you see an oxide has grown on the right, 00:56:51.872 --> 00:56:53.330 align:middle line:84% not on the left, which makes sense. 00:56:53.330 --> 00:56:54.247 align:middle line:90% But look at the boron. 00:56:54.247 --> 00:56:57.380 align:middle line:84% The junction depth for the boron is much deeper here. 00:56:57.380 --> 00:56:59.790 align:middle line:84% You can get the junction depth, say, from, let's say, 00:56:59.790 --> 00:57:05.480 align:middle line:84% the region between the green contour and the blue. 00:57:05.480 --> 00:57:07.850 align:middle line:84% The junction depth is a lot deeper underneath the place 00:57:07.850 --> 00:57:08.960 align:middle line:90% where the oxide grew. 00:57:08.960 --> 00:57:10.820 align:middle line:84% So this tells us away that we are 00:57:10.820 --> 00:57:16.010 align:middle line:84% getting oxidation enhanced diffusion of that boron. 00:57:16.010 --> 00:57:20.120 align:middle line:84% This is something that's been simulated. 00:57:20.120 --> 00:57:22.550 align:middle line:90% Doping effects on oxidation. 00:57:22.550 --> 00:57:26.120 align:middle line:84% So besides the oxidation injecting interstitials, 00:57:26.120 --> 00:57:29.240 align:middle line:84% people also noticed that at low temperatures, 00:57:29.240 --> 00:57:31.520 align:middle line:84% depending on the wafer doping, they 00:57:31.520 --> 00:57:33.330 align:middle line:84% would get a different oxidation rate. 00:57:33.330 --> 00:57:35.690 align:middle line:90% So this is kind of interesting. 00:57:35.690 --> 00:57:38.070 align:middle line:84% If you look at A, A is a very light-- 00:57:38.070 --> 00:57:41.070 align:middle line:84% so I'm plotting oxide thickness versus time. 00:57:41.070 --> 00:57:43.500 align:middle line:84% A is lightly doped silicon, kind of the 15th. 00:57:43.500 --> 00:57:48.305 align:middle line:84% F is 10 to the 20 or 3 times the 20. 00:57:48.305 --> 00:57:52.123 align:middle line:84% Much more heavily doped with n type, with phosphorus. 00:57:52.123 --> 00:57:53.540 align:middle line:84% So especially at low temperature-- 00:57:53.540 --> 00:57:55.940 align:middle line:84% at high temperature, not so much of a dispersion, 00:57:55.940 --> 00:57:58.782 align:middle line:84% but at low temperature, a large dispersion. 00:57:58.782 --> 00:58:00.740 align:middle line:84% So this needs to be accounted for in our models 00:58:00.740 --> 00:58:03.290 align:middle line:84% because depending on you have lots of different doping 00:58:03.290 --> 00:58:04.970 align:middle line:84% concentrations on the surface of a wafer 00:58:04.970 --> 00:58:07.280 align:middle line:84% when you're making a chip, you'll get different oxide 00:58:07.280 --> 00:58:08.300 align:middle line:90% thicknesses. 00:58:08.300 --> 00:58:10.190 align:middle line:84% You need to take this into account. 00:58:10.190 --> 00:58:14.650 align:middle line:90% 00:58:14.650 --> 00:58:17.510 align:middle line:84% People, in fact, not only found the thickness to be different. 00:58:17.510 --> 00:58:21.380 align:middle line:84% They extracted the B and the B over A parameter. 00:58:21.380 --> 00:58:23.510 align:middle line:84% So as a function of boron doping-- 00:58:23.510 --> 00:58:26.470 align:middle line:84% so this is the oxidation rate, or these rate 00:58:26.470 --> 00:58:29.530 align:middle line:84% constants B over A and B as a function 00:58:29.530 --> 00:58:33.237 align:middle line:90% of boron concentration. 00:58:33.237 --> 00:58:33.820 align:middle line:90% Oh, I'm sorry. 00:58:33.820 --> 00:58:37.120 align:middle line:84% It's actually phosphorus doping in this reference. 00:58:37.120 --> 00:58:39.760 align:middle line:84% The B over A parameter is the one impacted. 00:58:39.760 --> 00:58:44.140 align:middle line:84% So above a certain concentration of dopants, here above about 10 00:58:44.140 --> 00:58:46.360 align:middle line:90% of the 19, B over A takes off. 00:58:46.360 --> 00:58:49.450 align:middle line:90% So it's going up quite rapidly. 00:58:49.450 --> 00:58:52.695 align:middle line:84% The diffusion through the oxide doesn't seem to be affected. 00:58:52.695 --> 00:58:54.070 align:middle line:84% You can kind of imagine that I've 00:58:54.070 --> 00:58:55.480 align:middle line:90% got doping at the interface. 00:58:55.480 --> 00:59:00.550 align:middle line:84% You expect there's some effect of that doping on the reaction 00:59:00.550 --> 00:59:03.150 align:middle line:90% rate. 00:59:03.150 --> 00:59:06.660 align:middle line:84% And in fact, if we to slide 32, people 00:59:06.660 --> 00:59:08.550 align:middle line:90% had an idea of what it might be. 00:59:08.550 --> 00:59:11.280 align:middle line:84% Maybe it's a chemical effect of the dopant, perhaps, 00:59:11.280 --> 00:59:13.650 align:middle line:84% but people liked this model better. 00:59:13.650 --> 00:59:15.990 align:middle line:84% People liked the idea that the vacancies 00:59:15.990 --> 00:59:18.210 align:middle line:84% available in the near surface region 00:59:18.210 --> 00:59:20.890 align:middle line:84% are important to the oxidation process. 00:59:20.890 --> 00:59:24.420 align:middle line:84% So anything that caused vacancy concentration to go up 00:59:24.420 --> 00:59:27.940 align:middle line:84% might cause the oxidation rate to also go up. 00:59:27.940 --> 00:59:30.120 align:middle line:84% But what do we know about highly doped regions 00:59:30.120 --> 00:59:31.650 align:middle line:90% from your homework that you did? 00:59:31.650 --> 00:59:34.260 align:middle line:84% When you increase the doping in a certain region, 00:59:34.260 --> 00:59:36.420 align:middle line:84% the total number of vacancies go up, right? 00:59:36.420 --> 00:59:39.267 align:middle line:84% Because you move the Fermi level and you create charge vacancies 00:59:39.267 --> 00:59:41.350 align:middle line:84% and you count up all the vacancies and they go up. 00:59:41.350 --> 00:59:43.830 align:middle line:84% So remember these equations you used on your homework 00:59:43.830 --> 00:59:45.630 align:middle line:90% that you're handing in today. 00:59:45.630 --> 00:59:48.090 align:middle line:90% If I'm intrinsic, I'm here. 00:59:48.090 --> 00:59:49.590 align:middle line:90% My Fermi level is at mid-gap. 00:59:49.590 --> 00:59:51.900 align:middle line:84% I'm dominated by neutral vacancies. 00:59:51.900 --> 00:59:53.700 align:middle line:84% They have a certain concentration. 00:59:53.700 --> 00:59:55.350 align:middle line:90% 10 to the 13 or whatever. 00:59:55.350 --> 00:59:57.300 align:middle line:84% As I move the Fermi level up so I 00:59:57.300 --> 00:59:58.980 align:middle line:84% make it more and more heavily doped, 00:59:58.980 --> 01:00:01.560 align:middle line:84% I create a lot of v minus and v double minus. 01:00:01.560 --> 01:00:04.060 align:middle line:84% So I could have a hundred times the number 01:00:04.060 --> 01:00:06.160 align:middle line:90% of vacancies at the surface. 01:00:06.160 --> 01:00:08.750 align:middle line:84% Oxidation needs vacancies to take place. 01:00:08.750 --> 01:00:10.670 align:middle line:90% That's the hypothesis. 01:00:10.670 --> 01:00:13.750 align:middle line:84% So people explain the enhancement 01:00:13.750 --> 01:00:17.860 align:middle line:84% of the high doped oxidation case due to extra vacancies that 01:00:17.860 --> 01:00:20.680 align:middle line:90% are around at the surface. 01:00:20.680 --> 01:00:23.920 align:middle line:84% That's one potential explanation. 01:00:23.920 --> 01:00:27.370 align:middle line:84% And in fact, I just reminded, you did this in your homework. 01:00:27.370 --> 01:00:29.680 align:middle line:84% You did some calculations on slide 33. 01:00:29.680 --> 01:00:32.757 align:middle line:84% The vacancy concentration that's been calculated for you here 01:00:32.757 --> 01:00:34.340 align:middle line:84% as a function of doping concentration, 01:00:34.340 --> 01:00:35.650 align:middle line:90% look where it takes off. 01:00:35.650 --> 01:00:38.252 align:middle line:84% Of course, somewhere right around 10 to the 19 01:00:38.252 --> 01:00:39.460 align:middle line:90% it really starts to take off. 01:00:39.460 --> 01:00:42.940 align:middle line:84% That's exactly where Hogh found that the B over A coefficient 01:00:42.940 --> 01:00:47.350 align:middle line:84% took off is right where the vacancy concentration is going 01:00:47.350 --> 01:00:49.840 align:middle line:90% up as a function of doping. 01:00:49.840 --> 01:00:53.140 align:middle line:90% 01:00:53.140 --> 01:00:57.430 align:middle line:84% So let's just say on slide 34 that these vacancies are 01:00:57.430 --> 01:01:01.660 align:middle line:84% available to provide sites, these extra vacancies, 01:01:01.660 --> 01:01:03.940 align:middle line:90% for the oxidation reaction. 01:01:03.940 --> 01:01:07.690 align:middle line:84% Then we can imagine writing B over A equal to something, 01:01:07.690 --> 01:01:10.390 align:middle line:84% some number R1 where R1 has to do 01:01:10.390 --> 01:01:13.150 align:middle line:84% with all the mechanisms other than vacancy driven processes. 01:01:13.150 --> 01:01:16.252 align:middle line:84% Obviously oxidation is not only just controlled by vacancies. 01:01:16.252 --> 01:01:17.710 align:middle line:84% It's controlled by a lot of things. 01:01:17.710 --> 01:01:21.610 align:middle line:84% But there must be some term added in this k times 01:01:21.610 --> 01:01:24.080 align:middle line:90% V total, CV total. 01:01:24.080 --> 01:01:26.480 align:middle line:84% So the term on the right represents 01:01:26.480 --> 01:01:29.380 align:middle line:90% the vacancy driven process. 01:01:29.380 --> 01:01:31.250 align:middle line:84% It's, I believe, to be directly proportional 01:01:31.250 --> 01:01:33.458 align:middle line:84% to the concentration of the total number of vacancies 01:01:33.458 --> 01:01:35.090 align:middle line:90% in all charge states. 01:01:35.090 --> 01:01:38.720 align:middle line:84% So doing that, you can then rewrite the B over A parameter 01:01:38.720 --> 01:01:41.360 align:middle line:84% to be B over A in intrinsic material, 01:01:41.360 --> 01:01:44.780 align:middle line:84% lightly doped, when you're not extrinsic, 01:01:44.780 --> 01:01:49.020 align:middle line:84% times this 1 plus some exponential type of factor-- 01:01:49.020 --> 01:01:50.900 align:middle line:84% so this is empirically determined here 01:01:50.900 --> 01:01:53.060 align:middle line:84% as a function on temperature-- times something 01:01:53.060 --> 01:01:56.030 align:middle line:90% that goes like CV over CVI. 01:01:56.030 --> 01:01:58.100 align:middle line:84% So the concentration of vacancies 01:01:58.100 --> 01:02:00.320 align:middle line:84% in this heavily doped material divided 01:02:00.320 --> 01:02:03.680 align:middle line:84% by that intrinsic material minus one. 01:02:03.680 --> 01:02:06.110 align:middle line:84% So if I'm intrinsic, this goes to zero 01:02:06.110 --> 01:02:07.190 align:middle line:90% and this term goes away. 01:02:07.190 --> 01:02:10.310 align:middle line:84% As this increases, this ratio goes up above one. 01:02:10.310 --> 01:02:13.790 align:middle line:90% This term starts to kick in. 01:02:13.790 --> 01:02:16.520 align:middle line:84% Interestingly, though, when we think about this now, 01:02:16.520 --> 01:02:18.577 align:middle line:84% this model for the vacancy dependence 01:02:18.577 --> 01:02:20.660 align:middle line:84% is only going to depend on the electrically active 01:02:20.660 --> 01:02:21.592 align:middle line:90% concentration. 01:02:21.592 --> 01:02:23.800 align:middle line:84% Because remember, we said we move the Fermi level up. 01:02:23.800 --> 01:02:26.850 align:middle line:84% So if we add phosphorus to the wafer, 01:02:26.850 --> 01:02:30.480 align:middle line:84% it's only the amount of phosphorus 01:02:30.480 --> 01:02:32.370 align:middle line:84% that contributes electrons that's going 01:02:32.370 --> 01:02:35.220 align:middle line:90% to move the Fermi level around. 01:02:35.220 --> 01:02:37.920 align:middle line:84% So people had an idea on how to test this. 01:02:37.920 --> 01:02:40.410 align:middle line:90% What if I compensate the wafer? 01:02:40.410 --> 01:02:42.540 align:middle line:90% OK. 01:02:42.540 --> 01:02:44.877 align:middle line:90% Well, actually let me go on. 01:02:44.877 --> 01:02:45.960 align:middle line:90% I'll talk about that next. 01:02:45.960 --> 01:02:50.400 align:middle line:84% Let me just show slide 35 just to show, 01:02:50.400 --> 01:02:53.370 align:middle line:90% compare boron and phosphorus. 01:02:53.370 --> 01:02:56.040 align:middle line:84% We know from our calculations, actually, 01:02:56.040 --> 01:03:00.180 align:middle line:84% that n type regions, just because of the point defect 01:03:00.180 --> 01:03:04.260 align:middle line:84% statistics, n type regions tend to have higher charge vacancy 01:03:04.260 --> 01:03:07.500 align:middle line:84% concentrations at a given doping than p type, than boron. 01:03:07.500 --> 01:03:10.590 align:middle line:84% And in fact, what we see is that the B over A parameter 01:03:10.590 --> 01:03:14.760 align:middle line:84% is more impacted in n type silicon, very heavily doped, 01:03:14.760 --> 01:03:15.780 align:middle line:90% than it is in p type. 01:03:15.780 --> 01:03:17.520 align:middle line:84% That's the difference between the right 01:03:17.520 --> 01:03:21.840 align:middle line:84% and the left hand side on page 35. 01:03:21.840 --> 01:03:23.442 align:middle line:90% That's some hint. 01:03:23.442 --> 01:03:24.900 align:middle line:84% What I started to talk about, which 01:03:24.900 --> 01:03:28.020 align:middle line:84% I think to me is a little more convincing, that this is really 01:03:28.020 --> 01:03:29.910 align:middle line:90% maybe is due to vacancies. 01:03:29.910 --> 01:03:32.250 align:middle line:84% If you look at the oxide thickness 01:03:32.250 --> 01:03:34.500 align:middle line:84% as a function of the electrically active 01:03:34.500 --> 01:03:36.760 align:middle line:90% phosphorus-- 01:03:36.760 --> 01:03:39.150 align:middle line:84% and this solid line is for uncompensated. 01:03:39.150 --> 01:03:40.140 align:middle line:90% What is uncompensated? 01:03:40.140 --> 01:03:42.360 align:middle line:84% I just add phosphorus at certain level. 01:03:42.360 --> 01:03:46.200 align:middle line:84% Here I add it to 10 of the 19, 5 to the 19, 10 of the 20, 01:03:46.200 --> 01:03:48.520 align:middle line:90% and 5, 10 of the 20. 01:03:48.520 --> 01:03:49.950 align:middle line:84% And I don't add any other dopants. 01:03:49.950 --> 01:03:51.900 align:middle line:90% No boron. 01:03:51.900 --> 01:03:55.350 align:middle line:84% As the phosphorus goes up, the vacancy concentration 01:03:55.350 --> 01:03:57.760 align:middle line:84% goes up and so does the thickness. 01:03:57.760 --> 01:04:00.150 align:middle line:84% Now the dashed line is material that's been compensated. 01:04:00.150 --> 01:04:01.120 align:middle line:90% What does that mean? 01:04:01.120 --> 01:04:05.100 align:middle line:84% Well, what that means is when I add 10 of the 20 phosphorus, 01:04:05.100 --> 01:04:07.137 align:middle line:90% I add 10 of the 20 boron. 01:04:07.137 --> 01:04:08.970 align:middle line:84% There's still 10 of the 20 phosphorous there 01:04:08.970 --> 01:04:11.130 align:middle line:84% and 10 of the 20 boron, but electrically I 01:04:11.130 --> 01:04:14.040 align:middle line:84% haven't added any extra electrons or holes 01:04:14.040 --> 01:04:17.970 align:middle line:84% because they compensate each other in terms of dopants. 01:04:17.970 --> 01:04:20.490 align:middle line:84% So the electrically active concentration 01:04:20.490 --> 01:04:24.180 align:middle line:84% has not gone up when I add as much boron as I add phosphorus. 01:04:24.180 --> 01:04:26.310 align:middle line:84% You add equal amount of p and n type doping. 01:04:26.310 --> 01:04:28.920 align:middle line:90% I still end up with fairly-- 01:04:28.920 --> 01:04:32.650 align:middle line:84% I end up with material that's compensated. 01:04:32.650 --> 01:04:34.020 align:middle line:90% I haven't moved the Fermi level. 01:04:34.020 --> 01:04:36.240 align:middle line:84% I haven't created a lot of excess charge defects. 01:04:36.240 --> 01:04:39.750 align:middle line:84% And in fact, he found that there wasn't that much increase 01:04:39.750 --> 01:04:42.810 align:middle line:84% in the oxidation rate when he added a lot of phosphorus 01:04:42.810 --> 01:04:45.743 align:middle line:90% but compensated with boron. 01:04:45.743 --> 01:04:46.660 align:middle line:90% Equal amount of boron. 01:04:46.660 --> 01:04:48.390 align:middle line:84% So if it were a chemical effect, just the amount 01:04:48.390 --> 01:04:50.890 align:middle line:84% of phosphorus atoms-- let's say a phosphorus atom chemically 01:04:50.890 --> 01:04:52.960 align:middle line:90% causes oxide to grow faster. 01:04:52.960 --> 01:04:55.890 align:middle line:84% It wouldn't matter whether you added an equal amount of boron. 01:04:55.890 --> 01:04:57.960 align:middle line:84% But if you add an equal amount of boron, 01:04:57.960 --> 01:04:59.500 align:middle line:84% you don't change the Fermi level. 01:04:59.500 --> 01:05:01.860 align:middle line:84% So this is an argument that says to me that it really 01:05:01.860 --> 01:05:04.420 align:middle line:84% is the electrical concentration that's changing. 01:05:04.420 --> 01:05:06.150 align:middle line:84% It's the movement of the Fermi level. 01:05:06.150 --> 01:05:07.830 align:middle line:84% It doesn't prove it's a vacancy effect, 01:05:07.830 --> 01:05:12.370 align:middle line:84% but it gives a little bit more weight. 01:05:12.370 --> 01:05:16.120 align:middle line:84% Here on slide 37, it's just a simulation 01:05:16.120 --> 01:05:18.080 align:middle line:90% of this type of effect. 01:05:18.080 --> 01:05:21.190 align:middle line:84% What we have on the left hand side 01:05:21.190 --> 01:05:27.010 align:middle line:84% is local oxidation where it's taking place here 01:05:27.010 --> 01:05:31.360 align:middle line:84% at 800 degrees for 30 minutes and you're 01:05:31.360 --> 01:05:34.300 align:middle line:84% seeing get a little thicker oxide 01:05:34.300 --> 01:05:37.480 align:middle line:84% or get a five times thicker oxide growing 01:05:37.480 --> 01:05:39.200 align:middle line:90% in this heavily doped region. 01:05:39.200 --> 01:05:41.110 align:middle line:84% So this region over here on the right 01:05:41.110 --> 01:05:44.530 align:middle line:90% is heavily phosphorus doped. 01:05:44.530 --> 01:05:47.920 align:middle line:84% So it was I implanted prior to oxidation. 01:05:47.920 --> 01:05:49.510 align:middle line:84% So you need to take this into account. 01:05:49.510 --> 01:05:51.040 align:middle line:84% If you're making a device and you 01:05:51.040 --> 01:05:53.182 align:middle line:84% implant a region for a source drain 01:05:53.182 --> 01:05:54.640 align:middle line:84% and then you go to oxidize it, it's 01:05:54.640 --> 01:05:56.770 align:middle line:84% going to have a different oxidation rate compared 01:05:56.770 --> 01:05:59.410 align:middle line:84% to those parts of the chip that are being oxidized that don't 01:05:59.410 --> 01:06:01.070 align:middle line:90% have high doping in them. 01:06:01.070 --> 01:06:03.535 align:middle line:84% So in SUPREM you can easily take that into account. 01:06:03.535 --> 01:06:06.590 align:middle line:90% 01:06:06.590 --> 01:06:10.570 align:middle line:84% And the last thing I want to cover is here on slide 38, 01:06:10.570 --> 01:06:12.040 align:middle line:90% we talked about oxidation. 01:06:12.040 --> 01:06:16.190 align:middle line:84% It has all these point defect effects. 01:06:16.190 --> 01:06:20.110 align:middle line:84% We also talked a couple last time or the time 01:06:20.110 --> 01:06:24.550 align:middle line:84% before about interface charges that come along with oxidation. 01:06:24.550 --> 01:06:28.413 align:middle line:84% And this model on slide 38, is while we're 01:06:28.413 --> 01:06:29.830 align:middle line:84% talking about atomic level things, 01:06:29.830 --> 01:06:34.000 align:middle line:84% this is an atomic level cartoon or picture of what 01:06:34.000 --> 01:06:37.030 align:middle line:84% is actually causing the fixed charge and the density 01:06:37.030 --> 01:06:38.890 align:middle line:90% of interface states. 01:06:38.890 --> 01:06:41.830 align:middle line:84% And what it supposedly is here, in the near-- 01:06:41.830 --> 01:06:45.580 align:middle line:84% here's a miamorphous SiO2 up here and my crystal silicon. 01:06:45.580 --> 01:06:47.620 align:middle line:84% In the near interfacial region, there 01:06:47.620 --> 01:06:50.957 align:middle line:84% are believed to be a few interstitial silicon atoms. 01:06:50.957 --> 01:06:53.290 align:middle line:84% Remember, some interstitials get injected into the bulk. 01:06:53.290 --> 01:06:54.670 align:middle line:90% Some go up, out. 01:06:54.670 --> 01:06:57.160 align:middle line:84% There's a few that get stuck near the interface. 01:06:57.160 --> 01:06:58.930 align:middle line:84% This little extra silicon interstitial 01:06:58.930 --> 01:07:00.700 align:middle line:90% with a positive charge on it. 01:07:00.700 --> 01:07:04.150 align:middle line:84% It's not bonded to oxygen atoms, so has 01:07:04.150 --> 01:07:05.480 align:middle line:90% an excess positive charge. 01:07:05.480 --> 01:07:08.570 align:middle line:84% That unbonded silicon interstitial, 01:07:08.570 --> 01:07:11.090 align:middle line:84% the extra dangling silicon interstitials 01:07:11.090 --> 01:07:13.100 align:middle line:84% that are there that are positively charged 01:07:13.100 --> 01:07:14.855 align:middle line:84% are believed to be responsible for QM. 01:07:14.855 --> 01:07:18.650 align:middle line:90% 01:07:18.650 --> 01:07:20.650 align:middle line:84% Remember, fixed charge doesn't change with bias. 01:07:20.650 --> 01:07:22.700 align:middle line:84% The density of interface states that change 01:07:22.700 --> 01:07:27.110 align:middle line:84% is believed to be associated with dangling silicone 01:07:27.110 --> 01:07:30.470 align:middle line:84% bonds right at the interface, and we 01:07:30.470 --> 01:07:33.830 align:middle line:84% can passivate these because they're not in the oxide. 01:07:33.830 --> 01:07:37.040 align:middle line:84% They're right at the silicon dioxide surface. 01:07:37.040 --> 01:07:40.100 align:middle line:84% We can passivate these by adding hydrogen. 01:07:40.100 --> 01:07:43.100 align:middle line:84% So when you build a CMOS or you build a MOSFET, 01:07:43.100 --> 01:07:47.060 align:middle line:84% the last step is 450 degrees in hydrogen ambient. 01:07:47.060 --> 01:07:48.800 align:middle line:90% The hydrogen diffuses in. 01:07:48.800 --> 01:07:51.050 align:middle line:84% It satisfies these dangling bonds. 01:07:51.050 --> 01:07:54.860 align:middle line:84% It bonds here and it covers up that charge. 01:07:54.860 --> 01:07:58.310 align:middle line:84% The hydrogen can't do anything about this silicon. 01:07:58.310 --> 01:07:59.990 align:middle line:90% It's unoxidized. 01:07:59.990 --> 01:08:01.980 align:middle line:90% It needs oxygen atoms. 01:08:01.980 --> 01:08:04.920 align:middle line:84% So that's why when you do a forming gas anneal here, 01:08:04.920 --> 01:08:07.220 align:middle line:84% the last step, you can reduce the density 01:08:07.220 --> 01:08:08.075 align:middle line:90% of interface states. 01:08:08.075 --> 01:08:10.820 align:middle line:90% 01:08:10.820 --> 01:08:12.620 align:middle line:84% The fixed charge is a little different. 01:08:12.620 --> 01:08:16.100 align:middle line:84% To reduce it you don't anneal on hydrogen. 01:08:16.100 --> 01:08:18.260 align:middle line:90% This is so-called Deal triangle. 01:08:18.260 --> 01:08:20.390 align:middle line:84% What you do is you anneal the oxide 01:08:20.390 --> 01:08:25.109 align:middle line:84% after oxidizing it in an inert ambient at high temperature. 01:08:25.109 --> 01:08:27.560 align:middle line:84% So the idea would be that you try 01:08:27.560 --> 01:08:29.870 align:middle line:84% to get some of those silicon interstitials 01:08:29.870 --> 01:08:31.129 align:middle line:90% that were stuck there. 01:08:31.129 --> 01:08:32.359 align:middle line:90% They're unoxidized. 01:08:32.359 --> 01:08:35.750 align:middle line:84% You give them time to diffuse away from that interface 01:08:35.750 --> 01:08:38.623 align:middle line:84% so you can get rid of that fixed charge. 01:08:38.623 --> 01:08:40.040 align:middle line:84% That's exactly what you will find. 01:08:40.040 --> 01:08:43.460 align:middle line:84% If you run an oxidation program in the furnace, 01:08:43.460 --> 01:08:46.850 align:middle line:84% what people typically do, they grow an oxide, 800 or 900 01:08:46.850 --> 01:08:48.140 align:middle line:90% for an hour. 01:08:48.140 --> 01:08:50.540 align:middle line:84% The last step of the program is to let 01:08:50.540 --> 01:08:53.180 align:middle line:84% the wafer sit at 800 or 900 in the tube 01:08:53.180 --> 01:08:54.410 align:middle line:90% with no oxygen flowing. 01:08:54.410 --> 01:08:58.609 align:middle line:84% Just argon or nitrogen. And that last half hour anneal 01:08:58.609 --> 01:08:59.970 align:middle line:90% is the QF anneal. 01:08:59.970 --> 01:09:03.319 align:middle line:84% So if you grew in dry oxygen at 800, an oxide, 01:09:03.319 --> 01:09:06.770 align:middle line:84% and you didn't anneal it, you might have on 111 silicon 01:09:06.770 --> 01:09:10.580 align:middle line:84% about t 6 times to the 11 for your QF per square centimeter. 01:09:10.580 --> 01:09:13.827 align:middle line:84% If you anneal it, you can get it down at 800. 01:09:13.827 --> 01:09:15.410 align:middle line:84% If you anneal it in nitrogen or argon, 01:09:15.410 --> 01:09:17.899 align:middle line:84% you can get it down to about one times 10 to the 11 01:09:17.899 --> 01:09:19.279 align:middle line:90% on 111 silicon. 01:09:19.279 --> 01:09:20.100 align:middle line:90% 100. 01:09:20.100 --> 01:09:22.100 align:middle line:84% All the numbers will be about three times lower. 01:09:22.100 --> 01:09:24.949 align:middle line:84% If you anneal it at a higher temperature, 01:09:24.949 --> 01:09:26.220 align:middle line:90% it'll also come down. 01:09:26.220 --> 01:09:29.510 align:middle line:84% So the whole idea is to anneal that in inert ambient 01:09:29.510 --> 01:09:35.080 align:middle line:84% at a relatively high temperature right after your oxidation. 01:09:35.080 --> 01:09:35.580 align:middle line:90% OK. 01:09:35.580 --> 01:09:36.689 align:middle line:90% This is the last slide. 01:09:36.689 --> 01:09:42.651 align:middle line:84% Just to summarize chapter 6 and what we've talked about today. 01:09:42.651 --> 01:09:46.229 align:middle line:84% We know have to control the oxide thickness to atomic level 01:09:46.229 --> 01:09:48.720 align:middle line:90% for a thin gate dielectrics. 01:09:48.720 --> 01:09:50.550 align:middle line:84% There's a basic Deal-Grove mechanism 01:09:50.550 --> 01:09:53.790 align:middle line:84% which involves diffusion through the oxide, reaction 01:09:53.790 --> 01:09:55.530 align:middle line:90% at the interface. 01:09:55.530 --> 01:09:57.510 align:middle line:84% This doesn't work perfectly, but there's 01:09:57.510 --> 01:09:59.640 align:middle line:84% been a lot of situations that people can 01:09:59.640 --> 01:10:02.250 align:middle line:90% correct the Deal-Grove model. 01:10:02.250 --> 01:10:04.650 align:middle line:84% This non-planar oxidation is impacted 01:10:04.650 --> 01:10:06.720 align:middle line:84% by the orientation of the surface, two 01:10:06.720 --> 01:10:09.720 align:middle line:84% dimensional diffusion through a shaped oxide, 01:10:09.720 --> 01:10:11.250 align:middle line:90% and the stress effects. 01:10:11.250 --> 01:10:14.640 align:middle line:84% Stress retards or slows down the oxidation 01:10:14.640 --> 01:10:17.640 align:middle line:84% by slowing down the interface reaction k sub a 01:10:17.640 --> 01:10:20.980 align:middle line:84% and slowing down the diffusion through the oxide. 01:10:20.980 --> 01:10:23.820 align:middle line:84% And these silicon interstitials are injected into the bulk 01:10:23.820 --> 01:10:25.560 align:middle line:84% during oxidation, and we're going 01:10:25.560 --> 01:10:28.740 align:middle line:84% to come back and revisit them in the next few lectures 01:10:28.740 --> 01:10:31.800 align:middle line:84% or over the next several weeks because they're 01:10:31.800 --> 01:10:33.720 align:middle line:84% going to dramatically impact diffusion 01:10:33.720 --> 01:10:36.340 align:middle line:90% of boron and other dopants. 01:10:36.340 --> 01:10:38.080 align:middle line:90% So that's about what I have. 01:10:38.080 --> 01:10:41.770 align:middle line:84% I believe, for today, your homework is due. 01:10:41.770 --> 01:10:44.560 align:middle line:84% You can bring that up right now and put it 01:10:44.560 --> 01:10:45.700 align:middle line:90% in this orange folder. 01:10:45.700 --> 01:10:47.617 align:middle line:84% If you didn't pick up your homework last time, 01:10:47.617 --> 01:10:49.510 align:middle line:84% it's in the back with the handouts. 01:10:49.510 --> 01:10:50.890 align:middle line:90% I'm going to be out next week. 01:10:50.890 --> 01:10:51.950 align:middle line:90% Your TA, Maggie-- 01:10:51.950 --> 01:10:53.950 align:middle line:84% Maggie, you want to raise your hand in the back? 01:10:53.950 --> 01:10:57.070 align:middle line:84% Is going to give lectures as usual. 01:10:57.070 --> 01:10:59.860 align:middle line:84% So we'll start diffusion on Tuesday. 01:10:59.860 --> 01:11:03.570 align:middle line:84% And on Tuesday, then your new homework goes out. 01:11:03.570 --> 01:11:07.000 align:middle line:90%