WEBVTT 00:00:00.000 --> 00:00:00.958 align:middle line:90% JUDY HOYT: Get started. 00:00:00.958 --> 00:00:02.990 align:middle line:90% Just a couple of reminders. 00:00:02.990 --> 00:00:05.870 align:middle line:84% Homework number 4 is due today up front. 00:00:05.870 --> 00:00:08.045 align:middle line:84% You can bring that up at the end of the hour. 00:00:08.045 --> 00:00:10.910 align:middle line:90% 00:00:10.910 --> 00:00:12.860 align:middle line:90% Homework number 3 is going back. 00:00:12.860 --> 00:00:15.530 align:middle line:84% I've got it in the back there in that orange folder. 00:00:15.530 --> 00:00:20.065 align:middle line:84% So you can pick up your homework before you leave today. 00:00:20.065 --> 00:00:21.440 align:middle line:84% And there are a couple people who 00:00:21.440 --> 00:00:22.940 align:middle line:84% haven't picked up prior homework, 00:00:22.940 --> 00:00:26.330 align:middle line:84% so they're in the back of that orange folder. 00:00:26.330 --> 00:00:28.250 align:middle line:90% We've got one handout for today. 00:00:28.250 --> 00:00:30.683 align:middle line:90% That's handout 27. 00:00:30.683 --> 00:00:33.350 align:middle line:84% Today we're going to be talking about the SUPREM-IV four process 00:00:33.350 --> 00:00:34.280 align:middle line:90% simulator. 00:00:34.280 --> 00:00:38.270 align:middle line:84% This is where we are, November 2 here on election day. 00:00:38.270 --> 00:00:40.940 align:middle line:84% And you're handing in homework number 4. 00:00:40.940 --> 00:00:42.625 align:middle line:84% The solutions to homework number 3, 00:00:42.625 --> 00:00:44.750 align:middle line:84% you're getting your graded homework number 3s back. 00:00:44.750 --> 00:00:46.790 align:middle line:84% The solutions will be on the web later today, 00:00:46.790 --> 00:00:49.310 align:middle line:84% and I'll bring hard copies next time. 00:00:49.310 --> 00:00:54.810 align:middle line:84% They didn't get printed or xeroxed in time. 00:00:54.810 --> 00:00:58.560 align:middle line:84% OK, so let's go on to handout number 00:00:58.560 --> 00:01:01.560 align:middle line:84% 27, which is today's handout for today's lecture. 00:01:01.560 --> 00:01:06.740 align:middle line:90% 00:01:06.740 --> 00:01:12.560 align:middle line:84% This handout or this lecture is on so-called quote unquote 00:01:12.560 --> 00:01:14.780 align:middle line:84% "introduction to the "SUPREM-IV process simulator." 00:01:14.780 --> 00:01:17.270 align:middle line:84% Now, you've already really been introduced to it. 00:01:17.270 --> 00:01:19.760 align:middle line:84% Fortunately, you've been using it 00:01:19.760 --> 00:01:22.430 align:middle line:84% for the last couple of homeworks in your homework sets. 00:01:22.430 --> 00:01:28.790 align:middle line:84% And your TA introduced it in a practical sense 00:01:28.790 --> 00:01:31.940 align:middle line:84% a number of weeks ago when she gave a lecture when you 00:01:31.940 --> 00:01:34.310 align:middle line:90% had your first SUPREM homework. 00:01:34.310 --> 00:01:37.670 align:middle line:84% But I want to talk about it in a little more detail 00:01:37.670 --> 00:01:40.100 align:middle line:84% and maybe give you some examples of what 00:01:40.100 --> 00:01:42.065 align:middle line:90% it can do, to give you an idea. 00:01:42.065 --> 00:01:44.600 align:middle line:90% 00:01:44.600 --> 00:01:48.170 align:middle line:84% So in the very first lecture, we discussed an example 00:01:48.170 --> 00:01:50.030 align:middle line:90% of a CMOS process flow. 00:01:50.030 --> 00:01:52.490 align:middle line:84% And we drew a series of cartoons, 00:01:52.490 --> 00:01:57.740 align:middle line:84% PowerPoint artist's conception of what things would look like. 00:01:57.740 --> 00:02:01.130 align:middle line:84% At this point in the class, at this point in the class, 00:02:01.130 --> 00:02:03.260 align:middle line:84% we've covered some of the fundamentals 00:02:03.260 --> 00:02:07.320 align:middle line:84% and the modeling of a lot of the key processes in that CMOS flow. 00:02:07.320 --> 00:02:11.039 align:middle line:84% We've talked about thermal oxidation, diffusion of dopants. 00:02:11.039 --> 00:02:14.370 align:middle line:84% We just had four lectures on ion implantation 00:02:14.370 --> 00:02:16.080 align:middle line:90% and transit-enhanced diffusion. 00:02:16.080 --> 00:02:17.940 align:middle line:84% And we've talked a little bit about how 00:02:17.940 --> 00:02:19.710 align:middle line:90% these processes fit together. 00:02:19.710 --> 00:02:23.670 align:middle line:84% So today I want to review some of these concepts 00:02:23.670 --> 00:02:28.290 align:middle line:84% and some CMOS process modules and CMOS flows now 00:02:28.290 --> 00:02:31.140 align:middle line:84% in the context of what we know about models 00:02:31.140 --> 00:02:34.390 align:middle line:84% and in the context of using the SUPREM-IV simulator. 00:02:34.390 --> 00:02:39.540 align:middle line:84% This is just an example of an n MOSFET 00:02:39.540 --> 00:02:42.570 align:middle line:84% and a p MOSFET fabricated on a wafer. 00:02:42.570 --> 00:02:47.757 align:middle line:84% This was the cartoon that we showed earlier in this course, 00:02:47.757 --> 00:02:48.840 align:middle line:90% in the very first lecture. 00:02:48.840 --> 00:02:50.340 align:middle line:84% And what we can do today is we don't 00:02:50.340 --> 00:02:51.600 align:middle line:90% need to use cartoons anymore. 00:02:51.600 --> 00:02:53.700 align:middle line:84% As you know, you can use the SUPREM-IV simulator 00:02:53.700 --> 00:02:57.600 align:middle line:84% to generate this type of information reasonably 00:02:57.600 --> 00:02:59.650 align:middle line:90% accurately. 00:02:59.650 --> 00:03:01.890 align:middle line:84% So let's go on to slide number 2. 00:03:01.890 --> 00:03:04.593 align:middle line:84% And I have a series of examples I wanted to go through. 00:03:04.593 --> 00:03:06.510 align:middle line:84% And some of these things, since you've already 00:03:06.510 --> 00:03:08.370 align:middle line:84% run the simulator, you may be familiar with. 00:03:08.370 --> 00:03:11.790 align:middle line:84% But we'll go into some things you may not have seen. 00:03:11.790 --> 00:03:13.477 align:middle line:90% The first three are 1D. 00:03:13.477 --> 00:03:16.170 align:middle line:90% 00:03:16.170 --> 00:03:19.558 align:middle line:84% SUPREM-IV, it's a two-dimensional simulator. 00:03:19.558 --> 00:03:21.600 align:middle line:84% But it can be run in a one-dimensional mode where 00:03:21.600 --> 00:03:25.540 align:middle line:84% you're primarily interested in what happens in depth. 00:03:25.540 --> 00:03:27.770 align:middle line:84% But in general, you run it in 2D mode. 00:03:27.770 --> 00:03:30.730 align:middle line:84% But to run things quickly, to understand basic physics 00:03:30.730 --> 00:03:34.630 align:middle line:84% concepts, it's often easier to run the simulator in 1D mode 00:03:34.630 --> 00:03:37.780 align:middle line:84% and then later on run a two-dimensional example. 00:03:37.780 --> 00:03:39.700 align:middle line:84% Obviously, when you do it in 2D mode 00:03:39.700 --> 00:03:43.400 align:middle line:84% you have a lot more calculations to do and it's a lot slower. 00:03:43.400 --> 00:03:45.460 align:middle line:84% So first, I want to talk a little bit about boron 00:03:45.460 --> 00:03:48.400 align:middle line:84% segregation at a gate oxide silicon interface 00:03:48.400 --> 00:03:49.960 align:middle line:90% and compare a couple issues. 00:03:49.960 --> 00:03:52.240 align:middle line:84% And I'm using this as an example for you 00:03:52.240 --> 00:03:53.950 align:middle line:84% to understand a little bit about gridding 00:03:53.950 --> 00:03:56.200 align:middle line:84% issues that can take place when you use the simulator. 00:03:56.200 --> 00:04:03.100 align:middle line:84% So a very practical issue of keeping an eye on how fine 00:04:03.100 --> 00:04:05.040 align:middle line:90% your grid is. 00:04:05.040 --> 00:04:08.640 align:middle line:84% The second example, we'll talk about comparing some arsenic 00:04:08.640 --> 00:04:13.050 align:middle line:84% and phosphorus implant profiles, the different types of profiles, 00:04:13.050 --> 00:04:14.700 align:middle line:90% to actual SIMS data. 00:04:14.700 --> 00:04:17.190 align:middle line:84% We'll see which ones do a better job 00:04:17.190 --> 00:04:21.060 align:middle line:84% at actually simulating or reproducing 00:04:21.060 --> 00:04:22.170 align:middle line:90% what the data looks like. 00:04:22.170 --> 00:04:25.080 align:middle line:84% And then we'll talk about diffusion models, 00:04:25.080 --> 00:04:28.520 align:middle line:84% introduce the different types of diffusion models 00:04:28.520 --> 00:04:29.950 align:middle line:90% that are available. 00:04:29.950 --> 00:04:32.250 align:middle line:90% And again, these three are 1D. 00:04:32.250 --> 00:04:34.080 align:middle line:84% The last example-- I hope we have time-- 00:04:34.080 --> 00:04:38.280 align:middle line:84% is 2D example of an older technology. 00:04:38.280 --> 00:04:41.010 align:middle line:84% It's a 200-nanometer gate length n 00:04:41.010 --> 00:04:48.000 align:middle line:84% MOSFET just to go through some of the processing of that. 00:04:48.000 --> 00:04:49.700 align:middle line:84% So let's go on to slide number 3. 00:04:49.700 --> 00:04:52.100 align:middle line:84% This is an example, and if you've already 00:04:52.100 --> 00:04:54.525 align:middle line:84% done the homeworks, you've probably seen this. 00:04:54.525 --> 00:04:58.220 align:middle line:84% This is an example of what's called the mesh or the grid. 00:04:58.220 --> 00:05:04.250 align:middle line:84% And SUPREM-IV uses a triangular grid, triangular mesh. 00:05:04.250 --> 00:05:07.470 align:middle line:84% So it's a series of nonoverlapping triangular 00:05:07.470 --> 00:05:07.970 align:middle line:90% elements. 00:05:07.970 --> 00:05:09.053 align:middle line:90% You see each one of these. 00:05:09.053 --> 00:05:14.830 align:middle line:84% If you look at it carefully, it can be drawn as a triangle. 00:05:14.830 --> 00:05:18.850 align:middle line:84% The mesh is very important because what 00:05:18.850 --> 00:05:21.220 align:middle line:84% we do in the mesh or the grid is that we-- 00:05:21.220 --> 00:05:24.010 align:middle line:84% to frame four tracks, the values of all parameters 00:05:24.010 --> 00:05:28.210 align:middle line:84% numerically along a moving adjustable grid. 00:05:28.210 --> 00:05:31.180 align:middle line:84% So at these nodes, at these grid points 00:05:31.180 --> 00:05:34.840 align:middle line:84% is tracking the values of all-important parameters, 00:05:34.840 --> 00:05:38.620 align:middle line:84% the interface-- what the material is, the interface 00:05:38.620 --> 00:05:39.820 align:middle line:90% between two materials. 00:05:39.820 --> 00:05:43.420 align:middle line:84% It tracks the dopant concentration value. 00:05:43.420 --> 00:05:47.985 align:middle line:84% A lot of different things are tracked at each grid point. 00:05:47.985 --> 00:05:49.360 align:middle line:84% If you look at this grid, so this 00:05:49.360 --> 00:05:52.100 align:middle line:84% is a grid that belongs to a device structure. 00:05:52.100 --> 00:05:54.920 align:middle line:84% Let me see if I have in the next slide. 00:05:54.920 --> 00:05:55.930 align:middle line:90% Oh, that's a blow up. 00:05:55.930 --> 00:05:58.900 align:middle line:84% Yeah, let's just go a little bit of a blow up on slide number 4. 00:05:58.900 --> 00:06:01.270 align:middle line:84% You can see the grid a little bit better. 00:06:01.270 --> 00:06:03.310 align:middle line:84% This belongs to a device structure 00:06:03.310 --> 00:06:07.160 align:middle line:84% where over here on the left-hand side there's actually a gate. 00:06:07.160 --> 00:06:08.303 align:middle line:90% There's a polysilicon gate. 00:06:08.303 --> 00:06:09.220 align:middle line:90% There's a space there. 00:06:09.220 --> 00:06:09.790 align:middle line:90% I don't know if you can see. 00:06:09.790 --> 00:06:11.020 align:middle line:90% Here's a gate oxide. 00:06:11.020 --> 00:06:12.760 align:middle line:90% And there's some metal up here. 00:06:12.760 --> 00:06:15.060 align:middle line:90% And this is the silicon. 00:06:15.060 --> 00:06:17.240 align:middle line:84% This is an ion implanted and diffused junction. 00:06:17.240 --> 00:06:20.683 align:middle line:84% And you say, why does the grid look like it does? 00:06:20.683 --> 00:06:21.850 align:middle line:90% Why would it look like this? 00:06:21.850 --> 00:06:22.350 align:middle line:90% You see? 00:06:22.350 --> 00:06:26.960 align:middle line:84% In this region here, near the channel and near the surface, 00:06:26.960 --> 00:06:28.720 align:middle line:90% you see a very fine grid. 00:06:28.720 --> 00:06:31.270 align:middle line:90% The mesh is very fine. 00:06:31.270 --> 00:06:33.620 align:middle line:84% The spacing between the grid points is very small. 00:06:33.620 --> 00:06:35.860 align:middle line:84% Whereas, you get down here in the silicon substrate, 00:06:35.860 --> 00:06:37.630 align:middle line:90% and the mesh is very coarse. 00:06:37.630 --> 00:06:40.720 align:middle line:84% The spacing between points on which you do the solution 00:06:40.720 --> 00:06:42.430 align:middle line:90% is very large. 00:06:42.430 --> 00:06:43.870 align:middle line:90% So why would that be? 00:06:43.870 --> 00:06:47.000 align:middle line:84% Does anybody have any idea why up here I 00:06:47.000 --> 00:06:50.338 align:middle line:84% would have a fine grid, and down here I have a coarse grid? 00:06:50.338 --> 00:06:52.630 align:middle line:84% AUDIENCE: Because the term-- does this change anything? 00:06:52.630 --> 00:06:55.570 align:middle line:84% JUDY HOYT: Right, because up here at the junction 00:06:55.570 --> 00:06:57.820 align:middle line:84% you probably have some doping profiles that are 00:06:57.820 --> 00:06:59.440 align:middle line:90% changing very rapidly in depth. 00:06:59.440 --> 00:07:01.900 align:middle line:84% So you need a lot of grid points to maintain 00:07:01.900 --> 00:07:07.150 align:middle line:84% good accuracy and good fidelity in reproducing those profiles. 00:07:07.150 --> 00:07:08.980 align:middle line:84% Down here, there's really not much action. 00:07:08.980 --> 00:07:11.240 align:middle line:84% Probably just constant doping of boron. 00:07:11.240 --> 00:07:14.320 align:middle line:84% So there's no need to have very fine mesh. 00:07:14.320 --> 00:07:16.640 align:middle line:84% Even today's computers are only so powerful. 00:07:16.640 --> 00:07:19.030 align:middle line:84% So SUPREM tries to do intelligent gridding 00:07:19.030 --> 00:07:20.910 align:middle line:84% and not waste and put in a lot of grid points 00:07:20.910 --> 00:07:21.910 align:middle line:90% where you don't need it. 00:07:21.910 --> 00:07:23.663 align:middle line:84% If you use a very fine mesh everywhere, 00:07:23.663 --> 00:07:25.330 align:middle line:84% you'd say, well, that's the best, right? 00:07:25.330 --> 00:07:26.500 align:middle line:90% Make it fine everywhere. 00:07:26.500 --> 00:07:29.650 align:middle line:84% The problem with that is it would take too long 00:07:29.650 --> 00:07:33.970 align:middle line:84% to the solution, and it's really a waste of time. 00:07:33.970 --> 00:07:35.452 align:middle line:84% So just going back to slide 3, this 00:07:35.452 --> 00:07:36.660 align:middle line:90% is exactly what we just said. 00:07:36.660 --> 00:07:39.390 align:middle line:84% The grid needs to be fine near the active device region, 00:07:39.390 --> 00:07:42.750 align:middle line:84% near the interface, or where any quantity of interest 00:07:42.750 --> 00:07:44.700 align:middle line:84% is changing very rapidly, depending 00:07:44.700 --> 00:07:47.640 align:middle line:84% on what you're interested in knowing about, doping profiles, 00:07:47.640 --> 00:07:49.230 align:middle line:90% interstitials, vacancies. 00:07:49.230 --> 00:07:54.720 align:middle line:84% Wherever they change rapidly, you need a fine grid. 00:07:54.720 --> 00:07:58.740 align:middle line:84% For diffusion, very often SUPREM does a finite difference 00:07:58.740 --> 00:08:00.390 align:middle line:90% solution for oxidation. 00:08:00.390 --> 00:08:03.670 align:middle line:84% It uses a finite element type of technique, again, 00:08:03.670 --> 00:08:04.770 align:middle line:90% both on this grid. 00:08:04.770 --> 00:08:07.440 align:middle line:90% 00:08:07.440 --> 00:08:09.690 align:middle line:84% OK, let's go on to slide number 5. 00:08:09.690 --> 00:08:12.960 align:middle line:84% Actually, I probably should have shown this first. 00:08:12.960 --> 00:08:15.570 align:middle line:84% Of that mesh that we just showed, 00:08:15.570 --> 00:08:17.790 align:middle line:84% this is the actual device structure. 00:08:17.790 --> 00:08:19.920 align:middle line:90% It's half of an n MOSFET. 00:08:19.920 --> 00:08:21.660 align:middle line:84% And now it becomes pretty obvious 00:08:21.660 --> 00:08:23.577 align:middle line:84% what you were looking at in terms of the grid. 00:08:23.577 --> 00:08:25.090 align:middle line:90% The yellow here is the silicon. 00:08:25.090 --> 00:08:28.340 align:middle line:90% This is the polysilicon gate. 00:08:28.340 --> 00:08:30.030 align:middle line:90% The blue is the gate oxide. 00:08:30.030 --> 00:08:33.409 align:middle line:84% This is an oxide a spacer, a sidewall spacer 00:08:33.409 --> 00:08:34.669 align:middle line:90% right next to it. 00:08:34.669 --> 00:08:37.370 align:middle line:84% There's a metal contact here to the silicon. 00:08:37.370 --> 00:08:41.690 align:middle line:84% And this is the sort of LOCOS region, or the isolation region, 00:08:41.690 --> 00:08:42.919 align:middle line:90% in between devices. 00:08:42.919 --> 00:08:45.080 align:middle line:84% This is either the source or the drain, 00:08:45.080 --> 00:08:49.440 align:middle line:84% and you can see contours here corresponding to arsenic doping. 00:08:49.440 --> 00:08:51.830 align:middle line:90% So this is the actual structure. 00:08:51.830 --> 00:08:55.280 align:middle line:84% Very often, in SUPREM, things are symmetrical. 00:08:55.280 --> 00:08:58.520 align:middle line:84% So another way, besides doing intelligent gridding, 00:08:58.520 --> 00:09:00.110 align:middle line:84% if you have a symmetrical device, 00:09:00.110 --> 00:09:02.840 align:middle line:84% you can often only simulate half of it and then just 00:09:02.840 --> 00:09:06.290 align:middle line:84% project that about the center line 00:09:06.290 --> 00:09:10.400 align:middle line:84% here because the MOSFETs are very often symmetrical 00:09:10.400 --> 00:09:11.900 align:middle line:90% in the way they're fabricated. 00:09:11.900 --> 00:09:15.780 align:middle line:84% So sometimes that helps you save time in the simulation. 00:09:15.780 --> 00:09:17.060 align:middle line:90% So again, once you start-- 00:09:17.060 --> 00:09:19.868 align:middle line:84% we haven't had do very many two-dimensional simulations. 00:09:19.868 --> 00:09:21.410 align:middle line:84% But once you start doing them, you'll 00:09:21.410 --> 00:09:25.520 align:middle line:84% see how much time it can take to simulate all these diffusion, 00:09:25.520 --> 00:09:27.960 align:middle line:84% especially if you're doing the most sophisticated models. 00:09:27.960 --> 00:09:31.605 align:middle line:84% So you have to take advantage of intelligent gridding. 00:09:31.605 --> 00:09:34.490 align:middle line:90% 00:09:34.490 --> 00:09:37.100 align:middle line:84% I want to start on slide number 6 00:09:37.100 --> 00:09:42.380 align:middle line:84% the first example now, which is a simple one-dimensional example 00:09:42.380 --> 00:09:44.480 align:middle line:84% that looks at the behavior of boron 00:09:44.480 --> 00:09:46.610 align:middle line:84% and how it behaves and segregates 00:09:46.610 --> 00:09:49.040 align:middle line:90% at an oxide silicon interface. 00:09:49.040 --> 00:09:52.200 align:middle line:84% And this is a SUPREM-IV input file. 00:09:52.200 --> 00:09:54.200 align:middle line:84% As you know, all of you have done your homework, 00:09:54.200 --> 00:09:57.410 align:middle line:84% SUPREM-IV takes text files as input. 00:09:57.410 --> 00:09:59.150 align:middle line:84% You can write the text in any text editor 00:09:59.150 --> 00:10:00.650 align:middle line:84% you want as long as it doesn't leave 00:10:00.650 --> 00:10:03.890 align:middle line:84% a lot of spurious characters that confuse the program. 00:10:03.890 --> 00:10:06.980 align:middle line:84% And the program has a parser that reads in the text 00:10:06.980 --> 00:10:09.650 align:middle line:84% and interprets the text in terms of the commands 00:10:09.650 --> 00:10:12.850 align:middle line:90% that SUPREM-IV knows about. 00:10:12.850 --> 00:10:14.600 align:middle line:84% Whenever you see a dollar sign, of course, 00:10:14.600 --> 00:10:15.650 align:middle line:90% that's just a comment. 00:10:15.650 --> 00:10:16.980 align:middle line:90% So SUPREM ignores that. 00:10:16.980 --> 00:10:19.730 align:middle line:84% So this is only for your own notation. 00:10:19.730 --> 00:10:23.660 align:middle line:84% So the first command here is called the mesh command, 00:10:23.660 --> 00:10:25.970 align:middle line:84% and it tells you that a parameter called 00:10:25.970 --> 00:10:28.580 align:middle line:90% grid dot fac equals 0.04. 00:10:28.580 --> 00:10:30.090 align:middle line:84% And then the grid factor has to do 00:10:30.090 --> 00:10:31.340 align:middle line:90% with the fineness of the grid. 00:10:31.340 --> 00:10:35.260 align:middle line:84% The smaller the number, the finer the mesh. 00:10:35.260 --> 00:10:39.480 align:middle line:84% And so we're defining in this one statement 00:10:39.480 --> 00:10:44.940 align:middle line:84% what we want to make the default mesh size in a uniform sense. 00:10:44.940 --> 00:10:46.890 align:middle line:90% AUDIENCE: [INAUDIBLE]. 00:10:46.890 --> 00:10:49.920 align:middle line:84% JUDY HOYT: 0.04, it's not in units. 00:10:49.920 --> 00:10:51.730 align:middle line:90% I think it's a multiplier. 00:10:51.730 --> 00:10:57.300 align:middle line:84% So I think it can be anything from 1 to 10 to maybe 0.01. 00:10:57.300 --> 00:10:59.550 align:middle line:84% If you look up in the SUPREM manual, it will tell you. 00:10:59.550 --> 00:11:02.130 align:middle line:84% I don't remember the exact definition of grid dot fact, 00:11:02.130 --> 00:11:06.300 align:middle line:84% but when grid dot fac equals 1, it gives you 00:11:06.300 --> 00:11:07.680 align:middle line:90% a certain grid spacing. 00:11:07.680 --> 00:11:10.080 align:middle line:84% My experience has been that, for most modern devices, 00:11:10.080 --> 00:11:13.140 align:middle line:84% a grid dot fac of 1, it's a pretty large spacing, 00:11:13.140 --> 00:11:15.943 align:middle line:84% larger than we would probably want to have. 00:11:15.943 --> 00:11:18.360 align:middle line:84% I think I have some examples where we varied this in here, 00:11:18.360 --> 00:11:20.460 align:middle line:84% so we can look for the different grid dot facs. 00:11:20.460 --> 00:11:23.040 align:middle line:90% And you'll see how it comes out. 00:11:23.040 --> 00:11:24.712 align:middle line:90% The next thing is pretty simple. 00:11:24.712 --> 00:11:26.670 align:middle line:84% It says an initialize statement, this says just 00:11:26.670 --> 00:11:27.930 align:middle line:90% create the silicon substrate. 00:11:27.930 --> 00:11:28.810 align:middle line:90% That's all it does. 00:11:28.810 --> 00:11:32.800 align:middle line:84% It tells you the orientation in the z-axis is 100. 00:11:32.800 --> 00:11:35.250 align:middle line:84% So that's the orientation of the wafer. 00:11:35.250 --> 00:11:39.690 align:middle line:84% And it's boron doped, and it's 1E17 per cubic centimeter. 00:11:39.690 --> 00:11:43.320 align:middle line:84% And again, if you need to know things like boron equals 1E17, 00:11:43.320 --> 00:11:45.990 align:middle line:84% you wonder what the units are, if you go into the SUPREM-IV 00:11:45.990 --> 00:11:48.480 align:middle line:84% manual, it'll tell you what the default 00:11:48.480 --> 00:11:51.430 align:middle line:90% units are for any variable. 00:11:51.430 --> 00:11:52.680 align:middle line:90% This is an implant statement. 00:11:52.680 --> 00:11:56.400 align:middle line:84% You just did a homework on ion implantation, 00:11:56.400 --> 00:11:58.300 align:middle line:90% where we're implanting boron. 00:11:58.300 --> 00:11:59.370 align:middle line:90% So this is the species. 00:11:59.370 --> 00:12:02.050 align:middle line:84% This is the dose, the number of atoms per square centimeter, 00:12:02.050 --> 00:12:03.270 align:middle line:90% the energy. 00:12:03.270 --> 00:12:07.540 align:middle line:84% And you're specifying a tilt angle of seven degrees. 00:12:07.540 --> 00:12:09.150 align:middle line:84% Depending on the dopants, there will 00:12:09.150 --> 00:12:11.800 align:middle line:90% be different default models. 00:12:11.800 --> 00:12:14.885 align:middle line:84% A lot of the dopants default to the Pearson-IV model. 00:12:14.885 --> 00:12:17.010 align:middle line:84% In other words, if you don't say what model to use, 00:12:17.010 --> 00:12:19.590 align:middle line:84% it automatically uses default Pearson-IV. 00:12:19.590 --> 00:12:22.650 align:middle line:84% I think arsenic defaults to dual Pearson. 00:12:22.650 --> 00:12:24.330 align:middle line:84% You have to look up in the SUPREM manual 00:12:24.330 --> 00:12:27.640 align:middle line:84% and figure out what the default is if you don't specify. 00:12:27.640 --> 00:12:29.950 align:middle line:84% So remember, just because you didn't say a model, 00:12:29.950 --> 00:12:32.150 align:middle line:84% SUPREM has to assume one of the models. 00:12:32.150 --> 00:12:33.500 align:middle line:90% And so that's the default model. 00:12:33.500 --> 00:12:35.417 align:middle line:84% You need to become familiar with what that is. 00:12:35.417 --> 00:12:37.960 align:middle line:90% 00:12:37.960 --> 00:12:39.983 align:middle line:90% This statement here is deposing. 00:12:39.983 --> 00:12:41.150 align:middle line:90% It's a deposition statement. 00:12:41.150 --> 00:12:42.580 align:middle line:90% So it's not a growth. 00:12:42.580 --> 00:12:44.050 align:middle line:90% It's not thermal oxidation. 00:12:44.050 --> 00:12:48.220 align:middle line:84% It's literally depositing oxide just by plunking it, (WHIT), 00:12:48.220 --> 00:12:48.940 align:middle line:90% on the surface. 00:12:48.940 --> 00:12:50.410 align:middle line:84% You're not consuming any silicon. 00:12:50.410 --> 00:12:53.170 align:middle line:84% And we haven't talked about this type of process before, 00:12:53.170 --> 00:12:55.060 align:middle line:84% but it's called chemical vapor deposition. 00:12:55.060 --> 00:12:57.195 align:middle line:84% The next several lectures we'll talk about that. 00:12:57.195 --> 00:12:58.570 align:middle line:84% So this is not thermal oxidation. 00:12:58.570 --> 00:13:00.550 align:middle line:84% We're just plunking an oxide down 00:13:00.550 --> 00:13:03.243 align:middle line:90% at a certain temperature, 600. 00:13:03.243 --> 00:13:04.660 align:middle line:84% And we're telling it the thickness 00:13:04.660 --> 00:13:08.150 align:middle line:90% in microns, 0.005 microns. 00:13:08.150 --> 00:13:08.950 align:middle line:90% So it's very thin. 00:13:08.950 --> 00:13:10.330 align:middle line:90% It's 50 angstroms. 00:13:10.330 --> 00:13:14.460 align:middle line:84% DY, if you look in the SUPREM manual 00:13:14.460 --> 00:13:16.800 align:middle line:84% under the deposition statement for oxide, 00:13:16.800 --> 00:13:21.330 align:middle line:84% DY tells the grid spacing in the y direction 00:13:21.330 --> 00:13:23.110 align:middle line:84% that you want to use in the oxide. 00:13:23.110 --> 00:13:26.610 align:middle line:84% So DY is the grid spacing specific to the oxide. 00:13:26.610 --> 00:13:27.990 align:middle line:90% And look how small it is. 00:13:27.990 --> 00:13:30.450 align:middle line:90% It's 0.0001. 00:13:30.450 --> 00:13:33.030 align:middle line:84% Well, the reason it's so small is because the thickness 00:13:33.030 --> 00:13:34.120 align:middle line:90% is so small. 00:13:34.120 --> 00:13:36.000 align:middle line:84% You don't want one grid point in the oxide. 00:13:36.000 --> 00:13:37.140 align:middle line:90% You'd like to have several. 00:13:37.140 --> 00:13:39.720 align:middle line:84% So we're telling it what the gridding to use just 00:13:39.720 --> 00:13:42.090 align:middle line:90% in the oxide. 00:13:42.090 --> 00:13:44.010 align:middle line:84% And then there's a simple statement 00:13:44.010 --> 00:13:46.170 align:middle line:84% in which you select what variable you want to plot. 00:13:46.170 --> 00:13:48.840 align:middle line:84% In this case, you want to plot the log10 of boron. 00:13:48.840 --> 00:13:50.640 align:middle line:90% You can do some fancy-- 00:13:50.640 --> 00:13:53.980 align:middle line:84% put in titles and all that stuff. 00:13:53.980 --> 00:13:58.450 align:middle line:84% The Save File command saves the file in a format 00:13:58.450 --> 00:14:01.987 align:middle line:84% that SUPREM can then later use, read in and use later on. 00:14:01.987 --> 00:14:04.570 align:middle line:84% It's not necessarily a format that you would find very useful. 00:14:04.570 --> 00:14:06.070 align:middle line:90% But SUPREM uses it. 00:14:06.070 --> 00:14:09.790 align:middle line:84% So Save File into an out dot file is-- 00:14:09.790 --> 00:14:11.680 align:middle line:84% defines which file you want it to write to. 00:14:11.680 --> 00:14:14.860 align:middle line:84% And this is the particular name we chose, B1 dot inp. 00:14:14.860 --> 00:14:17.470 align:middle line:84% You can put it wherever you want. 00:14:17.470 --> 00:14:20.890 align:middle line:84% Here, you're simulating the diffusion. 00:14:20.890 --> 00:14:26.850 align:middle line:84% So this is simulating diffusion for a time of 17 minutes 00:14:26.850 --> 00:14:28.440 align:middle line:90% at a temperature of 850. 00:14:28.440 --> 00:14:30.930 align:middle line:84% And if I specify an inner ambient, 00:14:30.930 --> 00:14:33.870 align:middle line:84% it's just like a diffusion, like an argon or nitrogen. 00:14:33.870 --> 00:14:38.160 align:middle line:84% If I specify dry O2, then you're actually doing an oxidizing. 00:14:38.160 --> 00:14:41.400 align:middle line:84% So oxidation is done, as you know now, 00:14:41.400 --> 00:14:43.920 align:middle line:84% by using the diffusion statement and specifying 00:14:43.920 --> 00:14:47.550 align:middle line:84% what type of [? ambient ?]---- wet O2, dry O2, whichever. 00:14:47.550 --> 00:14:53.160 align:middle line:84% And then, after this amount of diffusion, 850 for 17 minutes, 00:14:53.160 --> 00:14:55.080 align:middle line:84% you're going to plot what the profile looks 00:14:55.080 --> 00:14:56.530 align:middle line:90% like of the boron after that. 00:14:56.530 --> 00:14:59.800 align:middle line:90% So we have an as-implanted plot. 00:14:59.800 --> 00:15:03.480 align:middle line:84% And we have a certain line type and color, here, color number 00:15:03.480 --> 00:15:04.080 align:middle line:90% one. 00:15:04.080 --> 00:15:08.880 align:middle line:84% And we have after diffusion at 17 minutes at 850. 00:15:08.880 --> 00:15:10.080 align:middle line:90% This is a simple file. 00:15:10.080 --> 00:15:12.912 align:middle line:84% It's a type of file that you've been using that your TAs been 00:15:12.912 --> 00:15:13.620 align:middle line:90% creating for you. 00:15:13.620 --> 00:15:14.370 align:middle line:90% And you've been running them. 00:15:14.370 --> 00:15:15.600 align:middle line:84% Now we want you to start to understand 00:15:15.600 --> 00:15:18.150 align:middle line:84% what those files do so you can create them and modify them 00:15:18.150 --> 00:15:19.638 align:middle line:90% yourself. 00:15:19.638 --> 00:15:21.930 align:middle line:84% And remember, you can always go to the SUPREM-IV manual 00:15:21.930 --> 00:15:24.360 align:middle line:84% to look up any of these commands, deposition, 00:15:24.360 --> 00:15:25.920 align:middle line:90% select, Save file. 00:15:25.920 --> 00:15:28.120 align:middle line:84% All those are explained in the manual. 00:15:28.120 --> 00:15:31.950 align:middle line:84% So this is the output of that file on slide number 7. 00:15:31.950 --> 00:15:33.300 align:middle line:90% This is what it looks like. 00:15:33.300 --> 00:15:34.920 align:middle line:90% Of course, I've doctored it up. 00:15:34.920 --> 00:15:36.870 align:middle line:84% I've added a little color and things 00:15:36.870 --> 00:15:40.470 align:middle line:84% to make it look a little better, but it's the basic output. 00:15:40.470 --> 00:15:42.540 align:middle line:84% What that file puts out is a plot. 00:15:42.540 --> 00:15:45.750 align:middle line:84% On the y-axis is the boron concentration. 00:15:45.750 --> 00:15:48.090 align:middle line:90% The x-axis is distance. 00:15:48.090 --> 00:15:50.820 align:middle line:90% And SUPREM assumes that 0-- 00:15:50.820 --> 00:15:54.750 align:middle line:84% it puts 0 at the original silicon interface, or surface, 00:15:54.750 --> 00:15:55.510 align:middle line:90% I should say. 00:15:55.510 --> 00:15:58.770 align:middle line:84% So if you deposit layers on top of that, 00:15:58.770 --> 00:16:02.020 align:middle line:84% they will be sort of on the negative x-axis. 00:16:02.020 --> 00:16:04.500 align:middle line:84% Remember, we deposited 50 angstroms of oxide. 00:16:04.500 --> 00:16:06.330 align:middle line:84% That's what I'm showing in the yellow here. 00:16:06.330 --> 00:16:09.360 align:middle line:84% If you go back one minute and you look at deposition oxide, 00:16:09.360 --> 00:16:12.180 align:middle line:90% thickness is 0.005 microns. 00:16:12.180 --> 00:16:14.770 align:middle line:90% That's 50 angstroms. 00:16:14.770 --> 00:16:16.980 align:middle line:84% So that's what this is, from here 00:16:16.980 --> 00:16:18.970 align:middle line:90% to here in this yellow region. 00:16:18.970 --> 00:16:23.300 align:middle line:84% And there are two profiles that are shown here. 00:16:23.300 --> 00:16:26.580 align:middle line:84% The black one, the dark one is SUPREM 00:16:26.580 --> 00:16:28.290 align:middle line:84% plotted the boron profile immediately 00:16:28.290 --> 00:16:30.190 align:middle line:90% after ion implantation. 00:16:30.190 --> 00:16:32.730 align:middle line:90% So that's the as-implanted. 00:16:32.730 --> 00:16:37.470 align:middle line:84% And the red one is the boron profile after that anneal 00:16:37.470 --> 00:16:40.780 align:middle line:84% that we did, 17 minutes inert at 850. 00:16:40.780 --> 00:16:41.980 align:middle line:90% Of course, it's diffused. 00:16:41.980 --> 00:16:43.510 align:middle line:90% The peak has gone down a little. 00:16:43.510 --> 00:16:44.750 align:middle line:90% It's broadened out. 00:16:44.750 --> 00:16:47.410 align:middle line:84% And there has been a little bit of segregation 00:16:47.410 --> 00:16:51.280 align:middle line:90% at the oxide silicon interface. 00:16:51.280 --> 00:16:54.070 align:middle line:84% The diffusion of boron and oxide is relatively slow. 00:16:54.070 --> 00:16:57.550 align:middle line:84% It's about a factor of 2,000, roughly, a factor of 1,000 00:16:57.550 --> 00:16:59.380 align:middle line:84% slower than diffusion in silicon. 00:16:59.380 --> 00:17:02.020 align:middle line:84% So you see a fair amount of motion of the boron and silicon. 00:17:02.020 --> 00:17:04.119 align:middle line:84% But when it hits that wall, hits that oxide, 00:17:04.119 --> 00:17:05.990 align:middle line:90% it doesn't diffuse very fast. 00:17:05.990 --> 00:17:08.200 align:middle line:84% So you're not going to get-- you see 00:17:08.200 --> 00:17:12.099 align:middle line:84% in the silicon there's very little boron after 17 minutes 00:17:12.099 --> 00:17:15.079 align:middle line:90% has actually diffused in. 00:17:15.079 --> 00:17:17.190 align:middle line:84% If you want to know the segregation coefficient, 00:17:17.190 --> 00:17:18.890 align:middle line:84% you can look it up in the appendix. 00:17:18.890 --> 00:17:20.599 align:middle line:84% They have an appendix in SUPREM-IV 00:17:20.599 --> 00:17:22.579 align:middle line:90% for each of the dopants. 00:17:22.579 --> 00:17:25.250 align:middle line:84% For boron segregation, which is the ratio 00:17:25.250 --> 00:17:27.829 align:middle line:84% of the concentration of boron in the silicon to that 00:17:27.829 --> 00:17:29.540 align:middle line:90% in the oxide at 850. 00:17:29.540 --> 00:17:32.125 align:middle line:84% SUPREM calculates it from this formula. 00:17:32.125 --> 00:17:34.500 align:middle line:84% You can modify either one of these constants, by the way, 00:17:34.500 --> 00:17:35.300 align:middle line:90% if you want to. 00:17:35.300 --> 00:17:40.460 align:middle line:84% And I calculate it at about a factor of 10 to 1, or 0.1. 00:17:40.460 --> 00:17:45.670 align:middle line:84% So that would be this peak value to the value in the oxide. 00:17:45.670 --> 00:17:47.980 align:middle line:84% So that's what that simple file puts out. 00:17:47.980 --> 00:17:52.730 align:middle line:84% Now, let's look at a couple of other special cases. 00:17:52.730 --> 00:17:55.480 align:middle line:84% So we're going to now do something similar. 00:17:55.480 --> 00:17:57.580 align:middle line:84% But instead of depositing the oxide straight 00:17:57.580 --> 00:18:01.630 align:middle line:84% down by deposition, from chemical vapor deposition, 00:18:01.630 --> 00:18:04.810 align:middle line:84% we're going to form the oxide by thermal oxidation. 00:18:04.810 --> 00:18:07.750 align:middle line:84% This is the type of model that we've talked about before. 00:18:07.750 --> 00:18:09.520 align:middle line:84% And we're going to display the output. 00:18:09.520 --> 00:18:15.160 align:middle line:84% Here on slide number 8, I'm showing a new SUPREM input 00:18:15.160 --> 00:18:17.360 align:middle line:90% file, or command file. 00:18:17.360 --> 00:18:19.360 align:middle line:84% And this time, instead of saying deposition, 00:18:19.360 --> 00:18:21.310 align:middle line:84% we're going to say to do a diffusion. 00:18:21.310 --> 00:18:25.030 align:middle line:84% Time equals 5-- that's five minutes-- at temperature of 850 00:18:25.030 --> 00:18:25.780 align:middle line:90% in dry O2. 00:18:25.780 --> 00:18:29.620 align:middle line:84% So it's going to grow five minutes worth of oxide at 850 00:18:29.620 --> 00:18:31.090 align:middle line:90% in dry O2. 00:18:31.090 --> 00:18:33.880 align:middle line:84% And then we're going to plot it using color number 2. 00:18:33.880 --> 00:18:35.950 align:middle line:84% And then we're going to do another five minutes 00:18:35.950 --> 00:18:37.280 align:middle line:90% and plot it again. 00:18:37.280 --> 00:18:38.600 align:middle line:90% So we can do sequential. 00:18:38.600 --> 00:18:40.960 align:middle line:84% So that will be after 10 minutes. 00:18:40.960 --> 00:18:42.340 align:middle line:84% So you can look at it and see how 00:18:42.340 --> 00:18:46.160 align:middle line:84% the oxide grows and see what happens to the boron profile 00:18:46.160 --> 00:18:47.840 align:middle line:90% as the oxide grows. 00:18:47.840 --> 00:18:51.410 align:middle line:84% So we're going to display it at 5, 10, and 17 minutes. 00:18:51.410 --> 00:18:53.750 align:middle line:84% We're going to display the boron profile in the oxide. 00:18:53.750 --> 00:18:57.000 align:middle line:90% 00:18:57.000 --> 00:19:00.350 align:middle line:84% So on slide number 9, this is after five minutes 00:19:00.350 --> 00:19:02.840 align:middle line:90% of now thermal oxidation. 00:19:02.840 --> 00:19:06.170 align:middle line:84% We have an initial oxide that's formed. 00:19:06.170 --> 00:19:10.760 align:middle line:84% And this is what the boron looks like. 00:19:10.760 --> 00:19:12.710 align:middle line:84% And you notice the boron near the surface 00:19:12.710 --> 00:19:15.740 align:middle line:84% has been depleted, again, because of that segregation. 00:19:15.740 --> 00:19:19.100 align:middle line:84% And now you're oxidizing, so you're consuming silicon. 00:19:19.100 --> 00:19:22.460 align:middle line:84% So you're moving into the silicon. 00:19:22.460 --> 00:19:24.800 align:middle line:84% This interface is moving towards the right. 00:19:24.800 --> 00:19:28.310 align:middle line:84% And same segregation as the previous case. 00:19:28.310 --> 00:19:30.770 align:middle line:84% And now we have a little bit of boron 00:19:30.770 --> 00:19:33.290 align:middle line:84% that's been incorporated in the silicon, not 00:19:33.290 --> 00:19:36.350 align:middle line:84% so much by diffusion because, again, but by consumption 00:19:36.350 --> 00:19:38.990 align:middle line:84% because, remember, we initially implant it with boron. 00:19:38.990 --> 00:19:40.970 align:middle line:90% And then you start oxidizing. 00:19:40.970 --> 00:19:45.650 align:middle line:84% If you go back to the command file, 00:19:45.650 --> 00:19:48.540 align:middle line:84% read in the as-implanted structure. 00:19:48.540 --> 00:19:55.010 align:middle line:84% So remember, we did a save file of B1.imp. 00:19:55.010 --> 00:19:56.350 align:middle line:90% How did I get that? 00:19:56.350 --> 00:20:00.180 align:middle line:84% Well, that save file-- so this is a way to just save time. 00:20:00.180 --> 00:20:04.860 align:middle line:84% After we did the as-implanted, we saved the file in something 00:20:04.860 --> 00:20:06.420 align:middle line:90% called B1 dot inp. 00:20:06.420 --> 00:20:10.020 align:middle line:84% And then we're going to call it back later, that same file, 00:20:10.020 --> 00:20:13.680 align:middle line:90% in a subsequent SUPREM run. 00:20:13.680 --> 00:20:17.063 align:middle line:84% So here, I'm loading in a file called B1.imp. 00:20:17.063 --> 00:20:18.480 align:middle line:84% If you didn't want to do that, you 00:20:18.480 --> 00:20:20.350 align:middle line:84% could repeat the simulation of the implant. 00:20:20.350 --> 00:20:22.058 align:middle line:84% It's not going to take very many minutes. 00:20:22.058 --> 00:20:24.390 align:middle line:84% But if you have a complex two-dimensional simulation, 00:20:24.390 --> 00:20:25.890 align:middle line:90% you might as well save it. 00:20:25.890 --> 00:20:26.910 align:middle line:90% It takes an hour to run. 00:20:26.910 --> 00:20:27.410 align:middle line:90% Save it. 00:20:27.410 --> 00:20:29.280 align:middle line:84% And you can always use it again as the input 00:20:29.280 --> 00:20:32.640 align:middle line:84% to another SUPREM simulation, another diffusion or oxidation. 00:20:32.640 --> 00:20:35.450 align:middle line:90% 00:20:35.450 --> 00:20:41.840 align:middle line:84% OK, so we are going to-- in this here back on slide number 8, 00:20:41.840 --> 00:20:45.440 align:middle line:84% we're going to load that in, take whatever oxide 00:20:45.440 --> 00:20:47.540 align:middle line:84% was on there off just to make sure it's not there, 00:20:47.540 --> 00:20:50.960 align:middle line:84% and diffuse it at 850 for five minutes in dry O2. 00:20:50.960 --> 00:20:54.920 align:middle line:84% We're going to grow a couple of different thicknesses of oxide. 00:20:54.920 --> 00:20:57.687 align:middle line:84% So I've consumed this, and that's how a lot of the boron 00:20:57.687 --> 00:20:58.520 align:middle line:90% has gotten in there. 00:20:58.520 --> 00:21:00.470 align:middle line:84% And you see the segregation from this point, 00:21:00.470 --> 00:21:03.600 align:middle line:84% from here to here, that ratio is about a factor of 10. 00:21:03.600 --> 00:21:06.470 align:middle line:84% That's what we expect because we said 00:21:06.470 --> 00:21:09.560 align:middle line:84% k0 for segregation of boron in SUPREM-IV 00:21:09.560 --> 00:21:12.700 align:middle line:90% is about a factor of 10. 00:21:12.700 --> 00:21:14.830 align:middle line:84% Go to the next slide, slide number 10. 00:21:14.830 --> 00:21:16.750 align:middle line:84% After another five minutes, so I have a total 00:21:16.750 --> 00:21:18.820 align:middle line:90% of 10 minutes of oxidation. 00:21:18.820 --> 00:21:22.150 align:middle line:84% Again, you see, now you've grown a little more oxide. 00:21:22.150 --> 00:21:23.320 align:middle line:90% The yellow is thicker. 00:21:23.320 --> 00:21:25.420 align:middle line:84% And you've consumed more of the boron, 00:21:25.420 --> 00:21:27.800 align:middle line:84% and the segregation is causing this effect. 00:21:27.800 --> 00:21:30.370 align:middle line:84% So this is not really diffusion because the diffusion, look 00:21:30.370 --> 00:21:32.530 align:middle line:90% how rapidly the boron drops. 00:21:32.530 --> 00:21:36.460 align:middle line:84% If it were diffusing, this would go straight through the oxide. 00:21:36.460 --> 00:21:38.830 align:middle line:84% But it's not because there is slow boron 00:21:38.830 --> 00:21:41.525 align:middle line:84% diffusion in this assumption in the oxide. 00:21:41.525 --> 00:21:43.150 align:middle line:84% You see the boron profile in the oxide. 00:21:43.150 --> 00:21:47.170 align:middle line:84% And if we finally go to the last, which is after the full 17 00:21:47.170 --> 00:21:53.170 align:middle line:84% minutes, we have grown now 50 angstroms of gate oxide 00:21:53.170 --> 00:21:56.830 align:middle line:90% at 850 in 17 minutes. 00:21:56.830 --> 00:22:00.280 align:middle line:84% And the red is what it looks like, the boron looks 00:22:00.280 --> 00:22:02.110 align:middle line:90% like due to thermal oxidation. 00:22:02.110 --> 00:22:04.300 align:middle line:84% You see a lot of the boron has been incorporated 00:22:04.300 --> 00:22:06.040 align:middle line:84% into the oxide because of segregation 00:22:06.040 --> 00:22:09.970 align:middle line:90% effects and consumption. 00:22:09.970 --> 00:22:12.410 align:middle line:84% The black is as-implanted, so that 00:22:12.410 --> 00:22:13.760 align:middle line:90% was how it was as-implanted. 00:22:13.760 --> 00:22:16.530 align:middle line:84% If I thermally oxidize it, I get the red curve. 00:22:16.530 --> 00:22:18.770 align:middle line:84% You can see it's depleted at the surface, 00:22:18.770 --> 00:22:21.260 align:middle line:90% and it's piled up in the oxide. 00:22:21.260 --> 00:22:23.300 align:middle line:84% Instead of thermally oxidizing, let's say 00:22:23.300 --> 00:22:26.240 align:middle line:84% I didn't want to grow my gate oxide, I just deposit it, 00:22:26.240 --> 00:22:28.070 align:middle line:90% 50 angstroms at 850. 00:22:28.070 --> 00:22:29.670 align:middle line:84% The boron profile is quite different. 00:22:29.670 --> 00:22:32.340 align:middle line:90% It's the blue. 00:22:32.340 --> 00:22:33.450 align:middle line:90% So why is that? 00:22:33.450 --> 00:22:35.040 align:middle line:84% So there's two different cases here. 00:22:35.040 --> 00:22:36.810 align:middle line:84% In the case of if I do an implant 00:22:36.810 --> 00:22:39.570 align:middle line:84% and I thermally oxidize it, I consume silicon. 00:22:39.570 --> 00:22:41.760 align:middle line:84% And I suck some of that boron into the oxide, 00:22:41.760 --> 00:22:44.010 align:middle line:84% and then it segregates by segregation. 00:22:44.010 --> 00:22:45.030 align:middle line:90% So you see? 00:22:45.030 --> 00:22:48.310 align:middle line:84% The boron concentration is relatively low at the surface. 00:22:48.310 --> 00:22:50.130 align:middle line:84% If instead of thermally oxidizing 00:22:50.130 --> 00:22:53.130 align:middle line:84% I just implant the boron, place an oxide 00:22:53.130 --> 00:22:57.150 align:middle line:84% down there by a chemical vapor deposition process, 00:22:57.150 --> 00:23:00.220 align:middle line:84% then I don't consume any of the silicon. 00:23:00.220 --> 00:23:01.750 align:middle line:90% So the boron stays a lot higher. 00:23:01.750 --> 00:23:05.170 align:middle line:84% So I have about a factor of two more boron at the surface. 00:23:05.170 --> 00:23:08.460 align:middle line:84% So you can imagine you would get different device characteristics 00:23:08.460 --> 00:23:11.940 align:middle line:84% due to this slightly different ways of forming 00:23:11.940 --> 00:23:16.080 align:middle line:84% this oxide because it has an impact on the boron profile 00:23:16.080 --> 00:23:17.738 align:middle line:90% by different diffusion. 00:23:17.738 --> 00:23:19.530 align:middle line:84% Well, the diffusion coefficient's the same, 00:23:19.530 --> 00:23:21.810 align:middle line:84% but it's really different segregation and the fact 00:23:21.810 --> 00:23:23.730 align:middle line:84% that in one case we're consuming silicon. 00:23:23.730 --> 00:23:26.750 align:middle line:90% In the other case, we're not. 00:23:26.750 --> 00:23:28.430 align:middle line:84% OK, that's just a simple example. 00:23:28.430 --> 00:23:30.400 align:middle line:84% I had to use-- you see all these points. 00:23:30.400 --> 00:23:32.400 align:middle line:84% I had to use a really fine grid in order for you 00:23:32.400 --> 00:23:35.840 align:middle line:90% to see this effect. 00:23:35.840 --> 00:23:37.910 align:middle line:84% So let me talk a little bit about that grid dot 00:23:37.910 --> 00:23:41.330 align:middle line:90% fac and the gridding. 00:23:41.330 --> 00:23:43.080 align:middle line:84% When you have very shallow profiles, 00:23:43.080 --> 00:23:46.430 align:middle line:84% which is mostly what you're doing these days for CMOS, 00:23:46.430 --> 00:23:48.170 align:middle line:90% and thin oxides-- 00:23:48.170 --> 00:23:49.430 align:middle line:90% oh, OK, here we go. 00:23:49.430 --> 00:23:51.260 align:middle line:90% This gives us the answer. 00:23:51.260 --> 00:23:56.720 align:middle line:84% If you use the default grid, then the grid spacing 00:23:56.720 --> 00:23:58.400 align:middle line:90% is about 0.1 micron. 00:23:58.400 --> 00:24:00.350 align:middle line:90% So there's about 0.1 micron. 00:24:00.350 --> 00:24:04.530 align:middle line:84% That's actually bigger than a lot of devices. 00:24:04.530 --> 00:24:07.190 align:middle line:90% So it's much too crude. 00:24:07.190 --> 00:24:11.510 align:middle line:84% So we can multiply that by a small number, 00:24:11.510 --> 00:24:13.820 align:middle line:84% like the grid dot fac and make that grid 00:24:13.820 --> 00:24:15.360 align:middle line:90% spacing much, much smaller. 00:24:15.360 --> 00:24:17.660 align:middle line:84% So in this particular example, I'm 00:24:17.660 --> 00:24:22.070 align:middle line:84% showing you where we did that same implant of boron 00:24:22.070 --> 00:24:25.175 align:middle line:84% and then 850, 17-minute oxidation, with a grid 00:24:25.175 --> 00:24:28.420 align:middle line:90% dot fac of 0.4. 00:24:28.420 --> 00:24:29.510 align:middle line:90% It's still very coarse. 00:24:29.510 --> 00:24:30.010 align:middle line:90% You can see. 00:24:30.010 --> 00:24:32.420 align:middle line:84% This doesn't really have much of a shape to it. 00:24:32.420 --> 00:24:34.490 align:middle line:84% There's a point here where we've solved. 00:24:34.490 --> 00:24:35.510 align:middle line:90% There's a point here. 00:24:35.510 --> 00:24:36.385 align:middle line:90% There's a point here. 00:24:36.385 --> 00:24:38.650 align:middle line:84% But it's just very jagged looking 00:24:38.650 --> 00:24:40.330 align:middle line:90% because the grid is too coarse. 00:24:40.330 --> 00:24:43.810 align:middle line:84% And in fact, in the oxide you're getting almost no detail 00:24:43.810 --> 00:24:45.593 align:middle line:84% or information at all because you really 00:24:45.593 --> 00:24:47.260 align:middle line:84% don't have any grid points in the oxide. 00:24:47.260 --> 00:24:50.360 align:middle line:84% You've got one at the interface and one at the surface. 00:24:50.360 --> 00:24:53.470 align:middle line:84% So for this particular example, a grid fac of 0.4 00:24:53.470 --> 00:24:54.280 align:middle line:90% is way too crude. 00:24:54.280 --> 00:24:55.780 align:middle line:84% So this is something you always want 00:24:55.780 --> 00:24:57.820 align:middle line:90% to check by doing a quick test. 00:24:57.820 --> 00:25:01.430 align:middle line:90% What kind of a grid do I want? 00:25:01.430 --> 00:25:04.790 align:middle line:84% If you go on to the next slide, slide number 13, here's 00:25:04.790 --> 00:25:10.490 align:middle line:84% a factor of 10, finer grid, grid dot fac of 0.04. 00:25:10.490 --> 00:25:12.090 align:middle line:90% Still, it looks pretty good. 00:25:12.090 --> 00:25:14.390 align:middle line:84% Now, it's starting to look like a smooth curve. 00:25:14.390 --> 00:25:15.685 align:middle line:90% It's not all jagged looking. 00:25:15.685 --> 00:25:17.060 align:middle line:84% And it has a shape that you would 00:25:17.060 --> 00:25:20.930 align:middle line:84% imagine could be associated with diffusion and-- 00:25:20.930 --> 00:25:22.790 align:middle line:90% ion implantation and diffusion. 00:25:22.790 --> 00:25:24.890 align:middle line:84% But the oxide grid, in the oxide, 00:25:24.890 --> 00:25:27.770 align:middle line:84% it's still too coarse because the oxide is thin. 00:25:27.770 --> 00:25:30.320 align:middle line:84% It's only on the order of 10 to 50 angstroms 00:25:30.320 --> 00:25:31.640 align:middle line:90% during this oxidation. 00:25:31.640 --> 00:25:33.620 align:middle line:90% So I'm only getting what? 00:25:33.620 --> 00:25:34.880 align:middle line:90% One point in the oxide. 00:25:34.880 --> 00:25:37.490 align:middle line:84% I have one at the interface, one in the oxide, and one here. 00:25:37.490 --> 00:25:40.720 align:middle line:90% That's still not enough. 00:25:40.720 --> 00:25:45.130 align:middle line:84% So if we go to the next slide, slide 00:25:45.130 --> 00:25:48.500 align:middle line:84% number 14, and as I mentioned before, 00:25:48.500 --> 00:25:51.700 align:middle line:84% we use this particular statement to make a very fine grid, 00:25:51.700 --> 00:25:53.230 align:middle line:90% and just inside the oxide. 00:25:53.230 --> 00:25:56.410 align:middle line:84% We don't want to make it that fine everywhere. 00:25:56.410 --> 00:26:01.300 align:middle line:84% Then we use this-- you can insert 00:26:01.300 --> 00:26:05.680 align:middle line:84% a statement called methoddy dot oxide equals 0.0005. 00:26:05.680 --> 00:26:09.770 align:middle line:84% That puts in a 5 angstrom grid in the oxide, 00:26:09.770 --> 00:26:10.900 align:middle line:90% which kind of makes sense. 00:26:10.900 --> 00:26:14.740 align:middle line:84% You have a 50 angstrom oxide you're growing. 00:26:14.740 --> 00:26:18.730 align:middle line:84% You'd like to have something like a 5 angstrom grid 00:26:18.730 --> 00:26:19.960 align:middle line:90% in that oxide. 00:26:19.960 --> 00:26:23.810 align:middle line:84% And now you can start to see what it really looks like. 00:26:23.810 --> 00:26:25.510 align:middle line:84% Remember, before I had one point here, 00:26:25.510 --> 00:26:27.280 align:middle line:90% one point here, and one here. 00:26:27.280 --> 00:26:29.775 align:middle line:84% It's a solution looked completely different. 00:26:29.775 --> 00:26:31.150 align:middle line:84% This is what it would have looked 00:26:31.150 --> 00:26:32.948 align:middle line:84% like if you just hadn't bothered to put 00:26:32.948 --> 00:26:34.240 align:middle line:90% in a special grid in the oxide. 00:26:34.240 --> 00:26:36.370 align:middle line:84% You'd think it was a triangular sort of profile. 00:26:36.370 --> 00:26:37.600 align:middle line:90% It's not triangular at all. 00:26:37.600 --> 00:26:40.520 align:middle line:84% The boron actually looks sort of flat topped in the oxide 00:26:40.520 --> 00:26:42.270 align:middle line:90% and then goes down very rapidly. 00:26:42.270 --> 00:26:44.960 align:middle line:84% So you really need to keep an eye on your grid. 00:26:44.960 --> 00:26:49.490 align:middle line:84% And you typically want to run problems or solutions 00:26:49.490 --> 00:26:51.980 align:middle line:84% for a number of different grid spacings 00:26:51.980 --> 00:26:54.350 align:middle line:84% to get a feel for what grid is appropriate 00:26:54.350 --> 00:27:01.470 align:middle line:84% and what grid gives you physically realistic results. 00:27:01.470 --> 00:27:04.710 align:middle line:84% OK, so that's a little bit of a trivial example on oxidation, 00:27:04.710 --> 00:27:06.240 align:middle line:84% but I just wanted to make the point 00:27:06.240 --> 00:27:08.550 align:middle line:84% of how important knowing your grid 00:27:08.550 --> 00:27:10.360 align:middle line:84% is because you can get a solution. 00:27:10.360 --> 00:27:12.660 align:middle line:84% You look at it, and you think it shows you something. 00:27:12.660 --> 00:27:15.288 align:middle line:84% And in fact, it could be physically not meaningful if you 00:27:15.288 --> 00:27:16.455 align:middle line:90% haven't used the right grid. 00:27:16.455 --> 00:27:19.570 align:middle line:90% 00:27:19.570 --> 00:27:22.860 align:middle line:84% So let's go on to slide number 15 and example number two. 00:27:22.860 --> 00:27:24.918 align:middle line:84% I want to talk about ion implant modeling. 00:27:24.918 --> 00:27:27.210 align:middle line:84% Remember, there's a number of different analytic models 00:27:27.210 --> 00:27:29.500 align:middle line:90% that we talked about. 00:27:29.500 --> 00:27:31.200 align:middle line:90% We said, you can use a Gaussian. 00:27:31.200 --> 00:27:32.730 align:middle line:90% You can use a Pearson-IV. 00:27:32.730 --> 00:27:35.370 align:middle line:90% You can use dual Pearson. 00:27:35.370 --> 00:27:38.760 align:middle line:84% And that there are tables in the literature that 00:27:38.760 --> 00:27:42.510 align:middle line:84% give the first three moments, the RP, the delta RP, 00:27:42.510 --> 00:27:43.230 align:middle line:90% and the skewness. 00:27:43.230 --> 00:27:47.580 align:middle line:84% Those tables have been tabulated since the 1970s or so. 00:27:47.580 --> 00:27:50.280 align:middle line:84% And they've been updated over time-- 00:27:50.280 --> 00:27:51.780 align:middle line:90% of the implant distributions. 00:27:51.780 --> 00:27:54.210 align:middle line:84% And they've been produced by theory and also 00:27:54.210 --> 00:27:56.450 align:middle line:90% by fitting to experiment. 00:27:56.450 --> 00:27:58.750 align:middle line:84% This is the most common analytic formulation 00:27:58.750 --> 00:28:00.580 align:middle line:90% that you will see in simulators. 00:28:00.580 --> 00:28:01.750 align:middle line:90% It's called the Pearson-IV. 00:28:01.750 --> 00:28:04.330 align:middle line:84% We talked about that distribution. 00:28:04.330 --> 00:28:09.850 align:middle line:84% And we said it does a pretty good job of replicating profiles 00:28:09.850 --> 00:28:11.050 align:middle line:90% into amorphous case. 00:28:11.050 --> 00:28:13.870 align:middle line:84% It does not model channeling all that well. 00:28:13.870 --> 00:28:16.660 align:middle line:84% Channeling requires the use of more parameters, 00:28:16.660 --> 00:28:19.490 align:middle line:90% generally by curve fitting. 00:28:19.490 --> 00:28:22.960 align:middle line:84% So a Pearson-IV is a single Pearson distribution. 00:28:22.960 --> 00:28:24.460 align:middle line:84% If you want to do channeling, you'll 00:28:24.460 --> 00:28:27.190 align:middle line:84% often see dual Pearson, meaning two different Pearson-IV 00:28:27.190 --> 00:28:29.710 align:middle line:90% profiles are added together. 00:28:29.710 --> 00:28:32.900 align:middle line:90% And lookup tables are used. 00:28:32.900 --> 00:28:34.990 align:middle line:84% One of the things you need to be aware of 00:28:34.990 --> 00:28:37.750 align:middle line:84% are inaccuracies in stopping powers, particularly 00:28:37.750 --> 00:28:42.730 align:middle line:84% the electronic stopping at low energies and even 00:28:42.730 --> 00:28:44.050 align:middle line:90% nuclear stopping powers. 00:28:44.050 --> 00:28:46.270 align:middle line:84% At very low energies, a lot of these stopping powers 00:28:46.270 --> 00:28:47.750 align:middle line:90% are not known that well. 00:28:47.750 --> 00:28:51.550 align:middle line:84% So if you're trying to simulate energies below, say, 10 keV, 00:28:51.550 --> 00:28:55.120 align:middle line:84% 20 keV, depending on the dopant, you 00:28:55.120 --> 00:28:57.800 align:middle line:90% may not get very good results. 00:28:57.800 --> 00:29:00.800 align:middle line:84% Dopants that are widely used in Silicon technology, 00:29:00.800 --> 00:29:02.480 align:middle line:84% they're constantly updating these tables 00:29:02.480 --> 00:29:03.980 align:middle line:84% to try to make them a little better, 00:29:03.980 --> 00:29:06.540 align:middle line:84% more accurately represent reality in experiments. 00:29:06.540 --> 00:29:09.710 align:middle line:84% But some dopants that are not used that often at low energies, 00:29:09.710 --> 00:29:12.140 align:middle line:84% you'll simulate, and you won't get a very good simulation 00:29:12.140 --> 00:29:13.430 align:middle line:90% compared to the actual data. 00:29:13.430 --> 00:29:16.140 align:middle line:90% 00:29:16.140 --> 00:29:18.140 align:middle line:84% So let's go on to slide number 16, 00:29:18.140 --> 00:29:21.470 align:middle line:84% and this is an example of a SUPREM input file 00:29:21.470 --> 00:29:23.180 align:middle line:84% where we're using tabulated moments. 00:29:23.180 --> 00:29:28.160 align:middle line:84% So SUPREM has in its database tables of all these moments. 00:29:28.160 --> 00:29:31.580 align:middle line:84% And it's using tabulated moments to evaluate and to plot for you 00:29:31.580 --> 00:29:33.150 align:middle line:90% an implant distribution. 00:29:33.150 --> 00:29:36.980 align:middle line:84% So here's an example of ion-implanted phosphorus. 00:29:36.980 --> 00:29:38.630 align:middle line:90% So we define our mesh. 00:29:38.630 --> 00:29:41.210 align:middle line:90% Here's a grid fac equal 0.1. 00:29:41.210 --> 00:29:44.600 align:middle line:84% So it's not all that fine of a mesh. 00:29:44.600 --> 00:29:45.800 align:middle line:90% The substrate is boron. 00:29:45.800 --> 00:29:47.330 align:middle line:84% We're going to implant phosphorus 00:29:47.330 --> 00:29:49.820 align:middle line:84% at a particular dose, [? 8014. ?] 00:29:49.820 --> 00:29:54.590 align:middle line:84% Energy, not all that low, about 30 keV, but somewhat low, 00:29:54.590 --> 00:29:57.590 align:middle line:84% a 7-degree tilt. And we're saying to use a Gaussian model, 00:29:57.590 --> 00:29:59.780 align:middle line:90% very, very simple model. 00:29:59.780 --> 00:30:02.930 align:middle line:84% Save it in a file called Gaussian.saved, 00:30:02.930 --> 00:30:04.837 align:middle line:90% and then plot it. 00:30:04.837 --> 00:30:06.170 align:middle line:90% And then we're going to plot it. 00:30:06.170 --> 00:30:10.490 align:middle line:84% In addition to plotting the phosphorus that it calculated, 00:30:10.490 --> 00:30:13.810 align:middle line:90% SUPREM has a nice way of-- 00:30:13.810 --> 00:30:15.660 align:middle line:84% using a command called profile, you 00:30:15.660 --> 00:30:20.700 align:middle line:84% can input into SUPREM xy data, a matrix of concentration 00:30:20.700 --> 00:30:21.510 align:middle line:90% versus depth. 00:30:21.510 --> 00:30:23.250 align:middle line:84% And that's called the profile statement. 00:30:23.250 --> 00:30:25.650 align:middle line:90% It reads in phosphorus. 00:30:25.650 --> 00:30:26.943 align:middle line:90% And the name of the file-- 00:30:26.943 --> 00:30:28.110 align:middle line:90% this is the particular name. 00:30:28.110 --> 00:30:31.260 align:middle line:84% It's kind of ugly, but you can call it whatever you want-- 00:30:31.260 --> 00:30:36.600 align:middle line:84% and use that to plot the phosphorus. 00:30:36.600 --> 00:30:39.730 align:middle line:84% So we can not only plot the SUPREM simulation, 00:30:39.730 --> 00:30:42.250 align:middle line:84% we can also plot the actual SIMS data. 00:30:42.250 --> 00:30:45.570 align:middle line:84% So this happens to be an implant for which I have actual SIMS 00:30:45.570 --> 00:30:47.680 align:middle line:84% data that I have entered into this. 00:30:47.680 --> 00:30:52.960 align:middle line:84% And we're going to compare that data to the Gaussian solution. 00:30:52.960 --> 00:30:56.220 align:middle line:84% So if you go on to slide number 17, you'll see. 00:30:56.220 --> 00:31:00.900 align:middle line:84% This is for that relatively low-energy phosphorus implant. 00:31:00.900 --> 00:31:03.280 align:middle line:84% The SIMS data here is shown in red. 00:31:03.280 --> 00:31:05.730 align:middle line:84% This is actual data obtained from a company called 00:31:05.730 --> 00:31:06.390 align:middle line:90% Charles Evans. 00:31:06.390 --> 00:31:10.080 align:middle line:84% I think we've talked about Evans Associates in this class. 00:31:10.080 --> 00:31:14.670 align:middle line:84% They used a cesium beam with a primary beam energy of 2 keV 00:31:14.670 --> 00:31:15.730 align:middle line:90% to profile. 00:31:15.730 --> 00:31:17.920 align:middle line:90% And this is what they got. 00:31:17.920 --> 00:31:20.530 align:middle line:84% And the black line is what SUPREM 00:31:20.530 --> 00:31:24.970 align:middle line:84% says it should look like using a Gaussian simulation. 00:31:24.970 --> 00:31:27.460 align:middle line:84% And these are default values that 00:31:27.460 --> 00:31:30.820 align:middle line:84% are tabulated inside the SUPREM-IV simulator using 00:31:30.820 --> 00:31:33.690 align:middle line:84% the standard LSS range parameters, RP and delta RP. 00:31:33.690 --> 00:31:35.440 align:middle line:84% Remember, you only get two for a Gaussian. 00:31:35.440 --> 00:31:39.410 align:middle line:84% So compared to the SIMS, it's not all that great. 00:31:39.410 --> 00:31:42.170 align:middle line:84% The simulation does not do a very good job. 00:31:42.170 --> 00:31:45.770 align:middle line:90% It's too simplified of a model. 00:31:45.770 --> 00:31:49.390 align:middle line:84% The stopping power is apparently too small, 00:31:49.390 --> 00:31:53.590 align:middle line:84% and the range is overestimated by the theoretical calculation 00:31:53.590 --> 00:31:57.780 align:middle line:84% compared to what was experimentally observed. 00:31:57.780 --> 00:32:00.390 align:middle line:90% Well, we can try another model. 00:32:00.390 --> 00:32:01.560 align:middle line:90% Same SIMS data. 00:32:01.560 --> 00:32:03.520 align:middle line:84% This time, we try the black model, 00:32:03.520 --> 00:32:06.630 align:middle line:90% which is the dual Pearson, OK? 00:32:06.630 --> 00:32:09.750 align:middle line:84% Dual Pearson uses more parameters. 00:32:09.750 --> 00:32:12.107 align:middle line:84% It simulates this sort of channel tail 00:32:12.107 --> 00:32:12.940 align:middle line:90% a little bit better. 00:32:12.940 --> 00:32:14.610 align:middle line:84% But again, the range and the skewness 00:32:14.610 --> 00:32:18.210 align:middle line:84% are both overestimated, way too much skewness. 00:32:18.210 --> 00:32:19.410 align:middle line:90% So what does this tell us? 00:32:19.410 --> 00:32:23.970 align:middle line:84% Well, this basically tells us that these tabulated moments are 00:32:23.970 --> 00:32:25.830 align:middle line:90% not perfect for all dopants. 00:32:25.830 --> 00:32:28.920 align:middle line:84% So take them with a grain of salt, especially phosphorus. 00:32:28.920 --> 00:32:31.620 align:middle line:84% Phosphorus [INAUDIBLE] used that much in CMOS technology, right? 00:32:31.620 --> 00:32:33.720 align:middle line:84% Shallow junctions for n-type are typically 00:32:33.720 --> 00:32:37.380 align:middle line:84% made by arsenic because phosphorus diffuses too fast. 00:32:37.380 --> 00:32:40.620 align:middle line:84% So not a whole lot of energy and manpower 00:32:40.620 --> 00:32:43.710 align:middle line:84% has gone into improving the moment 00:32:43.710 --> 00:32:45.870 align:middle line:90% tables for shallow phosphorus. 00:32:45.870 --> 00:32:49.470 align:middle line:84% So if you happen to be using phosphorus in your experiments, 00:32:49.470 --> 00:32:51.270 align:middle line:84% take it with a grain of salt if you're 00:32:51.270 --> 00:32:53.770 align:middle line:90% using these tabulated values. 00:32:53.770 --> 00:32:55.898 align:middle line:84% According to my data, it doesn't fit that well. 00:32:55.898 --> 00:32:58.190 align:middle line:84% Of course, there might be something wrong with my data. 00:32:58.190 --> 00:32:59.330 align:middle line:90% That's another possibility. 00:32:59.330 --> 00:33:03.250 align:middle line:84% So exactly what's reality, it's not always clear. 00:33:03.250 --> 00:33:05.620 align:middle line:84% But interestingly, on the next slide, 00:33:05.620 --> 00:33:08.350 align:middle line:84% exact same data this time compared 00:33:08.350 --> 00:33:10.150 align:middle line:90% to Monte Carlo simulation. 00:33:10.150 --> 00:33:12.910 align:middle line:84% Now, these Monte Carlo simulations take a lot longer. 00:33:12.910 --> 00:33:14.230 align:middle line:90% It's not an analytic. 00:33:14.230 --> 00:33:15.370 align:middle line:90% It's a numerical solution. 00:33:15.370 --> 00:33:17.830 align:middle line:84% You have to follow each ion into the silicon 00:33:17.830 --> 00:33:20.380 align:middle line:84% and see where it lands up and then statistically 00:33:20.380 --> 00:33:21.590 align:middle line:90% create a profile. 00:33:21.590 --> 00:33:24.920 align:middle line:84% And that's what the black is, OK? 00:33:24.920 --> 00:33:27.110 align:middle line:84% And you can see the black looks kind of jaggy 00:33:27.110 --> 00:33:29.270 align:middle line:84% because, again, you're statistically 00:33:29.270 --> 00:33:30.800 align:middle line:90% creating this profile. 00:33:30.800 --> 00:33:36.350 align:middle line:84% But interestingly, the black does a reasonably good job 00:33:36.350 --> 00:33:38.020 align:middle line:90% of agreeing with the SIMS. 00:33:38.020 --> 00:33:39.770 align:middle line:84% Doesn't mean they're right, but it kind of 00:33:39.770 --> 00:33:41.690 align:middle line:84% gives me some confidence that, when 00:33:41.690 --> 00:33:44.880 align:middle line:84% I see a simulation agrees pretty well with the SIMS data, 00:33:44.880 --> 00:33:49.160 align:middle line:84% at least in terms of its range and its delta RP, its width 00:33:49.160 --> 00:33:52.230 align:middle line:84% here, that's probably a pretty good sign. 00:33:52.230 --> 00:33:55.640 align:middle line:84% So there's enough physics in this Monte Carlo simulation 00:33:55.640 --> 00:33:58.290 align:middle line:84% to reproduce the data pretty well. 00:33:58.290 --> 00:34:02.360 align:middle line:84% But notice, this is how you tell SUPREM 00:34:02.360 --> 00:34:03.530 align:middle line:90% to do a Monte Carlo implant. 00:34:03.530 --> 00:34:05.340 align:middle line:84% You give it the implant statement. 00:34:05.340 --> 00:34:07.580 align:middle line:84% You tell it the species, the dose, the energy. 00:34:07.580 --> 00:34:09.570 align:middle line:84% Now, here you tell it how you want to do it. 00:34:09.570 --> 00:34:10.070 align:middle line:90% You see? 00:34:10.070 --> 00:34:13.310 align:middle line:84% In prior, I had said Gaussian or dual Pearson, 00:34:13.310 --> 00:34:17.570 align:middle line:84% or if you say nothing, it defaults probably to Pearson-IV. 00:34:17.570 --> 00:34:19.250 align:middle line:90% Here, you said Monte Carlo. 00:34:19.250 --> 00:34:20.900 align:middle line:84% Now, an important thing when you do 00:34:20.900 --> 00:34:23.250 align:middle line:84% Monte Carlo is to specify the number of ions you 00:34:23.250 --> 00:34:25.810 align:middle line:84% want it to shoot into the sample and to follow. 00:34:25.810 --> 00:34:31.270 align:middle line:84% So this is 25,000 ions, not very many, and tilt and rotation. 00:34:31.270 --> 00:34:35.280 align:middle line:84% You need at least 10,000 to get good accuracy. 00:34:35.280 --> 00:34:38.850 align:middle line:84% Probably, 100,000 is better because you can see here, 00:34:38.850 --> 00:34:41.580 align:middle line:84% after I'm two decades beyond the peak, 00:34:41.580 --> 00:34:43.230 align:middle line:90% I start to get a lot of noise. 00:34:43.230 --> 00:34:45.429 align:middle line:90% So the noise comes in. 00:34:45.429 --> 00:34:48.270 align:middle line:84% So I'm really only getting two decades of smooth data. 00:34:48.270 --> 00:34:50.639 align:middle line:90% That's with 25,000 ions. 00:34:50.639 --> 00:34:52.920 align:middle line:84% In a one-dimensional simulation, this isn't too bad. 00:34:52.920 --> 00:34:54.510 align:middle line:84% It might only take 5, 10 minutes. 00:34:54.510 --> 00:34:55.270 align:middle line:90% No big deal. 00:34:55.270 --> 00:34:56.699 align:middle line:90% But now this is just 1D. 00:34:56.699 --> 00:35:01.300 align:middle line:84% If I'm doing this in 2D across the channel of a MOSFET, that 5, 00:35:01.300 --> 00:35:04.540 align:middle line:84% 10 minutes goes to be three or four hours or more. 00:35:04.540 --> 00:35:09.130 align:middle line:84% So you have to really compromise to a certain extent. 00:35:09.130 --> 00:35:11.160 align:middle line:84% So what people will do sometimes is 00:35:11.160 --> 00:35:15.030 align:middle line:84% fit the 1D profile using Monte Carlo and then fit to it 00:35:15.030 --> 00:35:19.680 align:middle line:84% an analytic solution by changing some of the default parameters. 00:35:19.680 --> 00:35:22.890 align:middle line:84% And then use that analytic in the two-dimensional simulation 00:35:22.890 --> 00:35:26.640 align:middle line:84% as a way of getting around spending so much time. 00:35:26.640 --> 00:35:32.210 align:middle line:84% So that's just an example of how actual SUPREM outputs compare 00:35:32.210 --> 00:35:36.640 align:middle line:90% to real data. 00:35:36.640 --> 00:35:37.640 align:middle line:90% Let's take another case. 00:35:37.640 --> 00:35:38.598 align:middle line:90% So that was phosphorus. 00:35:38.598 --> 00:35:41.700 align:middle line:84% As I said, phosphorus is not all that widely used. 00:35:41.700 --> 00:35:43.910 align:middle line:84% Here's the case for 30 kilovolts, same energy, 00:35:43.910 --> 00:35:48.380 align:middle line:84% but this time arsenic, using the dual Pearson analytic. 00:35:48.380 --> 00:35:52.880 align:middle line:84% Again, the SIMS data is the red line. 00:35:52.880 --> 00:35:54.950 align:middle line:90% And the calculated is the black. 00:35:54.950 --> 00:35:56.570 align:middle line:84% And this time, with the dual Pearson, 00:35:56.570 --> 00:35:57.740 align:middle line:90% it does a pretty good job. 00:35:57.740 --> 00:36:01.460 align:middle line:84% You got the range almost just right and even the broadening, 00:36:01.460 --> 00:36:03.080 align:middle line:90% not quite, but pretty close. 00:36:03.080 --> 00:36:06.020 align:middle line:84% The shape are much more accurate than for phosphorus. 00:36:06.020 --> 00:36:08.300 align:middle line:84% I don't know why, but I'm guessing 00:36:08.300 --> 00:36:11.150 align:middle line:84% it's because shallow arsenic junctions are the way people 00:36:11.150 --> 00:36:12.270 align:middle line:90% make junctions these days. 00:36:12.270 --> 00:36:15.140 align:middle line:84% And so people have updated in SUPREM 00:36:15.140 --> 00:36:17.930 align:middle line:84% these tables, these tables of moments 00:36:17.930 --> 00:36:22.310 align:middle line:84% to fit better the experimental data. 00:36:22.310 --> 00:36:25.680 align:middle line:84% Arsenic is also heavier, dominated by nuclear stopping, 00:36:25.680 --> 00:36:27.290 align:middle line:84% which is a little better understood 00:36:27.290 --> 00:36:29.390 align:middle line:90% than electronic stopping. 00:36:29.390 --> 00:36:31.520 align:middle line:84% Remember, we said electronic stopping powers always 00:36:31.520 --> 00:36:33.740 align:middle line:84% have a little bit of fuzziness about them. 00:36:33.740 --> 00:36:35.210 align:middle line:84% Phosphorus is going to be lighter. 00:36:35.210 --> 00:36:36.570 align:middle line:90% It has more electronic stopping. 00:36:36.570 --> 00:36:37.530 align:middle line:90% Maybe that's why. 00:36:37.530 --> 00:36:40.230 align:middle line:84% But again, take the profiles you get with a grain of salt 00:36:40.230 --> 00:36:42.180 align:middle line:84% unless you've checked them out experimentally. 00:36:42.180 --> 00:36:45.040 align:middle line:90% 00:36:45.040 --> 00:36:48.910 align:middle line:84% Slide 21, that same data, this time with Monte Carlo. 00:36:48.910 --> 00:36:50.292 align:middle line:90% Excellent job. 00:36:50.292 --> 00:36:51.250 align:middle line:90% This looks even better. 00:36:51.250 --> 00:36:54.530 align:middle line:84% It's got the broadening a little better. 00:36:54.530 --> 00:36:58.360 align:middle line:84% So it looks quite good using Monte Carlo. 00:36:58.360 --> 00:37:01.150 align:middle line:84% Of course, Monte Carlo took longer to generate, 00:37:01.150 --> 00:37:03.850 align:middle line:90% longer to simulate. 00:37:03.850 --> 00:37:05.380 align:middle line:84% Another advantage of the Monte Carlo 00:37:05.380 --> 00:37:07.588 align:middle line:84% besides the fact that it seems to be pretty accurate, 00:37:07.588 --> 00:37:09.760 align:middle line:84% it has pretty good comparison to data, 00:37:09.760 --> 00:37:11.740 align:middle line:84% you can generate profiles of [INAUDIBLE] 00:37:11.740 --> 00:37:13.270 align:middle line:90% interstitials and vacancies. 00:37:13.270 --> 00:37:14.890 align:middle line:90% And we need those profiles. 00:37:14.890 --> 00:37:16.960 align:middle line:84% If you want to use those profiles, 00:37:16.960 --> 00:37:21.700 align:middle line:84% you can use them in subsequent diffusion simulations in order 00:37:21.700 --> 00:37:23.740 align:middle line:90% to simulate TED. 00:37:23.740 --> 00:37:27.190 align:middle line:84% If you do an implant and you don't tell the simulator 00:37:27.190 --> 00:37:30.280 align:middle line:84% to use a damage model, it's not going to be able to simulate TED 00:37:30.280 --> 00:37:32.600 align:middle line:84% because it needs to have some damage model. 00:37:32.600 --> 00:37:36.040 align:middle line:84% So Monte Carlo is also very useful for that purpose. 00:37:36.040 --> 00:37:38.760 align:middle line:90% 00:37:38.760 --> 00:37:40.550 align:middle line:90% OK, so let's go on to slide 22. 00:37:40.550 --> 00:37:44.360 align:middle line:84% So those were a couple of examples of simple oxidation, 00:37:44.360 --> 00:37:45.890 align:middle line:90% simple ion implantation. 00:37:45.890 --> 00:37:49.670 align:middle line:90% Now we get to the more-- 00:37:49.670 --> 00:37:52.560 align:middle line:84% a little more exciting models, a little more complicated. 00:37:52.560 --> 00:37:56.650 align:middle line:84% There are three major diffusion models in SUPREM-IV. 00:37:56.650 --> 00:37:58.270 align:middle line:84% And these are what they're called. 00:37:58.270 --> 00:38:00.430 align:middle line:84% The way you invoke a different model in SUPREM-IV 00:38:00.430 --> 00:38:02.320 align:middle line:90% is you use the method statement. 00:38:02.320 --> 00:38:05.890 align:middle line:84% And there are three methods for solving the differential 00:38:05.890 --> 00:38:08.680 align:middle line:84% equations associated with diffusion in SUPREM-IV. 00:38:08.680 --> 00:38:09.550 align:middle line:90% There's PD. 00:38:09.550 --> 00:38:11.230 align:middle line:84% That's Partial Differential equation. 00:38:11.230 --> 00:38:15.617 align:middle line:84% PD dot Fermi is a method that takes into account the impact 00:38:15.617 --> 00:38:17.200 align:middle line:84% of the Fermi level, just like the name 00:38:17.200 --> 00:38:20.560 align:middle line:84% would suggest on the dopant diffusion coefficient. 00:38:20.560 --> 00:38:23.290 align:middle line:84% For example, this is an equation that SUPREM 00:38:23.290 --> 00:38:25.178 align:middle line:90% uses for n-type dopants. 00:38:25.178 --> 00:38:26.470 align:middle line:90% You're very familiar with this. 00:38:26.470 --> 00:38:28.270 align:middle line:84% It has an n over ni, and it may be 00:38:28.270 --> 00:38:31.120 align:middle line:84% an n over ni squared depending on the dopant. 00:38:31.120 --> 00:38:34.000 align:middle line:84% And it has stored in it values for D0, D 00:38:34.000 --> 00:38:37.330 align:middle line:84% minus, D double minus for all the dopants. 00:38:37.330 --> 00:38:38.980 align:middle line:84% It models concentration-dependent 00:38:38.980 --> 00:38:42.190 align:middle line:84% diffusion, but it does not model TED or OED. 00:38:42.190 --> 00:38:44.120 align:middle line:90% So it's quite simple. 00:38:44.120 --> 00:38:45.430 align:middle line:90% It's very fast. 00:38:45.430 --> 00:38:48.400 align:middle line:84% And the nice thing is there's relatively few parameters, 00:38:48.400 --> 00:38:50.970 align:middle line:90% just a few D parameters. 00:38:50.970 --> 00:38:52.910 align:middle line:90% And you can get your answer. 00:38:52.910 --> 00:38:55.640 align:middle line:84% So that's the nice thing about it. 00:38:55.640 --> 00:38:57.450 align:middle line:84% And it does model concentration-dependent. 00:38:57.450 --> 00:38:59.117 align:middle line:84% So this is the first thing you would use 00:38:59.117 --> 00:39:00.940 align:middle line:90% because it's relatively fast. 00:39:00.940 --> 00:39:05.860 align:middle line:84% The next method, the next level of complication is PD.trans. 00:39:05.860 --> 00:39:09.320 align:middle line:84% And it takes into account the Fermi level. 00:39:09.320 --> 00:39:12.280 align:middle line:84% So it already has this model built in, the Fermi model, 00:39:12.280 --> 00:39:18.070 align:middle line:84% as well as the impact of nonequilibrium interstitial 00:39:18.070 --> 00:39:22.630 align:middle line:84% and vacancy profiles on the diffusion, but not the other way 00:39:22.630 --> 00:39:23.890 align:middle line:90% around. 00:39:23.890 --> 00:39:28.180 align:middle line:84% It will not show you the impact of dopant diffusion 00:39:28.180 --> 00:39:32.660 align:middle line:84% on the motion of interstitials and vacancies. 00:39:32.660 --> 00:39:34.810 align:middle line:84% So it's not fully coupled, but it 00:39:34.810 --> 00:39:38.800 align:middle line:84% does take into account the impact of the ions 00:39:38.800 --> 00:39:41.170 align:middle line:90% on the dopant diffusion. 00:39:41.170 --> 00:39:44.170 align:middle line:90% It's very useful for OED. 00:39:44.170 --> 00:39:47.230 align:middle line:84% So people tend to use it for oxidation-enhanced diffusion. 00:39:47.230 --> 00:39:50.770 align:middle line:84% Or you can use it for transient-enhanced diffusion 00:39:50.770 --> 00:39:54.950 align:middle line:84% with relatively low doping concentrations. 00:39:54.950 --> 00:39:57.520 align:middle line:84% If you have higher doping concentrations, then 00:39:57.520 --> 00:40:02.410 align:middle line:84% the coupled diffusion of the pair, of the dopant plus 00:40:02.410 --> 00:40:04.510 align:middle line:84% the point defect actually affects the point defect 00:40:04.510 --> 00:40:05.300 align:middle line:90% profile. 00:40:05.300 --> 00:40:07.990 align:middle line:84% And so you need something more sophisticated than PD.trans. 00:40:07.990 --> 00:40:11.783 align:middle line:84% And this is the model it uses, the basic concept that you 00:40:11.783 --> 00:40:14.200 align:middle line:84% should be familiar with now, if you've done your homework, 00:40:14.200 --> 00:40:19.630 align:middle line:84% is that the diffusivity is the unperturbed diffusivity times 00:40:19.630 --> 00:40:24.560 align:middle line:84% something in parentheses, where it depends on F sub i. 00:40:24.560 --> 00:40:29.330 align:middle line:84% So it takes into account the enhancement in CI over CI star 00:40:29.330 --> 00:40:33.350 align:middle line:84% that can happen when you do an oxidation or the enhancement 00:40:33.350 --> 00:40:36.830 align:middle line:84% in CV or CV star if you do a nitridation 00:40:36.830 --> 00:40:38.130 align:middle line:90% or if you do an implant. 00:40:38.130 --> 00:40:40.380 align:middle line:84% And these things get enlarged or suppressed. 00:40:40.380 --> 00:40:43.750 align:middle line:90% So it'll take that into account. 00:40:43.750 --> 00:40:47.130 align:middle line:84% The last method you can use takes the longest in terms of it 00:40:47.130 --> 00:40:51.000 align:middle line:84% can be very long simulation times, PD.full. 00:40:51.000 --> 00:40:52.620 align:middle line:84% The name comes from the fact that it 00:40:52.620 --> 00:40:54.150 align:middle line:90% is fully-coupled diffusion. 00:40:54.150 --> 00:40:55.480 align:middle line:90% And what does that mean? 00:40:55.480 --> 00:41:00.330 align:middle line:84% It means that the interstitials and vacancies impact the flux 00:41:00.330 --> 00:41:02.610 align:middle line:90% of the dopants, and vice versa. 00:41:02.610 --> 00:41:05.490 align:middle line:84% The flux of the dopants impact the interstitials 00:41:05.490 --> 00:41:08.130 align:middle line:90% and vacancy diffusion. 00:41:08.130 --> 00:41:10.320 align:middle line:84% This should be used most of the time 00:41:10.320 --> 00:41:14.730 align:middle line:84% if you are interested in transient-enhanced diffusion. 00:41:14.730 --> 00:41:17.385 align:middle line:84% And in high concentrations, in general, 00:41:17.385 --> 00:41:19.620 align:middle line:84% for instance, the emitter push effect, 00:41:19.620 --> 00:41:24.330 align:middle line:84% we talked about how phosphorus can pair with interstitials 00:41:24.330 --> 00:41:26.580 align:middle line:84% and, by diffusion, drag the interstitials 00:41:26.580 --> 00:41:27.720 align:middle line:90% into the substrate. 00:41:27.720 --> 00:41:30.420 align:middle line:84% When it goes substitutional, it releases those interstitials. 00:41:30.420 --> 00:41:34.110 align:middle line:84% Those interstitials then cause the boron base to broaden. 00:41:34.110 --> 00:41:35.490 align:middle line:84% That was the emitter push effect. 00:41:35.490 --> 00:41:37.920 align:middle line:84% You'll never be able to get that out of the Fermi model. 00:41:37.920 --> 00:41:38.760 align:middle line:90% There's no way. 00:41:38.760 --> 00:41:41.290 align:middle line:84% And even the trans model won't have that. 00:41:41.290 --> 00:41:44.230 align:middle line:84% So to simulate the things like emitter push 00:41:44.230 --> 00:41:47.080 align:middle line:84% effect, those types of fully coupled cases, 00:41:47.080 --> 00:41:50.313 align:middle line:90% you need this PD.full. 00:41:50.313 --> 00:41:51.980 align:middle line:84% So those are three different statements, 00:41:51.980 --> 00:41:57.020 align:middle line:84% and you can compare the results from all three statements 00:41:57.020 --> 00:41:58.520 align:middle line:90% for a given situation. 00:41:58.520 --> 00:41:59.780 align:middle line:90% Now, one thing I should say. 00:41:59.780 --> 00:42:02.450 align:middle line:84% As you go from here to here to here, 00:42:02.450 --> 00:42:06.200 align:middle line:84% you're invoking more physics, more chemistry, and also 00:42:06.200 --> 00:42:07.910 align:middle line:90% more parameters. 00:42:07.910 --> 00:42:09.260 align:middle line:90% That's one problem with this. 00:42:09.260 --> 00:42:13.230 align:middle line:84% Sure, you can model the emitter push effect. 00:42:13.230 --> 00:42:15.440 align:middle line:84% But there's a lot of parameters you need to know, 00:42:15.440 --> 00:42:19.040 align:middle line:84% the diffusion rates of the vacancies and the interstitials, 00:42:19.040 --> 00:42:20.630 align:middle line:84% their recombination of interfaces. 00:42:20.630 --> 00:42:22.820 align:middle line:84% If those parameters are in SUPREM, 00:42:22.820 --> 00:42:24.320 align:middle line:84% they always have default parameters. 00:42:24.320 --> 00:42:25.980 align:middle line:90% But they may not be accurate. 00:42:25.980 --> 00:42:28.953 align:middle line:84% So again, take everything you simulate with a grain of salt. 00:42:28.953 --> 00:42:30.620 align:middle line:84% If somebody has input something in there 00:42:30.620 --> 00:42:32.120 align:middle line:84% to the best of their knowledge, that 00:42:32.120 --> 00:42:34.910 align:middle line:84% might be the interstitial diffusion diffusivity. 00:42:34.910 --> 00:42:36.800 align:middle line:84% But maybe it wasn't well characterized 00:42:36.800 --> 00:42:39.390 align:middle line:84% at the temperature under the conditions that you're using. 00:42:39.390 --> 00:42:41.750 align:middle line:84% So you can get lots of different profiles, 00:42:41.750 --> 00:42:43.460 align:middle line:84% but how accurate are they represent 00:42:43.460 --> 00:42:48.635 align:middle line:84% experiment is something that you need to determine for yourself. 00:42:48.635 --> 00:42:51.340 align:middle line:90% 00:42:51.340 --> 00:42:52.255 align:middle line:90% So here's an example. 00:42:52.255 --> 00:42:54.630 align:middle line:84% Let's look at a couple of examples of different diffusion 00:42:54.630 --> 00:42:56.670 align:middle line:90% profiles. 00:42:56.670 --> 00:43:00.780 align:middle line:84% On page 23, this is a simple 1D arsenic implant, 00:43:00.780 --> 00:43:03.160 align:middle line:84% and then we're going to do a long-time diffusion. 00:43:03.160 --> 00:43:04.140 align:middle line:90% So the simplest thing-- 00:43:04.140 --> 00:43:07.380 align:middle line:84% is an implant arsenic at 30 keV, the certain dose, 00:43:07.380 --> 00:43:12.210 align:middle line:84% intermediate dose, 2.6e14, and diffuse it in a furnace at 1,000 00:43:12.210 --> 00:43:14.250 align:middle line:90% degrees for 30 minutes. 00:43:14.250 --> 00:43:17.010 align:middle line:84% Before we did the diffusion, we put down an oxide cap, 00:43:17.010 --> 00:43:19.140 align:middle line:84% not by oxidation, but by deposition. 00:43:19.140 --> 00:43:21.870 align:middle line:84% And so that oxide cap helps prevent the arsenic 00:43:21.870 --> 00:43:25.620 align:middle line:84% from evaporating out of the wafer. 00:43:25.620 --> 00:43:30.140 align:middle line:84% So this is the actual command that 00:43:30.140 --> 00:43:35.020 align:middle line:84% was used that specifies the point defect model, methodpd dot 00:43:35.020 --> 00:43:35.520 align:middle line:90% Fermi. 00:43:35.520 --> 00:43:37.280 align:middle line:84% So we're using the Fermi diffusion model. 00:43:37.280 --> 00:43:39.650 align:middle line:84% And this is the diffusion statement, 1,000 degrees 00:43:39.650 --> 00:43:41.540 align:middle line:90% for 30 minutes inert. 00:43:41.540 --> 00:43:44.480 align:middle line:84% I have some SIMS data here, which is shown in the red. 00:43:44.480 --> 00:43:47.420 align:middle line:84% And then the black dots or the black circles 00:43:47.420 --> 00:43:50.190 align:middle line:84% are the simulation using a Gaussian model. 00:43:50.190 --> 00:43:52.670 align:middle line:90% Here's a Gaussian and a Fermi. 00:43:52.670 --> 00:43:54.380 align:middle line:84% And you can see it actually reproduces 00:43:54.380 --> 00:43:56.030 align:middle line:90% the data amazingly well. 00:43:56.030 --> 00:43:58.010 align:middle line:84% It means that somebody calibrated 00:43:58.010 --> 00:44:01.800 align:middle line:84% SUPREM pretty accurately to this particular furnace. 00:44:01.800 --> 00:44:03.920 align:middle line:90% The Fermi models good enough. 00:44:03.920 --> 00:44:06.200 align:middle line:84% And you say, well, how about all that-- why didn't you 00:44:06.200 --> 00:44:08.510 align:middle line:90% use PD.trans or fully coupled? 00:44:08.510 --> 00:44:11.480 align:middle line:84% How did you get away with such a simple 30-second simulation? 00:44:11.480 --> 00:44:13.310 align:middle line:90% Well, the anneal is long. 00:44:13.310 --> 00:44:16.430 align:middle line:84% So there's not going to be any big damage or TED 00:44:16.430 --> 00:44:17.930 align:middle line:90% effects after-- 00:44:17.930 --> 00:44:19.970 align:middle line:84% normal diffusion is going to dominate this. 00:44:19.970 --> 00:44:22.710 align:middle line:84% You need concentration dependency how box-like it is. 00:44:22.710 --> 00:44:25.920 align:middle line:84% So you need to use the Fermi model for Fermi-level effects. 00:44:25.920 --> 00:44:28.830 align:middle line:84% But you don't need to go and invoking necessarily 00:44:28.830 --> 00:44:30.040 align:middle line:90% TED or anything. 00:44:30.040 --> 00:44:33.420 align:middle line:84% So if you know you have a case where TED effects are not 00:44:33.420 --> 00:44:35.670 align:middle line:84% that prevalent, you might as well use the Fermi model. 00:44:35.670 --> 00:44:39.300 align:middle line:84% It'll go a lot faster, and in a two-dimensional simulation, 00:44:39.300 --> 00:44:43.210 align:middle line:84% that could save you a lot of time. 00:44:43.210 --> 00:44:47.250 align:middle line:84% Here's that same simulation here on slide 24. 00:44:47.250 --> 00:44:51.970 align:middle line:84% But this time, we used the fully coupled model. 00:44:51.970 --> 00:44:55.770 align:middle line:84% So when we did the simulation, instead of methodpd dot Fermi, 00:44:55.770 --> 00:44:58.470 align:middle line:90% we said methodpd dot full. 00:44:58.470 --> 00:45:01.050 align:middle line:84% And the simulation is the black line. 00:45:01.050 --> 00:45:02.760 align:middle line:84% It agrees still very well with the data, 00:45:02.760 --> 00:45:03.870 align:middle line:90% maybe not quite as well. 00:45:03.870 --> 00:45:08.220 align:middle line:84% But to within all the experimental uncertainties, 00:45:08.220 --> 00:45:10.090 align:middle line:90% it really does a good job. 00:45:10.090 --> 00:45:13.350 align:middle line:84% So the-fully coupled model doesn't make much difference-- 00:45:13.350 --> 00:45:16.590 align:middle line:84% that should say difference, not different, difference-- 00:45:16.590 --> 00:45:18.660 align:middle line:84% because 30 minutes, again, is much longer 00:45:18.660 --> 00:45:19.980 align:middle line:90% than the time scale for TED. 00:45:19.980 --> 00:45:21.720 align:middle line:84% At 1,000 degrees with this dose, I'm 00:45:21.720 --> 00:45:26.040 align:middle line:84% guessing TED time scale is probably less than a 00:45:26.040 --> 00:45:26.890 align:middle line:90% second or so. 00:45:26.890 --> 00:45:30.450 align:middle line:84% And in fact, if you go on to slide 25, 00:45:30.450 --> 00:45:32.910 align:middle line:84% we can even figure out roughly what it is. 00:45:32.910 --> 00:45:38.100 align:middle line:84% At 1,000 degrees, this was for 1e14 phosphorus. 00:45:38.100 --> 00:45:42.460 align:middle line:84% At 40 keV, TED is less than a second. 00:45:42.460 --> 00:45:47.130 align:middle line:84% So for 2.6e14, again, it's 2.6 times that, 00:45:47.130 --> 00:45:49.892 align:middle line:84% almost the same range, not much difference. 00:45:49.892 --> 00:45:51.850 align:middle line:84% So we're talking on the order of a few seconds. 00:45:51.850 --> 00:45:53.973 align:middle line:84% So a 30-minute anneal, clearly, TED 00:45:53.973 --> 00:45:55.390 align:middle line:84% is not going to be very important. 00:45:55.390 --> 00:45:56.880 align:middle line:84% That's why you get the same results. 00:45:56.880 --> 00:45:58.963 align:middle line:84% But this is a sanity check you can do with SUPREM. 00:45:58.963 --> 00:46:01.440 align:middle line:84% Make sure you get the same result for that 30-minute anneal 00:46:01.440 --> 00:46:04.140 align:middle line:90% with PD.Fermi and PD.full. 00:46:04.140 --> 00:46:06.240 align:middle line:84% If you don't, then there's something 00:46:06.240 --> 00:46:11.340 align:middle line:84% missing in your understanding of what SUPREM is doing. 00:46:11.340 --> 00:46:12.555 align:middle line:90% So there's a simple example. 00:46:12.555 --> 00:46:14.985 align:middle line:90% 00:46:14.985 --> 00:46:16.360 align:middle line:84% Now let's do something different. 00:46:16.360 --> 00:46:19.630 align:middle line:84% On slide 26, instead of 1,000 degrees for 30 minutes, 00:46:19.630 --> 00:46:23.710 align:middle line:84% I'm going to do ten seconds, 1,000 degrees for ten seconds. 00:46:23.710 --> 00:46:26.560 align:middle line:84% OK, now you're in a range where you can imagine there 00:46:26.560 --> 00:46:29.560 align:middle line:90% might be some TED effects. 00:46:29.560 --> 00:46:32.890 align:middle line:84% And this is the simulation we're getting. 00:46:32.890 --> 00:46:36.790 align:middle line:84% The as-implanted here is shown in the black. 00:46:36.790 --> 00:46:39.130 align:middle line:84% And I believe this was implanted with a Monte Carlo. 00:46:39.130 --> 00:46:41.200 align:middle line:84% In fact, you can tell it's Monte Carlo because, 00:46:41.200 --> 00:46:43.750 align:middle line:84% see all this jaggedness in the as-implanted? 00:46:43.750 --> 00:46:46.160 align:middle line:84% You'd never get that out of an analytic solution, right? 00:46:46.160 --> 00:46:48.160 align:middle line:84% The jaggedness comes from the statistical nature 00:46:48.160 --> 00:46:49.160 align:middle line:90% of Monte Carlo. 00:46:49.160 --> 00:46:52.700 align:middle line:84% So that's the Monte Carlo as-implanted in black. 00:46:52.700 --> 00:46:57.440 align:middle line:84% PD.Fermi, which does no TED, shows not very much motion 00:46:57.440 --> 00:47:01.040 align:middle line:84% of the arsenic for a 10-second RTA. 00:47:01.040 --> 00:47:04.970 align:middle line:84% PD.full you get a fair amount of broadening. 00:47:04.970 --> 00:47:07.490 align:middle line:84% So that's got the TED built into it. 00:47:07.490 --> 00:47:10.820 align:middle line:84% Now, exactly which one is more accurate, I really can't say. 00:47:10.820 --> 00:47:12.710 align:middle line:84% In fact, my experience with arsenic 00:47:12.710 --> 00:47:15.200 align:middle line:84% has been on our rapid thermal anneals, 00:47:15.200 --> 00:47:16.940 align:middle line:84% that SUPREM tends to overestimate 00:47:16.940 --> 00:47:18.410 align:middle line:84% a little bit the amount of motion 00:47:18.410 --> 00:47:20.900 align:middle line:84% compared to what we see by SIMS, if you actually 00:47:20.900 --> 00:47:24.087 align:middle line:84% do a SIMS profile of 1,000 degree 10 second. 00:47:24.087 --> 00:47:25.670 align:middle line:84% Of course, there's always the question 00:47:25.670 --> 00:47:27.590 align:middle line:84% of how well is your RTA calibrated. 00:47:27.590 --> 00:47:29.630 align:middle line:84% Rapid thermal annealing machines are extremely 00:47:29.630 --> 00:47:31.787 align:middle line:90% difficult to calibrate. 00:47:31.787 --> 00:47:33.620 align:middle line:84% They are nonequilibrium environments, right? 00:47:33.620 --> 00:47:35.100 align:middle line:84% It's not like a hot furnace where everything's 00:47:35.100 --> 00:47:35.990 align:middle line:90% the same temperature. 00:47:35.990 --> 00:47:38.090 align:middle line:84% The wafer is the only thing that gets hot. 00:47:38.090 --> 00:47:40.490 align:middle line:84% And measuring its temperature is kind of an art. 00:47:40.490 --> 00:47:41.870 align:middle line:90% So it's hard to say. 00:47:41.870 --> 00:47:44.720 align:middle line:84% So you have to be careful, and typically 00:47:44.720 --> 00:47:47.720 align:middle line:84% when you're doing rapid thermal annealing and simulating TED, 00:47:47.720 --> 00:47:50.510 align:middle line:84% you want to compare it to a couple of experimental results 00:47:50.510 --> 00:47:52.190 align:middle line:84% just to make sure that it makes sense. 00:47:52.190 --> 00:47:54.090 align:middle line:84% Now, I can make this PD.full look 00:47:54.090 --> 00:47:57.780 align:middle line:84% like whatever I want by changing some of the internal parameters 00:47:57.780 --> 00:48:00.120 align:middle line:90% in the SUPREM-IV simulator. 00:48:00.120 --> 00:48:08.010 align:middle line:84% If I change things like the diffusivity of the interstitials 00:48:08.010 --> 00:48:11.350 align:middle line:84% and vacancies, I can change what this profile looks like. 00:48:11.350 --> 00:48:13.110 align:middle line:90% So you have some latitude there. 00:48:13.110 --> 00:48:17.730 align:middle line:84% This is using default parameters that are built into SUPREM-IV. 00:48:17.730 --> 00:48:20.820 align:middle line:90% 00:48:20.820 --> 00:48:25.290 align:middle line:84% Let me go on slide 27 to another related example, where 00:48:25.290 --> 00:48:26.700 align:middle line:90% we're going to talk about how-- 00:48:26.700 --> 00:48:29.250 align:middle line:84% we talked last time about some clever experiments 00:48:29.250 --> 00:48:32.042 align:middle line:84% that have been done at the IEDM, where the order of the anneals-- 00:48:32.042 --> 00:48:33.750 align:middle line:84% we have the exact same anneals, but we're 00:48:33.750 --> 00:48:35.208 align:middle line:84% going to reverse the order which we 00:48:35.208 --> 00:48:39.270 align:middle line:84% do them makes a difference because that's going to dissolve 00:48:39.270 --> 00:48:41.920 align:middle line:90% a lot of 311 defects. 00:48:41.920 --> 00:48:44.340 align:middle line:84% We're going to show how SUPREM can actually simulate that. 00:48:44.340 --> 00:48:46.980 align:middle line:84% And first, I'm going to just remind you 00:48:46.980 --> 00:48:50.670 align:middle line:84% what the usual order of making a MOSFET is just from last time. 00:48:50.670 --> 00:48:52.110 align:middle line:90% Usually, you start with a wafer. 00:48:52.110 --> 00:48:53.220 align:middle line:90% You do your isolation. 00:48:53.220 --> 00:48:56.040 align:middle line:84% This could be your shallow trench isolation 00:48:56.040 --> 00:48:57.900 align:middle line:90% we talk about as oxide. 00:48:57.900 --> 00:48:59.400 align:middle line:84% Oop, that didn't come out very well. 00:48:59.400 --> 00:49:04.650 align:middle line:84% I'm going to implant the boron profile. 00:49:04.650 --> 00:49:05.850 align:middle line:90% This is an n MOSFET. 00:49:05.850 --> 00:49:11.880 align:middle line:84% I'm going to implant that early on the super steep retrograde. 00:49:11.880 --> 00:49:16.540 align:middle line:84% Grow the gate oxide, so we're growing the gate oxide. 00:49:16.540 --> 00:49:21.235 align:middle line:84% Form the polysilicon gate by deposition and etching. 00:49:21.235 --> 00:49:24.960 align:middle line:90% 00:49:24.960 --> 00:49:27.540 align:middle line:84% The source drain extensions are now implanted. 00:49:27.540 --> 00:49:30.240 align:middle line:84% Now, the source drain extensions bring with them-- remember, 00:49:30.240 --> 00:49:33.540 align:middle line:84% we talked about the reverse short channel effect. 00:49:33.540 --> 00:49:37.680 align:middle line:84% They're going to introduce a certain number of 311 defects 00:49:37.680 --> 00:49:40.140 align:middle line:84% that are going to cause TED of the boron that's 00:49:40.140 --> 00:49:43.050 align:middle line:90% underneath here. 00:49:43.050 --> 00:49:44.370 align:middle line:90% Put in the spacers. 00:49:44.370 --> 00:49:46.068 align:middle line:84% Now, the spacers, if they're nitride-- 00:49:46.068 --> 00:49:48.360 align:middle line:84% a lot of people today are using nitride and not oxide-- 00:49:48.360 --> 00:49:51.270 align:middle line:84% if they're nitride and they're LPCVD nitride, that's 00:49:51.270 --> 00:49:54.060 align:middle line:84% a pretty high-temperature process, about 800 degrees, one 00:49:54.060 --> 00:49:57.600 align:middle line:84% of the worst temperatures you can possibly use for TED. 00:49:57.600 --> 00:49:58.380 align:middle line:90% Why is that? 00:49:58.380 --> 00:50:01.230 align:middle line:84% Well, we saw that the time that TED lasts at 800, 00:50:01.230 --> 00:50:04.530 align:middle line:90% it can be quite long. 00:50:04.530 --> 00:50:07.660 align:middle line:84% And CI over CI star can be quite large. 00:50:07.660 --> 00:50:10.020 align:middle line:84% So an hour at 800 can cause a lot 00:50:10.020 --> 00:50:12.120 align:middle line:90% of transient-enhanced diffusion. 00:50:12.120 --> 00:50:15.300 align:middle line:84% And then after that, typically after you make the spacers, 00:50:15.300 --> 00:50:18.210 align:middle line:84% you do this deep source drain implant here 00:50:18.210 --> 00:50:19.900 align:middle line:90% for the contact regions. 00:50:19.900 --> 00:50:22.350 align:middle line:84% And then you do a final rapid thermal anneal, 00:50:22.350 --> 00:50:26.760 align:middle line:84% usually 1,000 to 1,050, maybe 10 to 30 seconds, something 00:50:26.760 --> 00:50:27.610 align:middle line:90% in that range. 00:50:27.610 --> 00:50:31.590 align:middle line:84% So the important thermal steps are, this nitride at 900, 00:50:31.590 --> 00:50:34.230 align:middle line:84% this nitride thermal budget at 800, 00:50:34.230 --> 00:50:36.990 align:middle line:84% and then a rapid thermal anneal at around 1,000. 00:50:36.990 --> 00:50:40.020 align:middle line:84% And now we're going to compare with SUPREM, how SUPREM thinks 00:50:40.020 --> 00:50:45.660 align:middle line:90% in 1D sense this would go. 00:50:45.660 --> 00:50:49.500 align:middle line:84% So we're going to just look at the effect of TED 00:50:49.500 --> 00:50:51.510 align:middle line:84% just on the arsenic diffusion itself. 00:50:51.510 --> 00:50:54.360 align:middle line:84% We're not going to look at the effect on the boron diffusivity. 00:50:54.360 --> 00:50:56.710 align:middle line:84% That would require a two-dimensional simulation. 00:50:56.710 --> 00:50:59.670 align:middle line:84% So this is a simulation of an arsenic source drain extension 00:50:59.670 --> 00:51:03.810 align:middle line:84% implant and diffusing it using the usual order, which 00:51:03.810 --> 00:51:08.790 align:middle line:84% is 800 degrees C step for the nitride deposition, followed 00:51:08.790 --> 00:51:13.650 align:middle line:84% by a deep source drain, and then followed by 1,000 degrees 00:51:13.650 --> 00:51:17.220 align:middle line:84% C 5-second RTA to activate everything. 00:51:17.220 --> 00:51:19.430 align:middle line:84% So here's the arsenic Monte Carlo implant. 00:51:19.430 --> 00:51:21.800 align:middle line:90% And you can see on this scale-- 00:51:21.800 --> 00:51:23.810 align:middle line:84% this is the distance in microns and the-- 00:51:23.810 --> 00:51:27.200 align:middle line:84% they didn't use a very fine grid. 00:51:27.200 --> 00:51:31.280 align:middle line:84% Good enough to get this profile, but for the as-implanted, 00:51:31.280 --> 00:51:32.870 align:middle line:84% you can see it's kind of ugly looking. 00:51:32.870 --> 00:51:34.970 align:middle line:84% And that's part of the reason the grid was 00:51:34.970 --> 00:51:35.990 align:middle line:90% a little bit too coarse. 00:51:35.990 --> 00:51:37.820 align:middle line:84% But again, with Monte Carlo, we were 00:51:37.820 --> 00:51:40.250 align:middle line:84% trying to speed up the processing a little. 00:51:40.250 --> 00:51:46.190 align:middle line:84% So here's 2 keV, very shallow peaks, at 1e15. 00:51:46.190 --> 00:51:52.480 align:middle line:84% And then the blue line is 800 degrees, 30 minutes anneal. 00:51:52.480 --> 00:51:54.970 align:middle line:84% And then the red line is 800 degrees, 30 minutes, 00:51:54.970 --> 00:51:57.190 align:middle line:84% followed by a rapid thermal anneal at 1,000 00:51:57.190 --> 00:52:00.470 align:middle line:84% for five seconds, so 1D simulation. 00:52:00.470 --> 00:52:03.100 align:middle line:84% So almost all the diffusion really 00:52:03.100 --> 00:52:06.230 align:middle line:84% takes place at the 800 degree 30 minutes. 00:52:06.230 --> 00:52:07.480 align:middle line:90% And that's because of TED. 00:52:07.480 --> 00:52:11.875 align:middle line:84% That's not ordinary 800 degrees C. So in order to model this, 00:52:11.875 --> 00:52:14.020 align:middle line:90% this must be a PD.full. 00:52:14.020 --> 00:52:15.580 align:middle line:84% So it must be taking into account 00:52:15.580 --> 00:52:18.377 align:middle line:90% the transit-enhanced diffusion. 00:52:18.377 --> 00:52:19.210 align:middle line:90% Well, this is trans. 00:52:19.210 --> 00:52:19.710 align:middle line:90% I'm sorry. 00:52:19.710 --> 00:52:20.475 align:middle line:90% This is PD.trans. 00:52:20.475 --> 00:52:21.850 align:middle line:84% You can either use trans or full. 00:52:21.850 --> 00:52:24.923 align:middle line:90% This particular one is PD.trans. 00:52:24.923 --> 00:52:27.340 align:middle line:84% OK, so that's what it looks like, very little contribution 00:52:27.340 --> 00:52:28.240 align:middle line:90% of the RTA. 00:52:28.240 --> 00:52:30.850 align:middle line:84% And your junction depth here is about 0.1 micron. 00:52:30.850 --> 00:52:33.640 align:middle line:90% That's what this predicts. 00:52:33.640 --> 00:52:34.960 align:middle line:90% So why is that? 00:52:34.960 --> 00:52:38.470 align:middle line:84% If you look on slide 29, this is a plot-- 00:52:38.470 --> 00:52:40.000 align:middle line:84% SUPREM, in addition to the dopants, 00:52:40.000 --> 00:52:42.830 align:middle line:84% you can output the interstitials and vacancies. 00:52:42.830 --> 00:52:47.170 align:middle line:84% So here's a normalized concentration versus depth. 00:52:47.170 --> 00:52:51.710 align:middle line:84% And the blue line here is for CI over CI star at 800 degrees. 00:52:51.710 --> 00:52:57.820 align:middle line:84% And so you can see, CI over CI star after this anneal, 00:52:57.820 --> 00:53:00.440 align:middle line:84% at the 800-degree anneal, that is, it's pretty large. 00:53:00.440 --> 00:53:03.200 align:middle line:84% It's somewhere between 20 and 40 and near the surface. 00:53:03.200 --> 00:53:05.345 align:middle line:90% It's pretty big. 00:53:05.345 --> 00:53:06.220 align:middle line:90% That's the blue line. 00:53:06.220 --> 00:53:08.137 align:middle line:84% The red line, which I apologize you can't see. 00:53:08.137 --> 00:53:11.200 align:middle line:84% You have to write it on your-- it's the same as your x-axis, 00:53:11.200 --> 00:53:11.830 align:middle line:90% basically. 00:53:11.830 --> 00:53:13.472 align:middle line:90% The red line is just about one. 00:53:13.472 --> 00:53:14.930 align:middle line:84% You can't even see it on the scale. 00:53:14.930 --> 00:53:18.770 align:middle line:84% So CI over CI star at 1,000 degrees, 00:53:18.770 --> 00:53:22.040 align:middle line:84% after five seconds is one, which there isn't much enhancement 00:53:22.040 --> 00:53:22.610 align:middle line:90% left. 00:53:22.610 --> 00:53:23.480 align:middle line:90% And we saw that. 00:53:23.480 --> 00:53:27.560 align:middle line:84% Remember, the enhancement time, the amount of time that it lasts 00:53:27.560 --> 00:53:28.590 align:middle line:90% is relatively short. 00:53:28.590 --> 00:53:30.920 align:middle line:90% So this is after five seconds. 00:53:30.920 --> 00:53:32.420 align:middle line:90% So TED is already-- 00:53:32.420 --> 00:53:35.533 align:middle line:84% the interstitial concentration-- the 311s have all dissolved. 00:53:35.533 --> 00:53:36.950 align:middle line:84% And the interstitial concentration 00:53:36.950 --> 00:53:39.290 align:middle line:90% has gone back to CI star. 00:53:39.290 --> 00:53:41.480 align:middle line:84% So after five seconds, we still have 00:53:41.480 --> 00:53:43.730 align:middle line:84% a lot of enhancement at 800 degrees, 00:53:43.730 --> 00:53:49.960 align:middle line:84% and that's why we get all that TED in the 800-degree C profile. 00:53:49.960 --> 00:53:52.050 align:middle line:84% So now, on slide 30, we're going to do 00:53:52.050 --> 00:53:54.280 align:middle line:90% something a little different. 00:53:54.280 --> 00:53:57.390 align:middle line:84% We're going to activate the source drain extension 00:53:57.390 --> 00:54:01.140 align:middle line:84% implant prior to the nitride spacer at 800. 00:54:01.140 --> 00:54:02.880 align:middle line:84% So what we're going to do instead 00:54:02.880 --> 00:54:05.370 align:middle line:84% of doing the usual put the nitride down right 00:54:05.370 --> 00:54:07.830 align:middle line:84% after implanting, we're going to do 00:54:07.830 --> 00:54:11.470 align:middle line:84% a 5-second rapid thermal anneal after implanting. 00:54:11.470 --> 00:54:12.810 align:middle line:90% So we send it out for implant. 00:54:12.810 --> 00:54:13.770 align:middle line:90% We bring it back. 00:54:13.770 --> 00:54:19.410 align:middle line:84% We do a 5-second RTA first, and then we do another five 00:54:19.410 --> 00:54:20.220 align:middle line:90% seconds-- 00:54:20.220 --> 00:54:23.100 align:middle line:84% and then we do 800 degrees plus 30 minutes. 00:54:23.100 --> 00:54:24.160 align:middle line:90% We do that second. 00:54:24.160 --> 00:54:25.710 align:middle line:90% So we're changing the order. 00:54:25.710 --> 00:54:28.650 align:middle line:84% And you can see the red line now is 00:54:28.650 --> 00:54:34.630 align:middle line:84% after the implant plus 5 seconds at 1,000, you diffuse this far. 00:54:34.630 --> 00:54:37.740 align:middle line:84% And then if you add the 830 minutes, it only goes that far. 00:54:37.740 --> 00:54:40.590 align:middle line:84% It doesn't go very far because five seconds is 00:54:40.590 --> 00:54:43.710 align:middle line:84% 1,000 is really enough to dissolve all the 311s, right? 00:54:43.710 --> 00:54:48.790 align:middle line:84% The enhancement-- the TED time at 1,000 degrees 00:54:48.790 --> 00:54:50.360 align:middle line:90% is only a couple of seconds. 00:54:50.360 --> 00:54:51.880 align:middle line:90% So I dissolve all the 311s. 00:54:51.880 --> 00:54:54.010 align:middle line:84% Then you put it back in the furnace at 800, 00:54:54.010 --> 00:54:55.120 align:middle line:90% and you don't get-- 00:54:55.120 --> 00:54:57.100 align:middle line:84% this is just normal 800 degrees C diffusion. 00:54:57.100 --> 00:55:00.580 align:middle line:84% There's not many CI over CI*s now been reduced because I got 00:55:00.580 --> 00:55:01.870 align:middle line:90% rid of all those 311s. 00:55:01.870 --> 00:55:05.380 align:middle line:84% So the junction depth now, here, instead of being out here 00:55:05.380 --> 00:55:08.540 align:middle line:90% at 0.1, is only 0.07. 00:55:08.540 --> 00:55:13.460 align:middle line:84% So the exact same amount of time that the wafer spent in the RTA 00:55:13.460 --> 00:55:16.310 align:middle line:84% and in the furnace is just that the order of the operations 00:55:16.310 --> 00:55:18.240 align:middle line:90% was changed. 00:55:18.240 --> 00:55:20.400 align:middle line:84% And this is reduced, the junction depth is reduced 00:55:20.400 --> 00:55:24.420 align:middle line:84% because CI over CI star is reduced by about 3x during 00:55:24.420 --> 00:55:26.650 align:middle line:90% the 1,000-degree C step. 00:55:26.650 --> 00:55:29.220 align:middle line:84% So this is something-- these are all simulations. 00:55:29.220 --> 00:55:33.000 align:middle line:84% SUPREM can simulate the fact that 311s are generated 00:55:33.000 --> 00:55:35.373 align:middle line:84% by the damage and that they dissolve 00:55:35.373 --> 00:55:37.290 align:middle line:84% at different temperatures, at different rates. 00:55:37.290 --> 00:55:41.040 align:middle line:84% And then so it can take into account these types of effects. 00:55:41.040 --> 00:55:43.140 align:middle line:84% Again, it's using the trans model. 00:55:43.140 --> 00:55:44.772 align:middle line:84% The accuracy, of course, you always 00:55:44.772 --> 00:55:46.980 align:middle line:84% take with a grain of salt, whether the junction depth 00:55:46.980 --> 00:55:50.810 align:middle line:84% is really 0.06, 0.04, I wouldn't put my life on it, 00:55:50.810 --> 00:55:53.310 align:middle line:84% to be honest, because there's a lot of parameters in SUPREM. 00:55:53.310 --> 00:55:56.250 align:middle line:84% But the point is, you can fit your data-- assuming 00:55:56.250 --> 00:55:58.110 align:middle line:84% you get some data, you can fit data 00:55:58.110 --> 00:56:01.350 align:middle line:84% that shows the effect of the order of the different implants. 00:56:01.350 --> 00:56:02.960 align:middle line:84% There's enough physics built into it. 00:56:02.960 --> 00:56:04.710 align:middle line:84% You just have to get the right parameters. 00:56:04.710 --> 00:56:07.790 align:middle line:90% 00:56:07.790 --> 00:56:12.830 align:middle line:84% So the final process, just to show that what we did 00:56:12.830 --> 00:56:14.690 align:middle line:84% is the extension implant followed 00:56:14.690 --> 00:56:18.710 align:middle line:84% by 5-second RTA and 1,000 high-temperature nitride spacer 00:56:18.710 --> 00:56:21.020 align:middle line:84% at 800 and then another 5-second RTA. 00:56:21.020 --> 00:56:23.030 align:middle line:84% The reason I needed to add the last RTA 00:56:23.030 --> 00:56:25.040 align:middle line:84% is to activate the deep source drain. 00:56:25.040 --> 00:56:27.080 align:middle line:84% Remember, the deep source drain goes in last 00:56:27.080 --> 00:56:28.490 align:middle line:90% after the nitride spacer. 00:56:28.490 --> 00:56:32.900 align:middle line:84% So this is five seconds at 1,800, five seconds at 1,000. 00:56:32.900 --> 00:56:34.580 align:middle line:90% This is what it would look like. 00:56:34.580 --> 00:56:36.410 align:middle line:84% The final five seconds at 1,000 really 00:56:36.410 --> 00:56:38.600 align:middle line:84% doesn't contribute much to further motion. 00:56:38.600 --> 00:56:41.210 align:middle line:84% It's already done most of its motion 00:56:41.210 --> 00:56:44.930 align:middle line:84% because TED, at that point, is completely over. 00:56:44.930 --> 00:56:48.630 align:middle line:84% And if you've done it the usual order, 800 degrees, 00:56:48.630 --> 00:56:51.500 align:middle line:84% 30-minutes spacer, plus 5 seconds to anneal everything, 00:56:51.500 --> 00:56:53.070 align:middle line:90% you get this junction depth. 00:56:53.070 --> 00:56:57.947 align:middle line:84% So this the SUPREM predicts that reversing 00:56:57.947 --> 00:56:59.780 align:middle line:84% the order of the anneals makes a difference. 00:56:59.780 --> 00:57:02.510 align:middle line:84% Again, it's something you'd have to check experimentally just 00:57:02.510 --> 00:57:03.020 align:middle line:90% to be sure. 00:57:03.020 --> 00:57:05.910 align:middle line:90% 00:57:05.910 --> 00:57:07.890 align:middle line:90% OK, so that was a TED example. 00:57:07.890 --> 00:57:10.230 align:middle line:84% And then for the fourth example, this 00:57:10.230 --> 00:57:15.060 align:middle line:84% will be the only example I'll do where we do two-dimensional, 00:57:15.060 --> 00:57:16.590 align:middle line:90% a two-dimensional simulation. 00:57:16.590 --> 00:57:18.870 align:middle line:84% And this I took this right from the SUPREM manual. 00:57:18.870 --> 00:57:20.250 align:middle line:90% So you can run this yourself. 00:57:20.250 --> 00:57:22.333 align:middle line:84% This particular example is not one that I made up. 00:57:22.333 --> 00:57:24.513 align:middle line:84% I took it out of the example file, 00:57:24.513 --> 00:57:25.680 align:middle line:90% and this is on the computer. 00:57:25.680 --> 00:57:28.980 align:middle line:84% This example is one of the canned examples 00:57:28.980 --> 00:57:31.250 align:middle line:90% that comes with SUPREM-IV. 00:57:31.250 --> 00:57:33.910 align:middle line:84% It's a 200-nanometer gate length, 00:57:33.910 --> 00:57:36.660 align:middle line:84% so that's 0.2 micron gate length and MOSFET. 00:57:36.660 --> 00:57:38.280 align:middle line:90% It's kind of old fashioned now. 00:57:38.280 --> 00:57:40.435 align:middle line:84% But it uses something called self-aligned silicide, 00:57:40.435 --> 00:57:42.060 align:middle line:84% and we're going to talk about silicides 00:57:42.060 --> 00:57:42.990 align:middle line:90% in the next few lectures. 00:57:42.990 --> 00:57:43.907 align:middle line:90% So you'll get an idea. 00:57:43.907 --> 00:57:47.370 align:middle line:84% And I've put the [? input ?] file and columns. 00:57:47.370 --> 00:57:49.493 align:middle line:84% So this is the first part of the file, 00:57:49.493 --> 00:57:50.660 align:middle line:90% and this is the second part. 00:57:50.660 --> 00:57:52.660 align:middle line:84% Again, if you want to see the [? actual ?] file, 00:57:52.660 --> 00:57:56.480 align:middle line:90% it's in the SUPREM-IV directory. 00:57:56.480 --> 00:57:58.220 align:middle line:90% So this is TSUPREM-4. 00:57:58.220 --> 00:58:01.730 align:middle line:84% So example, here's the mesh using a fairly coarse grid, 00:58:01.730 --> 00:58:05.090 align:middle line:90% grid dot fac of 0.9. 00:58:05.090 --> 00:58:08.840 align:middle line:84% You can also define the mesh in different directions, 00:58:08.840 --> 00:58:12.090 align:middle line:84% in the y direction, in the x direction, 00:58:12.090 --> 00:58:16.790 align:middle line:84% so a little more sophisticated definition of the grid. 00:58:16.790 --> 00:58:19.250 align:middle line:84% We start out by growing the gate oxide. 00:58:19.250 --> 00:58:21.800 align:middle line:90% So here's 850, 25 minutes. 00:58:21.800 --> 00:58:25.070 align:middle line:84% And it's in dry O2 plus a little bit of HCl. 00:58:25.070 --> 00:58:28.130 align:middle line:90% So there's an HCl oxidation. 00:58:28.130 --> 00:58:29.630 align:middle line:84% And then you can plot the structure 00:58:29.630 --> 00:58:32.960 align:middle line:84% to see what it looks like after gate oxidation. 00:58:32.960 --> 00:58:37.400 align:middle line:84% And this says, whenever you see the word source, what that says 00:58:37.400 --> 00:58:38.070 align:middle line:90% is-- 00:58:38.070 --> 00:58:39.590 align:middle line:84% and then you have the file name that 00:58:39.590 --> 00:58:43.160 align:middle line:84% says, take all the commands from that file name and run them now. 00:58:43.160 --> 00:58:46.325 align:middle line:90% So TSUPREM-IV, in order to-- 00:58:46.325 --> 00:58:48.200 align:middle line:84% let's say you do the same operation in SUPREM 00:58:48.200 --> 00:58:49.100 align:middle line:90% over and over again. 00:58:49.100 --> 00:58:50.180 align:middle line:90% You always do a plot. 00:58:50.180 --> 00:58:52.070 align:middle line:84% You want to do several different plots. 00:58:52.070 --> 00:58:55.010 align:middle line:84% Rather than putting all those plot commands in the main file 00:58:55.010 --> 00:58:57.800 align:middle line:84% over and over again, you create sub files. 00:58:57.800 --> 00:59:03.230 align:middle line:84% And this file called S4EX10P.input is just 00:59:03.230 --> 00:59:05.240 align:middle line:84% a series of commands that defines the colors 00:59:05.240 --> 00:59:06.390 align:middle line:90% and things like that. 00:59:06.390 --> 00:59:08.810 align:middle line:84% So it enables you to call this file whenever 00:59:08.810 --> 00:59:11.300 align:middle line:90% you want it to create a plot. 00:59:11.300 --> 00:59:14.370 align:middle line:84% So it's a way of cleaning up your SUPREM input file. 00:59:14.370 --> 00:59:16.620 align:middle line:84% So if you want to know what this particular file does, 00:59:16.620 --> 00:59:19.650 align:middle line:84% you have to go look at that command file. 00:59:19.650 --> 00:59:22.740 align:middle line:84% Then you deposit some materials or depositing polysilicon. 00:59:22.740 --> 00:59:24.090 align:middle line:90% You tell it the thickness. 00:59:24.090 --> 00:59:26.890 align:middle line:84% You're etching the polysilicon to the right and to the left. 00:59:26.890 --> 00:59:27.855 align:middle line:90% This is to make a gate. 00:59:27.855 --> 00:59:30.500 align:middle line:90% 00:59:30.500 --> 00:59:31.980 align:middle line:84% And here you're calling that file, 00:59:31.980 --> 00:59:35.160 align:middle line:90% again, that plotting file, OK? 00:59:35.160 --> 00:59:38.370 align:middle line:84% Now, one of your homework problems that you just handed in 00:59:38.370 --> 00:59:39.870 align:middle line:90% was on-- 00:59:39.870 --> 00:59:41.880 align:middle line:84% or we just handed back to you, homework 3, 00:59:41.880 --> 00:59:42.960 align:middle line:90% was different methods. 00:59:42.960 --> 00:59:45.420 align:middle line:84% This is using the compressed method 00:59:45.420 --> 00:59:48.570 align:middle line:84% as far as solving for the oxidation rate goes. 00:59:48.570 --> 00:59:52.930 align:middle line:84% Here's 850 oxidation in dry O2 and HCl again. 00:59:52.930 --> 00:59:54.930 align:middle line:90% So this is what we call re-ox. 00:59:54.930 --> 00:59:59.160 align:middle line:90% So you formed the gate now. 00:59:59.160 --> 01:00:01.290 align:middle line:84% So the initial gate oxidation is just 01:00:01.290 --> 01:00:03.090 align:middle line:90% to grow a very thin gate oxide. 01:00:03.090 --> 01:00:06.750 align:middle line:84% You formed the gate, and you etched it. 01:00:06.750 --> 01:00:08.370 align:middle line:84% And then, in the course of etching, 01:00:08.370 --> 01:00:10.252 align:middle line:84% sometimes you introduce damage right 01:00:10.252 --> 01:00:12.460 align:middle line:84% near the corners of where the channel is going to be. 01:00:12.460 --> 01:00:12.660 align:middle line:90% Remember? 01:00:12.660 --> 01:00:13.952 align:middle line:90% This is going to be my channel. 01:00:13.952 --> 01:00:17.760 align:middle line:84% So you often introduce some plasma damage down here. 01:00:17.760 --> 01:00:19.230 align:middle line:90% You SiN up the oxide there. 01:00:19.230 --> 01:00:21.600 align:middle line:84% You do things that are not necessarily good for the gate 01:00:21.600 --> 01:00:22.100 align:middle line:90% oxide. 01:00:22.100 --> 01:00:26.220 align:middle line:84% So people often, at this point, will, in this technology, 01:00:26.220 --> 01:00:29.100 align:middle line:84% say, 0.18 micron technology, would take the wafer 01:00:29.100 --> 01:00:31.560 align:middle line:84% at this point, put it back in the oxidation furnace, 01:00:31.560 --> 01:00:33.660 align:middle line:84% and do what's called a gate re-ox. 01:00:33.660 --> 01:00:39.237 align:middle line:84% And this looks like we're doing a re-ox here at 850 for 25 01:00:39.237 --> 01:00:39.820 align:middle line:90% [? minutes. ?] 01:00:39.820 --> 01:00:42.900 align:middle line:84% So that re-ox is going to help-- if you thinned the oxide at all 01:00:42.900 --> 01:00:44.400 align:middle line:84% in the course of your etching, it's 01:00:44.400 --> 01:00:47.190 align:middle line:84% going to help boost that back up. 01:00:47.190 --> 01:00:52.800 align:middle line:84% And it's going to grow a little oxide along the edge here 01:00:52.800 --> 01:00:54.180 align:middle line:90% of the polysilicon. 01:00:54.180 --> 01:00:56.940 align:middle line:84% So it's a way of dealing-- it was introduced 01:00:56.940 --> 01:00:59.460 align:middle line:90% for reliability considerations. 01:00:59.460 --> 01:01:01.810 align:middle line:84% Re-ex, as it says, take the gate, 01:01:01.810 --> 01:01:05.790 align:middle line:84% put it back in, and do another oxidation. 01:01:05.790 --> 01:01:08.070 align:middle line:84% Then you form the sidewall spacer. 01:01:08.070 --> 01:01:10.020 align:middle line:90% Here, we're depositing oxide. 01:01:10.020 --> 01:01:14.730 align:middle line:84% It's using an oxidation-- a deposition process. 01:01:14.730 --> 01:01:18.060 align:middle line:84% And then you do the deep source drain implant. 01:01:18.060 --> 01:01:23.438 align:middle line:84% This is 1e15 energy of 60 keVs, fairly deep. 01:01:23.438 --> 01:01:25.230 align:middle line:84% Then you're going to deposit some titanium. 01:01:25.230 --> 01:01:26.855 align:middle line:84% I won't go through the details of this. 01:01:26.855 --> 01:01:30.018 align:middle line:84% But in the next few lectures, we'll talk about the fact that-- 01:01:30.018 --> 01:01:31.560 align:middle line:84% when we get to the silicide lecture-- 01:01:31.560 --> 01:01:33.985 align:middle line:84% you can take metals, put them on silicon, 01:01:33.985 --> 01:01:35.610 align:middle line:84% and react them at a certain temperature 01:01:35.610 --> 01:01:37.502 align:middle line:90% and form a metallic phase. 01:01:37.502 --> 01:01:38.460 align:middle line:90% It's called a silicide. 01:01:38.460 --> 01:01:42.570 align:middle line:84% This is titanium disilicide, which forms a good contact. 01:01:42.570 --> 01:01:45.220 align:middle line:84% And then we look at the final structure. 01:01:45.220 --> 01:01:47.470 align:middle line:84% So that's the file, and you can go ahead and run that. 01:01:47.470 --> 01:01:48.845 align:middle line:84% And let me show you some examples 01:01:48.845 --> 01:01:50.700 align:middle line:90% of what comes out of that. 01:01:50.700 --> 01:01:53.460 align:middle line:84% Oh, actually, this slide, number 33, 01:01:53.460 --> 01:01:55.770 align:middle line:84% remember, I was saying you wanted to repeat this over 01:01:55.770 --> 01:01:57.062 align:middle line:90% and over again, these commands? 01:01:57.062 --> 01:02:00.600 align:middle line:84% This was that file, S4EX10P.input. 01:02:00.600 --> 01:02:02.790 align:middle line:84% This is just a plotting sequence file 01:02:02.790 --> 01:02:07.260 align:middle line:84% that you use over and over again that tells it what colors to use 01:02:07.260 --> 01:02:10.810 align:middle line:84% and how to label things according to what material is. 01:02:10.810 --> 01:02:14.430 align:middle line:84% So rather than type this in every time in your SUPREM-IV, 01:02:14.430 --> 01:02:16.050 align:middle line:84% you can put it in a separate file. 01:02:16.050 --> 01:02:19.180 align:middle line:90% That's just for your reference. 01:02:19.180 --> 01:02:21.390 align:middle line:90% So let's go to slide 34. 01:02:21.390 --> 01:02:22.830 align:middle line:84% This is from running that example. 01:02:22.830 --> 01:02:24.705 align:middle line:84% After gate oxidation, what does it look like? 01:02:24.705 --> 01:02:27.810 align:middle line:84% Well, you have an xy structure, so y is in depth. 01:02:27.810 --> 01:02:30.750 align:middle line:84% And there's a certain gate oxide grown everywhere 01:02:30.750 --> 01:02:32.850 align:middle line:90% across on the structure. 01:02:32.850 --> 01:02:35.310 align:middle line:90% That's relatively simple. 01:02:35.310 --> 01:02:38.280 align:middle line:84% After gate patterning, you have deposited this green layer 01:02:38.280 --> 01:02:40.080 align:middle line:84% everywhere and then etched it off 01:02:40.080 --> 01:02:43.000 align:middle line:84% everywhere to the right of this one line. 01:02:43.000 --> 01:02:44.940 align:middle line:84% So now, that's what the polysilicon gate 01:02:44.940 --> 01:02:47.070 align:middle line:84% looks like with the gate oxide underneath it 01:02:47.070 --> 01:02:49.860 align:middle line:90% after gate patterning, OK? 01:02:49.860 --> 01:02:51.940 align:middle line:84% And notice, it doesn't model etching 01:02:51.940 --> 01:02:53.320 align:middle line:90% in any sophisticated way. 01:02:53.320 --> 01:02:55.150 align:middle line:84% You just tell it where to cut it off. 01:02:55.150 --> 01:02:58.150 align:middle line:84% So it's not modeling any shape effects of the etch or anything 01:02:58.150 --> 01:02:58.780 align:middle line:90% like that. 01:02:58.780 --> 01:03:00.575 align:middle line:84% That's not that type of a program. 01:03:00.575 --> 01:03:02.200 align:middle line:84% We'll talk more about that when we talk 01:03:02.200 --> 01:03:05.240 align:middle line:90% about how to model etching. 01:03:05.240 --> 01:03:06.952 align:middle line:84% After gate re-ox, remember, I said 01:03:06.952 --> 01:03:08.660 align:middle line:84% we're going to put it back in the furnace 01:03:08.660 --> 01:03:12.020 align:middle line:84% and subject it to a step of 850 for 25 minutes 01:03:12.020 --> 01:03:14.990 align:middle line:84% to oxidize all around and to beef up this oxide. 01:03:14.990 --> 01:03:18.320 align:middle line:84% But interestingly, look what it's done to the gate oxide. 01:03:18.320 --> 01:03:19.628 align:middle line:90% You see what it's done. 01:03:19.628 --> 01:03:21.170 align:middle line:84% There's like a bird's beaking effect. 01:03:21.170 --> 01:03:24.590 align:middle line:84% Some of the oxidant has diffused underneath the polysilicon 01:03:24.590 --> 01:03:25.520 align:middle line:90% and diffused it. 01:03:25.520 --> 01:03:27.950 align:middle line:84% And you've got a little thicker oxide now here 01:03:27.950 --> 01:03:29.810 align:middle line:84% on this part underneath the gate compared 01:03:29.810 --> 01:03:31.730 align:middle line:90% to the center of the gate. 01:03:31.730 --> 01:03:33.290 align:middle line:84% That's not necessarily a good thing 01:03:33.290 --> 01:03:35.062 align:middle line:84% if it's too thick here because what 01:03:35.062 --> 01:03:36.770 align:middle line:84% do we know about the gate oxide thickness 01:03:36.770 --> 01:03:39.450 align:middle line:84% determines the threshold voltage in the device. 01:03:39.450 --> 01:03:42.620 align:middle line:84% So if I have a thicker gate oxide here than here, 01:03:42.620 --> 01:03:49.255 align:middle line:84% I'm going to turn on my channel at different places differently. 01:03:49.255 --> 01:03:50.880 align:middle line:84% So that's not necessarily a good thing. 01:03:50.880 --> 01:03:54.340 align:middle line:90% We want a uniform gate oxide. 01:03:54.340 --> 01:03:59.050 align:middle line:84% Here, of that same structure now, after I've formed a spacer, 01:03:59.050 --> 01:04:00.850 align:middle line:90% here's the oxide spacer. 01:04:00.850 --> 01:04:04.980 align:middle line:84% And we did a source drain extension. 01:04:04.980 --> 01:04:05.480 align:middle line:90% I'm sorry. 01:04:05.480 --> 01:04:07.763 align:middle line:84% I first did the 5e13 source drain extension. 01:04:07.763 --> 01:04:09.430 align:middle line:84% Then you formed the spacer, and then you 01:04:09.430 --> 01:04:11.080 align:middle line:90% did the deep source drain. 01:04:11.080 --> 01:04:14.560 align:middle line:84% These contours, by the way, correspond 01:04:14.560 --> 01:04:16.693 align:middle line:90% to arsenic in the substrate. 01:04:16.693 --> 01:04:18.610 align:middle line:84% Each one is a different arsenic concentration. 01:04:18.610 --> 01:04:19.990 align:middle line:90% You can see how it's shaped. 01:04:19.990 --> 01:04:23.140 align:middle line:84% This region here corresponds to the extension, This. 01:04:23.140 --> 01:04:26.460 align:middle line:90% Region to the deep source drain. 01:04:26.460 --> 01:04:30.630 align:middle line:84% This particular simulation uses PD.Fermi, so the Fermi model, 01:04:30.630 --> 01:04:31.155 align:middle line:90% very simple. 01:04:31.155 --> 01:04:33.960 align:middle line:90% 01:04:33.960 --> 01:04:36.270 align:middle line:84% And then after silicidation, SUPREM doesn't necessarily 01:04:36.270 --> 01:04:37.853 align:middle line:84% go to a very good model of siliciding, 01:04:37.853 --> 01:04:39.330 align:middle line:90% but you see it's reacted. 01:04:39.330 --> 01:04:40.705 align:middle line:84% We'll talk in subsequent lectures 01:04:40.705 --> 01:04:42.247 align:middle line:84% about what silicidation is all about. 01:04:42.247 --> 01:04:43.980 align:middle line:84% I mostly want you to get a feel for how 01:04:43.980 --> 01:04:48.000 align:middle line:84% a two-dimensional simulation of such a structure actually looks. 01:04:48.000 --> 01:04:50.590 align:middle line:90% 01:04:50.590 --> 01:04:54.700 align:middle line:84% There's the final structure here on slide 39, 01:04:54.700 --> 01:04:57.040 align:middle line:84% showing the arsenic concentration contours. 01:04:57.040 --> 01:05:00.820 align:middle line:84% So here, this region in between is your channel. 01:05:00.820 --> 01:05:03.130 align:middle line:84% Here's the polysilicon gate in green, 01:05:03.130 --> 01:05:04.870 align:middle line:90% and this is your gate oxide. 01:05:04.870 --> 01:05:06.467 align:middle line:84% And because of the reoxidation, we 01:05:06.467 --> 01:05:08.050 align:middle line:84% have a little bit of the smile effect. 01:05:08.050 --> 01:05:10.030 align:middle line:84% And see the way the device seems to be smiling, 01:05:10.030 --> 01:05:11.290 align:middle line:90% has a little smile here? 01:05:11.290 --> 01:05:13.930 align:middle line:84% That's actually a very happy device. 01:05:13.930 --> 01:05:15.790 align:middle line:84% But it's not actually all that good. 01:05:15.790 --> 01:05:17.060 align:middle line:90% Smiling is not very good. 01:05:17.060 --> 01:05:19.870 align:middle line:84% In fact, you want a gate oxide that's perfectly flat, 01:05:19.870 --> 01:05:21.790 align:middle line:84% has the same concentration-- same thickness 01:05:21.790 --> 01:05:23.680 align:middle line:84% as much as possible all the way across it. 01:05:23.680 --> 01:05:26.470 align:middle line:84% It got unflat because we did the re-ox, 01:05:26.470 --> 01:05:29.050 align:middle line:84% and a little bit of oxidation took place from the corners. 01:05:29.050 --> 01:05:33.040 align:middle line:84% So these days, for the very shallow, the very shortest 01:05:33.040 --> 01:05:35.830 align:middle line:84% channel devices, reoxidation is not such a popular thing 01:05:35.830 --> 01:05:41.650 align:middle line:84% to do because it does create this nonuniformity in the oxide 01:05:41.650 --> 01:05:43.750 align:middle line:90% thickness. 01:05:43.750 --> 01:05:45.210 align:middle line:90% In fact, you can go back. 01:05:45.210 --> 01:05:49.810 align:middle line:84% What I did was I went back here on page 40 and said, oh, OK, 01:05:49.810 --> 01:05:52.290 align:middle line:84% is there some way we can deal with this and maybe make 01:05:52.290 --> 01:05:58.290 align:middle line:84% the re-ox have a little bit less effect on the-- 01:05:58.290 --> 01:06:02.400 align:middle line:84% a little less of this extra oxidation right at the corner? 01:06:02.400 --> 01:06:04.800 align:middle line:84% So this is really what we're talking about just 01:06:04.800 --> 01:06:06.802 align:middle line:84% to give you a zoom in on slide 40. 01:06:06.802 --> 01:06:08.760 align:middle line:84% Here, actually, here's a zoom in of what it is. 01:06:08.760 --> 01:06:13.020 align:middle line:84% This nonuniform oxide thickness took place under the gate. 01:06:13.020 --> 01:06:15.900 align:middle line:84% And the real origin is the lateral oxidation 01:06:15.900 --> 01:06:19.560 align:middle line:84% under the poly during the gate re-ox step. 01:06:19.560 --> 01:06:22.605 align:middle line:84% This is when the gate re-ox was done at 850 for 25 minutes. 01:06:22.605 --> 01:06:25.800 align:middle line:90% 01:06:25.800 --> 01:06:28.830 align:middle line:84% So in fact, from the very first lecture of class, 01:06:28.830 --> 01:06:31.337 align:middle line:84% this is a real TDM of a real device 01:06:31.337 --> 01:06:32.670 align:middle line:90% to show you this effect is real. 01:06:32.670 --> 01:06:35.460 align:middle line:84% See the way this device is smiling also? 01:06:35.460 --> 01:06:37.470 align:middle line:84% So SUPREM didn't just make that up. 01:06:37.470 --> 01:06:42.300 align:middle line:84% The oxidant did get under here and oxidize in this region. 01:06:42.300 --> 01:06:46.658 align:middle line:84% So the vt is going to be nonuniform across this device. 01:06:46.658 --> 01:06:48.450 align:middle line:84% So you have to actually watch out for this. 01:06:48.450 --> 01:06:50.460 align:middle line:84% This does really happen, just to show you 01:06:50.460 --> 01:06:52.988 align:middle line:84% that SUPREM is based in reality with respect to some 01:06:52.988 --> 01:06:53.655 align:middle line:90% of these things. 01:06:53.655 --> 01:06:58.030 align:middle line:90% 01:06:58.030 --> 01:07:05.680 align:middle line:84% On slide 43, actually did a little example 01:07:05.680 --> 01:07:11.150 align:middle line:84% of how we can change the amount of this effect that happens, 01:07:11.150 --> 01:07:13.130 align:middle line:84% the amount of this nonuniformity. 01:07:13.130 --> 01:07:15.130 align:middle line:84% It may not be very clear if I'm looking at this. 01:07:15.130 --> 01:07:17.680 align:middle line:84% But on the left-hand side, what was done was, 01:07:17.680 --> 01:07:26.830 align:middle line:84% when we did the etch, we etched all the way down, OK, 01:07:26.830 --> 01:07:29.170 align:middle line:90% and then did a re-ox. 01:07:29.170 --> 01:07:33.630 align:middle line:84% So on the left-hand side, you're etching like this. 01:07:33.630 --> 01:07:38.150 align:middle line:84% And here's your gate oxide like this. 01:07:38.150 --> 01:07:41.180 align:middle line:84% And we're basically etching in the example 01:07:41.180 --> 01:07:47.150 align:middle line:84% where we told SUPREM we act as if we etched all the way down. 01:07:47.150 --> 01:07:49.430 align:middle line:84% So the oxide-- the prior to re-ox. 01:07:49.430 --> 01:07:52.010 align:middle line:90% 01:07:52.010 --> 01:07:54.950 align:middle line:84% So when we put it in the furnace on the left-hand side-- oh, 01:07:54.950 --> 01:07:55.550 align:middle line:90% sorry-- 01:07:55.550 --> 01:07:58.310 align:middle line:84% when we put it in the furnace for oxidation, 01:07:58.310 --> 01:08:01.160 align:middle line:84% this oxide was gone at that corner. 01:08:01.160 --> 01:08:04.110 align:middle line:90% 01:08:04.110 --> 01:08:06.360 align:middle line:84% And that's what you end up with on the left-hand side. 01:08:06.360 --> 01:08:08.440 align:middle line:84% So you can imagine, when that oxide is gone, 01:08:08.440 --> 01:08:11.050 align:middle line:84% then you can get a fair amount of attack 01:08:11.050 --> 01:08:17.939 align:middle line:84% in here laterally of the oxidant getting underneath there. 01:08:17.939 --> 01:08:23.130 align:middle line:84% So what was done here on the right-hand side instead, 01:08:23.130 --> 01:08:25.649 align:middle line:84% 50 of the 70 angstroms were etched. 01:08:25.649 --> 01:08:28.410 align:middle line:84% So we left a lot of the oxide on there 01:08:28.410 --> 01:08:32.175 align:middle line:90% and then just oxidized it-- 01:08:32.175 --> 01:08:35.340 align:middle line:84% re-ox was done at a lower temperature, 825 for a shorter 01:08:35.340 --> 01:08:37.720 align:middle line:90% time, 12 minutes. 01:08:37.720 --> 01:08:39.810 align:middle line:84% So you get some of the benefits of the re-ox 01:08:39.810 --> 01:08:41.189 align:middle line:90% without some of that extra. 01:08:41.189 --> 01:08:45.390 align:middle line:84% So in this example, when we did the etching, 01:08:45.390 --> 01:08:47.760 align:middle line:90% it didn't remove it all. 01:08:47.760 --> 01:08:52.840 align:middle line:84% It looked like this when it went into the oxidation furnace. 01:08:52.840 --> 01:08:54.930 align:middle line:84% It looked like that and then got oxidized. 01:08:54.930 --> 01:08:57.420 align:middle line:84% So, by refining the etch process, 01:08:57.420 --> 01:08:59.861 align:middle line:90% we didn't etch all the way down. 01:08:59.861 --> 01:09:02.069 align:middle line:84% And if you look carefully, I think on the next slide, 01:09:02.069 --> 01:09:05.430 align:middle line:84% it'll be maybe a little more obvious. 01:09:05.430 --> 01:09:09.420 align:middle line:84% The lateral nonuniformity is quite a bit reduced. 01:09:09.420 --> 01:09:13.020 align:middle line:84% This oxide thickness is much more uniform going across here. 01:09:13.020 --> 01:09:17.580 align:middle line:84% There's still a little bit of it, but 825 for 12 minutes, 01:09:17.580 --> 01:09:19.770 align:middle line:84% there's a lot less of that lateral nonuniformity. 01:09:19.770 --> 01:09:22.000 align:middle line:84% So it's an example of a process-- 01:09:22.000 --> 01:09:26.290 align:middle line:84% you can use SUPREM-IV to optimize reasonably efficiently 01:09:26.290 --> 01:09:30.175 align:middle line:84% and reasonably accurately in this two-dimensional model. 01:09:30.175 --> 01:09:33.130 align:middle line:90% 01:09:33.130 --> 01:09:35.560 align:middle line:84% OK, so let me go on to summarize. 01:09:35.560 --> 01:09:39.350 align:middle line:84% The simulators, like SUPREM-IV, which, by the way, 01:09:39.350 --> 01:09:42.520 align:middle line:84% I didn't tell you, but SUPREM-IV was originally 01:09:42.520 --> 01:09:45.410 align:middle line:90% written at Stanford. 01:09:45.410 --> 01:09:47.689 align:middle line:84% And sometimes that's called SSuprem 01:09:47.689 --> 01:09:50.120 align:middle line:84% That's where it came out of originally in the 1980s. 01:09:50.120 --> 01:09:53.990 align:middle line:84% It was then commercialized in a small company called TMA. 01:09:53.990 --> 01:09:57.290 align:middle line:84% That company was bought out by another company called Avant. 01:09:57.290 --> 01:10:01.940 align:middle line:84% And Avant later sold the technology 01:10:01.940 --> 01:10:04.190 align:middle line:90% to a company called Synopsis. 01:10:04.190 --> 01:10:05.750 align:middle line:90% Synopsis is a big design house. 01:10:05.750 --> 01:10:09.500 align:middle line:84% They make a lot of CAD software for designing chips. 01:10:09.500 --> 01:10:11.660 align:middle line:84% But they also support TSUPREM-IV. 01:10:11.660 --> 01:10:15.080 align:middle line:84% So if you want to have questions about the simulator, 01:10:15.080 --> 01:10:18.290 align:middle line:84% you need to contact people at this company called Synopsis. 01:10:18.290 --> 01:10:21.740 align:middle line:84% There are other simulators out there. 01:10:21.740 --> 01:10:23.840 align:middle line:84% SUPREM-IV is one of the most popular, 01:10:23.840 --> 01:10:25.490 align:middle line:84% but there's another company called 01:10:25.490 --> 01:10:29.840 align:middle line:84% Silvaco, which makes a competing product, which is called ATHENA. 01:10:29.840 --> 01:10:32.618 align:middle line:84% In any case, these simulators like this, 01:10:32.618 --> 01:10:34.160 align:middle line:84% they've been developed over the years 01:10:34.160 --> 01:10:36.290 align:middle line:84% to enable what we call physically 01:10:36.290 --> 01:10:38.720 align:middle line:84% accurate or robust or correct simulations 01:10:38.720 --> 01:10:40.027 align:middle line:90% of complex processes. 01:10:40.027 --> 01:10:42.110 align:middle line:84% However, I've tried to tell you to take everything 01:10:42.110 --> 01:10:44.277 align:middle line:84% with a grain of salt. Don't believe it 01:10:44.277 --> 01:10:46.860 align:middle line:84% just because you simulate it in SUPREM and it looks like that. 01:10:46.860 --> 01:10:48.960 align:middle line:84% One thing, you might have done it wrong. 01:10:48.960 --> 01:10:52.950 align:middle line:84% You might have used the wrong grid factor. 01:10:52.950 --> 01:10:54.840 align:middle line:84% SUPREM itself has parameters in it 01:10:54.840 --> 01:10:57.842 align:middle line:84% that are unknown, that somebody just stuck in there. 01:10:57.842 --> 01:10:59.550 align:middle line:84% Some graduate student writing the program 01:10:59.550 --> 01:11:01.320 align:middle line:84% said, oh, I don't know exactly what this parameter is. 01:11:01.320 --> 01:11:02.320 align:middle line:90% I'll put this number in. 01:11:02.320 --> 01:11:03.100 align:middle line:90% It's rough. 01:11:03.100 --> 01:11:05.350 align:middle line:84% Well, you need to find out what parameter is in there. 01:11:05.350 --> 01:11:07.607 align:middle line:84% And if you go to the appendix in the manual, 01:11:07.607 --> 01:11:09.690 align:middle line:84% it'll tell you in general what all the numbers are 01:11:09.690 --> 01:11:10.400 align:middle line:90% that it's using. 01:11:10.400 --> 01:11:12.150 align:middle line:84% And you can decide whether those numbers-- 01:11:12.150 --> 01:11:15.150 align:middle line:90% whether you like them or not. 01:11:15.150 --> 01:11:17.610 align:middle line:84% The nice thing about these simulations from SUPREM-IV 01:11:17.610 --> 01:11:20.580 align:middle line:84% or ATHENA, you can feed them into a device simulator, 01:11:20.580 --> 01:11:22.440 align:middle line:84% such as Medici or whatever, and then 01:11:22.440 --> 01:11:24.600 align:middle line:84% predict IV and CV characteristics, 01:11:24.600 --> 01:11:26.770 align:middle line:84% the actual electrical characteristics of the device. 01:11:26.770 --> 01:11:28.590 align:middle line:84% So they are designed to be coupled, 01:11:28.590 --> 01:11:33.060 align:middle line:84% the process simulator and the device simulator. 01:11:33.060 --> 01:11:36.513 align:middle line:90% And that's a very nice feature. 01:11:36.513 --> 01:11:37.930 align:middle line:84% There's a lot of new understanding 01:11:37.930 --> 01:11:41.860 align:middle line:84% that's being developed and issues related to TED and OED 01:11:41.860 --> 01:11:45.365 align:middle line:84% and other anomalous effects, low energy ion implants. 01:11:45.365 --> 01:11:46.990 align:middle line:84% I showed you some low-energy phosphorus 01:11:46.990 --> 01:11:51.100 align:middle line:84% where SUPREM doesn't do a very good job of modeling the data. 01:11:51.100 --> 01:11:53.860 align:middle line:84% Highly tilted implants with shadowing, 01:11:53.860 --> 01:11:55.240 align:middle line:90% oxidation of trenches. 01:11:55.240 --> 01:11:58.810 align:middle line:84% SUPREM can handle most of these situations 01:11:58.810 --> 01:12:00.310 align:middle line:84% because it has the physics built in. 01:12:00.310 --> 01:12:03.055 align:middle line:84% Exactly the accuracy, that's always the question mark. 01:12:03.055 --> 01:12:05.680 align:middle line:90% 01:12:05.680 --> 01:12:07.660 align:middle line:84% Whenever you're running these simulators, 01:12:07.660 --> 01:12:10.400 align:middle line:84% keep in mind the basic physical models. 01:12:10.400 --> 01:12:14.087 align:middle line:84% There are a lot of parameters that must be known accurately, 01:12:14.087 --> 01:12:15.670 align:middle line:84% and we don't know them all accurately. 01:12:15.670 --> 01:12:18.330 align:middle line:90% So keep your eyes open. 01:12:18.330 --> 01:12:23.760 align:middle line:84% A big caveat is keep the mesh or the grid in mind. 01:12:23.760 --> 01:12:25.590 align:middle line:84% When your grid is fine enough, you 01:12:25.590 --> 01:12:28.920 align:middle line:84% should be able to get the exact same solution with minor changes 01:12:28.920 --> 01:12:29.620 align:middle line:90% to the grid. 01:12:29.620 --> 01:12:31.260 align:middle line:90% So I run it for a certain grid. 01:12:31.260 --> 01:12:33.060 align:middle line:90% I cut the grid spacing in half. 01:12:33.060 --> 01:12:34.810 align:middle line:84% The solution should look identical. 01:12:34.810 --> 01:12:37.050 align:middle line:84% If all of a sudden the solution looks much smoother 01:12:37.050 --> 01:12:40.140 align:middle line:84% when I cut the grid in half, the spacing or much finer, 01:12:40.140 --> 01:12:41.820 align:middle line:84% then obviously the original solution 01:12:41.820 --> 01:12:43.650 align:middle line:90% didn't have a fine enough grid. 01:12:43.650 --> 01:12:45.420 align:middle line:84% So you've got to do these sanity checks. 01:12:45.420 --> 01:12:47.820 align:middle line:90% Run it once with a certain grid. 01:12:47.820 --> 01:12:49.260 align:middle line:90% Then use twice the grid points. 01:12:49.260 --> 01:12:50.610 align:middle line:90% Does it look a lot smoother? 01:12:50.610 --> 01:12:52.500 align:middle line:84% Well, that means you probably didn't do a good enough job 01:12:52.500 --> 01:12:53.280 align:middle line:90% the first time. 01:12:53.280 --> 01:12:54.180 align:middle line:90% And then do it again. 01:12:54.180 --> 01:12:55.650 align:middle line:90% And eventually, it'll converge. 01:12:55.650 --> 01:12:59.220 align:middle line:84% It has about the same smoothness regardless of your grid. 01:12:59.220 --> 01:13:01.470 align:middle line:84% In which case, your grid is probably fine enough. 01:13:01.470 --> 01:13:03.930 align:middle line:84% Your solution should be independent of the grid. 01:13:03.930 --> 01:13:10.018 align:middle line:84% Otherwise, that's not physically realistic. 01:13:10.018 --> 01:13:12.310 align:middle line:84% There's a lot of different process integration schemes. 01:13:12.310 --> 01:13:15.160 align:middle line:84% I just showed you one, just changing the order 01:13:15.160 --> 01:13:17.230 align:middle line:90% of the device, of the anneals. 01:13:17.230 --> 01:13:19.480 align:middle line:84% The simulators help us to understand these interaction 01:13:19.480 --> 01:13:21.310 align:middle line:84% between the various steps and how 01:13:21.310 --> 01:13:24.220 align:middle line:84% to optimize the overall technology in ways 01:13:24.220 --> 01:13:26.540 align:middle line:84% that you could never do if you just did this by hand. 01:13:26.540 --> 01:13:28.630 align:middle line:90% So you've used SUPREM-IV now. 01:13:28.630 --> 01:13:30.760 align:middle line:84% You're going to use it again in homework number 5. 01:13:30.760 --> 01:13:35.110 align:middle line:84% So I think gives you some idea of how powerful this tool is. 01:13:35.110 --> 01:13:37.550 align:middle line:84% But if you are going to really use it in your research, 01:13:37.550 --> 01:13:38.860 align:middle line:90% make sure you read the manual. 01:13:38.860 --> 01:13:40.360 align:middle line:84% It is very important because there's 01:13:40.360 --> 01:13:43.420 align:middle line:84% a lot of little things in there you need to know about. 01:13:43.420 --> 01:13:45.370 align:middle line:90% OK, that's it for today. 01:13:45.370 --> 01:13:47.770 align:middle line:84% Homework number 4, please bring up front. 01:13:47.770 --> 01:13:50.890 align:middle line:84% Homework number 3 is in the orange folder in the back 01:13:50.890 --> 01:13:52.180 align:middle line:90% by the TA. 01:13:52.180 --> 01:13:55.600 align:middle line:90% And I guess that's it. 01:13:55.600 --> 01:14:01.017 align:middle line:84% We'll meet on-- oh, somebody have the sign 01:14:01.017 --> 01:14:02.350 align:middle line:90% up sheet for your final project? 01:14:02.350 --> 01:14:03.760 align:middle line:90% Oh, great. 01:14:03.760 --> 01:14:04.630 align:middle line:90% Thanks. 01:14:04.630 --> 01:14:06.900 align:middle line:90% Make sure you sign up. 01:14:06.900 --> 01:14:08.000 align:middle line:90%