WEBVTT 00:00:00.000 --> 00:00:02.570 align:middle line:84% Started with a couple of announcements. 00:00:02.570 --> 00:00:04.490 align:middle line:84% There are two handouts for today. 00:00:04.490 --> 00:00:07.050 align:middle line:84% There are the lecture notes for today, 00:00:07.050 --> 00:00:09.470 align:middle line:90% which are this handout 25. 00:00:09.470 --> 00:00:12.497 align:middle line:84% And you also have an additional handout 26, which 00:00:12.497 --> 00:00:13.580 align:middle line:90% we're going to go through. 00:00:13.580 --> 00:00:15.050 align:middle line:84% It's a little calculation example 00:00:15.050 --> 00:00:18.140 align:middle line:84% I was hoping we could go through during today's lecture 00:00:18.140 --> 00:00:20.760 align:middle line:84% and have you do some calculations. 00:00:20.760 --> 00:00:22.250 align:middle line:90% So it's not a homework problem. 00:00:22.250 --> 00:00:24.000 align:middle line:84% We're going to do that during the lecture. 00:00:24.000 --> 00:00:26.000 align:middle line:90% 00:00:26.000 --> 00:00:26.500 align:middle line:90% Let's see. 00:00:26.500 --> 00:00:29.020 align:middle line:84% As far as homeworks, I don't have any to pass out quite yet. 00:00:29.020 --> 00:00:31.510 align:middle line:84% The TA is still working on those. 00:00:31.510 --> 00:00:33.190 align:middle line:90% And she's out today. 00:00:33.190 --> 00:00:36.280 align:middle line:84% But I do have the clipboard, which I will have 00:00:36.280 --> 00:00:40.240 align:middle line:84% up front here if you want to sign up for your final project. 00:00:40.240 --> 00:00:42.790 align:middle line:84% We started circulating that last time where you can-- 00:00:42.790 --> 00:00:44.540 align:middle line:84% I'm asking people to put your name down. 00:00:44.540 --> 00:00:46.130 align:middle line:84% And even if you don't know your topic, 00:00:46.130 --> 00:00:47.560 align:middle line:84% if you can check whether you want 00:00:47.560 --> 00:00:50.800 align:middle line:84% to do a written report or an oral presentation to the class, 00:00:50.800 --> 00:00:52.010 align:middle line:90% that would be helpful. 00:00:52.010 --> 00:00:54.190 align:middle line:84% And then once you know your topic, 00:00:54.190 --> 00:00:57.080 align:middle line:84% that would be good to fill that in, as well. 00:00:57.080 --> 00:00:57.580 align:middle line:90% All right. 00:00:57.580 --> 00:00:59.038 align:middle line:84% So I'll leave that up here for now. 00:00:59.038 --> 00:01:03.320 align:middle line:84% And you can come up towards the end and sign up. 00:01:03.320 --> 00:01:07.850 align:middle line:84% OK, so today's lecture is actually 00:01:07.850 --> 00:01:11.150 align:middle line:84% going to be the final lecture on ion implantation and transit 00:01:11.150 --> 00:01:12.800 align:middle line:90% enhanced diffusion. 00:01:12.800 --> 00:01:17.377 align:middle line:84% Let me review here on handout 25, the first page-- 00:01:17.377 --> 00:01:18.960 align:middle line:84% review what we've talked about so far. 00:01:18.960 --> 00:01:20.710 align:middle line:84% We've talked about ion implanted profiles. 00:01:20.710 --> 00:01:24.710 align:middle line:84% And we said you can model them very simply as a Gaussian, 00:01:24.710 --> 00:01:27.770 align:middle line:84% more accurately as a Pearson Ford or dual Pearson 00:01:27.770 --> 00:01:28.610 align:middle line:90% distribution. 00:01:28.610 --> 00:01:31.280 align:middle line:84% I'll show you some examples of real implants today. 00:01:31.280 --> 00:01:35.060 align:middle line:84% Or they can be simulated by numerical techniques 00:01:35.060 --> 00:01:38.820 align:middle line:90% such as Monte Carlo. 00:01:38.820 --> 00:01:41.580 align:middle line:84% Last time we talked a lot about the damage modeling. 00:01:41.580 --> 00:01:44.460 align:middle line:84% And we introduced something called the plus end model where 00:01:44.460 --> 00:01:46.350 align:middle line:90% n is a small number, about 1. 00:01:46.350 --> 00:01:48.900 align:middle line:84% And this is the model for residual damage 00:01:48.900 --> 00:01:53.910 align:middle line:84% that says that there's roughly n excess silicon 00:01:53.910 --> 00:01:56.130 align:middle line:84% interstitials injected per primary iron 00:01:56.130 --> 00:02:00.080 align:middle line:84% where n is a small number on the order of 1. 00:02:00.080 --> 00:02:03.410 align:middle line:84% Then last time we talked about how these excess interstitials 00:02:03.410 --> 00:02:07.400 align:middle line:84% cluster very, very quickly into defects called 311 defects. 00:02:07.400 --> 00:02:12.620 align:middle line:84% And later, these defects then dissolve during annealing. 00:02:12.620 --> 00:02:16.130 align:middle line:84% And their evaporation rate is really 00:02:16.130 --> 00:02:19.400 align:middle line:84% what determines the kinetics of transient enhanced diffusion, 00:02:19.400 --> 00:02:22.490 align:middle line:84% determines how long TED lasts, and also 00:02:22.490 --> 00:02:25.530 align:middle line:90% determines the magnitude of TED. 00:02:25.530 --> 00:02:28.530 align:middle line:84% In fact, we had a clustering slash evaporation model 00:02:28.530 --> 00:02:30.600 align:middle line:90% that we talked about last time. 00:02:30.600 --> 00:02:33.660 align:middle line:84% And that was used to explain the time and temperature dependence, 00:02:33.660 --> 00:02:37.350 align:middle line:84% and to a certain extent, the dose of energy dependence of TED 00:02:37.350 --> 00:02:38.920 align:middle line:90% in a rough way. 00:02:38.920 --> 00:02:40.440 align:middle line:84% So what I wanted to cover this time 00:02:40.440 --> 00:02:42.360 align:middle line:84% is to actually give you some real examples 00:02:42.360 --> 00:02:45.600 align:middle line:84% of ion implantation profiles, data, and a little bit 00:02:45.600 --> 00:02:47.580 align:middle line:90% of simulations. 00:02:47.580 --> 00:02:50.160 align:middle line:84% I want to spend a few minutes-- maybe 10, 15 minutes if we 00:02:50.160 --> 00:02:51.043 align:middle line:90% have time in class. 00:02:51.043 --> 00:02:53.460 align:middle line:84% I know a lot of people are sleeping in because the Red Sox 00:02:53.460 --> 00:02:55.770 align:middle line:90% won last night the World Series. 00:02:55.770 --> 00:02:57.360 align:middle line:84% But for those of you who are here, 00:02:57.360 --> 00:03:00.610 align:middle line:84% that's what the handout number 26 is all about. 00:03:00.610 --> 00:03:04.020 align:middle line:84% We'll calculate together a little simple calculation 00:03:04.020 --> 00:03:05.388 align:middle line:90% on transit enhanced diffusion. 00:03:05.388 --> 00:03:06.930 align:middle line:84% And then I want to finish the lecture 00:03:06.930 --> 00:03:10.830 align:middle line:84% and spend the rest of the time on the effect of TED on devices. 00:03:10.830 --> 00:03:13.650 align:middle line:84% Now that we have these models for TED, 00:03:13.650 --> 00:03:17.010 align:middle line:84% we can go back and revisit the reverse short channel effect 00:03:17.010 --> 00:03:19.080 align:middle line:84% and talk about that in more detail. 00:03:19.080 --> 00:03:23.590 align:middle line:84% OK, so let's go on to slide number 2. 00:03:23.590 --> 00:03:28.270 align:middle line:84% And what I'm showing here, these are some actual SUPREM core 00:03:28.270 --> 00:03:33.280 align:middle line:84% outputs, output files, modeling ion implant profiles, and just 00:03:33.280 --> 00:03:37.960 align:middle line:84% to show you a practical example of how you might use SUPREM-IV 00:03:37.960 --> 00:03:40.060 align:middle line:90% along with a dual Pearson. 00:03:40.060 --> 00:03:42.790 align:middle line:84% A dual Pearson means you take two Pearson four distributions 00:03:42.790 --> 00:03:43.960 align:middle line:90% and you add them together. 00:03:43.960 --> 00:03:47.500 align:middle line:84% And you can use this in a quasi empirical way 00:03:47.500 --> 00:03:50.690 align:middle line:84% to simulate the ion channeling tail. 00:03:50.690 --> 00:03:52.510 align:middle line:84% So there are four different plots here, 00:03:52.510 --> 00:03:53.710 align:middle line:90% four different panels. 00:03:53.710 --> 00:03:56.260 align:middle line:84% Look at the first one marked A. These are all 00:03:56.260 --> 00:03:58.540 align:middle line:84% for the same condition in terms. of energy. 00:03:58.540 --> 00:04:02.380 align:middle line:90% This is all boron, BF2, implant. 00:04:02.380 --> 00:04:06.160 align:middle line:84% And the energy, the primary energy, is 65 KEV. 00:04:06.160 --> 00:04:08.260 align:middle line:84% And then what we're looking at in each panel 00:04:08.260 --> 00:04:09.920 align:middle line:90% is a different dose. 00:04:09.920 --> 00:04:12.280 align:middle line:84% So before we get started with looking at those, 00:04:12.280 --> 00:04:13.690 align:middle line:90% let me just make a note. 00:04:13.690 --> 00:04:18.880 align:middle line:84% The total atomic mass of a molecule of BF2 is 49. 00:04:18.880 --> 00:04:20.680 align:middle line:84% You just add up two fluorines and a boron. 00:04:20.680 --> 00:04:21.940 align:middle line:90% You get mass 49. 00:04:21.940 --> 00:04:24.170 align:middle line:90% The mass of boron is 11. 00:04:24.170 --> 00:04:26.980 align:middle line:84% So as I mentioned last time, people sometimes 00:04:26.980 --> 00:04:30.700 align:middle line:84% do a BF2 to get a lower energy or more easily modifies 00:04:30.700 --> 00:04:31.580 align:middle line:90% the silicon. 00:04:31.580 --> 00:04:34.630 align:middle line:84% For example, look at this little simple calculation. 00:04:34.630 --> 00:04:40.700 align:middle line:84% 65 kilovolt BF2 is equivalent to 65 times 11 over 49. 00:04:40.700 --> 00:04:43.360 align:middle line:84% So you assume that the energy is partitioned 00:04:43.360 --> 00:04:48.350 align:middle line:84% according to the ratio of the masses for a boron implant. 00:04:48.350 --> 00:04:51.440 align:middle line:84% So if I do 65 KEV BF2, it's the same as 00:04:51.440 --> 00:04:56.660 align:middle line:84% if I implanted the boron atom itself at 15 KEV. 00:04:56.660 --> 00:05:00.620 align:middle line:84% So especially in the old days when it was very, very difficult 00:05:00.620 --> 00:05:03.020 align:middle line:84% to build ion implanters that could ion implant 00:05:03.020 --> 00:05:06.710 align:middle line:84% at low energies, now they have implanters that can get you down 00:05:06.710 --> 00:05:08.840 align:middle line:90% to 1, 2, 3, KEV. 00:05:08.840 --> 00:05:12.620 align:middle line:84% But the beam current tends to suffer when you go that low. 00:05:12.620 --> 00:05:14.810 align:middle line:84% So instead of doing very low energy boron, 00:05:14.810 --> 00:05:18.770 align:middle line:84% people would do a higher energy molecule, BF2. 00:05:18.770 --> 00:05:21.740 align:middle line:84% So you'll often see BF2 implants. 00:05:21.740 --> 00:05:26.120 align:middle line:84% And so now let's go ahead and look at these BF2 implants. 00:05:26.120 --> 00:05:27.740 align:middle line:84% And the couple of things being shown 00:05:27.740 --> 00:05:31.370 align:middle line:84% here-- the plus signs or the symbols are the actual data. 00:05:31.370 --> 00:05:35.630 align:middle line:84% And these data was taken from Tasha's work back in 1989. 00:05:35.630 --> 00:05:38.270 align:middle line:90% These are actual sims profiles. 00:05:38.270 --> 00:05:43.550 align:middle line:84% And the lines, the smooth lines, are SUPREM simulations. 00:05:43.550 --> 00:05:47.200 align:middle line:84% So for example, in part A, or plot number A, we have a dose 00:05:47.200 --> 00:05:49.820 align:middle line:90% that's very high of 5e15. 00:05:49.820 --> 00:05:54.190 align:middle line:84% And so what that means is with such a high dose of BF2, 00:05:54.190 --> 00:05:57.080 align:middle line:84% throughout most of the implant, of the time of the implant, 00:05:57.080 --> 00:05:58.810 align:middle line:84% you're really going to be amortized. 00:05:58.810 --> 00:06:02.853 align:middle line:84% You're going to be implanting into an amorphous crystalline 00:06:02.853 --> 00:06:05.020 align:middle line:84% because it's going to amort-- the BF2 will amorphous 00:06:05.020 --> 00:06:05.600 align:middle line:90% the silicon. 00:06:05.600 --> 00:06:07.600 align:middle line:84% So this is the profile that you get. 00:06:07.600 --> 00:06:11.410 align:middle line:84% Notice it looks pretty much like a standard Pearson 4. 00:06:11.410 --> 00:06:14.620 align:middle line:84% And there's this little small exponential tail. 00:06:14.620 --> 00:06:18.070 align:middle line:84% But that tail doesn't really become obvious or evident 00:06:18.070 --> 00:06:20.770 align:middle line:84% until you get about one, two, about three orders of magnitude 00:06:20.770 --> 00:06:21.850 align:middle line:90% down from the peak. 00:06:21.850 --> 00:06:23.540 align:middle line:84% And then you start to see the tail. 00:06:23.540 --> 00:06:25.360 align:middle line:84% So the way this profile is constructed 00:06:25.360 --> 00:06:26.890 align:middle line:90% is what dual Pearson means. 00:06:26.890 --> 00:06:28.880 align:middle line:84% It's two Pearson force added together. 00:06:28.880 --> 00:06:31.900 align:middle line:84% So there's one Pearson four here, 00:06:31.900 --> 00:06:33.990 align:middle line:84% which is with this dashed line that's labeled 00:06:33.990 --> 00:06:35.680 align:middle line:90% as the amorphous profile. 00:06:35.680 --> 00:06:37.810 align:middle line:84% And then there's a second Pearson four here, 00:06:37.810 --> 00:06:40.037 align:middle line:84% which has also, unfortunately, a dashed line, 00:06:40.037 --> 00:06:42.120 align:middle line:84% but you can see it has a slightly different slope. 00:06:42.120 --> 00:06:44.190 align:middle line:84% And that's labeled the channel profile. 00:06:44.190 --> 00:06:48.130 align:middle line:84% Those two Pearson fours are added up with a certain ratio. 00:06:48.130 --> 00:06:51.120 align:middle line:84% This ratio here, in this case, is 0.969. 00:06:51.120 --> 00:06:53.900 align:middle line:90% And you get the total profile. 00:06:53.900 --> 00:06:56.290 align:middle line:84% And you can see the simulation, obviously 00:06:56.290 --> 00:06:58.300 align:middle line:84% if you pick the right number for the ratio, 00:06:58.300 --> 00:07:00.740 align:middle line:84% the simulation then fits the entire profile, 00:07:00.740 --> 00:07:03.130 align:middle line:84% including this region here at the near surface 00:07:03.130 --> 00:07:07.210 align:middle line:84% and the amorphous-- and the tail, rather, the channel tail. 00:07:07.210 --> 00:07:10.330 align:middle line:84% Let's go to a lower dose, say 1.5e15. 00:07:10.330 --> 00:07:11.560 align:middle line:90% Now you do see-- 00:07:11.560 --> 00:07:14.560 align:middle line:84% again, we see two Pearsons, this one here 00:07:14.560 --> 00:07:17.620 align:middle line:84% and the one that's associated with the channel profile. 00:07:17.620 --> 00:07:22.420 align:middle line:84% But now the channel profile is a higher fraction of the total. 00:07:22.420 --> 00:07:26.980 align:middle line:84% And you see this exponential tail coming in a little bit more 00:07:26.980 --> 00:07:27.950 align:middle line:90% obviously. 00:07:27.950 --> 00:07:32.140 align:middle line:84% And if you go down to panel number C or plot number C, 00:07:32.140 --> 00:07:35.470 align:middle line:90% the dose is even lower, 5e14. 00:07:35.470 --> 00:07:37.610 align:middle line:84% Not that much amorphization going on. 00:07:37.610 --> 00:07:39.880 align:middle line:84% So you have a large, very prominent channel tail. 00:07:39.880 --> 00:07:41.890 align:middle line:84% So the second Pearson is really dominating. 00:07:41.890 --> 00:07:45.490 align:middle line:84% And finally, at a low dose of 2e13, there is no amorphization. 00:07:45.490 --> 00:07:49.000 align:middle line:84% So the implant that really dominates in the dual Pearson 00:07:49.000 --> 00:07:52.750 align:middle line:84% is primarily that associated with the channel profile. 00:07:52.750 --> 00:07:56.210 align:middle line:84% And the ratio is actually 0, so it's just the channel one. 00:07:56.210 --> 00:07:57.770 align:middle line:84% So that's how dual Pearson works. 00:07:57.770 --> 00:08:02.330 align:middle line:84% You can see it is two Pearson fours added together, one 00:08:02.330 --> 00:08:04.190 align:middle line:84% of which takes care of the channel profile, 00:08:04.190 --> 00:08:06.190 align:middle line:84% and the other which takes care of the amorphous. 00:08:06.190 --> 00:08:08.870 align:middle line:84% And you add them together depending on a certain number 00:08:08.870 --> 00:08:09.660 align:middle line:90% called the ratio. 00:08:09.660 --> 00:08:11.150 align:middle line:90% So it's quasi empirical. 00:08:11.150 --> 00:08:14.190 align:middle line:90% 00:08:14.190 --> 00:08:17.490 align:middle line:90% So that was for BF2. 00:08:17.490 --> 00:08:20.730 align:middle line:84% Let me just show you some simulations. 00:08:20.730 --> 00:08:23.500 align:middle line:84% The different symbols here are not data, so I apologize. 00:08:23.500 --> 00:08:25.417 align:middle line:84% These are actually different symbols represent 00:08:25.417 --> 00:08:26.820 align:middle line:90% different SUPREM simulations. 00:08:26.820 --> 00:08:29.790 align:middle line:84% Just to give you an idea when you do a simulation, depending 00:08:29.790 --> 00:08:32.010 align:middle line:84% on which model you choose, you'll 00:08:32.010 --> 00:08:33.450 align:middle line:90% get slightly different profiles. 00:08:33.450 --> 00:08:37.140 align:middle line:84% And it's up to you to figure out by comparing to data 00:08:37.140 --> 00:08:39.299 align:middle line:84% or by looking at the literature which one of these 00:08:39.299 --> 00:08:41.600 align:middle line:90% is closest to the truth. 00:08:41.600 --> 00:08:43.409 align:middle line:84% For example, just as a comparison, 00:08:43.409 --> 00:08:45.980 align:middle line:84% if you look at the open circles here, 00:08:45.980 --> 00:08:47.877 align:middle line:90% this profile marked Gaussian-- 00:08:47.877 --> 00:08:50.210 align:middle line:84% it's a little bit hard because there's a lot of symbols. 00:08:50.210 --> 00:08:52.010 align:middle line:84% But there's an open circle profile, 00:08:52.010 --> 00:08:53.930 align:middle line:84% and it's quite symmetric, as it has to be, 00:08:53.930 --> 00:08:54.920 align:middle line:90% because it's Gaussian. 00:08:54.920 --> 00:08:57.710 align:middle line:90% It has no channel tail. 00:08:57.710 --> 00:09:02.180 align:middle line:84% If you look at the open triangles, 00:09:02.180 --> 00:09:03.380 align:middle line:90% it's the Pearson four. 00:09:03.380 --> 00:09:05.990 align:middle line:84% A single Pearson looks a lot like the Gaussian, not much 00:09:05.990 --> 00:09:06.650 align:middle line:90% difference. 00:09:06.650 --> 00:09:09.050 align:middle line:90% A little bit skewed. 00:09:09.050 --> 00:09:12.500 align:middle line:84% The dual Pearson is these open boxes. 00:09:12.500 --> 00:09:14.000 align:middle line:84% And it does a little bit better job, 00:09:14.000 --> 00:09:15.583 align:middle line:84% you would think-- well, we don't know. 00:09:15.583 --> 00:09:17.510 align:middle line:84% Actually we haven't seen the real data here. 00:09:17.510 --> 00:09:19.880 align:middle line:84% It has the channel tail incorporated in it. 00:09:19.880 --> 00:09:21.140 align:middle line:90% And look at the Monte Carlo. 00:09:21.140 --> 00:09:23.360 align:middle line:84% Now presumably, the Monte Carlo has the most physics 00:09:23.360 --> 00:09:27.150 align:middle line:90% built in into the simulation. 00:09:27.150 --> 00:09:30.340 align:middle line:84% And it looks a little noisy because, of course, 00:09:30.340 --> 00:09:32.350 align:middle line:84% we only follow a certain number of ions. 00:09:32.350 --> 00:09:34.838 align:middle line:84% But in fact, the dual Pearson looks reasonably 00:09:34.838 --> 00:09:35.880 align:middle line:90% close to the Monte Carlo. 00:09:35.880 --> 00:09:38.670 align:middle line:84% The Monte Carlo is attempting here 00:09:38.670 --> 00:09:41.310 align:middle line:84% to include a little bit of the ion channeling. 00:09:41.310 --> 00:09:44.850 align:middle line:84% So you get different models give you different answers in SUPREM, 00:09:44.850 --> 00:09:46.350 align:middle line:84% and you have to figure out which one 00:09:46.350 --> 00:09:49.515 align:middle line:84% suits your particular application best. 00:09:49.515 --> 00:09:52.160 align:middle line:90% 00:09:52.160 --> 00:09:53.930 align:middle line:84% So now what I want to do is I want 00:09:53.930 --> 00:09:58.040 align:middle line:84% to take a few minutes here in class if we go on to slide 00:09:58.040 --> 00:10:00.320 align:middle line:90% number 4 in the handout. 00:10:00.320 --> 00:10:03.050 align:middle line:84% And this is also the first page of your handout 00:10:03.050 --> 00:10:05.150 align:middle line:84% number 26 if you want to look at that. 00:10:05.150 --> 00:10:06.200 align:middle line:90% They're identical. 00:10:06.200 --> 00:10:09.618 align:middle line:84% And what I was hoping you could do is get together as a group. 00:10:09.618 --> 00:10:11.660 align:middle line:84% There's not enough people here to really split up 00:10:11.660 --> 00:10:14.060 align:middle line:84% into individual groups because everyone, 00:10:14.060 --> 00:10:16.580 align:middle line:84% again, sleeping in because Boston won-- 00:10:16.580 --> 00:10:19.340 align:middle line:84% is to go through this and make some simple calculations 00:10:19.340 --> 00:10:23.930 align:middle line:84% for the next 5 or 10 minutes together on how you would 00:10:23.930 --> 00:10:25.200 align:middle line:90% do the simple calculation. 00:10:25.200 --> 00:10:27.710 align:middle line:84% Let me just read through it before we start. 00:10:27.710 --> 00:10:31.700 align:middle line:84% So we have an engineer wants to form a shallow boron dope source 00:10:31.700 --> 00:10:33.810 align:middle line:84% drain for an advanced technology. 00:10:33.810 --> 00:10:36.050 align:middle line:84% And the question, the manager is wondering 00:10:36.050 --> 00:10:41.390 align:middle line:84% whether to buy a batch furnace or to use 00:10:41.390 --> 00:10:42.500 align:middle line:90% a rapid thermal anneal. 00:10:42.500 --> 00:10:45.050 align:middle line:84% And the furnace anneal they are considering 00:10:45.050 --> 00:10:48.830 align:middle line:84% for this implant would be 800 degrees for one hour. 00:10:48.830 --> 00:10:51.680 align:middle line:84% So that's a relatively long time at low temperature. 00:10:51.680 --> 00:10:55.040 align:middle line:84% Or should they use a rapid thermal processing machine, 00:10:55.040 --> 00:10:58.430 align:middle line:90% or RTA, at 1,050 for one second? 00:10:58.430 --> 00:11:02.030 align:middle line:84% And the implant that they want to activate is boron, 00:11:02.030 --> 00:11:04.430 align:middle line:84% and it's 20 kilovolts, relatively low energy. 00:11:04.430 --> 00:11:08.540 align:middle line:90% 20 kV, and a 5e14 dose. 00:11:08.540 --> 00:11:11.830 align:middle line:84% So what we're asked to do here in A 00:11:11.830 --> 00:11:15.400 align:middle line:84% is to make a rough estimate using a square root of DT 00:11:15.400 --> 00:11:19.570 align:middle line:84% estimate of how far the dopants move during an 800 degree 00:11:19.570 --> 00:11:24.392 align:middle line:84% one-hour anneal versus 1,050 for one second. 00:11:24.392 --> 00:11:26.350 align:middle line:84% Again, it's not necessarily Gaussian diffusion, 00:11:26.350 --> 00:11:27.910 align:middle line:84% but you can calculate the root DT. 00:11:27.910 --> 00:11:29.980 align:middle line:84% And this is without any TED effects. 00:11:29.980 --> 00:11:31.750 align:middle line:84% And this is pretty simple because I've 00:11:31.750 --> 00:11:33.460 align:middle line:84% printed at the bottom of the slide 00:11:33.460 --> 00:11:38.080 align:middle line:84% the intrinsic diffusivities under equilibrium conditions. 00:11:38.080 --> 00:11:40.060 align:middle line:90% D at 800 is this number. 00:11:40.060 --> 00:11:42.395 align:middle line:84% And the diffusivity of boron at 1,050 is this number. 00:11:42.395 --> 00:11:44.270 align:middle line:84% So we have those numbers written right there. 00:11:44.270 --> 00:11:45.520 align:middle line:90% So part A is pretty easy. 00:11:45.520 --> 00:11:47.860 align:middle line:84% Now, part B, what we want to try to do 00:11:47.860 --> 00:11:50.260 align:middle line:90% is include the effects of TED. 00:11:50.260 --> 00:11:53.170 align:middle line:84% So we want to use the charts that we handed out last time. 00:11:53.170 --> 00:11:55.307 align:middle line:84% And in fact, they're in handout number 26. 00:11:55.307 --> 00:11:56.890 align:middle line:84% If you didn't bring your old handouts, 00:11:56.890 --> 00:12:00.730 align:middle line:84% I've included them again in handout number 26. 00:12:00.730 --> 00:12:03.748 align:middle line:84% Use these charts to figure out the expected enhancement 00:12:03.748 --> 00:12:04.540 align:middle line:90% of the diffusivity. 00:12:04.540 --> 00:12:07.780 align:middle line:84% Remember we said the diffusivity is now the equilibrium, D star, 00:12:07.780 --> 00:12:09.760 align:middle line:90% times CI over CI star. 00:12:09.760 --> 00:12:11.410 align:middle line:84% That's the effective diffusivity. 00:12:11.410 --> 00:12:16.630 align:middle line:84% And how long Ted lasts to calculate the real square root 00:12:16.630 --> 00:12:20.240 align:middle line:90% of DT, including TED effects. 00:12:20.240 --> 00:12:23.570 align:middle line:84% So these are the two things I'd like you to try to do. 00:12:23.570 --> 00:12:26.230 align:middle line:84% And I'm hoping you can work as a team on this. 00:12:26.230 --> 00:12:30.220 align:middle line:84% Maybe get together with a couple of folks who are near you. 00:12:30.220 --> 00:12:32.560 align:middle line:84% Somebody here hopefully has a calculator 00:12:32.560 --> 00:12:35.315 align:middle line:90% if you need one to do this. 00:12:35.315 --> 00:12:36.940 align:middle line:84% I'm not so interested in exact numbers. 00:12:36.940 --> 00:12:41.060 align:middle line:84% I just want to get an idea of how the answers play out. 00:12:41.060 --> 00:12:41.810 align:middle line:90% So why don't you-- 00:12:41.810 --> 00:12:44.090 align:middle line:84% I'm going to give you 5 or 10 minutes to get together and work 00:12:44.090 --> 00:12:44.660 align:middle line:90% on that. 00:12:44.660 --> 00:12:48.050 align:middle line:84% And we'll stop the lecture now, and then we'll come back 00:12:48.050 --> 00:12:50.960 align:middle line:84% and we'll see who has an answer that looks reasonable, 00:12:50.960 --> 00:12:52.500 align:middle line:90% and we'll talk it through. 00:12:52.500 --> 00:12:53.000 align:middle line:90% OK? 00:12:53.000 --> 00:12:56.540 align:middle line:84% So we can stop the recording at this point, 00:12:56.540 --> 00:12:59.330 align:middle line:84% as well, because it won't be that interesting to record 00:12:59.330 --> 00:13:01.670 align:middle line:90% you punching your calculators. 00:13:01.670 --> 00:13:04.050 align:middle line:90% 1,300. 00:13:04.050 --> 00:13:09.040 align:middle line:84% What can you say about the first 1,300 seconds? 00:13:09.040 --> 00:13:15.450 align:middle line:84% How high is the diffusivity during the first 1,300 seconds? 00:13:15.450 --> 00:13:20.280 align:middle line:84% It's going to be 7,000 times than it is in the next 1,300 00:13:20.280 --> 00:13:22.470 align:middle line:90% seconds or 2,000 seconds. 00:13:22.470 --> 00:13:25.410 align:middle line:84% So who cares, in some sense about, the next 2,000? 00:13:25.410 --> 00:13:28.440 align:middle line:84% Now, if this instead of being 3,600 seconds 00:13:28.440 --> 00:13:30.893 align:middle line:84% was orders of magnitude more seconds, yeah, then 00:13:30.893 --> 00:13:33.060 align:middle line:84% at some point you actually have to take into account 00:13:33.060 --> 00:13:34.660 align:middle line:90% the normal diffusion. 00:13:34.660 --> 00:13:38.190 align:middle line:84% So you don't completely ignore the actual time. 00:13:38.190 --> 00:13:40.530 align:middle line:84% The actual time that the thing is in the furnace 00:13:40.530 --> 00:13:41.790 align:middle line:90% is 3,600 seconds. 00:13:41.790 --> 00:13:46.740 align:middle line:84% It's just that for on the order of almost a third of that, 00:13:46.740 --> 00:13:48.847 align:middle line:84% it's enhanced diffusivity by 7,000. 00:13:48.847 --> 00:13:50.430 align:middle line:84% So we can ignore the rest of the time. 00:13:50.430 --> 00:13:53.080 align:middle line:84% But you want to keep that in the back of your mind. 00:13:53.080 --> 00:13:55.830 align:middle line:84% So then were you able to calculate either 00:13:55.830 --> 00:14:01.760 align:middle line:84% a DT enhanced or the square of DT enhanced? 00:14:01.760 --> 00:14:02.390 align:middle line:90% Either one. 00:14:02.390 --> 00:14:06.815 align:middle line:90% 00:14:06.815 --> 00:14:07.815 align:middle line:90% What'd you get for that? 00:14:07.815 --> 00:14:10.870 align:middle line:90% 00:14:10.870 --> 00:14:12.310 align:middle line:90% AUDIENCE: DT 1.06. 00:14:12.310 --> 00:14:15.010 align:middle line:90% Is that the unit? 00:14:15.010 --> 00:14:16.210 align:middle line:90% JUDY HOYT: The DT you got-- 00:14:16.210 --> 00:14:20.735 align:middle line:84% OK, I ended up somehow with a 3.5 times to the 10 00:14:20.735 --> 00:14:22.510 align:middle line:90% to the -10 centimeters squared. 00:14:22.510 --> 00:14:25.820 align:middle line:84% But doesn't mean-- oh, I multiplied this. 00:14:25.820 --> 00:14:30.160 align:middle line:84% So if we multiply the ordinary diffusivity times that, 00:14:30.160 --> 00:14:37.870 align:middle line:84% I got 7,000 times 3.7 times 10 to the -17. 00:14:37.870 --> 00:14:43.870 align:middle line:84% So the actual diffusivity I got was 2.6 times 10 00:14:43.870 --> 00:14:47.860 align:middle line:84% to the -13 centimeters squared per second during the transit, 00:14:47.860 --> 00:14:51.220 align:middle line:84% during the time when the diffusivity was enhanced. 00:14:51.220 --> 00:14:54.580 align:middle line:90% And then multiply that by 13. 00:14:54.580 --> 00:14:56.260 align:middle line:90% Oh, my numbers-- 00:14:56.260 --> 00:14:57.490 align:middle line:90% I had 1,333 here. 00:14:57.490 --> 00:15:01.200 align:middle line:84% Maybe I was carrying a few more significant digits. 00:15:01.200 --> 00:15:03.342 align:middle line:84% AUDIENCE: Well the numbers should be in the root. 00:15:03.342 --> 00:15:04.300 align:middle line:90% JUDY HOYT: Oh, root DT. 00:15:04.300 --> 00:15:05.230 align:middle line:90% OK. 00:15:05.230 --> 00:15:06.860 align:middle line:84% So for the square root you got what? 00:15:06.860 --> 00:15:07.360 align:middle line:90% 18? 00:15:07.360 --> 00:15:10.200 align:middle line:90% 00:15:10.200 --> 00:15:10.770 align:middle line:90% Right. 00:15:10.770 --> 00:15:11.910 align:middle line:90% 10 to -5. 00:15:11.910 --> 00:15:16.530 align:middle line:84% So in angstroms, to put it on the same-- that's 1,860, 00:15:16.530 --> 00:15:19.570 align:middle line:84% just to put it in the same units. 00:15:19.570 --> 00:15:23.170 align:middle line:84% So not even close in the intrinsic diffusion. 00:15:23.170 --> 00:15:27.750 align:middle line:84% So really, the TV clearly dominates in that case. 00:15:27.750 --> 00:15:29.432 align:middle line:90% And there's quite a bit. 00:15:29.432 --> 00:15:30.640 align:middle line:90% That's quite a bit of motion. 00:15:30.640 --> 00:15:32.820 align:middle line:84% Now that, you can easily see on a sims profile, 00:15:32.820 --> 00:15:35.790 align:middle line:84% and that could definitely affect your device performance junction 00:15:35.790 --> 00:15:38.350 align:middle line:84% depth, a difference of that much. 00:15:38.350 --> 00:15:42.540 align:middle line:90% So how about a 1,050? 00:15:42.540 --> 00:15:46.260 align:middle line:90% Exact same formulation applies. 00:15:46.260 --> 00:15:48.030 align:middle line:90% But the nice thing is-- 00:15:48.030 --> 00:15:54.850 align:middle line:84% how about what can we say about CI over CI star at 1,050? 00:15:54.850 --> 00:15:57.790 align:middle line:90% It's a lot lower, right? 00:15:57.790 --> 00:15:59.980 align:middle line:90% I got about 550. 00:15:59.980 --> 00:16:01.960 align:middle line:84% So the enhancement of the diffusivity 00:16:01.960 --> 00:16:04.600 align:middle line:90% is only 550 times larger. 00:16:04.600 --> 00:16:07.330 align:middle line:90% 00:16:07.330 --> 00:16:11.190 align:middle line:84% That's the diffusivity-- and how long does it last? 00:16:11.190 --> 00:16:13.870 align:middle line:90% And the TED at 1,050-- 00:16:13.870 --> 00:16:17.220 align:middle line:84% well, if you go back to your magic curve, phosphorus, 00:16:17.220 --> 00:16:19.590 align:middle line:84% it doesn't look like it's going to last very long. 00:16:19.590 --> 00:16:23.250 align:middle line:84% Again, you have to multiply by 5 and by 0.08 00:16:23.250 --> 00:16:26.590 align:middle line:90% over 0.06, the RP ratio. 00:16:26.590 --> 00:16:29.730 align:middle line:84% So the amount of time it looks like TED 00:16:29.730 --> 00:16:34.770 align:middle line:84% lasts under this dose condition, I got about 0.67 seconds. 00:16:34.770 --> 00:16:35.270 align:middle line:90% 2/3. 00:16:35.270 --> 00:16:38.040 align:middle line:90% 00:16:38.040 --> 00:16:39.660 align:middle line:90% 2/3 of a second. 00:16:39.660 --> 00:16:40.940 align:middle line:90% Basically, 2/3 of the anneal. 00:16:40.940 --> 00:16:43.170 align:middle line:84% Remember, the total anneal time was one second. 00:16:43.170 --> 00:16:46.470 align:middle line:84% So again, the last one third of the anneal, 00:16:46.470 --> 00:16:48.945 align:middle line:84% I'm going to ignore just because this is still 00:16:48.945 --> 00:16:50.070 align:middle line:90% a large enhancement factor. 00:16:50.070 --> 00:16:51.000 align:middle line:90% It's 550. 00:16:51.000 --> 00:16:51.900 align:middle line:90% So I'll ignore it. 00:16:51.900 --> 00:16:54.180 align:middle line:84% If it had been 1,050 for hours, then obviously 00:16:54.180 --> 00:16:57.400 align:middle line:84% we can't ignore the normal diffusion. 00:16:57.400 --> 00:16:57.900 align:middle line:90% OK. 00:16:57.900 --> 00:17:01.980 align:middle line:84% So then in that case, the root DT at 1,050, 00:17:01.980 --> 00:17:07.520 align:middle line:84% I ended up with something about 420 angstroms. 00:17:07.520 --> 00:17:11.160 align:middle line:84% Is that close to what you guys got? 00:17:11.160 --> 00:17:14.099 align:middle line:90% 4.2 times 10 to -6. 00:17:14.099 --> 00:17:18.579 align:middle line:84% Again, putting it back all into angstroms. 00:17:18.579 --> 00:17:23.490 align:middle line:84% So again, we see this kind of interesting anomalous, 00:17:23.490 --> 00:17:27.270 align:middle line:84% and maybe initially somewhat non-intuitive, result. 00:17:27.270 --> 00:17:29.758 align:middle line:84% Here we have a higher temperature. 00:17:29.758 --> 00:17:31.050 align:middle line:90% Admittedly a much shorter time. 00:17:31.050 --> 00:17:32.520 align:middle line:90% It's only a second. 00:17:32.520 --> 00:17:36.660 align:middle line:84% We can get a lot less root DT, a lot less motion or broadening, 00:17:36.660 --> 00:17:38.820 align:middle line:84% if we do a 1,050 rapid thermal anneal compared 00:17:38.820 --> 00:17:39.630 align:middle line:90% to putting in the furnace. 00:17:39.630 --> 00:17:41.190 align:middle line:84% You might say, well, 800 is really low. 00:17:41.190 --> 00:17:42.523 align:middle line:90% It should be a safe temperature. 00:17:42.523 --> 00:17:45.630 align:middle line:84% It's only a couple of angstroms motion. 00:17:45.630 --> 00:17:49.620 align:middle line:84% But in fact, you're much better off, for this implant, doing-- 00:17:49.620 --> 00:17:51.870 align:middle line:84% according to this calculation of TED, 00:17:51.870 --> 00:17:55.320 align:middle line:84% you're much better off doing a 1,050 anneal for a fraction 00:17:55.320 --> 00:17:56.675 align:middle line:90% or for a second. 00:17:56.675 --> 00:17:58.050 align:middle line:84% So that's just to give you a feel 00:17:58.050 --> 00:18:01.535 align:middle line:84% for where the numbers come from and how the whole thing works. 00:18:01.535 --> 00:18:03.660 align:middle line:84% I think once you work through an example like that, 00:18:03.660 --> 00:18:06.480 align:middle line:84% you have a much better feel for TED. 00:18:06.480 --> 00:18:07.170 align:middle line:90% OK. 00:18:07.170 --> 00:18:07.740 align:middle line:90% Good. 00:18:07.740 --> 00:18:10.390 align:middle line:84% Well, that seemed like everybody had a good handle on that. 00:18:10.390 --> 00:18:15.930 align:middle line:84% So let's go back to the regular handout, handout 25. 00:18:15.930 --> 00:18:20.220 align:middle line:84% And I think you've got a good feel for how simple 00:18:20.220 --> 00:18:22.350 align:middle line:84% calculations and how powerful it is to see you 00:18:22.350 --> 00:18:23.670 align:middle line:90% as those two or three charts. 00:18:23.670 --> 00:18:26.490 align:middle line:84% With two or three charts, you can say a lot about-- 00:18:26.490 --> 00:18:28.920 align:middle line:84% a rough back-of-the-envelope calculation for TED. 00:18:28.920 --> 00:18:32.380 align:middle line:84% Anything more sophisticated than what we just did in the last 10 00:18:32.380 --> 00:18:33.990 align:middle line:84% minutes, you probably should be using 00:18:33.990 --> 00:18:38.532 align:middle line:84% SUPREM at that point, or some simulator, I should say. 00:18:38.532 --> 00:18:39.490 align:middle line:90% All right, let's go on. 00:18:39.490 --> 00:18:43.890 align:middle line:84% So just to remind you ourselves, where those charts just 00:18:43.890 --> 00:18:48.270 align:middle line:84% came from-- if I'm on the slide number 5 now on the handout, 00:18:48.270 --> 00:18:51.638 align:middle line:84% they came from this observation that was made at Bell Labs 00:18:51.638 --> 00:18:53.430 align:middle line:84% and confirmed by a lot of different workers 00:18:53.430 --> 00:18:57.540 align:middle line:84% that the time scale of TED was the same as the time 00:18:57.540 --> 00:19:02.520 align:middle line:84% scale of the shrinkage of these now famous 311 defects, 00:19:02.520 --> 00:19:05.070 align:middle line:84% and that at low temperatures, the 311s ones 00:19:05.070 --> 00:19:07.590 align:middle line:84% hang around a lot longer than at high temperatures. 00:19:07.590 --> 00:19:11.970 align:middle line:84% And of course, that's why TED lasts a lot longer 00:19:11.970 --> 00:19:15.110 align:middle line:90% at low temperatures. 00:19:15.110 --> 00:19:19.790 align:middle line:84% OK, so just to remind ourselves and from the little hand 00:19:19.790 --> 00:19:21.830 align:middle line:84% calculation we just did, this is exactly 00:19:21.830 --> 00:19:24.830 align:middle line:84% representative of what we just did in class here. 00:19:24.830 --> 00:19:28.040 align:middle line:84% The general picture of TED that we showed last time, we have 00:19:28.040 --> 00:19:29.570 align:middle line:90% a certain enhancement factor-- 00:19:29.570 --> 00:19:31.490 align:middle line:90% CI over CI star max. 00:19:31.490 --> 00:19:32.750 align:middle line:90% Again, that's a ballpark. 00:19:32.750 --> 00:19:34.200 align:middle line:90% That's the maximum enhancement. 00:19:34.200 --> 00:19:36.740 align:middle line:84% So we assume it's just a constant enhancement 00:19:36.740 --> 00:19:40.370 align:middle line:84% throughout the entire period of the steady state 00:19:40.370 --> 00:19:42.140 align:middle line:84% while the 311s ones are decaying. 00:19:42.140 --> 00:19:45.230 align:middle line:84% And then we suddenly say there's a rapid exponential decay right 00:19:45.230 --> 00:19:48.530 align:middle line:84% after some period of time called tau enhanced that we just 00:19:48.530 --> 00:19:51.440 align:middle line:90% calculated in our example. 00:19:51.440 --> 00:19:53.320 align:middle line:90% So that's the simple model. 00:19:53.320 --> 00:19:55.420 align:middle line:84% The critical parameters that determine TED? 00:19:55.420 --> 00:19:57.310 align:middle line:84% Well, the amount of TED we just said 00:19:57.310 --> 00:20:01.210 align:middle line:84% was the supersaturation level, which we write as I over I star 00:20:01.210 --> 00:20:02.950 align:middle line:90% or CI over CI star. 00:20:02.950 --> 00:20:04.900 align:middle line:84% And we know that's a function of temperature. 00:20:04.900 --> 00:20:06.550 align:middle line:84% And this is the functional dependence, 00:20:06.550 --> 00:20:08.560 align:middle line:84% or you can read it right off that plot. 00:20:08.560 --> 00:20:11.630 align:middle line:84% And the duration of the steady-state condition, 00:20:11.630 --> 00:20:14.110 align:middle line:84% which is the so-called tau enhanced time-- 00:20:14.110 --> 00:20:17.170 align:middle line:84% now, that tau enhanced depends linearly on dose. 00:20:17.170 --> 00:20:18.710 align:middle line:90% And doses have a wide range. 00:20:18.710 --> 00:20:19.630 align:middle line:90% So this is the key. 00:20:19.630 --> 00:20:21.040 align:middle line:90% It depends linearly on q. 00:20:21.040 --> 00:20:24.430 align:middle line:84% So if you implant 20 times or 100 times more dose, 00:20:24.430 --> 00:20:27.670 align:middle line:84% you get 20 or 100 times more interstitials. 00:20:27.670 --> 00:20:29.750 align:middle line:90% So it'll last that much longer. 00:20:29.750 --> 00:20:33.190 align:middle line:84% So notice the dose unintuitively somewhat. 00:20:33.190 --> 00:20:36.160 align:middle line:84% The dose doesn't determine CI over CI star. 00:20:36.160 --> 00:20:38.530 align:middle line:84% The dose determines how long the thing lasts, 00:20:38.530 --> 00:20:40.420 align:middle line:90% how long the transient lasts. 00:20:40.420 --> 00:20:43.840 align:middle line:84% And there is a certain small dependence on RP. 00:20:43.840 --> 00:20:45.730 align:middle line:84% And the energy dependence comes from 00:20:45.730 --> 00:20:47.980 align:middle line:84% the dependent-- in the equation, the energy 00:20:47.980 --> 00:20:53.350 align:middle line:84% dependence is represented through the dependency on RP. 00:20:53.350 --> 00:20:54.600 align:middle line:90% So that's the general picture. 00:20:54.600 --> 00:20:57.062 align:middle line:84% And I think that example helps us understand it. 00:20:57.062 --> 00:20:59.270 align:middle line:84% Now, let's look-- anything a little more complicated, 00:20:59.270 --> 00:21:02.180 align:middle line:84% I said we should be using a simulator. 00:21:02.180 --> 00:21:05.090 align:middle line:84% So the next couple of slides here on slide 7, 00:21:05.090 --> 00:21:07.940 align:middle line:84% starting with slide 7, I'm going to show you some SUPREM-IV 00:21:07.940 --> 00:21:12.870 align:middle line:84% examples using more and more sophisticated models. 00:21:12.870 --> 00:21:16.460 align:middle line:84% So the first model, I'm going to use SUPREM-IV using a Gaussian 00:21:16.460 --> 00:21:18.740 align:middle line:84% implant assumption, which of course is not 00:21:18.740 --> 00:21:20.000 align:middle line:90% very sophisticated. 00:21:20.000 --> 00:21:25.850 align:middle line:84% And I'm going to be annealing this arsenic implant at 1,000 00:21:25.850 --> 00:21:28.140 align:middle line:90% degrees C for different times. 00:21:28.140 --> 00:21:30.770 align:middle line:84% And if you look at the open boxes, 00:21:30.770 --> 00:21:33.260 align:middle line:84% that's my initial ion implanted arsenic. 00:21:33.260 --> 00:21:35.240 align:middle line:84% And what's being shown here is a series 00:21:35.240 --> 00:21:38.990 align:middle line:84% of different curves, different symbols for different times. 00:21:38.990 --> 00:21:42.170 align:middle line:84% And what the model that was used in these particular simulations 00:21:42.170 --> 00:21:43.940 align:middle line:84% is called the Fermi model in SUPREM. 00:21:43.940 --> 00:21:46.670 align:middle line:84% When you do your next homework, there'll be different models. 00:21:46.670 --> 00:21:49.340 align:middle line:84% You have to invoke the one here is called Fermi. 00:21:49.340 --> 00:21:53.430 align:middle line:84% As the name suggests, it takes into account the Fermi level, 00:21:53.430 --> 00:21:56.130 align:middle line:84% and therefore the concentration dependence of the diffusivity. 00:21:56.130 --> 00:21:58.680 align:middle line:84% It does not take into account any TED. 00:21:58.680 --> 00:22:03.900 align:middle line:84% So when you tell it to use PD method equals Fermi, 00:22:03.900 --> 00:22:05.460 align:middle line:84% you're not taking into account TED. 00:22:05.460 --> 00:22:07.860 align:middle line:84% So this is just normal ordinary diffusion. 00:22:07.860 --> 00:22:11.070 align:middle line:84% And the junction depth proceeds according 00:22:11.070 --> 00:22:13.998 align:middle line:84% to a square root DT type of behavior, as you would expect. 00:22:13.998 --> 00:22:16.290 align:middle line:84% It does take into account the concentration dependence. 00:22:16.290 --> 00:22:19.770 align:middle line:84% And you see the box-like profile as a result. 00:22:19.770 --> 00:22:24.720 align:middle line:84% Now we go on to a little more sophisticated model. 00:22:24.720 --> 00:22:29.250 align:middle line:84% This is that same implant, 34 KEV 4014. 00:22:29.250 --> 00:22:32.340 align:middle line:84% But now this is on slide number 8. 00:22:32.340 --> 00:22:35.430 align:middle line:84% What we're doing is we're using a Monte Carlo implant 00:22:35.430 --> 00:22:38.100 align:middle line:84% model that changes the as implanted slightly. 00:22:38.100 --> 00:22:40.410 align:middle line:84% But the reason we use that is because the Monte Carlo 00:22:40.410 --> 00:22:43.650 align:middle line:84% model in SUPREM keeps track of the damage because not only does 00:22:43.650 --> 00:22:46.800 align:middle line:84% it track the incoming ions, but every time a silicon gets 00:22:46.800 --> 00:22:49.740 align:middle line:84% displaced, it can keep track of that. 00:22:49.740 --> 00:22:52.110 align:middle line:84% So it can keep track of the damage. 00:22:52.110 --> 00:22:55.350 align:middle line:84% And then we're using a model in SUPREM called the fully coupled 00:22:55.350 --> 00:22:56.820 align:middle line:90% or the full coupled model. 00:22:56.820 --> 00:22:59.490 align:middle line:84% It takes into account these following factors-- 00:22:59.490 --> 00:23:01.380 align:middle line:90% the impact of the Fermi level. 00:23:01.380 --> 00:23:05.910 align:middle line:84% It also takes into account point defect injection and 311 cluster 00:23:05.910 --> 00:23:06.640 align:middle line:90% dissolution. 00:23:06.640 --> 00:23:10.390 align:middle line:84% It has a model there for 311s and their kinetics. 00:23:10.390 --> 00:23:13.290 align:middle line:84% And it also, because it's fully coupled, 00:23:13.290 --> 00:23:16.410 align:middle line:84% defect gradients can drive diffusion. 00:23:16.410 --> 00:23:19.380 align:middle line:84% So if there's a two-dimensional situation 00:23:19.380 --> 00:23:21.420 align:middle line:84% like in the reverse short channel effect, 00:23:21.420 --> 00:23:24.510 align:middle line:84% actually the gradient in the defect can drive a diffusion. 00:23:24.510 --> 00:23:26.700 align:middle line:84% But this is what you would see if you 00:23:26.700 --> 00:23:28.200 align:middle line:90% do this fully coupled model. 00:23:28.200 --> 00:23:30.750 align:middle line:84% Look at-- these first four curves 00:23:30.750 --> 00:23:32.570 align:middle line:84% all lie straight on top of each other. 00:23:32.570 --> 00:23:34.590 align:middle line:90% 30 seconds to 120 seconds. 00:23:34.590 --> 00:23:37.740 align:middle line:90% They're all the same. 00:23:37.740 --> 00:23:38.600 align:middle line:90% Now, why is that? 00:23:38.600 --> 00:23:41.450 align:middle line:90% 00:23:41.450 --> 00:23:43.460 align:middle line:84% Well, I mean, essentially, it's because of TED 00:23:43.460 --> 00:23:47.930 align:middle line:84% because whether you're 30 seconds or 120 seconds, 00:23:47.930 --> 00:23:50.240 align:middle line:84% there's some portion of tau enhance. 00:23:50.240 --> 00:23:51.620 align:middle line:90% There's a certain enhance time. 00:23:51.620 --> 00:23:54.080 align:middle line:84% And during that period, CI over CI star 00:23:54.080 --> 00:23:57.980 align:middle line:84% is so large because of the amount of 311s 00:23:57.980 --> 00:24:00.770 align:middle line:84% that end up being formed that it doesn't matter whether you're 00:24:00.770 --> 00:24:04.250 align:middle line:90% doing seconds or 120. 00:24:04.250 --> 00:24:06.170 align:middle line:84% The diffusion is dominated by whatever 00:24:06.170 --> 00:24:07.940 align:middle line:84% happens during the transient when 00:24:07.940 --> 00:24:09.440 align:middle line:90% the 311s ones are evaporating. 00:24:09.440 --> 00:24:11.480 align:middle line:84% And then after that, if you only-- if you 00:24:11.480 --> 00:24:15.820 align:middle line:84% go to longer times, so say 600 seconds, 1,200, 00:24:15.820 --> 00:24:20.000 align:middle line:84% and 1,800 seconds, do you start to see normal diffusion being 00:24:20.000 --> 00:24:21.050 align:middle line:90% large enough? 00:24:21.050 --> 00:24:23.810 align:middle line:84% Because the time is large enough that normal diffusion now 00:24:23.810 --> 00:24:27.330 align:middle line:84% overtakes the TED and you can actually see that. 00:24:27.330 --> 00:24:30.440 align:middle line:84% And so this is an example of why people didn't discover TED 00:24:30.440 --> 00:24:34.220 align:middle line:84% until they had rapid thermal annealers because in a furnace, 00:24:34.220 --> 00:24:35.570 align:middle line:90% you can't anneal for very short. 00:24:35.570 --> 00:24:38.690 align:middle line:84% So you could never access anneal times that were this short. 00:24:38.690 --> 00:24:42.460 align:middle line:84% And you never would discover that, in fact, TED was going on. 00:24:42.460 --> 00:24:46.380 align:middle line:84% So a signature of TED, if you do an experiment on annealing 00:24:46.380 --> 00:24:47.940 align:middle line:84% and implant, a very clear signature-- 00:24:47.940 --> 00:24:50.730 align:middle line:84% and if you do sims on different times, 00:24:50.730 --> 00:24:52.230 align:middle line:90% is you see very short times. 00:24:52.230 --> 00:24:53.790 align:middle line:90% They all look the same profile. 00:24:53.790 --> 00:24:55.320 align:middle line:90% That's an immediate signature. 00:24:55.320 --> 00:24:57.690 align:middle line:84% Aha, there's some kind of TED going on. 00:24:57.690 --> 00:25:01.440 align:middle line:84% Some very short transient is dominating all the diffusion. 00:25:01.440 --> 00:25:03.360 align:middle line:84% And then you'll go to long enough times 00:25:03.360 --> 00:25:05.850 align:middle line:84% that you'll start to see normal diffusion taking place. 00:25:05.850 --> 00:25:07.380 align:middle line:84% And these are SUPREM simulations. 00:25:07.380 --> 00:25:10.680 align:middle line:84% There's no actual data here the different symbols are not data. 00:25:10.680 --> 00:25:11.340 align:middle line:90% I apologize. 00:25:11.340 --> 00:25:12.870 align:middle line:84% It's for different-- the symbols are 00:25:12.870 --> 00:25:17.020 align:middle line:90% for the different simulations. 00:25:17.020 --> 00:25:18.780 align:middle line:84% OK, so that was a couple of examples. 00:25:18.780 --> 00:25:20.200 align:middle line:90% Let's go on to slide 9. 00:25:20.200 --> 00:25:22.390 align:middle line:84% I want to give-- we're going to go back now. 00:25:22.390 --> 00:25:24.610 align:middle line:84% Now that we've talked about TED and we understand 00:25:24.610 --> 00:25:26.500 align:middle line:84% the models in much more detail, I 00:25:26.500 --> 00:25:28.960 align:middle line:84% want to remind ourselves why we want 00:25:28.960 --> 00:25:32.950 align:middle line:84% to limit diffusion and then talk about some particular device 00:25:32.950 --> 00:25:34.630 align:middle line:90% impacts of TED. 00:25:34.630 --> 00:25:37.240 align:middle line:84% Well, we already know we want to keep junctions shallow. 00:25:37.240 --> 00:25:40.990 align:middle line:84% These junctions here, XJ, either in the deep-source drain 00:25:40.990 --> 00:25:43.240 align:middle line:84% and particularly in the shallow-source drain extension 00:25:43.240 --> 00:25:47.740 align:middle line:84% has to be kept shallow in order to do gate length scaling. 00:25:47.740 --> 00:25:49.760 align:middle line:84% We need steeper lateral junction. 00:25:49.760 --> 00:25:52.300 align:middle line:84% So in this direction, in the L direction, 00:25:52.300 --> 00:25:54.790 align:middle line:84% we need the junctions to be very steep in order 00:25:54.790 --> 00:25:58.180 align:middle line:84% to lower the series resistance at a given effective channel 00:25:58.180 --> 00:25:59.640 align:middle line:90% length. 00:25:59.640 --> 00:26:01.530 align:middle line:84% We need a very small under diffusion 00:26:01.530 --> 00:26:04.920 align:middle line:84% under the gate to reduce the overlap capacitance. 00:26:04.920 --> 00:26:06.900 align:middle line:84% If this underneath the gate, this 00:26:06.900 --> 00:26:09.090 align:middle line:84% diffuses too far under the gate, then you'll 00:26:09.090 --> 00:26:10.540 align:middle line:84% have a lot of overlap capacitance. 00:26:10.540 --> 00:26:14.470 align:middle line:84% And that's going to slow down the circuit speed. 00:26:14.470 --> 00:26:17.290 align:middle line:84% We talked about needing a retrograde well in order 00:26:17.290 --> 00:26:19.660 align:middle line:84% to get better mobility while controlling short channel 00:26:19.660 --> 00:26:20.160 align:middle line:90% effect. 00:26:20.160 --> 00:26:21.850 align:middle line:84% We need this retrograde well in depth. 00:26:21.850 --> 00:26:23.650 align:middle line:84% We need to retrograde the profile 00:26:23.650 --> 00:26:26.860 align:middle line:84% and decrease it towards the surface. 00:26:26.860 --> 00:26:30.070 align:middle line:84% And there's these fancy little angled halo implants. 00:26:30.070 --> 00:26:32.500 align:middle line:84% These halos here are shown in the bright red. 00:26:32.500 --> 00:26:39.070 align:middle line:84% These are implants that are done at an angle to the gate in order 00:26:39.070 --> 00:26:41.910 align:middle line:84% to put the dope just right where we want it. 00:26:41.910 --> 00:26:44.500 align:middle line:84% And if we're using this angled implant 00:26:44.500 --> 00:26:45.937 align:middle line:84% to obtain the right profile, we'd 00:26:45.937 --> 00:26:48.520 align:middle line:84% like to not have that diffuse all over the device because then 00:26:48.520 --> 00:26:49.840 align:middle line:90% it kind of-- 00:26:49.840 --> 00:26:51.500 align:middle line:84% we don't get the shape that we want. 00:26:51.500 --> 00:26:56.030 align:middle line:84% So all of these reasons, we need to limit diffusion. 00:26:56.030 --> 00:26:59.010 align:middle line:84% So I want to talk about, now that we know TED 00:26:59.010 --> 00:27:01.780 align:middle line:84% and go back to this reverse short channel effect 00:27:01.780 --> 00:27:04.152 align:middle line:84% that I introduced three or four lectures ago, 00:27:04.152 --> 00:27:06.610 align:middle line:84% and I think we'll have a better understanding at this time. 00:27:06.610 --> 00:27:09.230 align:middle line:84% If you want to learn a little bit more about it, 00:27:09.230 --> 00:27:12.580 align:middle line:84% there's an article by Rafferty in IED of '93 00:27:12.580 --> 00:27:17.290 align:middle line:84% or Crowder at IBM of '95 where this was first explained. 00:27:17.290 --> 00:27:20.740 align:middle line:84% And what people were trying to explain is the device physicists 00:27:20.740 --> 00:27:23.470 align:middle line:84% in the early 90s were all finding if they-- 00:27:23.470 --> 00:27:24.520 align:middle line:90% and the circuit guys-- 00:27:24.520 --> 00:27:27.220 align:middle line:84% if they plotted their VT a function 00:27:27.220 --> 00:27:32.170 align:middle line:84% of the gate length of the transistors on a chip 00:27:32.170 --> 00:27:34.083 align:middle line:90% or on a wafer. 00:27:34.083 --> 00:27:36.250 align:middle line:84% What they expect is the normal short channel effect. 00:27:36.250 --> 00:27:38.920 align:middle line:84% VT is fairly constant, and you get to short channels, 00:27:38.920 --> 00:27:40.630 align:middle line:84% and the VT is supposed to roll off 00:27:40.630 --> 00:27:43.420 align:middle line:84% due to well-known electrostatics. 00:27:43.420 --> 00:27:46.550 align:middle line:84% What they found and said is the so-called reverse short channel 00:27:46.550 --> 00:27:47.050 align:middle line:90% effect. 00:27:47.050 --> 00:27:49.490 align:middle line:84% It's reversed to one's expectations. 00:27:49.490 --> 00:27:52.090 align:middle line:84% In fact, the VT, the threshold voltage on these devices, 00:27:52.090 --> 00:27:54.430 align:middle line:84% was going up as you made them shorter. 00:27:54.430 --> 00:27:56.803 align:middle line:84% And it was going up and peaking quite a bit. 00:27:56.803 --> 00:27:58.970 align:middle line:84% And then eventually, the normal short channel effect 00:27:58.970 --> 00:27:59.678 align:middle line:90% would take place. 00:27:59.678 --> 00:28:01.610 align:middle line:84% So this so-called reverse short channel effect 00:28:01.610 --> 00:28:03.290 align:middle line:84% was quite bothersome in the early days 00:28:03.290 --> 00:28:04.832 align:middle line:84% because people didn't understand when 00:28:04.832 --> 00:28:06.450 align:middle line:84% they scaled to these channel lengths 00:28:06.450 --> 00:28:08.360 align:middle line:90% why VT would be going up. 00:28:08.360 --> 00:28:11.100 align:middle line:90% How could it be? 00:28:11.100 --> 00:28:15.050 align:middle line:84% Well, we talked qualitatively about the reasoning 00:28:15.050 --> 00:28:16.300 align:middle line:90% for this several lectures ago. 00:28:16.300 --> 00:28:18.175 align:middle line:84% Let me review the qualitative, and then we'll 00:28:18.175 --> 00:28:19.220 align:middle line:90% do the more quantitative. 00:28:19.220 --> 00:28:22.038 align:middle line:84% This is an actual simulation from SUPREM 00:28:22.038 --> 00:28:24.330 align:middle line:84% of what's going on in the reverse short channel effect. 00:28:24.330 --> 00:28:26.420 align:middle line:90% So we have our gate. 00:28:26.420 --> 00:28:28.580 align:middle line:90% You see the side wall spacers. 00:28:28.580 --> 00:28:32.690 align:middle line:84% This little region under here is the source extension. 00:28:32.690 --> 00:28:35.270 align:middle line:84% This is the drain extension and the deep source 00:28:35.270 --> 00:28:36.440 align:middle line:90% and the deep drain. 00:28:36.440 --> 00:28:38.510 align:middle line:84% And what's happened is that we have 00:28:38.510 --> 00:28:42.310 align:middle line:84% implanted these source drain regions and their extensions. 00:28:42.310 --> 00:28:44.360 align:middle line:84% We've implanted, say, with arsenic. 00:28:44.360 --> 00:28:46.950 align:middle line:90% And we've generated a flux. 00:28:46.950 --> 00:28:48.650 align:middle line:84% In fact, these arrows are supposed 00:28:48.650 --> 00:28:50.640 align:middle line:84% to indicate the interstitial fluxes. 00:28:50.640 --> 00:28:52.760 align:middle line:84% So these are silicon interstitial atoms 00:28:52.760 --> 00:28:54.680 align:middle line:90% that are diffusing. 00:28:54.680 --> 00:28:56.900 align:middle line:84% And in fact, and they're then recombining. 00:28:56.900 --> 00:28:58.430 align:middle line:84% They can recombine the bulk or they 00:28:58.430 --> 00:29:00.873 align:middle line:90% can recombine in the surface. 00:29:00.873 --> 00:29:02.540 align:middle line:84% And because of the way the recombination 00:29:02.540 --> 00:29:04.910 align:middle line:84% goes at the surface, they end up creating 00:29:04.910 --> 00:29:07.700 align:middle line:90% a flux that goes like this. 00:29:07.700 --> 00:29:12.460 align:middle line:84% And now that flux and the gradient of the interstitials 00:29:12.460 --> 00:29:14.860 align:middle line:84% is so sharp that it actually can drive 00:29:14.860 --> 00:29:17.540 align:middle line:90% uphill diffusion of the boron. 00:29:17.540 --> 00:29:19.190 align:middle line:84% And you can see that uphill diffusion. 00:29:19.190 --> 00:29:20.607 align:middle line:84% You can't see it in this, but if I 00:29:20.607 --> 00:29:23.030 align:middle line:84% take a cut right through the center here at X 00:29:23.030 --> 00:29:25.280 align:middle line:84% equals 0 and plot it versus depth, 00:29:25.280 --> 00:29:27.800 align:middle line:84% so boron concentration versus depth 00:29:27.800 --> 00:29:30.140 align:middle line:90% at the center of the channel. 00:29:30.140 --> 00:29:33.650 align:middle line:84% If you have a 1-micron device on the chip it looks like this. 00:29:33.650 --> 00:29:35.930 align:middle line:84% If you have a 0.18-micron device on the chip, 00:29:35.930 --> 00:29:38.997 align:middle line:90% it looks like the green curve. 00:29:38.997 --> 00:29:39.830 align:middle line:90% And what do you see? 00:29:39.830 --> 00:29:41.420 align:middle line:84% In the green curve, well, for one thing, 00:29:41.420 --> 00:29:42.837 align:middle line:84% the surface concentration of boron 00:29:42.837 --> 00:29:45.620 align:middle line:84% is a lot higher than it is in the 1-micron device. 00:29:45.620 --> 00:29:47.900 align:middle line:84% So that explains to the circuit designer, 00:29:47.900 --> 00:29:49.330 align:middle line:84% oh, I have a higher concentration 00:29:49.330 --> 00:29:50.330 align:middle line:90% of boron on the surface. 00:29:50.330 --> 00:29:52.040 align:middle line:90% That makes my VT higher. 00:29:52.040 --> 00:29:54.260 align:middle line:90% So that explains the VT effect. 00:29:54.260 --> 00:29:57.440 align:middle line:84% How it got to be so high, and how did-- 00:29:57.440 --> 00:30:00.980 align:middle line:84% look at even the peak of the boron profile in the 0.18 micron 00:30:00.980 --> 00:30:01.700 align:middle line:90% device. 00:30:01.700 --> 00:30:03.560 align:middle line:84% Even the peak move towards the surface. 00:30:03.560 --> 00:30:05.480 align:middle line:84% Again, that's totally non-Fickian diffusion. 00:30:05.480 --> 00:30:06.950 align:middle line:84% If I give you a Gaussian diffusion 00:30:06.950 --> 00:30:10.160 align:middle line:84% and you diffuse it in your calculator, the peak stays put. 00:30:10.160 --> 00:30:11.780 align:middle line:90% It just goes down. 00:30:11.780 --> 00:30:14.270 align:middle line:84% Peak doesn't suddenly move over to the left or the right. 00:30:14.270 --> 00:30:15.440 align:middle line:90% That's not Gaussian. 00:30:15.440 --> 00:30:17.450 align:middle line:84% Well, what's pulling this peak over 00:30:17.450 --> 00:30:20.960 align:middle line:84% is the fact that there's these interstitials that 00:30:20.960 --> 00:30:21.830 align:middle line:90% have a gradient. 00:30:21.830 --> 00:30:23.540 align:middle line:84% And they drag with them because remember, 00:30:23.540 --> 00:30:26.630 align:middle line:84% boron likes to diffuse with interstitials. 00:30:26.630 --> 00:30:29.330 align:middle line:84% It's diffusing as a pair because of the fully coupled diffusion. 00:30:29.330 --> 00:30:32.190 align:middle line:84% They drag with it the boron peak. 00:30:32.190 --> 00:30:36.320 align:middle line:84% So this is the qualitative explanation, at least, 00:30:36.320 --> 00:30:40.540 align:middle line:84% of what's going on in the reverse short channel effect. 00:30:40.540 --> 00:30:42.550 align:middle line:84% And now let's go through a little bit more 00:30:42.550 --> 00:30:44.830 align:middle line:84% carefully the articles by Rafferty and Crowder. 00:30:44.830 --> 00:30:47.860 align:middle line:84% And you can see what it is that they-- how they explain 00:30:47.860 --> 00:30:50.140 align:middle line:90% this in a little more detail. 00:30:50.140 --> 00:30:52.630 align:middle line:84% These are some of the key process steps that 00:30:52.630 --> 00:30:54.550 align:middle line:90% influence the channel profile. 00:30:54.550 --> 00:30:56.230 align:middle line:84% So the reverse short channel effect 00:30:56.230 --> 00:31:00.010 align:middle line:84% ends up all up about being what determines the channel profile. 00:31:00.010 --> 00:31:03.250 align:middle line:84% So step number one, we know we do an ion implant 00:31:03.250 --> 00:31:04.310 align:middle line:90% and it has some shape. 00:31:04.310 --> 00:31:07.300 align:middle line:84% This is supposed to represent, if you turn your head sideways, 00:31:07.300 --> 00:31:08.740 align:middle line:90% the shape of on implant. 00:31:08.740 --> 00:31:10.948 align:middle line:84% So we do an implant and a couple of energies. 00:31:10.948 --> 00:31:13.240 align:middle line:84% It peaks here, and then there's another peak down here. 00:31:13.240 --> 00:31:17.940 align:middle line:84% That's what we expect it to look like in the as-implanted case. 00:31:17.940 --> 00:31:20.070 align:middle line:84% Step number two in making the device. 00:31:20.070 --> 00:31:21.280 align:middle line:90% Well, we grow a gate oxide. 00:31:21.280 --> 00:31:22.050 align:middle line:84% All right, there's a certain amount 00:31:22.050 --> 00:31:24.180 align:middle line:84% of thermal budget associated with the gate oxide-- 00:31:24.180 --> 00:31:25.350 align:middle line:90% 800 degrees, whatever. 00:31:25.350 --> 00:31:26.580 align:middle line:90% It'll diffuse a little. 00:31:26.580 --> 00:31:28.170 align:middle line:90% OK, we can understand that. 00:31:28.170 --> 00:31:29.550 align:middle line:90% That's normal diffusion. 00:31:29.550 --> 00:31:32.890 align:middle line:84% Step number three, we put down a gate and we pattern it. 00:31:32.890 --> 00:31:34.140 align:middle line:90% So it has a pattern like that. 00:31:34.140 --> 00:31:35.970 align:middle line:84% Usually polysilicon gates are put down 00:31:35.970 --> 00:31:38.970 align:middle line:84% at 600, 500-- very low temperatures. 00:31:38.970 --> 00:31:41.460 align:middle line:84% Not a whole lot of motion going on. 00:31:41.460 --> 00:31:45.330 align:middle line:84% All right, now the step number four 00:31:45.330 --> 00:31:47.310 align:middle line:84% is called the lightly doped drain. 00:31:47.310 --> 00:31:49.957 align:middle line:84% Actually, we don't use that terminology anymore in devices. 00:31:49.957 --> 00:31:51.540 align:middle line:84% We call it the source drain extension. 00:31:51.540 --> 00:31:53.340 align:middle line:84% It's a shallow source drain right here. 00:31:53.340 --> 00:31:56.310 align:middle line:84% That gets an ion implanted and it gets masked by the gate. 00:31:56.310 --> 00:32:00.060 align:middle line:84% So when you're doing step number four, the sidewall is not there. 00:32:00.060 --> 00:32:01.890 align:middle line:84% Step number five hasn't been done yet. 00:32:01.890 --> 00:32:04.920 align:middle line:84% So you have this shallow source drain extension. 00:32:04.920 --> 00:32:08.400 align:middle line:84% So there you're injecting a certain number of point defects. 00:32:08.400 --> 00:32:09.330 align:middle line:90% OK? 00:32:09.330 --> 00:32:10.503 align:middle line:90% Now I take that. 00:32:10.503 --> 00:32:11.920 align:middle line:84% I've injected these point defects, 00:32:11.920 --> 00:32:14.100 align:middle line:84% and now I have to form the sidewall spacer. 00:32:14.100 --> 00:32:16.470 align:middle line:84% Depending on what material form the spacer, 00:32:16.470 --> 00:32:18.600 align:middle line:84% if I form it of silicon nitride, that goes down 00:32:18.600 --> 00:32:19.950 align:middle line:90% at 800 for about an hour. 00:32:19.950 --> 00:32:20.880 align:middle line:90% Uh oh. 00:32:20.880 --> 00:32:22.650 align:middle line:90% Bad temperature. 00:32:22.650 --> 00:32:24.040 align:middle line:90% 800 for about an hour. 00:32:24.040 --> 00:32:25.410 align:middle line:90% We just did that calculation. 00:32:25.410 --> 00:32:27.810 align:middle line:84% 800 for about an hour, you can get a lot of TED 00:32:27.810 --> 00:32:31.840 align:middle line:84% can have a 7,000, perhaps, enhancement in your diffusivity. 00:32:31.840 --> 00:32:33.850 align:middle line:84% So you got to watch out for that. 00:32:33.850 --> 00:32:36.780 align:middle line:84% That's why nitride spacers are a little tricky. 00:32:36.780 --> 00:32:39.480 align:middle line:84% If you do a low-temperature oxide deposit in space-- 00:32:39.480 --> 00:32:42.010 align:middle line:84% or you can do that at 400 so it's not so bad. 00:32:42.010 --> 00:32:43.560 align:middle line:90% You don't have to worry so much. 00:32:43.560 --> 00:32:46.230 align:middle line:84% All right, so there's a possible indication of problem. 00:32:46.230 --> 00:32:50.280 align:middle line:84% And now number six, we do the deep source drain implant. 00:32:50.280 --> 00:32:52.080 align:middle line:84% Again, that's probably going to be arsenic. 00:32:52.080 --> 00:32:54.300 align:middle line:90% Pretty high dose, 10 to the 15. 00:32:54.300 --> 00:32:56.220 align:middle line:84% Not a great thing in terms of TED. 00:32:56.220 --> 00:32:58.980 align:middle line:84% I'm introducing a lot of point defects 00:32:58.980 --> 00:33:02.610 align:middle line:84% that can then come in here and enhance the boron 00:33:02.610 --> 00:33:05.580 align:middle line:90% implant, the boron profile. 00:33:05.580 --> 00:33:07.740 align:middle line:84% So I just wanted to give you the order of the steps 00:33:07.740 --> 00:33:10.860 align:middle line:84% so you can understand how all these things fit together 00:33:10.860 --> 00:33:14.692 align:middle line:84% to determine the channel profile. 00:33:14.692 --> 00:33:16.150 align:middle line:84% And if you want to go a little more 00:33:16.150 --> 00:33:19.240 align:middle line:84% sophisticated into Rafferty's IDM article that's 00:33:19.240 --> 00:33:23.830 align:middle line:84% being shown here on slide number 13 of your handout, 00:33:23.830 --> 00:33:27.330 align:middle line:84% what he calculated here-- is so this is the edge of the gate. 00:33:27.330 --> 00:33:29.250 align:middle line:84% This is a two-dimensional plot out 00:33:29.250 --> 00:33:30.960 align:middle line:90% of a two-dimensional simulator. 00:33:30.960 --> 00:33:32.760 align:middle line:90% And these are contours. 00:33:32.760 --> 00:33:35.520 align:middle line:84% And what he's done is he's ion implanted 00:33:35.520 --> 00:33:41.250 align:middle line:84% a certain amount of damage or dose into the drain region. 00:33:41.250 --> 00:33:44.520 align:middle line:84% And he's calculated something called the time integral 00:33:44.520 --> 00:33:46.720 align:middle line:90% of the supersaturation ratio. 00:33:46.720 --> 00:33:50.340 align:middle line:84% So it is essentially-- and he calls it the enhanced time, 00:33:50.340 --> 00:33:53.640 align:middle line:84% but it's the integral of CI over CI star, 00:33:53.640 --> 00:33:56.900 align:middle line:90% integrated over the anneal. 00:33:56.900 --> 00:33:59.730 align:middle line:84% And he's plotted in terms of contours. 00:33:59.730 --> 00:34:05.520 align:middle line:84% So these are profiles of damage caused by this shallow source 00:34:05.520 --> 00:34:07.800 align:middle line:90% drain implant. 00:34:07.800 --> 00:34:12.130 align:middle line:84% And he represents this damage by this integral of CI over CI 00:34:12.130 --> 00:34:17.550 align:middle line:84% star, which he calls the enhanced time, so to speak. 00:34:17.550 --> 00:34:20.969 align:middle line:84% And a number here that corresponds to 10 to 6 00:34:20.969 --> 00:34:22.290 align:middle line:90% has the units of seconds. 00:34:22.290 --> 00:34:26.400 align:middle line:84% So it's as if you did a 10 to the 6 second anneal, 00:34:26.400 --> 00:34:28.440 align:middle line:90% essentially. 00:34:28.440 --> 00:34:31.860 align:middle line:84% And that corresponds to an average supersaturation 00:34:31.860 --> 00:34:35.070 align:middle line:90% of about a factor of 1,000. 00:34:35.070 --> 00:34:36.983 align:middle line:84% If the real time-- if this were really-- 00:34:36.983 --> 00:34:38.400 align:middle line:84% the real time is about 20 minutes. 00:34:38.400 --> 00:34:40.800 align:middle line:84% This cutoff at the bottom of your 20-minute real time 00:34:40.800 --> 00:34:42.420 align:middle line:90% is equivalent to 1,200 seconds. 00:34:42.420 --> 00:34:44.440 align:middle line:84% So there's huge amount of enhancement. 00:34:44.440 --> 00:34:47.969 align:middle line:84% But look at these contours look how they go down. 00:34:47.969 --> 00:34:50.250 align:middle line:84% You're going from 10 to the 6 here down to 10 00:34:50.250 --> 00:34:51.460 align:middle line:90% to the 5 in this range. 00:34:51.460 --> 00:34:55.590 align:middle line:84% So again, I have a large gradient in this CI over CI 00:34:55.590 --> 00:34:56.100 align:middle line:90% star. 00:34:56.100 --> 00:35:00.240 align:middle line:84% And the gradient is pointing me, pushing me, towards the gate 00:35:00.240 --> 00:35:02.130 align:middle line:90% and towards the gate oxide. 00:35:02.130 --> 00:35:04.710 align:middle line:90% 00:35:04.710 --> 00:35:08.250 align:middle line:84% Don't forget he's saying that the oxide interface under here 00:35:08.250 --> 00:35:13.270 align:middle line:84% acts as a sink for recombination of interstitials. 00:35:13.270 --> 00:35:17.140 align:middle line:84% OK, so this is the origin of the reverse short channel effect 00:35:17.140 --> 00:35:19.120 align:middle line:90% as he explained it on slide 14. 00:35:19.120 --> 00:35:21.970 align:middle line:84% The implant damage from the shallow source 00:35:21.970 --> 00:35:25.270 align:middle line:84% drains sets up a retrograde CI over CI 00:35:25.270 --> 00:35:28.000 align:middle line:90% star profile under the gate. 00:35:28.000 --> 00:35:30.290 align:middle line:84% And the gradient in this profile, 00:35:30.290 --> 00:35:33.100 align:middle line:84% so this grad of I over I star-- because again, 00:35:33.100 --> 00:35:35.410 align:middle line:90% it's diffusing as a pair-- 00:35:35.410 --> 00:35:38.980 align:middle line:84% results in an extra flux in the boron diffusion that 00:35:38.980 --> 00:35:41.110 align:middle line:84% wouldn't be there if it were diffusing alone, 00:35:41.110 --> 00:35:44.200 align:middle line:84% but because it diffuses as a pair with interstitials. 00:35:44.200 --> 00:35:47.800 align:middle line:84% So that retrograde causes a boron pileup at the interface. 00:35:47.800 --> 00:35:52.060 align:middle line:84% And the shorter the channel, the higher the pileup 00:35:52.060 --> 00:35:54.610 align:middle line:84% because as you make the channel shorter, you bring in those 00:35:54.610 --> 00:35:58.310 align:middle line:84% source drains closer and closer into the center of the channel. 00:35:58.310 --> 00:36:01.060 align:middle line:84% So that explains the reverse short channel effect. 00:36:01.060 --> 00:36:04.810 align:middle line:84% Why does the boron pileup get bigger and bigger, and therefore 00:36:04.810 --> 00:36:08.740 align:middle line:84% the VT goes up and up for a shorter device. 00:36:08.740 --> 00:36:09.240 align:middle line:90% Oops. 00:36:09.240 --> 00:36:11.590 align:middle line:90% And in fact, here's an example,. 00:36:11.590 --> 00:36:16.320 align:middle line:84% If we go onto slide 15 from his article, 00:36:16.320 --> 00:36:19.440 align:middle line:84% this is boron concentration simulated versus depth. 00:36:19.440 --> 00:36:22.460 align:middle line:84% Now, he's doing this at the center of the channel. 00:36:22.460 --> 00:36:25.050 align:middle line:84% And say a long channel device looks like this. 00:36:25.050 --> 00:36:26.420 align:middle line:90% Here's the boron he simulated. 00:36:26.420 --> 00:36:28.820 align:middle line:84% It basically has the as-implanted shape. 00:36:28.820 --> 00:36:29.690 align:middle line:90% There two implants. 00:36:29.690 --> 00:36:31.610 align:middle line:84% There's a low energy implant-- you 00:36:31.610 --> 00:36:33.750 align:middle line:84% can see its peak-- and a slightly higher one. 00:36:33.750 --> 00:36:37.290 align:middle line:84% So this boron at the center looks sort of like this. 00:36:37.290 --> 00:36:39.800 align:middle line:84% Now, if you take the boron that's 00:36:39.800 --> 00:36:41.390 align:middle line:84% the center of a 2-micron long channel, 00:36:41.390 --> 00:36:44.540 align:middle line:84% if you go to a 0.45 micron device, it looks like this one. 00:36:44.540 --> 00:36:46.430 align:middle line:84% There's a huge pileup at the surface. 00:36:46.430 --> 00:36:48.930 align:middle line:84% And the peak is close to the surface. 00:36:48.930 --> 00:36:55.270 align:middle line:84% So the surface doping could be three to five times different 00:36:55.270 --> 00:36:57.460 align:middle line:84% when you have a very short channel device compared 00:36:57.460 --> 00:37:00.610 align:middle line:84% to a long channel device because of the influence of these point 00:37:00.610 --> 00:37:03.230 align:middle line:84% defect gradients that get set up. 00:37:03.230 --> 00:37:06.580 align:middle line:84% And so these very different surface dopings 00:37:06.580 --> 00:37:09.310 align:middle line:84% can then be used to calculate the VT difference 00:37:09.310 --> 00:37:10.330 align:middle line:90% between these devices. 00:37:10.330 --> 00:37:12.970 align:middle line:84% And that's exactly what Rafferty did. 00:37:12.970 --> 00:37:17.770 align:middle line:84% And this is taken, again, from his article, 1993. 00:37:17.770 --> 00:37:21.847 align:middle line:84% He calculated here, based on those profiles, 00:37:21.847 --> 00:37:23.680 align:middle line:84% what he thought the threshold voltage should 00:37:23.680 --> 00:37:26.620 align:middle line:84% be as a function of 1 over the channel length, 00:37:26.620 --> 00:37:30.340 align:middle line:84% or if you want to read the channel length on the top axis 00:37:30.340 --> 00:37:32.290 align:middle line:90% for different biases. 00:37:32.290 --> 00:37:35.290 align:middle line:84% And if you look at his calculations here, 00:37:35.290 --> 00:37:39.370 align:middle line:84% this smooth, solid line is, including the TED effect, 00:37:39.370 --> 00:37:41.560 align:middle line:84% he predicted that the channel length would go-- 00:37:41.560 --> 00:37:45.490 align:middle line:84% the VT would go up like this as I go to shorter channel lengths. 00:37:45.490 --> 00:37:47.710 align:middle line:84% And the data-- these are the experimental data 00:37:47.710 --> 00:37:49.720 align:middle line:84% they obtained from devices at Bell Labs. 00:37:49.720 --> 00:37:51.700 align:middle line:84% Indeed, you can see the VT going up. 00:37:51.700 --> 00:37:55.630 align:middle line:84% So look at the VT for a 2-micron device is over here. 00:37:55.630 --> 00:37:57.680 align:middle line:84% The threshold voltage is about a volt. 00:37:57.680 --> 00:38:00.850 align:middle line:84% And for a 0.45 micron device, which 00:38:00.850 --> 00:38:05.140 align:middle line:84% is, I think, right about here, the VT is about 1.3 volts. 00:38:05.140 --> 00:38:06.700 align:middle line:90% And that agrees very well. 00:38:06.700 --> 00:38:11.350 align:middle line:84% So that roll-up of the VT agreed very well with the simulation 00:38:11.350 --> 00:38:13.090 align:middle line:90% when he included the TED. 00:38:13.090 --> 00:38:16.510 align:middle line:84% The dashed line is a simulation without the TED. 00:38:16.510 --> 00:38:20.440 align:middle line:84% And indeed, you see the normal short channel effect going down 00:38:20.440 --> 00:38:21.380 align:middle line:90% like this. 00:38:21.380 --> 00:38:23.170 align:middle line:84% So the only way people could really 00:38:23.170 --> 00:38:24.730 align:middle line:84% understand the reverse short channel 00:38:24.730 --> 00:38:28.990 align:middle line:84% effect was to really understand in detail what was happening 00:38:28.990 --> 00:38:31.610 align:middle line:84% with the boron TED and the influence of the damage 00:38:31.610 --> 00:38:33.440 align:middle line:84% on each side of the channel on what 00:38:33.440 --> 00:38:36.830 align:middle line:84% was going on inside, underneath the gate. 00:38:36.830 --> 00:38:40.850 align:middle line:84% So that's kind of a famous paper on how understanding 00:38:40.850 --> 00:38:42.890 align:middle line:84% these process models end up influencing 00:38:42.890 --> 00:38:48.710 align:middle line:84% the device model, which ends up influencing the circuit model. 00:38:48.710 --> 00:38:49.507 align:middle line:90% So there's a lot-- 00:38:49.507 --> 00:38:51.340 align:middle line:84% in simulators, there's been a certain amount 00:38:51.340 --> 00:38:54.790 align:middle line:84% of work in the early 90s on how to model reverse short channel 00:38:54.790 --> 00:38:56.830 align:middle line:84% effect, what are important parameters. 00:38:56.830 --> 00:39:00.730 align:middle line:84% Well, obviously, the magnitude of the initial implanted damage. 00:39:00.730 --> 00:39:03.465 align:middle line:84% Well, you know the dose of the source and drain extensions 00:39:03.465 --> 00:39:04.340 align:middle line:90% and the source drain. 00:39:04.340 --> 00:39:06.190 align:middle line:90% So that's important. 00:39:06.190 --> 00:39:10.270 align:middle line:84% But you need to figure out exactly how much I over I 00:39:10.270 --> 00:39:11.260 align:middle line:90% star there is. 00:39:11.260 --> 00:39:13.150 align:middle line:84% We did it in our book, our textbook, 00:39:13.150 --> 00:39:15.460 align:middle line:84% using a very simple analytic calculation. 00:39:15.460 --> 00:39:18.130 align:middle line:84% We calculate the maximum CI over CI star. 00:39:18.130 --> 00:39:20.650 align:middle line:84% Remember, that was a ratio of k reverse 00:39:20.650 --> 00:39:22.840 align:middle line:90% to k forward in that equation. 00:39:22.840 --> 00:39:24.580 align:middle line:84% And we put some estimates down for that, 00:39:24.580 --> 00:39:26.390 align:middle line:84% but it's not clear exactly how accurate. 00:39:26.390 --> 00:39:29.350 align:middle line:84% So depending on your simulator, the way they calculate it 00:39:29.350 --> 00:39:30.830 align:middle line:90% will be slightly different. 00:39:30.830 --> 00:39:33.340 align:middle line:84% So these are different clustering factors 00:39:33.340 --> 00:39:34.667 align:middle line:90% in this simulator. 00:39:34.667 --> 00:39:36.250 align:middle line:84% They have a parameter called a cluster 00:39:36.250 --> 00:39:38.800 align:middle line:84% factor that will adjust, essentially, 00:39:38.800 --> 00:39:40.870 align:middle line:90% the CI over CI star max. 00:39:40.870 --> 00:39:45.047 align:middle line:84% And depending on how that is, that CI over CI star max, 00:39:45.047 --> 00:39:47.630 align:middle line:84% you'll get different amounts of reverse short channel effects. 00:39:47.630 --> 00:39:51.160 align:middle line:84% So I apologize here, the vertical axis here is VT. 00:39:51.160 --> 00:39:52.540 align:middle line:90% That got cut off somehow. 00:39:52.540 --> 00:39:54.250 align:middle line:84% So this is the threshold voltage. 00:39:54.250 --> 00:39:56.750 align:middle line:84% And this is the gate length calculated. 00:39:56.750 --> 00:39:58.858 align:middle line:84% And you can see in this particular simulator-- 00:39:58.858 --> 00:39:59.650 align:middle line:90% this is not SUPREM. 00:39:59.650 --> 00:40:02.450 align:middle line:90% This is a silvaco simulator. 00:40:02.450 --> 00:40:06.110 align:middle line:84% But for a cluster factor of 0, they 00:40:06.110 --> 00:40:08.660 align:middle line:84% didn't predict any reverse short channel effect. 00:40:08.660 --> 00:40:10.280 align:middle line:84% When they allow clustering and they 00:40:10.280 --> 00:40:14.350 align:middle line:84% allow these 311s to come in, that you do 00:40:14.350 --> 00:40:16.780 align:middle line:84% see a roll-up, a reverse short channel effect. 00:40:16.780 --> 00:40:19.720 align:middle line:84% And depending on the magnitude of the cluster factor, 00:40:19.720 --> 00:40:24.220 align:middle line:84% reverse short channel effect can be more or less prominent. 00:40:24.220 --> 00:40:26.553 align:middle line:84% So there will be usually-- depending 00:40:26.553 --> 00:40:27.970 align:middle line:84% on the simulator you use, there'll 00:40:27.970 --> 00:40:29.350 align:middle line:84% be a couple of parameters one can 00:40:29.350 --> 00:40:33.290 align:middle line:84% tweak to effect both the initial implant damage 00:40:33.290 --> 00:40:35.365 align:middle line:90% and the recombination rate. 00:40:35.365 --> 00:40:36.740 align:middle line:84% This is a two-dimensional effect. 00:40:36.740 --> 00:40:40.930 align:middle line:84% So not only how much damage and how many 311s do I end up with, 00:40:40.930 --> 00:40:42.400 align:middle line:84% and how many interstitial-- what's 00:40:42.400 --> 00:40:45.520 align:middle line:84% the interstitial concentration that's important, 00:40:45.520 --> 00:40:49.720 align:middle line:84% but how those fluxes diffuse and how they recombine at that oxide 00:40:49.720 --> 00:40:52.270 align:middle line:84% interface will determine the actual gradient 00:40:52.270 --> 00:40:53.380 align:middle line:90% of the interstitials. 00:40:53.380 --> 00:40:55.990 align:middle line:84% So the k sub s factor is also important, 00:40:55.990 --> 00:40:59.950 align:middle line:84% the recombination rate of interstitials at the gate oxide 00:40:59.950 --> 00:41:01.390 align:middle line:90% interface in the channel. 00:41:01.390 --> 00:41:04.155 align:middle line:84% And that's a parameter that one can adjust. 00:41:04.155 --> 00:41:05.530 align:middle line:84% Hopefully it's fairly well-known, 00:41:05.530 --> 00:41:08.380 align:middle line:84% but by adjusting that parameter, you can change these curves, 00:41:08.380 --> 00:41:08.980 align:middle line:90% as well. 00:41:08.980 --> 00:41:12.070 align:middle line:90% 00:41:12.070 --> 00:41:14.110 align:middle line:84% Annealing temperature is important, as you know. 00:41:14.110 --> 00:41:17.800 align:middle line:84% Again, this is from that silvaco simulator, as well. 00:41:17.800 --> 00:41:19.920 align:middle line:90% Again, VT versus gate length. 00:41:19.920 --> 00:41:21.690 align:middle line:84% And the different color curves here 00:41:21.690 --> 00:41:23.530 align:middle line:84% are for different annealing temperatures. 00:41:23.530 --> 00:41:26.910 align:middle line:84% And as you might imagine, low temperatures give you 00:41:26.910 --> 00:41:28.980 align:middle line:84% more reverse short channel effect 00:41:28.980 --> 00:41:30.900 align:middle line:84% because we know at low temperatures, 00:41:30.900 --> 00:41:32.710 align:middle line:90% TED is more prominent. 00:41:32.710 --> 00:41:36.070 align:middle line:84% The CI over CI star hangs around longer. 00:41:36.070 --> 00:41:37.990 align:middle line:84% The time of the enhancement is worse. 00:41:37.990 --> 00:41:41.400 align:middle line:84% So here at 850 in the red, you see more a little bit more 00:41:41.400 --> 00:41:43.290 align:middle line:84% of a roll-up than if you were to anneal 00:41:43.290 --> 00:41:45.060 align:middle line:90% at very high temperatures. 00:41:45.060 --> 00:41:47.940 align:middle line:84% In this particular anneal, I don't know what the dose was. 00:41:47.940 --> 00:41:50.770 align:middle line:84% There wasn't much reverse short channel effect. 00:41:50.770 --> 00:41:53.670 align:middle line:84% So lower temperatures tend to change that. 00:41:53.670 --> 00:41:57.060 align:middle line:84% And that's just because of the CI over CI star term, 00:41:57.060 --> 00:42:04.120 align:middle line:84% as well as the tau enhanced term, depend on temperature. 00:42:04.120 --> 00:42:08.830 align:middle line:84% And in fact, the next slide in your handout, slide number 19, 00:42:08.830 --> 00:42:11.200 align:middle line:84% is just that same plot that we've 00:42:11.200 --> 00:42:14.290 align:middle line:84% used to do the example a few minutes ago to remind you 00:42:14.290 --> 00:42:16.870 align:middle line:84% that CI over CI star is increasing 00:42:16.870 --> 00:42:18.560 align:middle line:90% as we go lower in temperature. 00:42:18.560 --> 00:42:22.860 align:middle line:84% And that's why you see more of a reverse short channel effect 00:42:22.860 --> 00:42:27.280 align:middle line:90% as you lower the temperature. 00:42:27.280 --> 00:42:29.613 align:middle line:84% So not only does the VT change, but just 00:42:29.613 --> 00:42:31.530 align:middle line:84% to give you an idea of other device parameters 00:42:31.530 --> 00:42:34.320 align:middle line:84% that can change depending on the channel profile, 00:42:34.320 --> 00:42:37.770 align:middle line:84% that boron pileup at the surface underneath the gate 00:42:37.770 --> 00:42:39.960 align:middle line:84% can decrease the channel mobility. 00:42:39.960 --> 00:42:42.900 align:middle line:84% Channel mobility, at very high concentrations 00:42:42.900 --> 00:42:46.198 align:middle line:84% of channel dopant, you get more scattering of the electrons, 00:42:46.198 --> 00:42:47.490 align:middle line:90% more scattering of the carrier. 00:42:47.490 --> 00:42:48.960 align:middle line:90% So the mobility can go down. 00:42:48.960 --> 00:42:52.140 align:middle line:84% And in fact, this is a plot of a calculation of the mobility 00:42:52.140 --> 00:42:55.380 align:middle line:84% that Rafferty made in that article here on slide 20. 00:42:55.380 --> 00:42:59.760 align:middle line:84% So he's calculated the mobility as a function of the channel 00:42:59.760 --> 00:43:03.380 align:middle line:90% length for two different doses. 00:43:03.380 --> 00:43:07.700 align:middle line:84% This EXT dose-- EXT stands for the source drain extension. 00:43:07.700 --> 00:43:09.680 align:middle line:84% So that's the dose that gets implanted 00:43:09.680 --> 00:43:10.910 align:middle line:90% right next to the gate. 00:43:10.910 --> 00:43:15.050 align:middle line:84% Right after you cut and define the gate poly, 00:43:15.050 --> 00:43:17.360 align:middle line:84% you implant the source drain extensions. 00:43:17.360 --> 00:43:20.450 align:middle line:84% And here this is for a 3e12, a relatively low dose. 00:43:20.450 --> 00:43:22.520 align:middle line:84% Not that many interstitials are injected. 00:43:22.520 --> 00:43:25.260 align:middle line:84% The boron profile looks pretty much as implanted, 00:43:25.260 --> 00:43:26.630 align:middle line:90% and you don't get much pileup. 00:43:26.630 --> 00:43:28.820 align:middle line:84% So the mobility stays pretty high 00:43:28.820 --> 00:43:31.190 align:middle line:84% until you get to very short channels. 00:43:31.190 --> 00:43:36.050 align:middle line:84% If you do an extension dose here that's about three times that-- 00:43:36.050 --> 00:43:40.100 align:middle line:84% say about 8e12-- what happens is you get this reverse short 00:43:40.100 --> 00:43:43.550 align:middle line:84% channel effect to get the boron being drawn to the surface. 00:43:43.550 --> 00:43:45.530 align:middle line:84% A very high amount of boron in the channel 00:43:45.530 --> 00:43:48.920 align:middle line:84% means you have a lot of ionized acceptors. 00:43:48.920 --> 00:43:51.260 align:middle line:84% And the electrons feel those ionized acceptors. 00:43:51.260 --> 00:43:53.120 align:middle line:84% They get scattered by coulombic scattering. 00:43:53.120 --> 00:43:55.710 align:middle line:84% And in fact, the mobility then can go down. 00:43:55.710 --> 00:43:59.502 align:middle line:84% So for a higher extension dose, different profile causes pile up 00:43:59.502 --> 00:44:00.710 align:middle line:90% and you get a lower mobility. 00:44:00.710 --> 00:44:02.850 align:middle line:84% Lower mobility can mean you can end up 00:44:02.850 --> 00:44:04.350 align:middle line:84% lowering your current drive compared 00:44:04.350 --> 00:44:06.180 align:middle line:90% to what it should have been. 00:44:06.180 --> 00:44:08.490 align:middle line:84% And that affects the overall circuit speed. 00:44:08.490 --> 00:44:12.570 align:middle line:84% So again, just subtle changes in the channel profile 00:44:12.570 --> 00:44:16.230 align:middle line:84% which have nothing to do with how I implanted the channel. 00:44:16.230 --> 00:44:18.180 align:middle line:84% It's how I implanted the neighboring regions. 00:44:18.180 --> 00:44:21.570 align:middle line:84% The source and drain end up impacting not only the VT, 00:44:21.570 --> 00:44:28.413 align:middle line:84% but the mobility of the device, and therefore the current drive. 00:44:28.413 --> 00:44:29.580 align:middle line:90% Well, this was a neat paper. 00:44:29.580 --> 00:44:33.900 align:middle line:84% Now, several years later, 1995, Scott Crowder 00:44:33.900 --> 00:44:36.630 align:middle line:84% came up with an interesting idea. 00:44:36.630 --> 00:44:40.510 align:middle line:84% Knowing what he knew about 311s and what we in our class, 00:44:40.510 --> 00:44:45.120 align:middle line:84% he said, OK, well, 311s, they anneal out a lot faster 00:44:45.120 --> 00:44:46.830 align:middle line:90% at high temperature. 00:44:46.830 --> 00:44:50.600 align:middle line:84% So let's say I have to do an 800-degree step because I'm 00:44:50.600 --> 00:44:54.260 align:middle line:84% going to have to put down those nitride spacers. 00:44:54.260 --> 00:44:56.960 align:middle line:84% Maybe I can still do that if before I do it, 00:44:56.960 --> 00:44:59.160 align:middle line:90% I do 1,000-degree short anneal. 00:44:59.160 --> 00:45:02.270 align:middle line:84% So let's do a very short anneal, one second at 1,000, 00:45:02.270 --> 00:45:05.990 align:middle line:84% evaporate all those 311s, and get rid of them. 00:45:05.990 --> 00:45:09.290 align:middle line:84% Then I can put the wafer in the furnace 00:45:09.290 --> 00:45:11.330 align:middle line:84% and put down my 800-degree nitride. 00:45:11.330 --> 00:45:13.310 align:middle line:90% All the TED will be over with. 00:45:13.310 --> 00:45:16.252 align:middle line:84% So I use a high temperature to cause the 311s to evaporate, 00:45:16.252 --> 00:45:17.960 align:middle line:84% get rid of all those excess interstitials 00:45:17.960 --> 00:45:20.210 align:middle line:90% in a very short time. 00:45:20.210 --> 00:45:24.170 align:middle line:84% I should have less of the reverse short channel effect, 00:45:24.170 --> 00:45:25.235 align:middle line:90% less total interstitials. 00:45:25.235 --> 00:45:26.735 align:middle line:84% And that's what, in fact, we showed. 00:45:26.735 --> 00:45:29.408 align:middle line:90% 00:45:29.408 --> 00:45:31.700 align:middle line:84% This starred region is when he did the high temperature 00:45:31.700 --> 00:45:35.810 align:middle line:84% rapid thermal anneal first before he did the 800-degree C, 00:45:35.810 --> 00:45:36.890 align:middle line:90% longer time anneal. 00:45:36.890 --> 00:45:40.040 align:middle line:84% And he saw he had less roll-up, less reverse short channel 00:45:40.040 --> 00:45:42.500 align:middle line:90% effect. 00:45:42.500 --> 00:45:44.990 align:middle line:84% The open squares represent the case 00:45:44.990 --> 00:45:48.470 align:middle line:84% where he did the 800-degree C anneal in the furnace first. 00:45:48.470 --> 00:45:49.790 align:middle line:90% And then did the 1,000. 00:45:49.790 --> 00:45:52.070 align:middle line:84% And you see a lot of reverse short channel effect. 00:45:52.070 --> 00:45:53.000 align:middle line:90% Interesting. 00:45:53.000 --> 00:45:56.020 align:middle line:84% Exact same amount of time in the furnace and in the RTA 00:45:56.020 --> 00:45:57.560 align:middle line:90% in both wafers. 00:45:57.560 --> 00:46:00.090 align:middle line:84% He just changed the order of the operations. 00:46:00.090 --> 00:46:02.660 align:middle line:84% So this is kind of interesting that what 00:46:02.660 --> 00:46:06.740 align:middle line:84% we know about 311 defects and how they anneal, 00:46:06.740 --> 00:46:08.690 align:middle line:84% it's important for us to think about the order 00:46:08.690 --> 00:46:11.600 align:middle line:84% of the steps in which we make a MOSFET 00:46:11.600 --> 00:46:14.100 align:middle line:90% where we insert the anneals. 00:46:14.100 --> 00:46:15.600 align:middle line:84% Now, you don't always have a choice. 00:46:15.600 --> 00:46:18.867 align:middle line:84% You have to cut the gate before you do the source drain 00:46:18.867 --> 00:46:20.700 align:middle line:84% extension so it's self-aligned and all that. 00:46:20.700 --> 00:46:22.430 align:middle line:84% But this was kind of a clever experiment 00:46:22.430 --> 00:46:27.360 align:middle line:84% just to show that the process order can have a big effect 00:46:27.360 --> 00:46:28.650 align:middle line:90% because of things like 311s. 00:46:28.650 --> 00:46:31.460 align:middle line:90% 00:46:31.460 --> 00:46:34.020 align:middle line:84% And again, this is just a reminder. 00:46:34.020 --> 00:46:35.720 align:middle line:84% We already saw this several times 00:46:35.720 --> 00:46:38.580 align:middle line:84% on slide 22 of the amount of time. 00:46:38.580 --> 00:46:41.180 align:middle line:84% And what he did was this 1,000-degree, 1-second anneal. 00:46:41.180 --> 00:46:42.810 align:middle line:90% He popped it up. 00:46:42.810 --> 00:46:45.410 align:middle line:84% And so he can get-- basically within a second or so, 00:46:45.410 --> 00:46:47.660 align:middle line:90% he can get rid of all the 311s. 00:46:47.660 --> 00:46:50.120 align:middle line:84% The tau enhance is only a few seconds. 00:46:50.120 --> 00:46:53.330 align:middle line:84% And then he could go on later, and without any 311s around, 00:46:53.330 --> 00:46:57.960 align:middle line:84% put it in the furnace, or with very few, and go down to 800. 00:46:57.960 --> 00:47:00.948 align:middle line:90% So that was his proposal. 00:47:00.948 --> 00:47:02.990 align:middle line:84% Another thing that he showed in that paper-- this 00:47:02.990 --> 00:47:06.630 align:middle line:84% is from that same paper by Scott Crowder, IEDM of 1995. 00:47:06.630 --> 00:47:09.320 align:middle line:84% He did an interesting comparison. 00:47:09.320 --> 00:47:12.920 align:middle line:84% He also compared devices made by various similar processes 00:47:12.920 --> 00:47:16.340 align:middle line:84% on bulk, so on regular Czochralski wafers, 00:47:16.340 --> 00:47:21.000 align:middle line:84% to the similar device made on an SOI wafer. 00:47:21.000 --> 00:47:23.000 align:middle line:84% Remember, we said there's this technology called 00:47:23.000 --> 00:47:27.230 align:middle line:84% SOI where you can have single-crystal silicon 00:47:27.230 --> 00:47:28.850 align:middle line:90% layer on insulator. 00:47:28.850 --> 00:47:31.130 align:middle line:90% Very high quality material. 00:47:31.130 --> 00:47:34.610 align:middle line:84% And you can make MOSFETs in that material. 00:47:34.610 --> 00:47:37.070 align:middle line:84% And what he did was now he found, 00:47:37.070 --> 00:47:40.310 align:middle line:84% when he did the source and drain, 00:47:40.310 --> 00:47:42.540 align:middle line:84% implants-- indeed, you of course, 00:47:42.540 --> 00:47:46.700 align:middle line:84% even in silicon insulator, you do inject interstitials up here. 00:47:46.700 --> 00:47:49.400 align:middle line:84% But interestingly, remember, the interstitials 00:47:49.400 --> 00:47:51.800 align:middle line:84% tend to get injected both down, and then they go up. 00:47:51.800 --> 00:47:54.680 align:middle line:84% But ones that went down were going up here now 00:47:54.680 --> 00:47:59.150 align:middle line:84% against an interface between oxide and silicon. 00:47:59.150 --> 00:48:03.080 align:middle line:84% And it turns out that interface is a very good sync. 00:48:03.080 --> 00:48:05.810 align:middle line:84% There's a lot of recombination that can take place 00:48:05.810 --> 00:48:07.340 align:middle line:90% at this interface. 00:48:07.340 --> 00:48:09.920 align:middle line:84% K sub s is fairly large between oxide 00:48:09.920 --> 00:48:12.320 align:middle line:90% and single-crystal silicon. 00:48:12.320 --> 00:48:14.750 align:middle line:84% So a lot of interstitials can be sunk 00:48:14.750 --> 00:48:20.350 align:middle line:84% or can be absorbed there, whereas in the bulk, 00:48:20.350 --> 00:48:22.510 align:middle line:84% we get these interstitial fluxes that come in 00:48:22.510 --> 00:48:24.520 align:middle line:84% and they don't get absorbed because there's 00:48:24.520 --> 00:48:26.180 align:middle line:90% no oxide down there. 00:48:26.180 --> 00:48:28.180 align:middle line:84% You can end up getting profiles that look like-- 00:48:28.180 --> 00:48:29.890 align:middle line:84% fluxes that look like these arrows, 00:48:29.890 --> 00:48:33.040 align:middle line:84% driving the boron to the surface. 00:48:33.040 --> 00:48:36.580 align:middle line:84% So what he showed was, in fact, on an SOI wafer, 00:48:36.580 --> 00:48:41.080 align:middle line:84% you don't get as much pileup of boron underneath the gate. 00:48:41.080 --> 00:48:44.950 align:middle line:84% And on the SOI wafer, this plot-- the y-axis didn't get 00:48:44.950 --> 00:48:48.370 align:middle line:84% showed up here, but the y-axis is the threshold voltage, VT. 00:48:48.370 --> 00:48:52.270 align:middle line:84% Again, the roll-up of the VT is a lot less on an SOI wafer, 00:48:52.270 --> 00:48:54.550 align:middle line:84% subject to the same kind of annealing, 00:48:54.550 --> 00:48:57.790 align:middle line:84% compared to this bulk wafer shown 00:48:57.790 --> 00:49:01.130 align:middle line:84% by the solid line, which had a lot more reverse short channel 00:49:01.130 --> 00:49:01.630 align:middle line:90% effect. 00:49:01.630 --> 00:49:02.830 align:middle line:90% Interesting idea. 00:49:02.830 --> 00:49:04.420 align:middle line:90% Use SOI. 00:49:04.420 --> 00:49:07.300 align:middle line:84% Use the property of the fact that there's 00:49:07.300 --> 00:49:10.120 align:middle line:84% a sink here for interstitials to sink out 00:49:10.120 --> 00:49:13.030 align:middle line:84% a lot of the interstitials that you implanted 00:49:13.030 --> 00:49:17.390 align:middle line:84% and get less motion of the channel doping. 00:49:17.390 --> 00:49:18.890 align:middle line:84% So this tells us right away, though, 00:49:18.890 --> 00:49:21.848 align:middle line:84% if we're doing a process in SOI-- 00:49:21.848 --> 00:49:23.390 align:middle line:84% and we can do the exact same process. 00:49:23.390 --> 00:49:25.460 align:middle line:84% We can get quite a different result 00:49:25.460 --> 00:49:28.490 align:middle line:84% in insulator in SOI compared to bulk 00:49:28.490 --> 00:49:32.018 align:middle line:84% because the channel doping profiles will not be the same. 00:49:32.018 --> 00:49:34.685 align:middle line:84% And so the simulators need to be able to simulate these effects. 00:49:34.685 --> 00:49:38.870 align:middle line:90% 00:49:38.870 --> 00:49:40.670 align:middle line:84% And another thing I want to talk about 00:49:40.670 --> 00:49:42.470 align:middle line:84% is I just want to remind you what's 00:49:42.470 --> 00:49:46.280 align:middle line:84% the usual order in which we form the channel, the gate, 00:49:46.280 --> 00:49:48.740 align:middle line:84% and the source drain in a MOSFET, 00:49:48.740 --> 00:49:54.980 align:middle line:84% and we are doing clever things with the order of things. 00:49:54.980 --> 00:49:56.780 align:middle line:84% This is sort of a cartoon in PowerPoint. 00:49:56.780 --> 00:49:59.850 align:middle line:84% And so this is supposed to be my wafer. 00:49:59.850 --> 00:50:02.000 align:middle line:84% These green regions here on the left and right 00:50:02.000 --> 00:50:05.660 align:middle line:84% are going to represent the shallow trench isolation. 00:50:05.660 --> 00:50:08.900 align:middle line:84% So green is my STI, which is my isolation region. 00:50:08.900 --> 00:50:11.060 align:middle line:84% And then typically, after you do shallow trench, 00:50:11.060 --> 00:50:15.500 align:middle line:84% you implant what they call SSR, super steep retrograde. 00:50:15.500 --> 00:50:18.710 align:middle line:84% It's just, in implant regions, if you're making an MOSFET, 00:50:18.710 --> 00:50:22.790 align:middle line:84% you have usually a shallow implant up here, or more lightly 00:50:22.790 --> 00:50:23.840 align:middle line:90% doping near the surface. 00:50:23.840 --> 00:50:26.580 align:middle line:84% And you have the peak of the implant a little bit deeper. 00:50:26.580 --> 00:50:28.190 align:middle line:84% So this could be a boron implant. 00:50:28.190 --> 00:50:30.410 align:middle line:84% So that boron implant usually goes 00:50:30.410 --> 00:50:32.360 align:middle line:90% in pretty early in the process. 00:50:32.360 --> 00:50:33.980 align:middle line:84% And it's that boron implant that's 00:50:33.980 --> 00:50:37.740 align:middle line:84% going to determine your channel profile, and therefore 00:50:37.740 --> 00:50:39.340 align:middle line:90% your VT and things like that. 00:50:39.340 --> 00:50:41.820 align:middle line:84% So it usually goes in fairly early, right about here. 00:50:41.820 --> 00:50:45.810 align:middle line:84% After that, it sees the thermal budget of the gate oxide growth. 00:50:45.810 --> 00:50:47.890 align:middle line:84% That could be typically 800 degrees. 00:50:47.890 --> 00:50:51.090 align:middle line:84% So it's got to go through that diffusion. 00:50:51.090 --> 00:50:52.290 align:middle line:90% And then you make the gate. 00:50:52.290 --> 00:50:54.790 align:middle line:84% The gate is usually a very low-temperature process, 00:50:54.790 --> 00:50:55.810 align:middle line:90% and it's just etching. 00:50:55.810 --> 00:50:58.810 align:middle line:84% So that's no motion of the boron there. 00:50:58.810 --> 00:51:00.960 align:middle line:84% Then you implant the shallow source drains 00:51:00.960 --> 00:51:03.040 align:middle line:90% and you use the gate as a mask. 00:51:03.040 --> 00:51:05.100 align:middle line:84% So now I'm doing implants of arsenic. 00:51:05.100 --> 00:51:06.720 align:middle line:84% And you're introducing a little bit 00:51:06.720 --> 00:51:08.628 align:middle line:84% of a certain number of point defects 00:51:08.628 --> 00:51:10.545 align:middle line:84% here on the left and the right of the channel. 00:51:10.545 --> 00:51:14.470 align:middle line:90% 00:51:14.470 --> 00:51:16.980 align:middle line:84% Now I put it in the furnace and put on spacers, 00:51:16.980 --> 00:51:17.932 align:middle line:90% these green spacers. 00:51:17.932 --> 00:51:19.890 align:middle line:84% If they're oxide, I do it at a low temperature. 00:51:19.890 --> 00:51:21.210 align:middle line:90% LTO goes down at 400. 00:51:21.210 --> 00:51:22.380 align:middle line:90% It's not usually a problem. 00:51:22.380 --> 00:51:24.510 align:middle line:84% If they're silicon nitride spacers, 00:51:24.510 --> 00:51:28.410 align:middle line:84% typically nitride LPCVD goes down at 800. 00:51:28.410 --> 00:51:32.010 align:middle line:84% So I could get a fair amount of diffusion at 800 of TED, 00:51:32.010 --> 00:51:35.160 align:middle line:84% especially because I have the implant damage introduced 00:51:35.160 --> 00:51:37.420 align:middle line:84% from the shallow source and drain. 00:51:37.420 --> 00:51:39.510 align:middle line:84% So watch out for nitride spacers. 00:51:39.510 --> 00:51:42.150 align:middle line:84% 800 for an hour to make these spacers could really-- 00:51:42.150 --> 00:51:44.880 align:middle line:84% as we saw in our example, can really cause a lot of motion 00:51:44.880 --> 00:51:47.270 align:middle line:90% at that boron. 00:51:47.270 --> 00:51:48.830 align:middle line:84% Then we do the deep source drains 00:51:48.830 --> 00:51:53.533 align:middle line:84% using the spacers as the mask, now, again, self-reliant. 00:51:53.533 --> 00:51:54.950 align:middle line:84% And then usually right after that, 00:51:54.950 --> 00:51:57.780 align:middle line:90% there's a final thermal anneal. 00:51:57.780 --> 00:52:01.790 align:middle line:84% So one idea people have had is, well, don't put the P type SSR 00:52:01.790 --> 00:52:03.740 align:middle line:90% implant in at the beginning. 00:52:03.740 --> 00:52:06.610 align:middle line:84% Why don't you put it at the end of the process? 00:52:06.610 --> 00:52:08.890 align:middle line:84% And a very radical idea is to put it in 00:52:08.890 --> 00:52:11.278 align:middle line:84% even after the gate has already been formed. 00:52:11.278 --> 00:52:12.820 align:middle line:84% This is not being done in production, 00:52:12.820 --> 00:52:15.910 align:middle line:84% but it was a neat idea that people had in research. 00:52:15.910 --> 00:52:18.610 align:middle line:84% OK, you have to deal with all these point defects. 00:52:18.610 --> 00:52:21.220 align:middle line:84% Well, don't put the boron in until you've already 00:52:21.220 --> 00:52:24.440 align:middle line:84% annealed out a lot of those point defects. 00:52:24.440 --> 00:52:26.920 align:middle line:84% So here is an alternate process to give you 00:52:26.920 --> 00:52:29.440 align:middle line:84% an idea of how device engineers were trying 00:52:29.440 --> 00:52:34.780 align:middle line:84% to get around TED, to a certain extent, on slide 25, 00:52:34.780 --> 00:52:38.260 align:middle line:84% and also process integration scheme for forming a MOSFET. 00:52:38.260 --> 00:52:40.420 align:middle line:84% And this was published back in 1998 00:52:40.420 --> 00:52:45.130 align:middle line:84% by Philips Corporation called channel profile engineering, 0.1 00:52:45.130 --> 00:52:48.100 align:middle line:84% micron MOSFETs, by doing through-the-gate ion 00:52:48.100 --> 00:52:49.780 align:middle line:90% implantation. 00:52:49.780 --> 00:52:54.000 align:middle line:84% So they were proposing a flow that goes like this. 00:52:54.000 --> 00:52:56.370 align:middle line:84% A conventional flow is on the right. 00:52:56.370 --> 00:52:59.310 align:middle line:84% Conventional usual sequence to make a MOSFET. 00:52:59.310 --> 00:53:03.240 align:middle line:84% As you form the P well, you implant the channel boron ion 00:53:03.240 --> 00:53:05.130 align:middle line:90% profile all the way up here. 00:53:05.130 --> 00:53:07.200 align:middle line:84% Then like we just said, we do gate ox. 00:53:07.200 --> 00:53:09.180 align:middle line:90% We form the gate. 00:53:09.180 --> 00:53:10.800 align:middle line:90% We make spacers. 00:53:10.800 --> 00:53:14.487 align:middle line:84% We form the source drain extensions. 00:53:14.487 --> 00:53:16.320 align:middle line:84% Oh, I'm sorry, that's some oxide deposition. 00:53:16.320 --> 00:53:17.700 align:middle line:90% Here's the sidewall spacer. 00:53:17.700 --> 00:53:22.350 align:middle line:84% Here's one of the killers, the nitride dip at 800. 00:53:22.350 --> 00:53:23.940 align:middle line:90% A lot of TED can happen there. 00:53:23.940 --> 00:53:25.230 align:middle line:90% Then the deep source drains. 00:53:25.230 --> 00:53:27.150 align:middle line:90% And then we do the RTA. 00:53:27.150 --> 00:53:29.640 align:middle line:84% So what they were saying is instead of putting the channel 00:53:29.640 --> 00:53:31.590 align:middle line:84% implants in here, where they can diffuse 00:53:31.590 --> 00:53:34.740 align:middle line:84% during this nitride spacer step, take the channel implant 00:53:34.740 --> 00:53:37.410 align:middle line:84% and put it in towards the end after you already 00:53:37.410 --> 00:53:39.522 align:middle line:90% have the nitride spacers in. 00:53:39.522 --> 00:53:41.980 align:middle line:84% The thing that's weird about that, though-- think about it. 00:53:41.980 --> 00:53:45.230 align:middle line:84% Now I have a topography that looks like this. 00:53:45.230 --> 00:53:48.260 align:middle line:84% Now I'm going to implant the channel. 00:53:48.260 --> 00:53:50.810 align:middle line:84% I have to make sure I give it enough energy so boron 00:53:50.810 --> 00:53:52.020 align:middle line:90% can get through the gate. 00:53:52.020 --> 00:53:53.730 align:middle line:84% So you have to calculate that energy. 00:53:53.730 --> 00:53:56.222 align:middle line:84% And in fact, your born profile now 00:53:56.222 --> 00:53:57.680 align:middle line:84% is going to look sort of like this. 00:53:57.680 --> 00:53:59.670 align:middle line:84% It's going to have this shape to it. 00:53:59.670 --> 00:54:02.210 align:middle line:84% It's going to be a little deeper here where the gate doesn't 00:54:02.210 --> 00:54:03.690 align:middle line:90% exist in the source and drain. 00:54:03.690 --> 00:54:05.780 align:middle line:84% It'll be a little shallower here. 00:54:05.780 --> 00:54:08.990 align:middle line:84% That might be OK from a electrical point of view, 00:54:08.990 --> 00:54:10.260 align:middle line:90% but it is kind of strange. 00:54:10.260 --> 00:54:12.680 align:middle line:84% But the advantage they have is that it doesn't go in-- 00:54:12.680 --> 00:54:16.400 align:middle line:84% the only thermal step it sees is the last high-temperature step. 00:54:16.400 --> 00:54:18.950 align:middle line:84% It never sees all the TED would happen 00:54:18.950 --> 00:54:22.310 align:middle line:84% during the sidewall spacer at 800. 00:54:22.310 --> 00:54:23.780 align:middle line:90% So that was an idea they had. 00:54:23.780 --> 00:54:25.620 align:middle line:84% One thing you might think, though-- what 00:54:25.620 --> 00:54:27.870 align:middle line:84% do you think about-- if you're an electrical engineer, 00:54:27.870 --> 00:54:30.287 align:middle line:84% what do you think about ion planting through a gate oxide? 00:54:30.287 --> 00:54:32.390 align:middle line:90% 00:54:32.390 --> 00:54:35.390 align:middle line:84% It's a little scary because that gate oxide might only 00:54:35.390 --> 00:54:37.160 align:middle line:90% be 20 angstroms thick, right? 00:54:37.160 --> 00:54:40.060 align:middle line:84% Are you going to implant a high energy ion through? 00:54:40.060 --> 00:54:41.905 align:middle line:84% It's not clear what damage takes place right 00:54:41.905 --> 00:54:44.030 align:middle line:84% at the interface between the oxide and the silicon, 00:54:44.030 --> 00:54:45.720 align:middle line:90% and interface states and things. 00:54:45.720 --> 00:54:47.270 align:middle line:84% So although this was a neat process, 00:54:47.270 --> 00:54:49.100 align:middle line:84% I don't think it was ever accepted in production. 00:54:49.100 --> 00:54:51.320 align:middle line:84% I'm not sure people thought it was reliable enough. 00:54:51.320 --> 00:54:52.670 align:middle line:90% But they did show-- 00:54:52.670 --> 00:54:57.500 align:middle line:84% on the next page, they did show in their IDM article 00:54:57.500 --> 00:54:59.570 align:middle line:84% that they could achieve a dramatic improvement 00:54:59.570 --> 00:55:00.590 align:middle line:90% over the boron control. 00:55:00.590 --> 00:55:03.770 align:middle line:84% This is dopant boron concentration versus depth. 00:55:03.770 --> 00:55:07.317 align:middle line:84% And the black line is the reference device. 00:55:07.317 --> 00:55:09.650 align:middle line:84% So that's the device that went through the ordinary flow 00:55:09.650 --> 00:55:12.018 align:middle line:84% where they put the boron in at the start of the process. 00:55:12.018 --> 00:55:13.310 align:middle line:90% And it goes through everything. 00:55:13.310 --> 00:55:16.340 align:middle line:84% It sees the sidewall spacer, nitride depth, all that. 00:55:16.340 --> 00:55:17.180 align:middle line:90% Lots of TED. 00:55:17.180 --> 00:55:19.140 align:middle line:84% The boron is essentially almost flat. 00:55:19.140 --> 00:55:22.610 align:middle line:84% You don't even see much of the implant, the initial implant, 00:55:22.610 --> 00:55:26.790 align:middle line:84% whereas when they did through-the-gate as implanted 00:55:26.790 --> 00:55:30.150 align:middle line:84% and after processing-- so look at after processing here-- 00:55:30.150 --> 00:55:32.920 align:middle line:84% you can see there wasn't that much diffusion at all. 00:55:32.920 --> 00:55:34.470 align:middle line:84% So they were really able to control 00:55:34.470 --> 00:55:38.070 align:middle line:84% much better because it didn't see any of the TED the nitride. 00:55:38.070 --> 00:55:40.500 align:middle line:84% All it saw was the last high-temperature, 00:55:40.500 --> 00:55:43.507 align:middle line:90% 1,000-degree step. 00:55:43.507 --> 00:55:45.090 align:middle line:84% But as I say, for reliability reasons, 00:55:45.090 --> 00:55:46.710 align:middle line:84% I don't know if it was ever really accepted. 00:55:46.710 --> 00:55:49.350 align:middle line:84% They did show that they could reduce the reverse short channel 00:55:49.350 --> 00:55:50.550 align:middle line:90% effect. 00:55:50.550 --> 00:55:56.250 align:middle line:84% This is now on the slide 27, VT as a function of gate length. 00:55:56.250 --> 00:56:01.220 align:middle line:84% And they have different things here. 00:56:01.220 --> 00:56:04.868 align:middle line:84% Here's a reference process, the black. 00:56:04.868 --> 00:56:07.160 align:middle line:84% That's when you put the boron in at the very beginning, 00:56:07.160 --> 00:56:08.600 align:middle line:90% the standard process flow. 00:56:08.600 --> 00:56:11.540 align:middle line:84% You see the roll-up, the reverse short channel effect. 00:56:11.540 --> 00:56:13.880 align:middle line:90% And through the gate is TGI. 00:56:13.880 --> 00:56:15.740 align:middle line:84% And they had several different doses. 00:56:15.740 --> 00:56:18.050 align:middle line:90% TGI, you see no roll-up in VT. 00:56:18.050 --> 00:56:20.300 align:middle line:84% The VT is very flat until it goes 00:56:20.300 --> 00:56:24.120 align:middle line:84% to the conventional short channel effect sort of effects. 00:56:24.120 --> 00:56:26.450 align:middle line:84% So they were able to eliminate the roll-up 00:56:26.450 --> 00:56:29.090 align:middle line:84% because they had better control over the profile. 00:56:29.090 --> 00:56:31.070 align:middle line:84% They essentially eliminated a lot of the TED 00:56:31.070 --> 00:56:34.380 align:middle line:90% in the boron profile. 00:56:34.380 --> 00:56:37.380 align:middle line:84% I just want to mention before we finish 00:56:37.380 --> 00:56:40.170 align:middle line:84% chapter 8 some new diffusion modeling issues that 00:56:40.170 --> 00:56:41.670 align:middle line:84% are in the literature right now that 00:56:41.670 --> 00:56:46.200 align:middle line:84% are people working on today, conditions where we have 00:56:46.200 --> 00:56:48.390 align:middle line:90% very high dose being implanted. 00:56:48.390 --> 00:56:51.270 align:middle line:84% The 311 model that we talked about 00:56:51.270 --> 00:56:53.550 align:middle line:84% is good in intermediate dose regime, 00:56:53.550 --> 00:56:55.540 align:middle line:84% but doesn't really work for very high doses. 00:56:55.540 --> 00:56:58.440 align:middle line:84% So we need to model the type of damage that takes 00:56:58.440 --> 00:56:59.880 align:middle line:90% place in a very high doses. 00:56:59.880 --> 00:57:02.580 align:middle line:84% It's being investigated by people today. 00:57:02.580 --> 00:57:03.750 align:middle line:90% Very wide energy range. 00:57:03.750 --> 00:57:05.820 align:middle line:84% There are people doing implants less than a kilovolt today. 00:57:05.820 --> 00:57:07.528 align:middle line:84% There are some really crazy people trying 00:57:07.528 --> 00:57:10.890 align:middle line:84% to do implants very shallow, and some very deep ones, even 00:57:10.890 --> 00:57:14.550 align:middle line:84% greater than a megavolt. The physics of stopping 00:57:14.550 --> 00:57:18.000 align:middle line:84% and the physics of how to model those implants is not really 00:57:18.000 --> 00:57:21.600 align:middle line:84% all that well-known in these two ranges. 00:57:21.600 --> 00:57:24.240 align:middle line:84% We talked about pre-amorphization the substrate. 00:57:24.240 --> 00:57:26.380 align:middle line:84% Prior to introducing boron we said, 00:57:26.380 --> 00:57:28.620 align:middle line:84% well, hit it with silicon at a high enough dose 00:57:28.620 --> 00:57:31.980 align:middle line:84% that you can amorphize it, and then you can avoid channeling. 00:57:31.980 --> 00:57:35.130 align:middle line:84% But what kind of damage is produced by a pre-amorphization 00:57:35.130 --> 00:57:39.120 align:middle line:84% by an amorphous implant and what effect 00:57:39.120 --> 00:57:41.548 align:middle line:84% exactly it has is not all that well understood. 00:57:41.548 --> 00:57:43.590 align:middle line:84% There's a whole bunch of new annealing techniques 00:57:43.590 --> 00:57:45.715 align:middle line:84% that have come out, something called a spike anneal 00:57:45.715 --> 00:57:47.550 align:middle line:84% where you take a RTA machine, you zap up 00:57:47.550 --> 00:57:51.030 align:middle line:84% the lamps really fast, and you zap them down immediately. 00:57:51.030 --> 00:57:53.880 align:middle line:84% And the wafer never even spends time at any one temperature. 00:57:53.880 --> 00:57:55.890 align:middle line:84% It just sort of goes up and down. 00:57:55.890 --> 00:57:59.940 align:middle line:84% You're accessing sub-1 second time regime. 00:57:59.940 --> 00:58:01.770 align:middle line:84% Exactly what happens during those ramps 00:58:01.770 --> 00:58:04.110 align:middle line:84% is not completely understood in spike annealing. 00:58:04.110 --> 00:58:06.330 align:middle line:84% Laser annealing, where we take a laser 00:58:06.330 --> 00:58:09.450 align:middle line:84% and we scan it across and we only 00:58:09.450 --> 00:58:12.090 align:middle line:90% heat the area for a nanosecond. 00:58:12.090 --> 00:58:14.455 align:middle line:84% Again, the kinetics of defect evolution 00:58:14.455 --> 00:58:16.080 align:middle line:84% in the nanosecond regime has not really 00:58:16.080 --> 00:58:17.530 align:middle line:90% been very well understood. 00:58:17.530 --> 00:58:21.030 align:middle line:84% So that's a very hot topic these days. 00:58:21.030 --> 00:58:23.460 align:middle line:84% So process conditions are changing 00:58:23.460 --> 00:58:24.690 align:middle line:90% from what they used to be? 00:58:24.690 --> 00:58:26.020 align:middle line:90% New mechanisms. 00:58:26.020 --> 00:58:27.450 align:middle line:90% Well, people already know 311s. 00:58:27.450 --> 00:58:28.810 align:middle line:84% That's kind of been beaten to death, 00:58:28.810 --> 00:58:29.760 align:middle line:90% I would say, at this point. 00:58:29.760 --> 00:58:31.020 align:middle line:90% A lot of work has been done. 00:58:31.020 --> 00:58:34.530 align:middle line:84% But end of range dislocation loops, which we never 00:58:34.530 --> 00:58:35.730 align:middle line:90% completely get rid of. 00:58:35.730 --> 00:58:37.260 align:middle line:90% And these are very important. 00:58:37.260 --> 00:58:39.660 align:middle line:84% If you do an amortizing implant, you can't avoid them. 00:58:39.660 --> 00:58:41.520 align:middle line:84% Their effect on diffusion has never 00:58:41.520 --> 00:58:44.710 align:middle line:90% been completely understood. 00:58:44.710 --> 00:58:48.280 align:middle line:84% The clustering of dopants with interstitials, 00:58:48.280 --> 00:58:51.340 align:middle line:84% which may restore a defect and may 00:58:51.340 --> 00:58:53.830 align:middle line:84% affect the electrical solubility of the dopant. 00:58:53.830 --> 00:58:55.505 align:middle line:84% How many electrons or holes you end up 00:58:55.505 --> 00:58:57.130 align:middle line:84% getting has not really been understood. 00:58:57.130 --> 00:58:58.880 align:middle line:90% So that's a new topic. 00:58:58.880 --> 00:59:02.440 align:middle line:84% And at interfaces, what the recombination rates 00:59:02.440 --> 00:59:04.025 align:middle line:90% are at interfaces. 00:59:04.025 --> 00:59:05.650 align:middle line:84% As we put more and different materials, 00:59:05.650 --> 00:59:07.870 align:middle line:84% we're putting high K into the problem now. 00:59:07.870 --> 00:59:10.240 align:middle line:84% What happens when you have a high K interface? 00:59:10.240 --> 00:59:15.160 align:middle line:84% How do point defects recombine at an interface between a high K 00:59:15.160 --> 00:59:17.200 align:middle line:90% and silicon as opposed to SiO2? 00:59:17.200 --> 00:59:19.280 align:middle line:84% A lot of interesting new research topics 00:59:19.280 --> 00:59:21.598 align:middle line:90% that's not covered. 00:59:21.598 --> 00:59:23.890 align:middle line:84% And I won't go through the summary in any great detail. 00:59:23.890 --> 00:59:27.650 align:middle line:84% I think we've gone through all this. 00:59:27.650 --> 00:59:30.300 align:middle line:84% I encourage you to read through chapter 8 carefully. 00:59:30.300 --> 00:59:32.978 align:middle line:84% And I think we're at a stage now-- 00:59:32.978 --> 00:59:35.270 align:middle line:84% next time I'm going to talk about SUPREM-IV in detail-- 00:59:35.270 --> 00:59:38.420 align:middle line:84% where we have enough tools that we can really understand 00:59:38.420 --> 00:59:42.140 align:middle line:84% to a reasonably high level how people put processes together 00:59:42.140 --> 00:59:45.650 align:middle line:84% to make devices and to minimize doping diffusion. 00:59:45.650 --> 00:59:48.050 align:middle line:90% OK, that's it. 00:59:48.050 --> 00:59:50.480 align:middle line:84% If you haven't signed up yet on the clipboard, 00:59:50.480 --> 00:59:51.620 align:middle line:90% I've got it up front. 00:59:51.620 --> 00:59:54.550 align:middle line:90% I'd be happy for you to sign up. 00:59:54.550 --> 01:00:00.000 align:middle line:90%