WEBVTT 00:00:00.000 --> 00:00:02.083 align:middle line:84% JUDY HOYT: And then on Thursday, we'll have these. 00:00:02.083 --> 00:00:04.460 align:middle line:84% I also want to mention, towards the end of lecture today, 00:00:04.460 --> 00:00:10.710 align:middle line:84% we'll have the course evaluations for you to fill out. 00:00:10.710 --> 00:00:14.850 align:middle line:84% OK, before we start the formal lecture for today, 00:00:14.850 --> 00:00:17.670 align:middle line:84% for those of you who were here last time, 00:00:17.670 --> 00:00:21.210 align:middle line:84% I wanted to go over something that we had talked about. 00:00:21.210 --> 00:00:23.280 align:middle line:84% We didn't have time to go through how 00:00:23.280 --> 00:00:27.990 align:middle line:84% you would do these calculations from the last handout, which 00:00:27.990 --> 00:00:31.530 align:middle line:90% was handout 36. 00:00:31.530 --> 00:00:34.440 align:middle line:84% On slide 27, there was a couple of questions 00:00:34.440 --> 00:00:35.280 align:middle line:90% that we were asking. 00:00:35.280 --> 00:00:38.100 align:middle line:84% And I thought we'd just spent about five minutes before we 00:00:38.100 --> 00:00:41.410 align:middle line:84% start today's formal lecture going through this, 00:00:41.410 --> 00:00:45.060 align:middle line:84% at least on the board, so people have an idea of how to make 00:00:45.060 --> 00:00:46.900 align:middle line:84% these back-of-the-envelope calculations. 00:00:46.900 --> 00:00:48.900 align:middle line:84% So what you're being asked to do here 00:00:48.900 --> 00:00:51.360 align:middle line:84% is you're looking at a silicon MOSFET 00:00:51.360 --> 00:00:53.400 align:middle line:84% with-- a these are the source drain 00:00:53.400 --> 00:00:55.260 align:middle line:84% extensions, these shallow extensions that 00:00:55.260 --> 00:00:57.390 align:middle line:90% have a certain depth XJ. 00:00:57.390 --> 00:01:01.443 align:middle line:84% These are the deep source and drain that have been silicided. 00:01:01.443 --> 00:01:03.610 align:middle line:84% Remember, the last lecture was all about siliciding. 00:01:03.610 --> 00:01:09.370 align:middle line:84% So this blue region is supposed to represent a silicided region. 00:01:09.370 --> 00:01:12.550 align:middle line:84% And then there's a metal contact and this is highly schematized 00:01:12.550 --> 00:01:14.830 align:middle line:84% but imagine this little thing coming down here 00:01:14.830 --> 00:01:18.910 align:middle line:84% as your aluminum metal or whatever coming down 00:01:18.910 --> 00:01:23.980 align:middle line:84% and touching the silicide and is spaced a certain distance 00:01:23.980 --> 00:01:29.290 align:middle line:84% the metal contact is by an insulator from this point here 00:01:29.290 --> 00:01:31.220 align:middle line:90% where the silicide starts. 00:01:31.220 --> 00:01:33.910 align:middle line:84% And so what we're asked to do is to calculate separately 00:01:33.910 --> 00:01:36.400 align:middle line:84% three resistors that are in this problem. 00:01:36.400 --> 00:01:37.900 align:middle line:84% There's this resistor here, which 00:01:37.900 --> 00:01:39.850 align:middle line:84% represents the resistance of the source drain 00:01:39.850 --> 00:01:42.460 align:middle line:84% extension as the current flows through the channel 00:01:42.460 --> 00:01:45.190 align:middle line:84% into the extension, which is underneath the spacer. 00:01:45.190 --> 00:01:48.280 align:middle line:84% And then there's a resistance associated with this sheet 00:01:48.280 --> 00:01:49.930 align:middle line:84% resistance of the silicide arrangement 00:01:49.930 --> 00:01:51.847 align:middle line:84% because the current has to get from this point 00:01:51.847 --> 00:01:54.730 align:middle line:84% here to the point where the metal contact is up here. 00:01:54.730 --> 00:01:57.040 align:middle line:84% And then there's the interfacial resistance 00:01:57.040 --> 00:01:58.540 align:middle line:84% or the contact resistance associated 00:01:58.540 --> 00:01:59.810 align:middle line:90% with this metal contact. 00:01:59.810 --> 00:02:01.420 align:middle line:90% So those are the three. 00:02:01.420 --> 00:02:04.780 align:middle line:84% And in a few minutes here, we can just go through that. 00:02:04.780 --> 00:02:08.120 align:middle line:84% And I thought it might be useful for us to calculate those. 00:02:08.120 --> 00:02:13.690 align:middle line:84% So the first one, we're asked to calculate the-- 00:02:13.690 --> 00:02:15.740 align:middle line:84% let's figure out the contact resistance. 00:02:15.740 --> 00:02:18.910 align:middle line:84% So you're given the specific contact 00:02:18.910 --> 00:02:21.910 align:middle line:84% resistivity, which we talked about in the last lecture as 2 00:02:21.910 --> 00:02:24.538 align:middle line:90% and 1/2 times 10 to the minus 7. 00:02:24.538 --> 00:02:27.220 align:middle line:90% 00:02:27.220 --> 00:02:31.650 align:middle line:84% And that has units of ohm centimeter squared. 00:02:31.650 --> 00:02:34.310 align:middle line:90% It's a resistance times an area. 00:02:34.310 --> 00:02:37.660 align:middle line:84% And if I take that-- if I were to get the contact resistance 00:02:37.660 --> 00:02:41.560 align:middle line:84% R of that contact, I just have to take that rho sub S value 00:02:41.560 --> 00:02:44.770 align:middle line:84% and divide it by the area of the contact. 00:02:44.770 --> 00:02:47.740 align:middle line:84% And are told that the area of the contact in number one 00:02:47.740 --> 00:02:52.270 align:middle line:84% there on the slide is 0.2 by 1 square micron. 00:02:52.270 --> 00:02:57.260 align:middle line:84% So that's just going to be 2.5 times 10 to the minus 7 00:02:57.260 --> 00:02:59.850 align:middle line:90% over 0.25-- 00:02:59.850 --> 00:03:02.060 align:middle line:90% so the numbers are easy-- 00:03:02.060 --> 00:03:06.080 align:middle line:84% times 10 to the minus 8 square centimeters 00:03:06.080 --> 00:03:10.190 align:middle line:84% because 1 micron is 10 to the minus 4 centimeters. 00:03:10.190 --> 00:03:12.490 align:middle line:90% And you have to square that. 00:03:12.490 --> 00:03:14.610 align:middle line:90% So you just calculate this out. 00:03:14.610 --> 00:03:19.870 align:middle line:84% And you end up with 100 ohms for the contact resistance. 00:03:19.870 --> 00:03:22.250 align:middle line:84% So that's the resistance between the metal 00:03:22.250 --> 00:03:25.820 align:middle line:84% and this contact right here flowing through this face, 00:03:25.820 --> 00:03:29.310 align:middle line:84% given a contact area of that size. 00:03:29.310 --> 00:03:31.730 align:middle line:84% So that'd be kind of a state of the art contact. 00:03:31.730 --> 00:03:35.340 align:middle line:84% So you add about 100 ohms associated with that. 00:03:35.340 --> 00:03:38.280 align:middle line:84% You also want to calculate the resistor shown here 00:03:38.280 --> 00:03:41.550 align:middle line:84% by this little resistor here, estimate what this resistor is 00:03:41.550 --> 00:03:43.350 align:middle line:84% for the current flowing from this point 00:03:43.350 --> 00:03:46.653 align:middle line:84% to the point where it just gets into the contact. 00:03:46.653 --> 00:03:48.070 align:middle line:84% Well, that's going to be dominated 00:03:48.070 --> 00:03:51.550 align:middle line:84% by the sheet resistance of the silicide of this blue region. 00:03:51.550 --> 00:03:53.470 align:middle line:84% So if we know the sheet resistance, 00:03:53.470 --> 00:03:56.050 align:middle line:84% when we divide by the thickness, we can calculate. 00:03:56.050 --> 00:04:00.620 align:middle line:84% For any sheet, you can always calculate the resistance. 00:04:00.620 --> 00:04:08.112 align:middle line:84% So to get the sheet resistance of the silicide, 00:04:08.112 --> 00:04:09.570 align:middle line:84% well, you're given its resistivity. 00:04:09.570 --> 00:04:12.210 align:middle line:84% It's 15 times 10 to the minus 6 ohm centimeter. 00:04:12.210 --> 00:04:15.360 align:middle line:84% So it's just going to be its resistive rho divided 00:04:15.360 --> 00:04:18.160 align:middle line:90% by the thickness of the sheet. 00:04:18.160 --> 00:04:21.732 align:middle line:84% So that's 15 times 10 to the minus 6 ohm centimeter. 00:04:21.732 --> 00:04:23.190 align:middle line:84% Remember, that's a resistivity now. 00:04:23.190 --> 00:04:25.560 align:middle line:84% That's not a specific contact resistance. 00:04:25.560 --> 00:04:28.560 align:middle line:84% And you are told the thickness of that sheet, roughly. 00:04:28.560 --> 00:04:30.600 align:middle line:84% And I'm going to ignore the doped region down 00:04:30.600 --> 00:04:32.063 align:middle line:90% here because this is a metal. 00:04:32.063 --> 00:04:32.730 align:middle line:90% It's a silicide. 00:04:32.730 --> 00:04:34.980 align:middle line:84% It has a much lower resistance than the silicon. 00:04:34.980 --> 00:04:35.800 align:middle line:90% So I'm going to ignore that. 00:04:35.800 --> 00:04:38.258 align:middle line:84% We're just going to calculate the resistivity of that metal 00:04:38.258 --> 00:04:40.000 align:middle line:90% sheet. 00:04:40.000 --> 00:04:45.280 align:middle line:84% And that thickness is 4 times 10 to the minus 6 centimeters. 00:04:45.280 --> 00:04:46.560 align:middle line:90% So the centimeters go out. 00:04:46.560 --> 00:04:49.290 align:middle line:84% And you end up with a certain resistance, 00:04:49.290 --> 00:04:53.200 align:middle line:90% about 3.75 ohms per square. 00:04:53.200 --> 00:04:54.700 align:middle line:84% Remember, that's a sheet resistance. 00:04:54.700 --> 00:04:57.980 align:middle line:84% So that's an ohms per square sheet 00:04:57.980 --> 00:04:59.930 align:middle line:84% As for current flowing through a sheet, 00:04:59.930 --> 00:05:04.430 align:middle line:84% we talked about this, that is a square, so going in this space, 00:05:04.430 --> 00:05:08.670 align:middle line:84% coming out that face, where this length and this width are equal. 00:05:08.670 --> 00:05:10.770 align:middle line:84% That's the definition of the sheet resistance. 00:05:10.770 --> 00:05:13.505 align:middle line:84% So it's about 4 ohms per square, something like that. 00:05:13.505 --> 00:05:16.020 align:middle line:90% 00:05:16.020 --> 00:05:18.420 align:middle line:84% Then we need to know-- once you know the sheet 00:05:18.420 --> 00:05:21.280 align:middle line:84% resistance, the nice thing about once the sheet resistance, that 00:05:21.280 --> 00:05:23.280 align:middle line:84% tells you the resistance going through a square. 00:05:23.280 --> 00:05:25.230 align:middle line:84% You just use geometry considerations, 00:05:25.230 --> 00:05:26.820 align:middle line:84% how many squares does the current go 00:05:26.820 --> 00:05:31.160 align:middle line:84% through in flowing from this point here to the contact? 00:05:31.160 --> 00:05:33.530 align:middle line:90% Well, it's a half micron wall. 00:05:33.530 --> 00:05:36.180 align:middle line:90% And it's a micron into the page. 00:05:36.180 --> 00:05:40.870 align:middle line:84% So that's a half a square, basically, 00:05:40.870 --> 00:05:43.100 align:middle line:84% because it looks something like this. 00:05:43.100 --> 00:05:44.170 align:middle line:90% It's a half micron long. 00:05:44.170 --> 00:05:47.690 align:middle line:84% But it goes a micron into the board, into the page. 00:05:47.690 --> 00:05:48.980 align:middle line:90% So this is 0.5. 00:05:48.980 --> 00:05:50.110 align:middle line:90% And this is 1. 00:05:50.110 --> 00:05:53.080 align:middle line:84% So by definition, that's half a square 00:05:53.080 --> 00:05:54.460 align:middle line:84% because if it were a full square, 00:05:54.460 --> 00:05:56.585 align:middle line:84% there'd be two of these right next to each other. 00:05:56.585 --> 00:05:58.210 align:middle line:84% So then you can just get the resistance 00:05:58.210 --> 00:06:07.680 align:middle line:84% of a silicide to be a 3.8 ohms per squared times 1/2 a square. 00:06:07.680 --> 00:06:12.390 align:middle line:84% And you end up with about 1.9 ohms, 00:06:12.390 --> 00:06:14.700 align:middle line:84% very small, comparatively speaking, 00:06:14.700 --> 00:06:17.340 align:middle line:84% couple of orders of magnitude, maybe a factor of 50 00:06:17.340 --> 00:06:20.250 align:middle line:84% less than the contact resistance. 00:06:20.250 --> 00:06:22.820 align:middle line:90% 00:06:22.820 --> 00:06:24.630 align:middle line:90% OK, so that seems reasonable. 00:06:24.630 --> 00:06:29.180 align:middle line:84% And then because it's a metal and so it has a very low sheet 00:06:29.180 --> 00:06:29.730 align:middle line:90% resistance. 00:06:29.730 --> 00:06:32.070 align:middle line:90% So the resistance is small. 00:06:32.070 --> 00:06:37.110 align:middle line:84% So on the final one, number three, what you're asked, now 00:06:37.110 --> 00:06:41.540 align:middle line:84% we're going to be calculating the source drain extension 00:06:41.540 --> 00:06:42.040 align:middle line:90% resistance. 00:06:42.040 --> 00:06:43.498 align:middle line:84% So what's the resistance associated 00:06:43.498 --> 00:06:47.880 align:middle line:84% with this very thin, shallow junction depth, XJ, so 00:06:47.880 --> 00:06:49.270 align:middle line:90% this little resistor right here? 00:06:49.270 --> 00:06:50.937 align:middle line:84% So the current comes out of the channel, 00:06:50.937 --> 00:06:52.980 align:middle line:84% has to flow in that source drain extension 00:06:52.980 --> 00:06:56.190 align:middle line:90% before it hits the silicide. 00:06:56.190 --> 00:06:58.505 align:middle line:84% Well, again, that's a very similar situation 00:06:58.505 --> 00:07:00.630 align:middle line:84% where we just need to calculate a sheet resistance. 00:07:00.630 --> 00:07:02.100 align:middle line:84% And from the sheet resistance, we 00:07:02.100 --> 00:07:04.290 align:middle line:84% figure out the number of squares from the geometry. 00:07:04.290 --> 00:07:05.440 align:middle line:90% And we get the resistance. 00:07:05.440 --> 00:07:14.690 align:middle line:84% So again, so for the source drain extension, 00:07:14.690 --> 00:07:20.990 align:middle line:84% for that reason, we're told the resistive rho is 5 times 10 00:07:20.990 --> 00:07:22.580 align:middle line:90% to the minus 4 ohm centimeter. 00:07:22.580 --> 00:07:27.530 align:middle line:84% And we know the number of squares. 00:07:27.530 --> 00:07:30.280 align:middle line:84% And we have a thickness of that region. 00:07:30.280 --> 00:07:34.370 align:middle line:84% So I can get the sheet resistance. 00:07:34.370 --> 00:07:36.080 align:middle line:84% Again, it's just a resistivity divided 00:07:36.080 --> 00:07:38.430 align:middle line:90% by the thickness of the region. 00:07:38.430 --> 00:07:43.420 align:middle line:84% So that's 5 times 10 to the minus 4 ohm centimeter 00:07:43.420 --> 00:07:44.740 align:middle line:90% divided by-- 00:07:44.740 --> 00:07:46.240 align:middle line:90% I think it was an easy number-- 00:07:46.240 --> 00:07:48.010 align:middle line:90% 500 angstrom. 00:07:48.010 --> 00:07:51.460 align:middle line:84% So that's 5 times 10 to the minus six centimeters. 00:07:51.460 --> 00:07:52.450 align:middle line:90% And that goes out. 00:07:52.450 --> 00:07:58.210 align:middle line:84% And we end up with about 100 ohms per square. 00:07:58.210 --> 00:08:02.110 align:middle line:84% That's a typical shallow junction for a shallow junction, 00:08:02.110 --> 00:08:05.170 align:middle line:84% shallow being 500 angstroms or less. 00:08:05.170 --> 00:08:06.670 align:middle line:90% You dope it very heavily. 00:08:06.670 --> 00:08:09.760 align:middle line:84% We said this is doped 210 to the 20 with arsenic or whatever. 00:08:09.760 --> 00:08:14.530 align:middle line:84% And it's about 100 ohms per square as the sheet resistance. 00:08:14.530 --> 00:08:15.430 align:middle line:90% Now, again, 00:08:15.430 --> 00:08:17.193 align:middle line:84% Looking at this, how many squares is this? 00:08:17.193 --> 00:08:18.610 align:middle line:84% Well, the distance from this point 00:08:18.610 --> 00:08:20.260 align:middle line:84% in the channel to the point where 00:08:20.260 --> 00:08:22.200 align:middle line:84% it contacts the silicide that we're told, 00:08:22.200 --> 00:08:24.410 align:middle line:84% that spacer width is 0.1 microns. 00:08:24.410 --> 00:08:25.300 align:middle line:90% So that's 0.1. 00:08:25.300 --> 00:08:27.920 align:middle line:90% And into the board is 1 again. 00:08:27.920 --> 00:08:30.990 align:middle line:90% So that's a tenth of a square. 00:08:30.990 --> 00:08:34.190 align:middle line:84% So then we can just get the resistance of the source drain 00:08:34.190 --> 00:08:35.809 align:middle line:90% extension. 00:08:35.809 --> 00:08:38.750 align:middle line:84% Then it's just the sheet resistance 00:08:38.750 --> 00:08:41.039 align:middle line:90% times the number of squares. 00:08:41.039 --> 00:08:47.490 align:middle line:84% So that's 100 ohms per squared times 0.1 squares. 00:08:47.490 --> 00:08:50.770 align:middle line:90% And then we end up with 10 ohms. 00:08:50.770 --> 00:08:52.495 align:middle line:90% So that resistors, roughly-- 00:08:52.495 --> 00:08:56.290 align:middle line:84% I mean, this is a very crude estimate, that resistor here. 00:08:56.290 --> 00:08:58.240 align:middle line:84% So this resistor here is about 10 ohms. 00:08:58.240 --> 00:09:00.730 align:middle line:84% This little sheet here represents about a couple 00:09:00.730 --> 00:09:01.660 align:middle line:90% of ohms. 00:09:01.660 --> 00:09:03.490 align:middle line:90% The big resistance is what? 00:09:03.490 --> 00:09:05.147 align:middle line:84% The big resistance is the 100 ohms. 00:09:05.147 --> 00:09:06.730 align:middle line:84% It's the contact resistance because we 00:09:06.730 --> 00:09:10.570 align:middle line:84% have a finite specific contact resistivity, 00:09:10.570 --> 00:09:13.240 align:middle line:84% 2 and 1/2 times 10 to the minus 7 ohm centimeter squared. 00:09:13.240 --> 00:09:16.040 align:middle line:84% And we're trying to make contact in a very small area. 00:09:16.040 --> 00:09:16.998 align:middle line:90% So what's the solution? 00:09:16.998 --> 00:09:18.373 align:middle line:84% Well, I can make the area bigger, 00:09:18.373 --> 00:09:19.930 align:middle line:84% make the devices bigger on the chip. 00:09:19.930 --> 00:09:20.980 align:middle line:90% I get fewer devices. 00:09:20.980 --> 00:09:22.240 align:middle line:90% That's not good. 00:09:22.240 --> 00:09:24.427 align:middle line:90% Or we have to scale this number. 00:09:24.427 --> 00:09:26.260 align:middle line:84% According to the ITRS, remember, this number 00:09:26.260 --> 00:09:27.490 align:middle line:90% is going to have to go down. 00:09:27.490 --> 00:09:31.690 align:middle line:84% In 2003, ITRS wants this number to be about 2 times 10 00:09:31.690 --> 00:09:33.400 align:middle line:90% to minus 7. 00:09:33.400 --> 00:09:35.390 align:middle line:84% So it's already supposed to be lower than that. 00:09:35.390 --> 00:09:37.098 align:middle line:84% And over the next 10 years, it's supposed 00:09:37.098 --> 00:09:38.920 align:middle line:84% to be lowered by a factor of 3 or 4. 00:09:38.920 --> 00:09:42.400 align:middle line:84% So this number of 100 ohms gets lower 00:09:42.400 --> 00:09:44.170 align:middle line:84% or it certainly does not increase 00:09:44.170 --> 00:09:45.635 align:middle line:90% as we scale the devices. 00:09:45.635 --> 00:09:47.260 align:middle line:84% But just kind of gives you a rough idea 00:09:47.260 --> 00:09:50.230 align:middle line:84% of what the big contributions are if you do it right. 00:09:50.230 --> 00:09:53.070 align:middle line:84% Again, if you didn't get enough doping in this shallow source 00:09:53.070 --> 00:09:55.160 align:middle line:84% drain extension, this 10 ohm number 00:09:55.160 --> 00:09:59.230 align:middle line:84% could go up, depending on your junction depth, how you 00:09:59.230 --> 00:10:01.210 align:middle line:90% scale that, and your doping. 00:10:01.210 --> 00:10:03.220 align:middle line:84% But these are simple calculations 00:10:03.220 --> 00:10:04.780 align:middle line:84% you can do to back of the envelope. 00:10:04.780 --> 00:10:07.430 align:middle line:84% And we didn't get to go through that last time we presented it. 00:10:07.430 --> 00:10:12.520 align:middle line:84% But this gives you an idea of how those numbers work out. 00:10:12.520 --> 00:10:15.380 align:middle line:84% Any questions on that or on the contact? 00:10:15.380 --> 00:10:17.380 align:middle line:84% We only spent one lecture on contacts. 00:10:17.380 --> 00:10:21.310 align:middle line:84% So there's a little bit of this in chapter 11, 00:10:21.310 --> 00:10:23.590 align:middle line:84% although I don't think there's a specific example 00:10:23.590 --> 00:10:24.625 align:middle line:90% of a MOSFET like this. 00:10:24.625 --> 00:10:28.000 align:middle line:90% 00:10:28.000 --> 00:10:28.510 align:middle line:90% OK. 00:10:28.510 --> 00:10:33.580 align:middle line:84% So let's go on then to today's actual material, which 00:10:33.580 --> 00:10:38.420 align:middle line:84% is going to be a little bit of a divergence, to a certain extent, 00:10:38.420 --> 00:10:40.300 align:middle line:90% from the other lectures. 00:10:40.300 --> 00:10:41.800 align:middle line:84% This material is not in the text. 00:10:41.800 --> 00:10:47.770 align:middle line:84% This is all more up-to-date research material 00:10:47.770 --> 00:10:50.650 align:middle line:84% that's come out and has not yet been put into a textbook. 00:10:50.650 --> 00:10:52.960 align:middle line:90% But you have the handouts. 00:10:52.960 --> 00:10:55.540 align:middle line:84% So what I want to talk about-- what we've talked about so 00:10:55.540 --> 00:11:00.370 align:middle line:84% far for semiconductors is just using pure silicon, pure silicon 00:11:00.370 --> 00:11:05.020 align:middle line:84% wafers and pure silicon materials to make CMOS, 00:11:05.020 --> 00:11:07.240 align:middle line:90% to make MOSFETs. 00:11:07.240 --> 00:11:09.183 align:middle line:84% It turns out, if we use epitaxial growth, 00:11:09.183 --> 00:11:10.600 align:middle line:84% we did spend a little time talking 00:11:10.600 --> 00:11:12.190 align:middle line:84% about epitaxial growth of silicon 00:11:12.190 --> 00:11:15.800 align:middle line:84% by CVD or some similar technique. 00:11:15.800 --> 00:11:17.320 align:middle line:84% If we use epitaxial growth, you can 00:11:17.320 --> 00:11:19.630 align:middle line:84% mix in small amounts of germanium 00:11:19.630 --> 00:11:21.790 align:middle line:84% into the silicon material and make an alloy 00:11:21.790 --> 00:11:22.900 align:middle line:90% called silicon germanium. 00:11:22.900 --> 00:11:25.570 align:middle line:84% Or you can mix an even smaller amounts 00:11:25.570 --> 00:11:28.910 align:middle line:84% of carbon, maybe 1% or so, and make an alloy of the silicon 00:11:28.910 --> 00:11:29.660 align:middle line:90% and carbon. 00:11:29.660 --> 00:11:31.220 align:middle line:90% And people are doing this. 00:11:31.220 --> 00:11:34.010 align:middle line:84% And some of these alloys are used today in modern production 00:11:34.010 --> 00:11:35.570 align:middle line:90% devices in manufacturing. 00:11:35.570 --> 00:11:39.170 align:middle line:84% So I just wanted to go through some of the important material 00:11:39.170 --> 00:11:42.380 align:middle line:84% electronic properties of these newer semiconductors 00:11:42.380 --> 00:11:45.290 align:middle line:84% and how the layers are processed to a certain extent 00:11:45.290 --> 00:11:47.900 align:middle line:84% and what some of the issues in research and development 00:11:47.900 --> 00:11:50.940 align:middle line:90% that people are considering. 00:11:50.940 --> 00:11:53.600 align:middle line:84% So here on slide two, what I'm showing 00:11:53.600 --> 00:11:57.380 align:middle line:84% are some applications of these silicon germanium 00:11:57.380 --> 00:12:01.760 align:middle line:84% alloys and related materials like silicon germanium carbon. 00:12:01.760 --> 00:12:05.450 align:middle line:84% The first bullet is for a number of years now, 00:12:05.450 --> 00:12:07.940 align:middle line:84% maybe the last 10 years, silicon germanium 00:12:07.940 --> 00:12:10.370 align:middle line:84% has been used in production as the base 00:12:10.370 --> 00:12:14.180 align:middle line:84% layer in a heterojunction bipolar transistor, 00:12:14.180 --> 00:12:17.570 align:middle line:84% in a bipolar transistor, the base layer being quite thin. 00:12:17.570 --> 00:12:19.860 align:middle line:84% We haven't talked much about bipolars in this class. 00:12:19.860 --> 00:12:22.850 align:middle line:84% This class is mostly about making CMOS devices. 00:12:22.850 --> 00:12:25.520 align:middle line:84% But it's a thin layer, a couple hundred angstroms, maybe 00:12:25.520 --> 00:12:27.760 align:middle line:90% 200 to 500 angstroms thick. 00:12:27.760 --> 00:12:31.330 align:middle line:84% This is used in high frequency analog applications, generally 00:12:31.330 --> 00:12:34.440 align:middle line:84% telecommunications applications, base stations, and things 00:12:34.440 --> 00:12:34.940 align:middle line:90% like that. 00:12:34.940 --> 00:12:37.540 align:middle line:84% And there are a number of companies, maybe 00:12:37.540 --> 00:12:40.210 align:middle line:84% half a dozen or more, that manufacture these and sell 00:12:40.210 --> 00:12:41.300 align:middle line:90% these devices. 00:12:41.300 --> 00:12:44.290 align:middle line:84% So for the purpose of growing thin films, 00:12:44.290 --> 00:12:46.450 align:middle line:84% this is something that has been done now 00:12:46.450 --> 00:12:49.160 align:middle line:90% in manufacturing for some time. 00:12:49.160 --> 00:12:52.930 align:middle line:84% The second bullet refers to a little more advanced type 00:12:52.930 --> 00:12:53.600 align:middle line:90% of material. 00:12:53.600 --> 00:12:56.270 align:middle line:84% And this just coming in to production now. 00:12:56.270 --> 00:12:59.818 align:middle line:84% And these are layers of relaxed silicon germanium or strained 00:12:59.818 --> 00:13:02.110 align:middle line:84% silicon germanium-- and I'll talk about what that means 00:13:02.110 --> 00:13:03.250 align:middle line:90% in a little while-- 00:13:03.250 --> 00:13:05.440 align:middle line:84% that are used to induce strain in silicon. 00:13:05.440 --> 00:13:09.280 align:middle line:84% So these layers are used to take silicon channel 00:13:09.280 --> 00:13:12.880 align:middle line:84% and to either stretch it or compress it mechanically. 00:13:12.880 --> 00:13:14.470 align:middle line:84% And by stretching and compressing 00:13:14.470 --> 00:13:17.410 align:middle line:84% the lattice in the channel, you can enhance the mobility 00:13:17.410 --> 00:13:18.280 align:middle line:90% as it turns out. 00:13:18.280 --> 00:13:20.738 align:middle line:84% Enhancing the mobility means you get a higher current drive 00:13:20.738 --> 00:13:23.930 align:middle line:84% for a given voltage so the devices will switch faster. 00:13:23.930 --> 00:13:28.040 align:middle line:84% So for enhancing CMOS, some of these materials 00:13:28.040 --> 00:13:29.240 align:middle line:90% are now in production. 00:13:29.240 --> 00:13:33.290 align:middle line:84% Intel has a new Pentium that's in production using this type 00:13:33.290 --> 00:13:36.530 align:middle line:90% of technique, these materials. 00:13:36.530 --> 00:13:39.260 align:middle line:84% Sometimes, silicon germanium can be used as a boron diffusion 00:13:39.260 --> 00:13:39.860 align:middle line:90% barrier. 00:13:39.860 --> 00:13:42.318 align:middle line:84% We spent a lot of time in this class talking about the fact 00:13:42.318 --> 00:13:44.625 align:middle line:84% that boron is a fast diffuser in silicon. 00:13:44.625 --> 00:13:47.000 align:middle line:84% That's one of the problems with making shallow junctions. 00:13:47.000 --> 00:13:48.980 align:middle line:84% Well, it turns out, in silicon germanium, 00:13:48.980 --> 00:13:51.380 align:middle line:84% for reasons that are not completely understood, 00:13:51.380 --> 00:13:54.320 align:middle line:84% the more germanium you add, the slower the boron diffusion 00:13:54.320 --> 00:13:54.990 align:middle line:90% coefficient. 00:13:54.990 --> 00:13:57.260 align:middle line:84% So if you put in 20% germanium the lattice, 00:13:57.260 --> 00:14:00.110 align:middle line:84% the boron diffusivity drops by about a factor of 10 00:14:00.110 --> 00:14:01.180 align:middle line:90% in silicon germanium. 00:14:01.180 --> 00:14:03.300 align:middle line:90% So that's a huge factor. 00:14:03.300 --> 00:14:05.810 align:middle line:84% So people may want to use silicon germanium, just as 00:14:05.810 --> 00:14:08.540 align:middle line:84% a means of slowing down boron diffusion. 00:14:08.540 --> 00:14:10.160 align:middle line:90% It's used in micromachining. 00:14:10.160 --> 00:14:13.070 align:middle line:84% In MEMS technology, it happens to be an excellent etch stop. 00:14:13.070 --> 00:14:17.780 align:middle line:84% So there are etches, wet chemical etches, primarily wet, 00:14:17.780 --> 00:14:21.410 align:middle line:84% maybe a few dry, that will stop on silicon germanium 00:14:21.410 --> 00:14:23.990 align:middle line:84% and etch through silicon and stop on silicon germanium 00:14:23.990 --> 00:14:25.110 align:middle line:90% and vice versa. 00:14:25.110 --> 00:14:29.910 align:middle line:84% So it can be used as a sort of an etch stop layer. 00:14:29.910 --> 00:14:31.417 align:middle line:84% Some people are advocating using-- 00:14:31.417 --> 00:14:33.000 align:middle line:84% I don't think it's in production but I 00:14:33.000 --> 00:14:35.130 align:middle line:84% know of-- but polycrystalline silicon 00:14:35.130 --> 00:14:37.270 align:middle line:90% germanium as the gate material. 00:14:37.270 --> 00:14:39.270 align:middle line:84% We've talked in this class the last few lectures 00:14:39.270 --> 00:14:40.380 align:middle line:90% about metal gates. 00:14:40.380 --> 00:14:42.060 align:middle line:84% Before they get to metal gates, people 00:14:42.060 --> 00:14:44.850 align:middle line:84% were advocating using a polycrystalline silicon 00:14:44.850 --> 00:14:47.190 align:middle line:84% germanium instead of polycrystalline silicon. 00:14:47.190 --> 00:14:48.720 align:middle line:84% I don't know, again, if that really 00:14:48.720 --> 00:14:50.100 align:middle line:90% is making it into production. 00:14:50.100 --> 00:14:52.680 align:middle line:84% Silicon germanium is also used as an infrared photo detector 00:14:52.680 --> 00:14:55.590 align:middle line:84% material for integrating photodetectors 00:14:55.590 --> 00:14:57.940 align:middle line:90% with silicon CMOS. 00:14:57.940 --> 00:15:00.997 align:middle line:84% Silicon germanium carbon is in production now, maybe not 00:15:00.997 --> 00:15:02.830 align:middle line:84% silicon carbon, but silicon germanium carbon 00:15:02.830 --> 00:15:05.740 align:middle line:84% is in production with very small amounts of carbon, maybe a half 00:15:05.740 --> 00:15:08.680 align:middle line:84% a percent, as a way of suppressing 00:15:08.680 --> 00:15:12.670 align:middle line:84% transient enhanced diffusion in heterojunction bipolar 00:15:12.670 --> 00:15:13.540 align:middle line:90% transistors. 00:15:13.540 --> 00:15:15.748 align:middle line:84% So think somebody's doing a report on that. 00:15:15.748 --> 00:15:17.290 align:middle line:84% One of the people signed up for that. 00:15:17.290 --> 00:15:22.330 align:middle line:84% So that's kind of a hot topic, the use of carbon to completely 00:15:22.330 --> 00:15:23.830 align:middle line:90% suck up all the interstitials. 00:15:23.830 --> 00:15:27.320 align:middle line:84% And then that completely changes the TD effect. 00:15:27.320 --> 00:15:30.610 align:middle line:84% So these materials have quite a bit of application. 00:15:30.610 --> 00:15:33.460 align:middle line:84% So on slide 3, this looks like a long outline. 00:15:33.460 --> 00:15:35.200 align:middle line:84% But I'm just going to touch on-- after I 00:15:35.200 --> 00:15:36.970 align:middle line:90% introduce these the materials. 00:15:36.970 --> 00:15:40.420 align:middle line:84% I'm going to touch a little bit on the MOSFETs 00:15:40.420 --> 00:15:43.090 align:middle line:84% and the applications of strained silicon and strained silicon 00:15:43.090 --> 00:15:44.957 align:middle line:90% germanium to CMOS. 00:15:44.957 --> 00:15:47.290 align:middle line:84% I'll talk a little bit about how the materials are grown 00:15:47.290 --> 00:15:49.060 align:middle line:84% and about how the dopants diffused 00:15:49.060 --> 00:15:52.450 align:middle line:84% in these materials and a little bit 00:15:52.450 --> 00:15:55.270 align:middle line:84% about some newer material where we're putting them on insulator. 00:15:55.270 --> 00:15:58.360 align:middle line:84% Strained silicon on insulator is a very kind 00:15:58.360 --> 00:16:02.860 align:middle line:84% of a hot topic of research these days. 00:16:02.860 --> 00:16:08.110 align:middle line:84% Well, if we go to slide 4, this is more of a motivational slide. 00:16:08.110 --> 00:16:09.760 align:middle line:84% The question we're trying to ask is, 00:16:09.760 --> 00:16:13.300 align:middle line:84% if you have a CMOS circuit, maybe inverter or something, 00:16:13.300 --> 00:16:16.180 align:middle line:84% how quickly can it switch from one state of the zero state 00:16:16.180 --> 00:16:20.260 align:middle line:84% to on state-- the zero state to the one state, for example? 00:16:20.260 --> 00:16:22.870 align:middle line:84% And this the delay in doing that is something 00:16:22.870 --> 00:16:24.640 align:middle line:90% called the gate delay. 00:16:24.640 --> 00:16:26.260 align:middle line:84% It's usually measured in picoseconds. 00:16:26.260 --> 00:16:28.695 align:middle line:84% And that gate delay is something we want to minimize. 00:16:28.695 --> 00:16:30.070 align:middle line:84% We want to minimize the switching 00:16:30.070 --> 00:16:33.340 align:middle line:84% of each gate in the logic sequence 00:16:33.340 --> 00:16:38.180 align:middle line:84% so that the overall circuit can do computation faster. 00:16:38.180 --> 00:16:41.890 align:middle line:84% So the gate delay generally, if you've taken classes like 6012 00:16:41.890 --> 00:16:45.970 align:middle line:84% or you've had some other undergraduate classes on CMOS, 00:16:45.970 --> 00:16:48.730 align:middle line:84% as you scale the gate length-- that's why we're making the gate 00:16:48.730 --> 00:16:49.960 align:middle line:90% length shorter and shorter-- 00:16:49.960 --> 00:16:51.170 align:middle line:90% the gate delay goes down. 00:16:51.170 --> 00:16:53.650 align:middle line:84% This is actually some data showing how it does indeed 00:16:53.650 --> 00:16:55.030 align:middle line:90% scale at 0.1 micron. 00:16:55.030 --> 00:16:59.140 align:middle line:84% The gate delay is down in the 5 or so picosecond 00:16:59.140 --> 00:17:01.940 align:middle line:90% or some 5 picosecond range. 00:17:01.940 --> 00:17:04.760 align:middle line:84% Well, it turns out that the gate delay is inversely 00:17:04.760 --> 00:17:06.119 align:middle line:90% proportional to the current. 00:17:06.119 --> 00:17:10.220 align:middle line:84% So if you can pump more current through each device, then 00:17:10.220 --> 00:17:13.010 align:middle line:84% as one device charges, the device after it 00:17:13.010 --> 00:17:14.093 align:middle line:90% has to charge up a device. 00:17:14.093 --> 00:17:15.802 align:middle line:84% The more currents going through, the more 00:17:15.802 --> 00:17:17.730 align:middle line:84% quickly it will charge its neighboring device. 00:17:17.730 --> 00:17:20.329 align:middle line:84% So the name of the game in getting higher speed in logic 00:17:20.329 --> 00:17:23.420 align:middle line:84% devices, at least for digital, is at a given voltage, 00:17:23.420 --> 00:17:26.849 align:middle line:84% to get more drain current out of the device. 00:17:26.849 --> 00:17:28.975 align:middle line:84% And so you say, OK, at a given voltage, 00:17:28.975 --> 00:17:30.350 align:middle line:84% how can I get more drain current? 00:17:30.350 --> 00:17:32.390 align:middle line:84% Well, this is what the drain current looks like. 00:17:32.390 --> 00:17:35.360 align:middle line:84% There are several physical variables it depends on. 00:17:35.360 --> 00:17:37.940 align:middle line:84% It depends on the mobility of the carriers, directly 00:17:37.940 --> 00:17:38.750 align:middle line:90% proportional. 00:17:38.750 --> 00:17:41.100 align:middle line:84% It's inversely proportional to the oxide thickness. 00:17:41.100 --> 00:17:43.970 align:middle line:84% So if I make the oxide thinner, I get a higher capacitance, 00:17:43.970 --> 00:17:45.020 align:middle line:90% and the current goes up. 00:17:45.020 --> 00:17:47.330 align:middle line:84% That's why we're talking about gate dielectric scaling 00:17:47.330 --> 00:17:51.318 align:middle line:84% all the way down to below 10 angstroms. 00:17:51.318 --> 00:17:53.610 align:middle line:84% So that's where the gate dielectric scaling comes from. 00:17:53.610 --> 00:17:57.090 align:middle line:84% And if I make the channel shorter, it does also go up. 00:17:57.090 --> 00:18:00.140 align:middle line:84% So how do people do conventional scaling 00:18:00.140 --> 00:18:02.820 align:middle line:84% to increase ID or speed for logic? 00:18:02.820 --> 00:18:05.148 align:middle line:84% They make the gate length shorter. 00:18:05.148 --> 00:18:06.690 align:middle line:84% They reduce the gate oxide thickness. 00:18:06.690 --> 00:18:09.780 align:middle line:84% The real question has been in the last five or 10 years, what 00:18:09.780 --> 00:18:11.310 align:middle line:90% about the mobility? 00:18:11.310 --> 00:18:13.617 align:middle line:84% Is that a variable that we can tweak? 00:18:13.617 --> 00:18:15.450 align:middle line:84% We're pretty much done with scaling the gate 00:18:15.450 --> 00:18:19.180 align:middle line:84% oxide because we can't get much thinner, only a few atoms thick. 00:18:19.180 --> 00:18:21.660 align:middle line:84% Making the transistor shorter is getting harder and harder 00:18:21.660 --> 00:18:24.060 align:middle line:84% lithographically and for electrostatic reasons, 00:18:24.060 --> 00:18:26.260 align:middle line:84% don't have too many more variables in this equation. 00:18:26.260 --> 00:18:30.510 align:middle line:84% So mobility is the last thing to touch. 00:18:30.510 --> 00:18:32.760 align:middle line:84% And that's where the silicon germanium alloys come in. 00:18:32.760 --> 00:18:35.370 align:middle line:84% Here on page five, what I'm showing 00:18:35.370 --> 00:18:38.820 align:middle line:84% is some classic data on the literature 00:18:38.820 --> 00:18:41.280 align:middle line:84% of what the electron mobility that number. 00:18:41.280 --> 00:18:43.800 align:middle line:84% Mu N is not a single number in a MOSFET. 00:18:43.800 --> 00:18:46.230 align:middle line:84% So this is a plot of the effective electron 00:18:46.230 --> 00:18:47.760 align:middle line:90% mobility in a channel. 00:18:47.760 --> 00:18:49.650 align:middle line:84% And again, as this number goes up, 00:18:49.650 --> 00:18:51.660 align:middle line:84% the current goes up at a given voltage. 00:18:51.660 --> 00:18:53.970 align:middle line:84% As a function of the vertical field in the device, 00:18:53.970 --> 00:18:55.800 align:middle line:84% the vertical field is induced by the gate. 00:18:55.800 --> 00:18:59.400 align:middle line:84% So as I put a gate bias on the device, 00:18:59.400 --> 00:19:01.830 align:middle line:84% as I bias the device, the gate bias, 00:19:01.830 --> 00:19:03.230 align:middle line:90% I get higher inversion charge. 00:19:03.230 --> 00:19:06.825 align:middle line:84% And so in fact, the mobility actually goes down. 00:19:06.825 --> 00:19:08.200 align:middle line:84% But so you notice it's a function 00:19:08.200 --> 00:19:11.170 align:middle line:84% of the vertical effective field in the silicon 00:19:11.170 --> 00:19:14.830 align:middle line:84% and that modern MOSFETs are operating somewhere 00:19:14.830 --> 00:19:16.010 align:middle line:90% in this range right now. 00:19:16.010 --> 00:19:19.420 align:middle line:84% I can just tell you, based on the gate oxide thickness 00:19:19.420 --> 00:19:21.850 align:middle line:84% and the typical voltages, 1 to 3 volts 00:19:21.850 --> 00:19:25.400 align:middle line:84% that MOSFETs are operating in, if you calculated out the effect 00:19:25.400 --> 00:19:27.400 align:middle line:84% the field in the channel, it's about a mega volt 00:19:27.400 --> 00:19:28.400 align:middle line:90% per centimeter. 00:19:28.400 --> 00:19:31.600 align:middle line:84% So modern MOSFETs these days, the electron mobility 00:19:31.600 --> 00:19:34.350 align:middle line:84% is right about on this curve, somewhere between 200 00:19:34.350 --> 00:19:37.180 align:middle line:90% and 300 but, closer to 200. 00:19:37.180 --> 00:19:39.940 align:middle line:84% A number of years ago when we used lighter doping, 00:19:39.940 --> 00:19:43.360 align:middle line:84% so let's say you're using much lighter doping like 317, 00:19:43.360 --> 00:19:46.570 align:middle line:84% and the oxides were thicker, the vertical field was much lower, 00:19:46.570 --> 00:19:49.000 align:middle line:90% maybe half of what it is today. 00:19:49.000 --> 00:19:50.125 align:middle line:90% And the mobility was twice. 00:19:50.125 --> 00:19:51.430 align:middle line:90% It was what 400. 00:19:51.430 --> 00:19:53.350 align:middle line:84% So this is the problem with mobility. 00:19:53.350 --> 00:19:57.992 align:middle line:84% As we march down the ITRS, we scale that gate oxide thickness. 00:19:57.992 --> 00:19:59.200 align:middle line:90% So we get higher capacitance. 00:19:59.200 --> 00:20:01.010 align:middle line:90% That's good. 00:20:01.010 --> 00:20:03.460 align:middle line:84% But we also increasing the doping in the channel, 00:20:03.460 --> 00:20:05.860 align:middle line:84% and we're increasing the vertical effective field. 00:20:05.860 --> 00:20:07.425 align:middle line:84% So we're marching down this curve. 00:20:07.425 --> 00:20:09.550 align:middle line:84% And the mobility is going lower and lower and lower 00:20:09.550 --> 00:20:10.510 align:middle line:90% with each generation. 00:20:10.510 --> 00:20:11.090 align:middle line:90% That's bad. 00:20:11.090 --> 00:20:16.567 align:middle line:84% That takes away from our current drive, to a certain extent. 00:20:16.567 --> 00:20:17.400 align:middle line:90% So that's a problem. 00:20:17.400 --> 00:20:18.775 align:middle line:84% So what people have decided to do 00:20:18.775 --> 00:20:23.010 align:middle line:84% is OK, it turns out, if I can change this mobility curve, 00:20:23.010 --> 00:20:25.880 align:middle line:84% in fact, I can push the whole curve up quite significantly 00:20:25.880 --> 00:20:28.790 align:middle line:90% if I put stress on the silicon. 00:20:28.790 --> 00:20:31.340 align:middle line:84% And I'll say a little bit more in the next few slides 00:20:31.340 --> 00:20:33.520 align:middle line:90% about how that works. 00:20:33.520 --> 00:20:35.790 align:middle line:84% So people have tried a number of different ways 00:20:35.790 --> 00:20:39.420 align:middle line:84% to stress out the devices or to put stress on the silicone. 00:20:39.420 --> 00:20:41.290 align:middle line:84% You have to stress it in a certain way. 00:20:41.290 --> 00:20:43.020 align:middle line:84% And this schematic kind of gives you 00:20:43.020 --> 00:20:45.600 align:middle line:84% an idea of a good way of inducing stress. 00:20:45.600 --> 00:20:49.050 align:middle line:84% For electrons, this looks like just like a regular nMOSFET. 00:20:49.050 --> 00:20:53.370 align:middle line:84% But you notice I've highlighted this channel region in yellow. 00:20:53.370 --> 00:20:54.800 align:middle line:84% And it's called strained silicon. 00:20:54.800 --> 00:20:56.460 align:middle line:84% A classically strained silicon was 00:20:56.460 --> 00:20:59.940 align:middle line:84% generated by taking a wafer of silicon germanium 00:20:59.940 --> 00:21:04.350 align:middle line:84% or a very thick layer of silicon germanium, say with 20% or 30% 00:21:04.350 --> 00:21:06.540 align:middle line:84% germanium, silicon germanium has a larger lattice 00:21:06.540 --> 00:21:08.630 align:middle line:90% parameter than silicon. 00:21:08.630 --> 00:21:10.800 align:middle line:84% Pure germanium is about 4% bigger. 00:21:10.800 --> 00:21:15.388 align:middle line:84% So this 30% layer is going to be a little bit larger 00:21:15.388 --> 00:21:16.430 align:middle line:90% in its lattice parameter. 00:21:16.430 --> 00:21:18.870 align:middle line:90% So if it's relaxed, it's bigger. 00:21:18.870 --> 00:21:20.900 align:middle line:84% And if the silicon layer is thin, 00:21:20.900 --> 00:21:23.060 align:middle line:84% when you grow it epitaxially, the silicon lattice 00:21:23.060 --> 00:21:25.940 align:middle line:84% will actually stretch in the X and the Y directions 00:21:25.940 --> 00:21:28.520 align:middle line:84% to match that of the substrate, as long as it's thin enough. 00:21:28.520 --> 00:21:30.680 align:middle line:84% It's energetically favorable to stretch the bonds 00:21:30.680 --> 00:21:32.460 align:middle line:90% instead of breaking them. 00:21:32.460 --> 00:21:33.830 align:middle line:90% So the silicon is stretched. 00:21:33.830 --> 00:21:37.610 align:middle line:84% This lattice parameter in the X and Y plane is about 1.5% 00:21:37.610 --> 00:21:40.400 align:middle line:84% bigger than it would be in normal silicon. 00:21:40.400 --> 00:21:44.070 align:middle line:84% That's a huge amount for it to take a crystal lattice 00:21:44.070 --> 00:21:45.380 align:middle line:90% and stretch it by 1.5%. 00:21:45.380 --> 00:21:46.850 align:middle line:90% That's a lot. 00:21:46.850 --> 00:21:49.280 align:middle line:84% As a result, when you stretch it in the X and Y, 00:21:49.280 --> 00:21:51.050 align:middle line:84% it actually compresses a little bit in Z. 00:21:51.050 --> 00:21:55.490 align:middle line:84% So in fact, we call this particular configuration uniform 00:21:55.490 --> 00:21:56.810 align:middle line:90% biaxial tensile stress. 00:21:56.810 --> 00:22:01.940 align:middle line:84% It's biaxial and it's tensile, is being pulled in X and Y. 00:22:01.940 --> 00:22:03.500 align:middle line:84% So this is sort of the classic way 00:22:03.500 --> 00:22:06.000 align:middle line:84% that people have introduced strain and made MOSFETs. 00:22:06.000 --> 00:22:08.000 align:middle line:84% And they've looked at the mobility as a function 00:22:08.000 --> 00:22:09.167 align:middle line:90% of the amount of the strain. 00:22:09.167 --> 00:22:11.360 align:middle line:84% There are other methods though for straining. 00:22:11.360 --> 00:22:13.220 align:middle line:84% Most of what I'll talk about in this lecture 00:22:13.220 --> 00:22:15.740 align:middle line:90% is this epitaxial growth method. 00:22:15.740 --> 00:22:18.443 align:middle line:84% It's the easiest one to get a handle on the physics. 00:22:18.443 --> 00:22:19.610 align:middle line:90% But there are other methods. 00:22:19.610 --> 00:22:22.280 align:middle line:84% People have used, for example, when you make a MOSFET, 00:22:22.280 --> 00:22:24.950 align:middle line:84% you don't just make an individual MOSFET on a device. 00:22:24.950 --> 00:22:28.490 align:middle line:84% All around the MOSFET, we do something called shallow trench 00:22:28.490 --> 00:22:30.510 align:middle line:84% isolation that we talked about in this course. 00:22:30.510 --> 00:22:33.980 align:middle line:84% We dig out a trench on all the way around the MOSFET. 00:22:33.980 --> 00:22:37.970 align:middle line:84% And we fill it up with oxide or something, some insulator. 00:22:37.970 --> 00:22:41.330 align:middle line:84% Well, IBM show back in '97, depending 00:22:41.330 --> 00:22:44.210 align:middle line:84% on how you fill that up, you can potentially induce stress, 00:22:44.210 --> 00:22:46.880 align:middle line:84% either compressive or tensile, depending on what material you 00:22:46.880 --> 00:22:50.505 align:middle line:84% put in because you're digging a moat around the device, 00:22:50.505 --> 00:22:52.130 align:middle line:84% and you're putting some material in it. 00:22:52.130 --> 00:22:58.080 align:middle line:84% So you can actually use the STI, so-called, to induce stress. 00:22:58.080 --> 00:23:00.690 align:middle line:84% There are other processes during fabrication 00:23:00.690 --> 00:23:05.460 align:middle line:84% that Intel is using now where they actually dig out the source 00:23:05.460 --> 00:23:06.090 align:middle line:90% and drains. 00:23:06.090 --> 00:23:07.547 align:middle line:90% They cut them out with etching. 00:23:07.547 --> 00:23:09.630 align:middle line:84% And they fill them back up with silicon germanium, 00:23:09.630 --> 00:23:12.390 align:middle line:84% which compresses the channel, which helps the hole mobility. 00:23:12.390 --> 00:23:14.670 align:middle line:84% So people are doing things like that. 00:23:14.670 --> 00:23:17.520 align:middle line:84% You can actually stress the device. 00:23:17.520 --> 00:23:20.070 align:middle line:84% On top of the device, they put a layer of high stress nitride 00:23:20.070 --> 00:23:21.810 align:middle line:84% and try to induce stress by applying 00:23:21.810 --> 00:23:23.355 align:middle line:90% films that are high stress. 00:23:23.355 --> 00:23:24.480 align:middle line:90% You can also bend the chip. 00:23:24.480 --> 00:23:27.868 align:middle line:84% There's a small startup company in the southeast of the United 00:23:27.868 --> 00:23:29.910 align:middle line:84% States where they're taking chips, thinning them, 00:23:29.910 --> 00:23:32.390 align:middle line:84% and actually trying to pin them mechanically during stress. 00:23:32.390 --> 00:23:33.390 align:middle line:90% It seems a little crazy. 00:23:33.390 --> 00:23:36.040 align:middle line:90% But it may be possible. 00:23:36.040 --> 00:23:40.060 align:middle line:84% But epitaxial growth is in some ways the most well controlled. 00:23:40.060 --> 00:23:45.181 align:middle line:84% And the physics is a little bit easier to get a handle on. 00:23:45.181 --> 00:23:47.180 align:middle line:84% So let me just mention here on slide 7. 00:23:47.180 --> 00:23:54.050 align:middle line:84% This is the IBM work in 1997 where they talked about in STI-- 00:23:54.050 --> 00:23:56.900 align:middle line:84% we learned in this course, you fill up the shallow trench 00:23:56.900 --> 00:23:57.710 align:middle line:90% with-- 00:23:57.710 --> 00:23:59.240 align:middle line:84% well, first, you oxidize the liner. 00:23:59.240 --> 00:24:02.180 align:middle line:84% Then you fill it up with LTO or some low temperature oxide. 00:24:02.180 --> 00:24:07.790 align:middle line:84% What they say was they grew an oxide liner for isolation. 00:24:07.790 --> 00:24:11.540 align:middle line:84% And then they fill the trench with polysilicon, 00:24:11.540 --> 00:24:14.120 align:middle line:84% which they thought would induce a lot of stress, 00:24:14.120 --> 00:24:16.310 align:middle line:84% compressive stress, pushing in all 00:24:16.310 --> 00:24:18.390 align:middle line:90% around the edges of the device. 00:24:18.390 --> 00:24:20.720 align:middle line:84% And then they looked at the hole mobility. 00:24:20.720 --> 00:24:23.270 align:middle line:84% And indeed, what they found, particularly 00:24:23.270 --> 00:24:27.410 align:middle line:84% on SOI layers, which are very thin, when they made the device 00:24:27.410 --> 00:24:30.872 align:middle line:84% width very small, say 1 micron compared to 10 microns, 00:24:30.872 --> 00:24:32.330 align:middle line:84% so you make the device width small, 00:24:32.330 --> 00:24:34.573 align:middle line:84% that brings the shallow trench in closer and closer 00:24:34.573 --> 00:24:36.740 align:middle line:84% to the center of the channel so it can have a bigger 00:24:36.740 --> 00:24:38.510 align:middle line:90% effect as it's pushing in. 00:24:38.510 --> 00:24:41.270 align:middle line:84% And they saw that the hole mobility went up quite a bit. 00:24:41.270 --> 00:24:43.290 align:middle line:84% On a 10-micron device, maybe the mobility 00:24:43.290 --> 00:24:45.330 align:middle line:90% was here, about 100 or 110. 00:24:45.330 --> 00:24:50.050 align:middle line:84% It went up by about 34% to about 140 just by doing this. 00:24:50.050 --> 00:24:51.720 align:middle line:84% Now, maybe it wasn't such a great idea 00:24:51.720 --> 00:24:54.420 align:middle line:84% to put that much stress and use polysilicon. 00:24:54.420 --> 00:24:56.100 align:middle line:84% It wasn't necessarily a practical thing. 00:24:56.100 --> 00:24:58.710 align:middle line:84% But it just shows that depending on how 00:24:58.710 --> 00:25:01.075 align:middle line:84% you do your processes, how you do your STI, 00:25:01.075 --> 00:25:03.450 align:middle line:84% it can have a pretty big effect, particularly on the hole 00:25:03.450 --> 00:25:05.540 align:middle line:90% mobility. 00:25:05.540 --> 00:25:08.960 align:middle line:84% So this was an indicator early on that people 00:25:08.960 --> 00:25:13.070 align:middle line:84% need to pay dramatic attention to exactly what thermal budget 00:25:13.070 --> 00:25:16.520 align:middle line:84% you use, what materials you use in STI every layer you 00:25:16.520 --> 00:25:18.060 align:middle line:90% put on that device. 00:25:18.060 --> 00:25:20.450 align:middle line:84% Now that the devices are small-- the width of devices 00:25:20.450 --> 00:25:22.560 align:middle line:90% today is less than a micron. 00:25:22.560 --> 00:25:24.410 align:middle line:84% Now that they're so small, edge effects 00:25:24.410 --> 00:25:26.382 align:middle line:84% can have a huge impact on the stress 00:25:26.382 --> 00:25:27.590 align:middle line:90% in the center of the channel. 00:25:27.590 --> 00:25:31.900 align:middle line:84% And that can end up impacting the mobility. 00:25:31.900 --> 00:25:33.900 align:middle line:84% So I refer to the fact that we have this strain. 00:25:33.900 --> 00:25:35.780 align:middle line:84% And I guess I should have introduced slide 8 a little bit 00:25:35.780 --> 00:25:36.500 align:middle line:90% earlier. 00:25:36.500 --> 00:25:40.190 align:middle line:84% But this just gives you a picture, a ball and stick 00:25:40.190 --> 00:25:41.760 align:middle line:84% picture of what we're talking about. 00:25:41.760 --> 00:25:44.420 align:middle line:84% So this ball and sticking up in the upper left 00:25:44.420 --> 00:25:46.550 align:middle line:84% is meant to represent cubic silicone 00:25:46.550 --> 00:25:49.370 align:middle line:84% because it has a lattice parameter of about 00:25:49.370 --> 00:25:50.720 align:middle line:90% 5.4 angstroms. 00:25:50.720 --> 00:25:52.250 align:middle line:84% It's the diamond cubic structure. 00:25:52.250 --> 00:25:55.460 align:middle line:84% But nevertheless, it has a cubic symmetry to it. 00:25:55.460 --> 00:25:57.920 align:middle line:84% X, Y and Z in those three dimensions, 00:25:57.920 --> 00:26:03.150 align:middle line:84% the lattice constant parameter is the same number. 00:26:03.150 --> 00:26:06.390 align:middle line:84% Cuprate germanium is larger, same cubic structure, 00:26:06.390 --> 00:26:08.280 align:middle line:90% but it's larger by about 4%. 00:26:08.280 --> 00:26:10.890 align:middle line:84% If you make an alloy anywhere in between, 00:26:10.890 --> 00:26:13.890 align:middle line:84% and you have it thick enough that it relaxes, 00:26:13.890 --> 00:26:15.842 align:middle line:84% pretty much can use a linear interpolation. 00:26:15.842 --> 00:26:16.800 align:middle line:90% It's not quite perfect. 00:26:16.800 --> 00:26:19.000 align:middle line:84% But you can use a linear interpolation. 00:26:19.000 --> 00:26:22.380 align:middle line:84% So at 50% silicon germanium or 50% germanium, 00:26:22.380 --> 00:26:25.782 align:middle line:84% the lattice mismatch is about 2%. 00:26:25.782 --> 00:26:27.990 align:middle line:84% So there are two different types of silicon germanium 00:26:27.990 --> 00:26:29.560 align:middle line:90% that people tend to talk about. 00:26:29.560 --> 00:26:31.950 align:middle line:84% One is when you start with a silicon substrate 00:26:31.950 --> 00:26:34.170 align:middle line:84% and you grow a very thin layer, say a few hundred 00:26:34.170 --> 00:26:37.170 align:middle line:84% angstroms of silicon germanium on silicon, what happens 00:26:37.170 --> 00:26:39.870 align:middle line:84% is you're taking this lattice here, 00:26:39.870 --> 00:26:43.290 align:middle line:84% this cubic silicon germanium wants to be cubic. 00:26:43.290 --> 00:26:44.280 align:middle line:90% It's larger. 00:26:44.280 --> 00:26:46.800 align:middle line:84% It wants to be larger in equilibrium than silicon. 00:26:46.800 --> 00:26:48.210 align:middle line:90% And you're squeezing it though. 00:26:48.210 --> 00:26:50.252 align:middle line:84% It turns out, again, if the layer is thin enough, 00:26:50.252 --> 00:26:53.580 align:middle line:84% it's energetically favorable for the in-plane lattice parameter 00:26:53.580 --> 00:26:55.140 align:middle line:84% of this material that you're growing 00:26:55.140 --> 00:26:56.950 align:middle line:90% to match that of the substrate. 00:26:56.950 --> 00:27:00.060 align:middle line:84% So the silicon germanium is kind of squeezed in X and y. 00:27:00.060 --> 00:27:03.770 align:middle line:84% And it pushes up in Z. So this layer then 00:27:03.770 --> 00:27:05.570 align:middle line:84% is what we call a strained silicon 00:27:05.570 --> 00:27:07.220 align:middle line:90% germanium on relaxed silicon. 00:27:07.220 --> 00:27:08.930 align:middle line:84% And it's under biaxial compressions. 00:27:08.930 --> 00:27:11.310 align:middle line:84% Being compressed in the XY plane. 00:27:11.310 --> 00:27:13.670 align:middle line:84% Now, if we make it thick enough, eventually, there's 00:27:13.670 --> 00:27:16.658 align:middle line:84% too much stored energy in those stretched bonds, 00:27:16.658 --> 00:27:18.950 align:middle line:84% and the bonds will start breaking and get dislocations, 00:27:18.950 --> 00:27:21.930 align:middle line:84% and it'll relax back to its cubic structure. 00:27:21.930 --> 00:27:24.080 align:middle line:84% But for thin layers, you can achieve this sort 00:27:24.080 --> 00:27:26.520 align:middle line:90% of pseudomorphic structure. 00:27:26.520 --> 00:27:28.430 align:middle line:84% You can also do kind of the opposite, 00:27:28.430 --> 00:27:30.600 align:middle line:90% or the upside-down experiment. 00:27:30.600 --> 00:27:32.480 align:middle line:84% You take a wafer of silicon germanium 00:27:32.480 --> 00:27:34.350 align:middle line:84% that's very thick, that's already relaxed, 00:27:34.350 --> 00:27:36.170 align:middle line:84% that has a larger lattice parameter 00:27:36.170 --> 00:27:40.010 align:middle line:84% and try to stick silicon on it in epitaxial growth. 00:27:40.010 --> 00:27:41.640 align:middle line:84% Again, if the silicon is thin enough, 00:27:41.640 --> 00:27:45.170 align:middle line:84% it will now stretch in X and Y to match the lattice 00:27:45.170 --> 00:27:46.400 align:middle line:90% parameter of the substrate. 00:27:46.400 --> 00:27:49.670 align:middle line:84% And it'll compress in Z. So we call this silicon layer, epi 00:27:49.670 --> 00:27:51.440 align:middle line:84% layer as being injured biaxial tension. 00:27:51.440 --> 00:27:53.210 align:middle line:84% And again, if you grow it too large. 00:27:53.210 --> 00:27:54.710 align:middle line:84% The bonds will no longer stretch. 00:27:54.710 --> 00:27:55.310 align:middle line:90% They'll break. 00:27:55.310 --> 00:27:56.640 align:middle line:90% You'll get dislocations. 00:27:56.640 --> 00:27:57.500 align:middle line:90% So it can go back. 00:27:57.500 --> 00:27:58.310 align:middle line:84% Instead of being stretched, it'll 00:27:58.310 --> 00:27:59.977 align:middle line:84% go back to its normal lattice parameter. 00:27:59.977 --> 00:28:02.110 align:middle line:90% At that point, it'd be relaxed. 00:28:02.110 --> 00:28:05.050 align:middle line:84% So this is the type of material you'll see a lot in FETs. 00:28:05.050 --> 00:28:08.170 align:middle line:84% And you see this material in bipolar transistors 00:28:08.170 --> 00:28:09.860 align:middle line:90% and some p-channel MOSFETs. 00:28:09.860 --> 00:28:12.690 align:middle line:90% 00:28:12.690 --> 00:28:15.240 align:middle line:84% So what are some of the important properties 00:28:15.240 --> 00:28:16.270 align:middle line:90% of this strain? 00:28:16.270 --> 00:28:18.450 align:middle line:84% I already mentioned to you, for MOSFETs, 00:28:18.450 --> 00:28:21.060 align:middle line:84% when you start stretching the silicon lattice parameter 00:28:21.060 --> 00:28:24.295 align:middle line:84% or compressing the parameter, you can improve the mobility. 00:28:24.295 --> 00:28:25.920 align:middle line:84% But there are other parameters that you 00:28:25.920 --> 00:28:27.720 align:middle line:84% change when you introduce strain or when 00:28:27.720 --> 00:28:29.160 align:middle line:90% you put silicon germanium in. 00:28:29.160 --> 00:28:32.760 align:middle line:84% Probably one of the most important parameters you need 00:28:32.760 --> 00:28:35.110 align:middle line:90% to know about is the band gap. 00:28:35.110 --> 00:28:38.243 align:middle line:84% So there is an energy band gap difference between silicon 00:28:38.243 --> 00:28:39.160 align:middle line:90% and silicon germanium. 00:28:39.160 --> 00:28:42.700 align:middle line:84% If this represents the band gap of relaxed silicon, 00:28:42.700 --> 00:28:45.700 align:middle line:84% so EC is the conduction band, EV is the valence band, 00:28:45.700 --> 00:28:48.070 align:middle line:84% this is the band gap of strained silicon germanium. 00:28:48.070 --> 00:28:49.420 align:middle line:90% You see it's much smaller. 00:28:49.420 --> 00:28:52.240 align:middle line:84% And most of the difference occurs in the valence band. 00:28:52.240 --> 00:28:54.520 align:middle line:84% The valence band energy is larger. 00:28:54.520 --> 00:28:56.710 align:middle line:84% So the silicon germanium has a smaller band gap. 00:28:56.710 --> 00:28:57.850 align:middle line:90% How much smaller? 00:28:57.850 --> 00:28:59.900 align:middle line:84% Well, you can look on this curve. 00:28:59.900 --> 00:29:02.350 align:middle line:84% This is the difference between the band gap of the two 00:29:02.350 --> 00:29:04.150 align:middle line:84% as a function of germanium content. 00:29:04.150 --> 00:29:06.760 align:middle line:84% So here, if I have strained silicon germanium, 00:29:06.760 --> 00:29:08.645 align:middle line:84% you should use its upper curve, not the one 00:29:08.645 --> 00:29:09.520 align:middle line:90% that's as unstrained. 00:29:09.520 --> 00:29:14.650 align:middle line:84% So a 20%, the band gap shrinks by about 150 millivolts roughly. 00:29:14.650 --> 00:29:16.720 align:middle line:84% So and you can read that right off the curve. 00:29:16.720 --> 00:29:19.630 align:middle line:84% Here at 50%, it's another number, maybe 350. 00:29:19.630 --> 00:29:22.240 align:middle line:84% So you have an idea of how it moves. 00:29:22.240 --> 00:29:25.030 align:middle line:84% All these data points are our data from the literature. 00:29:25.030 --> 00:29:28.030 align:middle line:84% There are people extracted the band gap by various devices, 00:29:28.030 --> 00:29:30.130 align:middle line:84% diodes, bipolar transistors, photodiodes. 00:29:30.130 --> 00:29:32.740 align:middle line:84% And the solid line is some theory that 00:29:32.740 --> 00:29:34.190 align:middle line:90% was published a long time ago. 00:29:34.190 --> 00:29:37.270 align:middle line:84% So now, the other thing, that's if it's strained. 00:29:37.270 --> 00:29:40.460 align:middle line:84% If it's unstrained, if you grow the silicon germanium 00:29:40.460 --> 00:29:44.020 align:middle line:84% so thick that it relaxes back, so it comes back 00:29:44.020 --> 00:29:47.530 align:middle line:84% to this cubic structure, then what happens to the band gap? 00:29:47.530 --> 00:29:50.500 align:middle line:84% Well, in fact, the band gap difference is a lot smaller. 00:29:50.500 --> 00:29:53.890 align:middle line:84% When it's unstrained, the band gap difference say a 20% 00:29:53.890 --> 00:29:55.820 align:middle line:90% is only about 100 millivolts. 00:29:55.820 --> 00:29:58.802 align:middle line:84% So a big fraction of this band gap offset, 00:29:58.802 --> 00:30:00.760 align:middle line:84% this band gap difference, is due to the strain. 00:30:00.760 --> 00:30:03.640 align:middle line:84% It's due to the lattice parameter change. 00:30:03.640 --> 00:30:05.200 align:middle line:84% A certain fraction is due to the fact 00:30:05.200 --> 00:30:07.520 align:middle line:84% that we're just adding germanium. 00:30:07.520 --> 00:30:10.000 align:middle line:84% So you need to know whether the material is 00:30:10.000 --> 00:30:12.610 align:middle line:84% unstrained or strained or somewhere in between in order 00:30:12.610 --> 00:30:17.050 align:middle line:84% to figure out what the band structure looks like. 00:30:17.050 --> 00:30:18.473 align:middle line:84% And in fact, on slide 10, I don't 00:30:18.473 --> 00:30:20.140 align:middle line:84% have to go through this in great detail. 00:30:20.140 --> 00:30:22.640 align:middle line:84% If you haven't had energy band theory, it's a little tricky. 00:30:22.640 --> 00:30:25.450 align:middle line:84% But slide 10 gives you an idea of how 00:30:25.450 --> 00:30:29.410 align:middle line:84% the strain does affect the energy bands and the lineup. 00:30:29.410 --> 00:30:32.980 align:middle line:84% And in fact, if we look at the case for strained silicon 00:30:32.980 --> 00:30:36.195 align:middle line:84% germanium on silicon, so it'll be on-- remember, 00:30:36.195 --> 00:30:38.320 align:middle line:84% we said when we grow strain, so if the germanium is 00:30:38.320 --> 00:30:42.340 align:middle line:84% under biaxial compression in the X and Y planes being compressed, 00:30:42.340 --> 00:30:44.170 align:middle line:84% strain the silicon germanium on silicon, 00:30:44.170 --> 00:30:46.280 align:middle line:84% the band lineup looks like this, as I mentioned. 00:30:46.280 --> 00:30:49.230 align:middle line:84% Most of the band offset is in the valence band. 00:30:49.230 --> 00:30:51.620 align:middle line:84% So the holes will tend to be confined in the silicon 00:30:51.620 --> 00:30:53.150 align:middle line:90% germanium. 00:30:53.150 --> 00:30:55.220 align:middle line:90% There's very little band offset. 00:30:55.220 --> 00:30:57.350 align:middle line:84% The lowest energy conduction band 00:30:57.350 --> 00:30:59.270 align:middle line:84% is about aligned in strained silicon 00:30:59.270 --> 00:31:01.820 align:middle line:90% germanium versus cubic silicon. 00:31:01.820 --> 00:31:04.710 align:middle line:84% So there's not much opportunity for confining electrons. 00:31:04.710 --> 00:31:06.620 align:middle line:84% So this material, it turns out, is 00:31:06.620 --> 00:31:09.447 align:middle line:84% ideal for a heterojunction bipolar transistor for an NPN. 00:31:09.447 --> 00:31:11.780 align:middle line:84% This material has been in production for about 10 years. 00:31:11.780 --> 00:31:16.450 align:middle line:84% People use it in high speed RF devices. 00:31:16.450 --> 00:31:18.810 align:middle line:84% So that's kind of a more well-characterized and 00:31:18.810 --> 00:31:21.150 align:middle line:84% well-known structure where the germanium 00:31:21.150 --> 00:31:24.540 align:middle line:84% fraction in a typical HBT you would buy off the shelf right 00:31:24.540 --> 00:31:28.470 align:middle line:84% now is probably about between 10% and 20% of the atoms 00:31:28.470 --> 00:31:31.650 align:middle line:90% would be germanium in an HBT. 00:31:31.650 --> 00:31:34.180 align:middle line:84% On the other hand, if you grow the opposite structure, 00:31:34.180 --> 00:31:36.990 align:middle line:84% so you have relaxed silicon germanium substrate and you grow 00:31:36.990 --> 00:31:38.100 align:middle line:84% a thin layer of silicon-- remember, 00:31:38.100 --> 00:31:39.540 align:middle line:84% we said this was biaxial tension-- 00:31:39.540 --> 00:31:41.500 align:middle line:84% the band line up is quite different. 00:31:41.500 --> 00:31:45.120 align:middle line:84% In fact, you do get a sizable difference in the conduction 00:31:45.120 --> 00:31:47.590 align:middle line:90% band position. 00:31:47.590 --> 00:31:49.960 align:middle line:84% And in fact, the conduction band is lower. 00:31:49.960 --> 00:31:51.210 align:middle line:90% The energy is lower in strain. 00:31:51.210 --> 00:31:54.700 align:middle line:84% So the electrons will tend to go into the strained silicon layer. 00:31:54.700 --> 00:31:57.275 align:middle line:84% So if you make a FET, and you make this as your channel, 00:31:57.275 --> 00:31:59.400 align:middle line:84% the electrons want to go into the strained silicon. 00:31:59.400 --> 00:32:02.657 align:middle line:84% The holes want to be in the silicon germanium side. 00:32:02.657 --> 00:32:04.740 align:middle line:84% So this is what's called a type II band alignment. 00:32:04.740 --> 00:32:06.990 align:middle line:84% So you can make lots of interesting devices 00:32:06.990 --> 00:32:09.870 align:middle line:84% just by taking silicon and straining it 00:32:09.870 --> 00:32:11.430 align:middle line:90% on unrelaxed silicon germanium. 00:32:11.430 --> 00:32:14.090 align:middle line:90% 00:32:14.090 --> 00:32:17.025 align:middle line:84% The other thing that happens here that's important, 00:32:17.025 --> 00:32:18.400 align:middle line:84% and we'll talk about when we talk 00:32:18.400 --> 00:32:20.710 align:middle line:84% about the mobility enhancement, you notice what's happened 00:32:20.710 --> 00:32:23.085 align:middle line:84% is the conduction band is actually split into two levels. 00:32:23.085 --> 00:32:25.690 align:middle line:84% Originally, there was only a single energy level. 00:32:25.690 --> 00:32:28.240 align:middle line:84% And that splitting is what's really responsible, 00:32:28.240 --> 00:32:32.042 align:middle line:84% to a large extent, for the mobility improvement. 00:32:32.042 --> 00:32:33.750 align:middle line:84% If you really wanted to get nitty gritty, 00:32:33.750 --> 00:32:34.667 align:middle line:90% I won't go through it. 00:32:34.667 --> 00:32:38.220 align:middle line:84% But here on page 11, I've made a reference to a famous paper 00:32:38.220 --> 00:32:41.250 align:middle line:84% by Chris Van de Walle and Richard Martin published 00:32:41.250 --> 00:32:42.780 align:middle line:90% in the mid 1980s. 00:32:42.780 --> 00:32:45.810 align:middle line:84% Van de Walle and Martin actually theoretically did 00:32:45.810 --> 00:32:50.250 align:middle line:84% a pretty good job of trying to calculate what the energy band 00:32:50.250 --> 00:32:52.920 align:middle line:84% structure looks like for strained silicon 00:32:52.920 --> 00:32:55.500 align:middle line:84% germanium on cubic silicon, which we're showing over 00:32:55.500 --> 00:32:58.830 align:middle line:84% here and also for strained silicon germanium 00:32:58.830 --> 00:33:00.200 align:middle line:90% on cubic germanium. 00:33:00.200 --> 00:33:01.200 align:middle line:90% They have both of these. 00:33:01.200 --> 00:33:03.970 align:middle line:84% And I won't go through it in any great detail. 00:33:03.970 --> 00:33:06.270 align:middle line:84% But if you're interested in the solid state physics 00:33:06.270 --> 00:33:07.930 align:middle line:84% and you end up working in this area, 00:33:07.930 --> 00:33:10.330 align:middle line:84% this is a good paper for you to look at. 00:33:10.330 --> 00:33:13.560 align:middle line:84% Again, the band gap would be the distance 00:33:13.560 --> 00:33:15.540 align:middle line:84% between this point here in the conduction band 00:33:15.540 --> 00:33:17.260 align:middle line:84% and a point here in the valence band. 00:33:17.260 --> 00:33:19.677 align:middle line:84% So they actually calculated how the valence and conduction 00:33:19.677 --> 00:33:23.445 align:middle line:84% bands move as you add strain and germanium to the system. 00:33:23.445 --> 00:33:26.040 align:middle line:90% 00:33:26.040 --> 00:33:31.083 align:middle line:84% On slide 12, I'm also giving you a reference for not only has 00:33:31.083 --> 00:33:33.000 align:middle line:84% the band gap changes, but people have actually 00:33:33.000 --> 00:33:36.330 align:middle line:84% made some reference some measurements of the energy band 00:33:36.330 --> 00:33:39.180 align:middle line:84% offsets when you have strained silicon on relaxed silicon 00:33:39.180 --> 00:33:41.650 align:middle line:90% germanium say for a MOSFET. 00:33:41.650 --> 00:33:44.540 align:middle line:84% This is taken out of Jeff Welzer's thesis from 1994. 00:33:44.540 --> 00:33:46.960 align:middle line:84% And what you can see here, this is the basic structure. 00:33:46.960 --> 00:33:48.580 align:middle line:90% You have an oxide. 00:33:48.580 --> 00:33:50.530 align:middle line:84% You have a thin layer of strained silicon. 00:33:50.530 --> 00:33:52.180 align:middle line:84% And you have these energy offsets here. 00:33:52.180 --> 00:33:54.400 align:middle line:84% The energy conduction band energy is a little lower. 00:33:54.400 --> 00:33:56.600 align:middle line:84% And the valence band energy is slightly 00:33:56.600 --> 00:33:57.850 align:middle line:90% different in strained silicon. 00:33:57.850 --> 00:34:00.520 align:middle line:84% And these numbers, delta, you see how different they are, 00:34:00.520 --> 00:34:02.380 align:middle line:84% these energy band discontinuities 00:34:02.380 --> 00:34:04.450 align:middle line:84% were actually measured and delta EV here 00:34:04.450 --> 00:34:08.264 align:middle line:84% as a function of the germanium fraction in the substrate. 00:34:08.264 --> 00:34:09.639 align:middle line:84% The bullets are the measurements. 00:34:09.639 --> 00:34:12.370 align:middle line:84% And the lines are the calculations 00:34:12.370 --> 00:34:13.780 align:middle line:90% from Van de Wall and Martin. 00:34:13.780 --> 00:34:15.530 align:middle line:84% And indeed, you get pretty good agreement. 00:34:15.530 --> 00:34:18.040 align:middle line:84% So we have a rough idea now of what these energy band 00:34:18.040 --> 00:34:20.155 align:middle line:84% structures look like when I add strain 00:34:20.155 --> 00:34:24.650 align:middle line:84% or when I add silicon germanium to the substrate. 00:34:24.650 --> 00:34:27.325 align:middle line:84% So that's important for figuring out how all these devices work. 00:34:27.325 --> 00:34:29.980 align:middle line:90% 00:34:29.980 --> 00:34:32.530 align:middle line:84% I mentioned that when you squeeze silicone, 00:34:32.530 --> 00:34:36.070 align:middle line:84% either you stretch it in a biaxial sense, 00:34:36.070 --> 00:34:37.330 align:middle line:90% you can improve the mobility. 00:34:37.330 --> 00:34:39.262 align:middle line:84% And the question is, well, why is that? 00:34:39.262 --> 00:34:41.679 align:middle line:84% I don't want to go into detail in the solid state physics. 00:34:41.679 --> 00:34:44.860 align:middle line:84% But it turns out, if you put silicon in biaxial tension, 00:34:44.860 --> 00:34:48.310 align:middle line:84% you break the cubic symmetry of the lattice. 00:34:48.310 --> 00:34:51.550 align:middle line:84% Ordinarily, in the lattice, the XYZ directions 00:34:51.550 --> 00:34:54.550 align:middle line:84% are all completely symmetric when silicon is not strained. 00:34:54.550 --> 00:34:57.640 align:middle line:84% When it strained, they actually become asymmetric. 00:34:57.640 --> 00:35:00.850 align:middle line:84% X and Y lattice parameter is larger than in the Z. 00:35:00.850 --> 00:35:02.770 align:middle line:84% And what happens by that breaking 00:35:02.770 --> 00:35:09.370 align:middle line:84% of the physical symmetry is that the sixth energy of ellipsoids 00:35:09.370 --> 00:35:12.190 align:middle line:84% in constant energy space that are all at the same energy 00:35:12.190 --> 00:35:15.050 align:middle line:84% actually split up when you introduce a strain. 00:35:15.050 --> 00:35:17.420 align:middle line:84% And two of them, these two ellipsoids here 00:35:17.420 --> 00:35:20.840 align:middle line:84% where the electrons can be, these two red ones, 00:35:20.840 --> 00:35:22.340 align:middle line:84% the so-called perpendicular valleys 00:35:22.340 --> 00:35:25.490 align:middle line:84% labeled Del 2, they actually end up being at a lower energy 00:35:25.490 --> 00:35:27.080 align:middle line:90% than these other four. 00:35:27.080 --> 00:35:31.460 align:middle line:84% So we get this strain-induced splitting in bulk silicon. 00:35:31.460 --> 00:35:34.548 align:middle line:84% All six of these valleys that represent 00:35:34.548 --> 00:35:36.590 align:middle line:84% states that electrons can occupy, all six of them 00:35:36.590 --> 00:35:37.730 align:middle line:90% are at the same energy level. 00:35:37.730 --> 00:35:39.730 align:middle line:84% But by group theory, when we break the symmetry, 00:35:39.730 --> 00:35:44.252 align:middle line:84% we break that up into a two-fold degenerate level, which is lower 00:35:44.252 --> 00:35:45.960 align:middle line:84% energy, and a four-fold degenerate level. 00:35:45.960 --> 00:35:49.220 align:middle line:84% So by breaking this up, actually, what happens is now, 00:35:49.220 --> 00:35:52.250 align:middle line:84% most of the electrons want to be in these red valleys. 00:35:52.250 --> 00:35:54.470 align:middle line:84% And the scattering between the red and the green 00:35:54.470 --> 00:35:56.460 align:middle line:84% is suppressed because it costs you energy. 00:35:56.460 --> 00:35:58.820 align:middle line:84% So instead of scattering between all six, 00:35:58.820 --> 00:36:01.010 align:middle line:84% you can only scatter very effectively 00:36:01.010 --> 00:36:02.310 align:middle line:90% between these two red ones. 00:36:02.310 --> 00:36:04.622 align:middle line:84% And so the scattering time goes up. 00:36:04.622 --> 00:36:07.080 align:middle line:84% And the mobility is directly proportional to the scattering 00:36:07.080 --> 00:36:08.070 align:middle line:90% time. 00:36:08.070 --> 00:36:11.355 align:middle line:84% The in-plane effective mass also goes down. 00:36:11.355 --> 00:36:13.230 align:middle line:84% So the mobility also goes up for that reason. 00:36:13.230 --> 00:36:15.330 align:middle line:84% So just by breaking the symmetry, 00:36:15.330 --> 00:36:18.630 align:middle line:84% we can change the occupation of the energy bands in silicon. 00:36:18.630 --> 00:36:21.783 align:middle line:84% And we can make the mobility higher for electrons. 00:36:21.783 --> 00:36:23.200 align:middle line:84% In the valence band, I'm not going 00:36:23.200 --> 00:36:24.450 align:middle line:90% to go through in great detail. 00:36:24.450 --> 00:36:26.070 align:middle line:84% But it's quite a bit more complicated. 00:36:26.070 --> 00:36:29.100 align:middle line:84% But it turns out, with strain, we also split the degeneracy 00:36:29.100 --> 00:36:30.310 align:middle line:90% in the valence band. 00:36:30.310 --> 00:36:32.130 align:middle line:84% We also suppress interband scattering. 00:36:32.130 --> 00:36:33.210 align:middle line:90% It is a lot more tricky. 00:36:33.210 --> 00:36:39.060 align:middle line:84% But it is possible to also induce improved hole mobility 00:36:39.060 --> 00:36:43.860 align:middle line:84% or PMOS mobility by introducing strain. 00:36:43.860 --> 00:36:45.542 align:middle line:84% And in fact, here on slide 14, I'll 00:36:45.542 --> 00:36:47.250 align:middle line:84% just walk through some of the older data. 00:36:47.250 --> 00:36:49.740 align:middle line:90% This was from 1992 by Welzer. 00:36:49.740 --> 00:36:51.510 align:middle line:84% These were the first strain of MOSFETs 00:36:51.510 --> 00:36:54.090 align:middle line:84% that were fabricated where he made 00:36:54.090 --> 00:36:56.920 align:middle line:84% a thin layer of strained silicon on relaxed silicon germanium 00:36:56.920 --> 00:36:59.640 align:middle line:84% and then just implanted it and made a silicon 00:36:59.640 --> 00:37:01.500 align:middle line:90% MOSFET with a source and drain. 00:37:01.500 --> 00:37:02.250 align:middle line:90% It was an nMOSFET. 00:37:02.250 --> 00:37:04.560 align:middle line:84% And this, he called the surface channel 00:37:04.560 --> 00:37:07.475 align:middle line:84% MOSFET because the bands aligned so that the electrons wanted 00:37:07.475 --> 00:37:08.350 align:middle line:90% to be at the surface. 00:37:08.350 --> 00:37:10.320 align:middle line:84% So the channel is at the surface. 00:37:10.320 --> 00:37:11.700 align:middle line:84% This he called the buried channel 00:37:11.700 --> 00:37:15.030 align:middle line:84% MOSFET because if you look at the energy band diagram here, 00:37:15.030 --> 00:37:17.160 align:middle line:84% there's a little bit of silicon germanium on top. 00:37:17.160 --> 00:37:20.070 align:middle line:84% And that confines the electrons to be a little bit buried 00:37:20.070 --> 00:37:21.400 align:middle line:90% below the surface. 00:37:21.400 --> 00:37:25.520 align:middle line:84% It's like a MOSFET but where the channel is slightly buried. 00:37:25.520 --> 00:37:27.808 align:middle line:84% And on slide 15, he actually measured 00:37:27.808 --> 00:37:28.850 align:middle line:90% some of those mobilities. 00:37:28.850 --> 00:37:30.558 align:middle line:84% These were the first mobilities measured. 00:37:30.558 --> 00:37:32.420 align:middle line:84% And if you look at this solid line, 00:37:32.420 --> 00:37:36.140 align:middle line:84% this is the mobility measured in a MOSFET that had no strain. 00:37:36.140 --> 00:37:38.870 align:middle line:84% And for a surface channel MOSFET, the mobility 00:37:38.870 --> 00:37:41.270 align:middle line:84% that he measured was about, oh, I 00:37:41.270 --> 00:37:44.510 align:middle line:84% don't know, 80% higher, about a factor of 1.8 improvement 00:37:44.510 --> 00:37:45.560 align:middle line:90% in the surface channel. 00:37:45.560 --> 00:37:48.110 align:middle line:84% So right off the bat, as a function of vertical field, 00:37:48.110 --> 00:37:49.880 align:middle line:90% you see we have parameter. 00:37:49.880 --> 00:37:52.945 align:middle line:84% By changing the strain, you can up the mobility by about 80%. 00:37:52.945 --> 00:37:54.320 align:middle line:84% If you do the buried channel, you 00:37:54.320 --> 00:37:57.530 align:middle line:84% can get even higher mobilities at low gate bias. 00:37:57.530 --> 00:37:59.780 align:middle line:84% But the mobility seems to go down 00:37:59.780 --> 00:38:02.760 align:middle line:84% at high gate bias for various reasons. 00:38:02.760 --> 00:38:04.970 align:middle line:84% So kind of the first evidence that strain 00:38:04.970 --> 00:38:07.113 align:middle line:90% makes a big difference. 00:38:07.113 --> 00:38:08.780 align:middle line:84% These are the actual structures that one 00:38:08.780 --> 00:38:12.170 align:middle line:90% can grow shown on slide 16. 00:38:12.170 --> 00:38:15.540 align:middle line:84% Again to make an nMOSFET, you can grow relaxed silicate 00:38:15.540 --> 00:38:17.540 align:middle line:84% germanium with a thin layer of strained silicon. 00:38:17.540 --> 00:38:19.970 align:middle line:84% And this becomes a high electron mobility channel. 00:38:19.970 --> 00:38:21.350 align:middle line:84% This slightly different structure 00:38:21.350 --> 00:38:23.930 align:middle line:84% is needed for the PMOS, very similar 00:38:23.930 --> 00:38:26.480 align:middle line:84% but with slight variations, so you 00:38:26.480 --> 00:38:31.890 align:middle line:84% can make CMOS devices using this type of technique. 00:38:31.890 --> 00:38:36.640 align:middle line:84% And I'll skip through the energy band diagrams for now. 00:38:36.640 --> 00:38:38.295 align:middle line:90% Slide 17 shows some metrology. 00:38:38.295 --> 00:38:39.920 align:middle line:84% The question you may be thinking, well, 00:38:39.920 --> 00:38:41.200 align:middle line:90% how do you really know? 00:38:41.200 --> 00:38:42.575 align:middle line:84% All right, your mobility went up. 00:38:42.575 --> 00:38:44.920 align:middle line:84% But how do you really know that the layers are strained? 00:38:44.920 --> 00:38:46.420 align:middle line:84% Especially when they're so thin, how 00:38:46.420 --> 00:38:49.210 align:middle line:84% do you measure strain in a layer that's only 100 angstroms thick? 00:38:49.210 --> 00:38:51.340 align:middle line:84% You really can't use X-ray diffraction. 00:38:51.340 --> 00:38:53.620 align:middle line:84% X-ray diffraction on 100 angstrom layer 00:38:53.620 --> 00:38:55.870 align:middle line:84% could be really tough because the layer is very thin. 00:38:55.870 --> 00:38:59.230 align:middle line:84% So how do you measure that change in the lattice parameter? 00:38:59.230 --> 00:39:01.210 align:middle line:84% Well, it turns out there's a technique called 00:39:01.210 --> 00:39:04.150 align:middle line:84% Raman scattering, which is an optical interaction. 00:39:04.150 --> 00:39:05.770 align:middle line:84% There are certain vibrational modes 00:39:05.770 --> 00:39:10.540 align:middle line:84% of a lattice that will scatter a scatter of photons. 00:39:10.540 --> 00:39:13.750 align:middle line:84% And the Raman effect can be used to measure 00:39:13.750 --> 00:39:16.480 align:middle line:84% the strain or the change in the lattice parameter, even 00:39:16.480 --> 00:39:18.020 align:middle line:90% a very, very thin layer. 00:39:18.020 --> 00:39:22.850 align:middle line:84% So here's an example of a MOSFET that was made, a Raman spectrum. 00:39:22.850 --> 00:39:25.840 align:middle line:84% So you're measuring the intensity of the scattered light 00:39:25.840 --> 00:39:30.280 align:middle line:84% as a function of the frequency in inverse centimeters. 00:39:30.280 --> 00:39:34.540 align:middle line:84% And you see a large peak here associated 00:39:34.540 --> 00:39:37.010 align:middle line:90% with relaxed silicon germanium. 00:39:37.010 --> 00:39:39.440 align:middle line:84% And there's a small peak here, which 00:39:39.440 --> 00:39:41.780 align:middle line:84% is the thin strained silicon layer. 00:39:41.780 --> 00:39:44.540 align:middle line:84% Now, if that layer were unstrained, 00:39:44.540 --> 00:39:47.000 align:middle line:84% bulk silicon, if you just take a wafer of bulk silicon 00:39:47.000 --> 00:39:48.560 align:middle line:84% and subject to Raman scattering, you 00:39:48.560 --> 00:39:51.450 align:middle line:84% get a line at 521 inverse centimeters. 00:39:51.450 --> 00:39:54.950 align:middle line:84% So it turns out the shift of the strained silicon 00:39:54.950 --> 00:39:59.030 align:middle line:84% peak, the shift between that peak and 521 is a measure 00:39:59.030 --> 00:40:00.540 align:middle line:90% of the amount of strain. 00:40:00.540 --> 00:40:04.130 align:middle line:84% So this is a very nice way you can oxidize a wafer, 00:40:04.130 --> 00:40:06.032 align:middle line:84% strip the oxide of strained silicon. 00:40:06.032 --> 00:40:07.740 align:middle line:84% You can see if the strain is still there. 00:40:07.740 --> 00:40:11.630 align:middle line:84% In fact, you can figure out the relaxed silicon germanium 00:40:11.630 --> 00:40:14.070 align:middle line:90% content as well using this. 00:40:14.070 --> 00:40:17.160 align:middle line:84% So slide 18 just shows you some measurements 00:40:17.160 --> 00:40:22.510 align:middle line:84% that people have made, the Raman peak shift and the frequency 00:40:22.510 --> 00:40:24.670 align:middle line:84% shift in inverse centimeter as a function 00:40:24.670 --> 00:40:27.280 align:middle line:84% of the amount of germanium fraction of the substrate. 00:40:27.280 --> 00:40:28.780 align:middle line:84% For strained silicon layers, you can 00:40:28.780 --> 00:40:31.780 align:middle line:84% see there's a theoretical line here. 00:40:31.780 --> 00:40:35.080 align:middle line:84% And the bullets represent measured points. 00:40:35.080 --> 00:40:40.180 align:middle line:84% So using Raman, if you have the right kind of Raman setup, 00:40:40.180 --> 00:40:42.080 align:middle line:84% you can actually get a peak shift. 00:40:42.080 --> 00:40:44.890 align:middle line:84% So for 20% germanium here, we expect a Raman peak shift 00:40:44.890 --> 00:40:49.310 align:middle line:84% of about 6 to 7 inverse centimeters. 00:40:49.310 --> 00:40:52.630 align:middle line:84% So that gives you an idea of how you can measure 00:40:52.630 --> 00:40:56.250 align:middle line:90% the strain in the material. 00:40:56.250 --> 00:40:59.070 align:middle line:84% Slide 19 shows you a material that's been oxidized. 00:40:59.070 --> 00:41:01.020 align:middle line:90% You might be thinking, OK, fine. 00:41:01.020 --> 00:41:02.850 align:middle line:84% You can grow these epitaxial layers. 00:41:02.850 --> 00:41:03.870 align:middle line:90% They'll be strained. 00:41:03.870 --> 00:41:06.240 align:middle line:84% What happens if you oxidize the silicon? 00:41:06.240 --> 00:41:08.970 align:middle line:84% Is it going to somehow relax the strain because of that oxidation 00:41:08.970 --> 00:41:11.740 align:middle line:84% process and the breaking the bonds and the oxidation? 00:41:11.740 --> 00:41:13.020 align:middle line:90% Well, it turns out it doesn't. 00:41:13.020 --> 00:41:16.230 align:middle line:84% In fact, the as-grown Raman spectrum 00:41:16.230 --> 00:41:20.160 align:middle line:84% here, this is for the epitaxial layer as it was first grown, 00:41:20.160 --> 00:41:24.060 align:middle line:84% you do get a strained silicon peak right here near to 510. 00:41:24.060 --> 00:41:26.910 align:middle line:84% If you oxidize it at 850 for 10 minutes, 00:41:26.910 --> 00:41:28.860 align:middle line:84% you still get that strained silicon peak in. 00:41:28.860 --> 00:41:31.097 align:middle line:84% Although, the peak is smaller height because, 00:41:31.097 --> 00:41:32.430 align:middle line:90% of course, the layer is thinner. 00:41:32.430 --> 00:41:34.830 align:middle line:84% When you oxidize, you consume silicon. 00:41:34.830 --> 00:41:37.095 align:middle line:84% But the peak position, it remains at 510. 00:41:37.095 --> 00:41:38.580 align:middle line:90% It remains at the same point. 00:41:38.580 --> 00:41:40.943 align:middle line:84% And this has been verified very many times. 00:41:40.943 --> 00:41:42.360 align:middle line:84% And, of course, in the MOSFET, you 00:41:42.360 --> 00:41:43.568 align:middle line:90% do see the enhanced mobility. 00:41:43.568 --> 00:41:46.500 align:middle line:84% So there doesn't seem to be any significant strain relaxation, 00:41:46.500 --> 00:41:49.380 align:middle line:84% at least when you oxidize in these type of temperatures. 00:41:49.380 --> 00:41:51.722 align:middle line:84% If you were to go up to 1,000 or more, 00:41:51.722 --> 00:41:53.430 align:middle line:84% then you could start running into trouble 00:41:53.430 --> 00:41:55.635 align:middle line:84% when the germanium starts diffusing around. 00:41:55.635 --> 00:41:58.480 align:middle line:90% 00:41:58.480 --> 00:42:01.323 align:middle line:84% So on slide 20, I'm just reviewing for you, 00:42:01.323 --> 00:42:03.490 align:middle line:84% giving you an idea of some of the early measurements 00:42:03.490 --> 00:42:05.740 align:middle line:84% people made of how the mobility was enhanced. 00:42:05.740 --> 00:42:08.500 align:middle line:84% And the left side is for nMOSFETs. 00:42:08.500 --> 00:42:11.650 align:middle line:84% And you see this black line, this dashed line 00:42:11.650 --> 00:42:13.240 align:middle line:90% is for control devices. 00:42:13.240 --> 00:42:17.200 align:middle line:84% And these other lines on top are for 10% germanium 00:42:17.200 --> 00:42:18.700 align:middle line:90% in the substrate, 20 and 30. 00:42:18.700 --> 00:42:20.530 align:middle line:84% You see, as you add more germanium 00:42:20.530 --> 00:42:23.110 align:middle line:84% in the substrate, the whole of the electron mobility 00:42:23.110 --> 00:42:24.460 align:middle line:90% curve all moves up. 00:42:24.460 --> 00:42:28.030 align:middle line:90% And you get a nice enhancement. 00:42:28.030 --> 00:42:32.650 align:middle line:84% On holes, for PMOS vets, there was some enhancement 00:42:32.650 --> 00:42:34.870 align:middle line:84% measured, although the enhancement is diminished. 00:42:34.870 --> 00:42:36.850 align:middle line:84% You see the enhancement factor is actually 00:42:36.850 --> 00:42:39.340 align:middle line:84% going down as you go to higher vertical fields here. 00:42:39.340 --> 00:42:43.890 align:middle line:84% So at higher vertical fields is where a lot of modern MOSFETs 00:42:43.890 --> 00:42:44.390 align:middle line:90% operate. 00:42:44.390 --> 00:42:47.350 align:middle line:84% So that's a bit of an issue for pMOSFETs 00:42:47.350 --> 00:42:50.080 align:middle line:90% for using this technique. 00:42:50.080 --> 00:42:51.760 align:middle line:90% Those are research devices. 00:42:51.760 --> 00:42:55.660 align:middle line:84% This on slide 21 is some more recent publications 00:42:55.660 --> 00:42:57.592 align:middle line:84% in a manufacturing style process. 00:42:57.592 --> 00:42:59.050 align:middle line:84% Although it's not in manufacturing, 00:42:59.050 --> 00:43:02.350 align:middle line:84% the process was a full manufacturing style process 00:43:02.350 --> 00:43:03.280 align:middle line:90% that IBM did. 00:43:03.280 --> 00:43:06.250 align:middle line:84% It had all the attributes of a real manufactured device. 00:43:06.250 --> 00:43:08.170 align:middle line:90% It had shallow trench isolation. 00:43:08.170 --> 00:43:12.340 align:middle line:84% Had it had weld implants, halo implants, 00:43:12.340 --> 00:43:15.520 align:middle line:84% raised source drains by EPI, and a thin gate oxide. 00:43:15.520 --> 00:43:19.030 align:middle line:84% So this is a real device as opposed to the university style. 00:43:19.030 --> 00:43:20.740 align:middle line:84% And again, you can see the mobility 00:43:20.740 --> 00:43:24.550 align:middle line:84% for electrons goes up with strain, about 100% improvement. 00:43:24.550 --> 00:43:25.910 align:middle line:90% That is a factor of 2. 00:43:25.910 --> 00:43:28.970 align:middle line:84% Holes, there is an improvement at low gate bias. 00:43:28.970 --> 00:43:31.550 align:middle line:84% But here at high effective field or high gate bias, 00:43:31.550 --> 00:43:35.360 align:middle line:84% the improvement is pretty small, only a few percent. 00:43:35.360 --> 00:43:37.760 align:middle line:84% That just shows the state of the art 00:43:37.760 --> 00:43:41.880 align:middle line:84% as it exists a couple of years ago. 00:43:41.880 --> 00:43:44.450 align:middle line:84% And if you were working at relatively low gate bias, 00:43:44.450 --> 00:43:47.240 align:middle line:84% and you just want to get an idea of how the mobility enhancement 00:43:47.240 --> 00:43:48.650 align:middle line:84% varies with the amount of strain, 00:43:48.650 --> 00:43:51.260 align:middle line:84% you can look at this curve on page 22. 00:43:51.260 --> 00:43:54.330 align:middle line:84% This is the electron mobility enhancement factor. 00:43:54.330 --> 00:43:56.720 align:middle line:84% So it's the peak mobility in a strained silicon 00:43:56.720 --> 00:43:59.240 align:middle line:84% FET divided the peak mobility in an unstrained FET. 00:43:59.240 --> 00:44:01.940 align:middle line:84% And you see here with no strain, that it's 1. 00:44:01.940 --> 00:44:04.700 align:middle line:84% And then it goes up to about a factor of 1.8, 00:44:04.700 --> 00:44:05.908 align:middle line:90% maybe 2, something like that. 00:44:05.908 --> 00:44:07.617 align:middle line:84% There's a lot of data have been published 00:44:07.617 --> 00:44:08.990 align:middle line:90% in the last 10 or 12 years. 00:44:08.990 --> 00:44:12.790 align:middle line:84% And they all seem to agree about this saturating behavior. 00:44:12.790 --> 00:44:15.160 align:middle line:84% On the right is for loss for hole mobility. 00:44:15.160 --> 00:44:17.890 align:middle line:84% The data does show an increase, especially as you 00:44:17.890 --> 00:44:20.020 align:middle line:90% get above about 30% germanium. 00:44:20.020 --> 00:44:22.450 align:middle line:84% You can get larger enhancement factors. 00:44:22.450 --> 00:44:24.108 align:middle line:84% And this is the theoretical curve. 00:44:24.108 --> 00:44:26.650 align:middle line:84% The data seems to be a little bit below the theoretical curve 00:44:26.650 --> 00:44:29.710 align:middle line:84% but shows roughly the kind of strain dependence 00:44:29.710 --> 00:44:30.950 align:middle line:90% you would expect. 00:44:30.950 --> 00:44:33.435 align:middle line:84% And so what this is telling you, if you're saturating out, 00:44:33.435 --> 00:44:34.810 align:middle line:84% you don't really need to put more 00:44:34.810 --> 00:44:37.450 align:middle line:84% for electrons, more than about 20% germanium in the substrate. 00:44:37.450 --> 00:44:38.600 align:middle line:90% That's enough strain. 00:44:38.600 --> 00:44:40.660 align:middle line:84% So the bands are split enough, if you go beyond 00:44:40.660 --> 00:44:43.355 align:middle line:84% that, you're inducing more strain and more splitting. 00:44:43.355 --> 00:44:44.980 align:middle line:84% But the mobility itself is not actually 00:44:44.980 --> 00:44:47.813 align:middle line:84% improving because something else is limiting the mobility. 00:44:47.813 --> 00:44:50.230 align:middle line:84% But it gives you a rough idea of the amount of strain need 00:44:50.230 --> 00:44:54.540 align:middle line:84% to induce in some of these devices. 00:44:54.540 --> 00:44:59.250 align:middle line:84% Page 23 is a summary of some of the mobility enhancements now 00:44:59.250 --> 00:45:01.590 align:middle line:90% for pMOSFETs. 00:45:01.590 --> 00:45:03.312 align:middle line:84% It's the mobility enhancement ratio. 00:45:03.312 --> 00:45:04.770 align:middle line:84% Again, this is the number you would 00:45:04.770 --> 00:45:07.230 align:middle line:84% like to be large as a function of vertical field. 00:45:07.230 --> 00:45:09.510 align:middle line:84% If you operate the device at very low vertical fields, 00:45:09.510 --> 00:45:11.100 align:middle line:90% you can get large ratios. 00:45:11.100 --> 00:45:13.510 align:middle line:84% But most modern devices operate over here. 00:45:13.510 --> 00:45:16.770 align:middle line:84% So the enhancement factor is only about 1.1 or 1.2. 00:45:16.770 --> 00:45:18.600 align:middle line:90% It's approaching 1. 00:45:18.600 --> 00:45:22.200 align:middle line:84% That's a problem, unless you go to very high germanium contents. 00:45:22.200 --> 00:45:25.890 align:middle line:84% There's some data here with a 40% substrate. 00:45:25.890 --> 00:45:28.410 align:middle line:84% So that means the strained silicon layer is very thin, 00:45:28.410 --> 00:45:30.270 align:middle line:84% and it's strained to a 40% substrate. 00:45:30.270 --> 00:45:32.080 align:middle line:84% Here, in mobility enhancement factor, 00:45:32.080 --> 00:45:33.570 align:middle line:84% there's not a full curve available. 00:45:33.570 --> 00:45:37.710 align:middle line:84% The highest field is people went to was about 0.6. 00:45:37.710 --> 00:45:40.060 align:middle line:84% But they did get pretty large enhancement factors. 00:45:40.060 --> 00:45:43.980 align:middle line:84% So it looks like if you're going to use biaxial tensile 00:45:43.980 --> 00:45:45.600 align:middle line:84% strain in the silicon, you need to go 00:45:45.600 --> 00:45:49.270 align:middle line:84% to very high levels of strain if you want this to work. 00:45:49.270 --> 00:45:52.350 align:middle line:84% So that's one big disadvantage for the pMOSFET 00:45:52.350 --> 00:45:53.580 align:middle line:90% in this technology. 00:45:53.580 --> 00:45:56.350 align:middle line:90% 00:45:56.350 --> 00:46:01.390 align:middle line:84% Oh, just another graph on page 24 to give you an idea. 00:46:01.390 --> 00:46:03.130 align:middle line:90% We talked about mobility. 00:46:03.130 --> 00:46:08.410 align:middle line:84% How about things that you care about electron velocity or GM 00:46:08.410 --> 00:46:10.840 align:middle line:84% that a lot of circuit designers are worried about? 00:46:10.840 --> 00:46:12.340 align:middle line:90% Well, indeed, that does improve. 00:46:12.340 --> 00:46:16.540 align:middle line:84% And even as you scale devices, this is data back from 1998 now. 00:46:16.540 --> 00:46:17.290 align:middle line:90% It's fairly old. 00:46:17.290 --> 00:46:21.100 align:middle line:84% But it still gives you an idea of if you look at the GM 00:46:21.100 --> 00:46:25.060 align:middle line:84% here for an unstrained device shown in red, as you scale, 00:46:25.060 --> 00:46:26.270 align:middle line:90% it goes up like this. 00:46:26.270 --> 00:46:29.950 align:middle line:84% For a strain nMOSFET shown in blue, it goes up. 00:46:29.950 --> 00:46:33.730 align:middle line:84% And this enhancement factor or the ratio between these two 00:46:33.730 --> 00:46:35.950 align:middle line:90% lines stays about constant. 00:46:35.950 --> 00:46:40.943 align:middle line:84% In this plot all the way down to about 90 nanometers or so 00:46:40.943 --> 00:46:42.610 align:middle line:84% was the shortest devices that were made, 00:46:42.610 --> 00:46:44.140 align:middle line:90% just around 100 nanometers. 00:46:44.140 --> 00:46:46.632 align:middle line:84% Since that time, shorter devices have been made. 00:46:46.632 --> 00:46:48.340 align:middle line:84% So the nice thing is that the enhancement 00:46:48.340 --> 00:46:50.590 align:middle line:84% doesn't seem to go away when you go to short channels. 00:46:50.590 --> 00:46:54.683 align:middle line:84% So that's very important for the technology. 00:46:54.683 --> 00:46:56.100 align:middle line:84% So let me talk a little bit now, I 00:46:56.100 --> 00:46:58.980 align:middle line:84% want to spend a few minutes on how the material is grown 00:46:58.980 --> 00:47:02.948 align:middle line:84% and some of the issues related to how dopants 00:47:02.948 --> 00:47:03.990 align:middle line:90% diffuse in this material. 00:47:03.990 --> 00:47:06.660 align:middle line:90% 00:47:06.660 --> 00:47:10.980 align:middle line:84% Slide 26 just shows some typical gas sources that are used. 00:47:10.980 --> 00:47:12.990 align:middle line:84% A very common way to grow this is 00:47:12.990 --> 00:47:17.520 align:middle line:84% by low pressure CVD or sometimes ultra high vacuum CVD, 00:47:17.520 --> 00:47:22.440 align:middle line:84% depending on the particular lab that's doing the work. 00:47:22.440 --> 00:47:25.590 align:middle line:84% For epitaxial silicon germanium, a very common silicon source 00:47:25.590 --> 00:47:27.000 align:middle line:90% is silane. 00:47:27.000 --> 00:47:28.708 align:middle line:84% Sometimes people use dichlorosilane. 00:47:28.708 --> 00:47:30.750 align:middle line:84% We've talked about dichloro earlier in the class. 00:47:30.750 --> 00:47:32.550 align:middle line:84% For germanium, people use germane. 00:47:32.550 --> 00:47:33.590 align:middle line:90% What is germane? 00:47:33.590 --> 00:47:34.590 align:middle line:90% It's the same as silane. 00:47:34.590 --> 00:47:38.380 align:middle line:84% But you take out the silicon, and you put in a germanium atom. 00:47:38.380 --> 00:47:39.000 align:middle line:90% So it's GeH4. 00:47:39.000 --> 00:47:41.520 align:middle line:90% 00:47:41.520 --> 00:47:43.530 align:middle line:84% Hydrogen is often used as a carrier gas, 00:47:43.530 --> 00:47:46.290 align:middle line:84% although not in UH BCVD but in low pressure CVD. 00:47:46.290 --> 00:47:47.700 align:middle line:90% And these are common dopants. 00:47:47.700 --> 00:47:50.010 align:middle line:84% Diborane and arsene are used for P and N 00:47:50.010 --> 00:47:54.300 align:middle line:84% type doping, sometimes phosphine. 00:47:54.300 --> 00:47:57.360 align:middle line:84% For silicon germanium in a selective process, 00:47:57.360 --> 00:47:58.570 align:middle line:90% you need to add-- 00:47:58.570 --> 00:48:00.930 align:middle line:84% so if you want to deposit silicon germanium 00:48:00.930 --> 00:48:05.430 align:middle line:84% on a wafer that has oxide on it and little openings in the oxide 00:48:05.430 --> 00:48:09.450 align:middle line:84% where there's windows that are patterned so that you can reach 00:48:09.450 --> 00:48:12.360 align:middle line:84% down into the silicon, this is called selective deposition. 00:48:12.360 --> 00:48:14.940 align:middle line:84% You need to add hydrogen chloride, HCl. 00:48:14.940 --> 00:48:18.930 align:middle line:84% And this is commonly done for both the bipolar and CMOS 00:48:18.930 --> 00:48:20.670 align:middle line:90% applications. 00:48:20.670 --> 00:48:24.150 align:middle line:84% For silicon germanium carbon, you do the same type of thing. 00:48:24.150 --> 00:48:26.850 align:middle line:84% But you add one more gas, which is called methyl silane. 00:48:26.850 --> 00:48:29.010 align:middle line:84% And that the methyl group, the CH3, 00:48:29.010 --> 00:48:32.190 align:middle line:84% brings in the carbon, which is used to dope those devices. 00:48:32.190 --> 00:48:35.380 align:middle line:90% 00:48:35.380 --> 00:48:37.000 align:middle line:84% We talked a little bit about strain. 00:48:37.000 --> 00:48:40.930 align:middle line:84% And here on slide 27, I'm showing a little bit of idea 00:48:40.930 --> 00:48:42.370 align:middle line:90% what that means. 00:48:42.370 --> 00:48:45.370 align:middle line:84% This ball and stick diagram in the lower right 00:48:45.370 --> 00:48:48.430 align:middle line:84% was our pictorial picture in our heads. 00:48:48.430 --> 00:48:51.520 align:middle line:84% When the EPI layer is thin, we said the silicon germanium, 00:48:51.520 --> 00:48:54.760 align:middle line:84% if it's thin enough, it will compress in X and Y. 00:48:54.760 --> 00:48:56.320 align:middle line:84% It will match the lattice parameter 00:48:56.320 --> 00:48:58.570 align:middle line:90% of the thick substrate. 00:48:58.570 --> 00:49:00.490 align:middle line:84% That's only when it's thin enough 00:49:00.490 --> 00:49:04.480 align:middle line:84% and it's energetically favorable to bend the bonds. 00:49:04.480 --> 00:49:06.897 align:middle line:84% In that case, we say it's fully strained. 00:49:06.897 --> 00:49:08.980 align:middle line:84% If you build up more and more layers, you build up 00:49:08.980 --> 00:49:10.820 align:middle line:84% more and more stress, at some point, 00:49:10.820 --> 00:49:13.070 align:middle line:84% it becomes energetically favorable to break bonds. 00:49:13.070 --> 00:49:16.000 align:middle line:84% So at this interface, bonds are actually broken. 00:49:16.000 --> 00:49:19.150 align:middle line:84% And it's represented here by these misfit dislocations 00:49:19.150 --> 00:49:24.860 align:middle line:84% that are going into the board or into the board here. 00:49:24.860 --> 00:49:27.910 align:middle line:84% And so these misfit dislocations end up 00:49:27.910 --> 00:49:30.733 align:middle line:84% relieving the strain on the thick layer. 00:49:30.733 --> 00:49:31.900 align:middle line:90% They have a certain spacing. 00:49:31.900 --> 00:49:33.900 align:middle line:84% And if you have enough of them and their spacing 00:49:33.900 --> 00:49:35.750 align:middle line:84% is tight enough, in fact, you will end up 00:49:35.750 --> 00:49:37.170 align:middle line:90% with a fully relaxed layer. 00:49:37.170 --> 00:49:39.170 align:middle line:84% The silicon germanium will have a last parameter 00:49:39.170 --> 00:49:41.550 align:middle line:90% that it has an equilibrium. 00:49:41.550 --> 00:49:44.040 align:middle line:84% And this is a cross-section view in the upper right. 00:49:44.040 --> 00:49:47.700 align:middle line:84% And this transmission electron micrograph is a planar view. 00:49:47.700 --> 00:49:50.600 align:middle line:84% So if you look down through an epitaxial layer that's 00:49:50.600 --> 00:49:54.380 align:middle line:84% been grown to be too thick, and here is a bird's eye view. 00:49:54.380 --> 00:49:57.440 align:middle line:84% This is a misfit dislocation segment along the interface. 00:49:57.440 --> 00:49:59.960 align:middle line:84% And there's a dislocation arm that threads to the surface. 00:49:59.960 --> 00:50:03.950 align:middle line:84% And these dark lines end up having extra electron 00:50:03.950 --> 00:50:05.110 align:middle line:90% diffraction contrast. 00:50:05.110 --> 00:50:08.510 align:middle line:84% And these dark lines correspond to misfit dislocations 00:50:08.510 --> 00:50:09.800 align:middle line:90% at this interface. 00:50:09.800 --> 00:50:12.620 align:middle line:84% And they run along the easy slip planes 00:50:12.620 --> 00:50:14.480 align:middle line:84% in silicon and silicon germanium, 00:50:14.480 --> 00:50:16.410 align:middle line:84% which they run along the 110 direction, 00:50:16.410 --> 00:50:20.160 align:middle line:84% so just typically parallel and perpendicular to the flat. 00:50:20.160 --> 00:50:22.970 align:middle line:84% So this film was grown to be a little bit too thick. 00:50:22.970 --> 00:50:25.020 align:middle line:84% And it started to form dislocations. 00:50:25.020 --> 00:50:26.270 align:middle line:90% It's no longer fully strained. 00:50:26.270 --> 00:50:28.850 align:middle line:90% 00:50:28.850 --> 00:50:31.100 align:middle line:84% In fact, you may be wondering, oh, well, 00:50:31.100 --> 00:50:33.500 align:middle line:90% how thick is too thick? 00:50:33.500 --> 00:50:35.450 align:middle line:84% Well, there is no one number, unfortunately. 00:50:35.450 --> 00:50:37.430 align:middle line:84% There for a long time, people thought 00:50:37.430 --> 00:50:39.980 align:middle line:84% there was a quote unquote critical thickness so that 00:50:39.980 --> 00:50:41.730 align:middle line:84% at a particular germanium fraction, 00:50:41.730 --> 00:50:43.830 align:middle line:84% when you go above that thickness, bingo. 00:50:43.830 --> 00:50:45.080 align:middle line:90% You get dislocations. 00:50:45.080 --> 00:50:47.620 align:middle line:90% Doesn't quite happen that way. 00:50:47.620 --> 00:50:50.970 align:middle line:84% In fact, there is a whole metastable regime. 00:50:50.970 --> 00:50:52.470 align:middle line:84% And the critical thickness depends 00:50:52.470 --> 00:50:53.580 align:middle line:84% on the temperature at which you're 00:50:53.580 --> 00:50:56.220 align:middle line:84% doing the growth because it's a kinetically limited process. 00:50:56.220 --> 00:50:59.910 align:middle line:84% And this is a good paper, Derek Houghton's paper from 1991, 00:50:59.910 --> 00:51:03.692 align:middle line:84% who describes the kinetic model and explains the fact 00:51:03.692 --> 00:51:06.150 align:middle line:84% that there's no one critical thickness at a given germanium 00:51:06.150 --> 00:51:07.470 align:middle line:90% fraction. 00:51:07.470 --> 00:51:09.150 align:middle line:84% What he plotted here was a summary 00:51:09.150 --> 00:51:12.960 align:middle line:84% of the critical thickness above which dislocations 00:51:12.960 --> 00:51:16.410 align:middle line:84% form as a function of germanium fraction of the silicon 00:51:16.410 --> 00:51:19.150 align:middle line:84% germanium layer for different growth temperatures. 00:51:19.150 --> 00:51:22.200 align:middle line:84% So if you're growing at 900 degrees, you're very hot. 00:51:22.200 --> 00:51:24.300 align:middle line:84% And it's very easy to nucleate dislocations. 00:51:24.300 --> 00:51:25.950 align:middle line:90% And they travel very rapidly. 00:51:25.950 --> 00:51:28.410 align:middle line:84% You tend to then get the equilibrium critical thickness 00:51:28.410 --> 00:51:30.370 align:middle line:90% that that's this line here. 00:51:30.370 --> 00:51:32.130 align:middle line:84% So if you're growing at high temperature, 00:51:32.130 --> 00:51:34.950 align:middle line:84% or you're annealing at high temperature, if you're at 20%, 00:51:34.950 --> 00:51:37.350 align:middle line:84% the equilibrium critical thickness for silicon germanium 00:51:37.350 --> 00:51:41.320 align:middle line:84% is about 120 angstroms, 110 angstroms, something like that. 00:51:41.320 --> 00:51:43.950 align:middle line:84% So it says that if you're growing in high temperature, 00:51:43.950 --> 00:51:47.520 align:middle line:84% and you grow above about 120 angstroms of 20% silicon 00:51:47.520 --> 00:51:51.220 align:middle line:84% germanium, it'll start to form dislocations at that point. 00:51:51.220 --> 00:51:55.650 align:middle line:84% However, if you are growing that same germanium fraction at 550, 00:51:55.650 --> 00:51:58.140 align:middle line:84% so you're in a very low temperature epitaxial growth, 00:51:58.140 --> 00:51:59.850 align:middle line:84% you're are kinetically suppressing 00:51:59.850 --> 00:52:01.700 align:middle line:90% the formation of dislocations. 00:52:01.700 --> 00:52:04.420 align:middle line:84% In fact, the critical thickness is like this curve here. 00:52:04.420 --> 00:52:06.450 align:middle line:90% It's about 700 angstroms. 00:52:06.450 --> 00:52:09.240 align:middle line:84% So there's a metastable regime, if you 00:52:09.240 --> 00:52:12.000 align:middle line:84% grow the layers thin enough, you can keep them strained. 00:52:12.000 --> 00:52:15.270 align:middle line:84% And they won't form dislocations until a much thicker layer. 00:52:15.270 --> 00:52:16.800 align:middle line:84% The problem with that is, you can 00:52:16.800 --> 00:52:19.500 align:middle line:84% grow them very thin at high temperature, very thick, 00:52:19.500 --> 00:52:20.910 align:middle line:84% and they won't have dislocations. 00:52:20.910 --> 00:52:24.180 align:middle line:84% But if you go to process them it to make a device, at a higher 00:52:24.180 --> 00:52:26.055 align:middle line:84% temperature, dislocations can pop in 00:52:26.055 --> 00:52:28.180 align:middle line:84% if you're above the equilibrium critical thickness. 00:52:28.180 --> 00:52:30.330 align:middle line:84% So this region is called metastable 00:52:30.330 --> 00:52:32.280 align:middle line:84% because you can grow them at low temperatures 00:52:32.280 --> 00:52:33.420 align:middle line:90% without dislocations. 00:52:33.420 --> 00:52:35.670 align:middle line:84% But if you were to anneal them at a higher temperature 00:52:35.670 --> 00:52:38.230 align:middle line:84% dislocation, it would fall, and it would start to relax. 00:52:38.230 --> 00:52:41.060 align:middle line:84% Generally, this region up here, up in the upper right, 00:52:41.060 --> 00:52:43.840 align:middle line:84% so it's very thick films, are almost always relaxed. 00:52:43.840 --> 00:52:46.240 align:middle line:84% They'll always have dislocations. 00:52:46.240 --> 00:52:49.180 align:middle line:84% I refer you to this 1991 paper for a good reference 00:52:49.180 --> 00:52:54.610 align:middle line:84% on understanding the formation of misfits in strained silicon 00:52:54.610 --> 00:52:57.130 align:middle line:90% germanium layers. 00:52:57.130 --> 00:53:04.370 align:middle line:84% Slide 29, it turns out, in the early 1990s, 00:53:04.370 --> 00:53:06.115 align:middle line:84% there was some breakthroughs on, you 00:53:06.115 --> 00:53:07.990 align:middle line:84% say, well, growing strains silicon germanium. 00:53:07.990 --> 00:53:09.070 align:middle line:90% That's very interesting. 00:53:09.070 --> 00:53:10.180 align:middle line:90% People want to do that. 00:53:10.180 --> 00:53:11.763 align:middle line:84% There were actually some breakthroughs 00:53:11.763 --> 00:53:13.790 align:middle line:84% on how to grow relaxed silicon germanium. 00:53:13.790 --> 00:53:16.720 align:middle line:84% This seems like it should be very easy to make the silicon 00:53:16.720 --> 00:53:17.420 align:middle line:90% germanium relax. 00:53:17.420 --> 00:53:18.670 align:middle line:90% You just grow it thick, right? 00:53:18.670 --> 00:53:21.400 align:middle line:84% We just saw that paper from Derek Houghton. 00:53:21.400 --> 00:53:26.238 align:middle line:84% Well, if I'm at 30%, I just grow it to be 3 or 4 microns up here. 00:53:26.238 --> 00:53:28.280 align:middle line:84% It'll definitely be above the critical thickness. 00:53:28.280 --> 00:53:29.072 align:middle line:90% It will be relaxed. 00:53:29.072 --> 00:53:30.880 align:middle line:84% It will have its own lattice parameter. 00:53:30.880 --> 00:53:32.380 align:middle line:84% The problem with that, if you just 00:53:32.380 --> 00:53:35.740 align:middle line:84% grow it right off the bat like that, a box layer, 00:53:35.740 --> 00:53:40.490 align:middle line:84% you end up with a huge number of threading dislocations. 00:53:40.490 --> 00:53:42.668 align:middle line:84% So if you don't grade the profile, 00:53:42.668 --> 00:53:44.710 align:middle line:84% you end up with more than 10 to the 9th, maybe 10 00:53:44.710 --> 00:53:47.920 align:middle line:84% to the 10th of these threading arms. 00:53:47.920 --> 00:53:51.760 align:middle line:84% For every misfit, you have two arms that go up to the surface. 00:53:51.760 --> 00:53:54.070 align:middle line:84% And you have so many of them that the layers 00:53:54.070 --> 00:53:56.230 align:middle line:84% aren't very useful for very many devices 00:53:56.230 --> 00:53:58.520 align:middle line:84% because there's too many defects. 00:53:58.520 --> 00:54:00.680 align:middle line:84% But in the early '90s, there were breakthroughs 00:54:00.680 --> 00:54:03.350 align:middle line:84% on how to grow these layers, how to grow them 00:54:03.350 --> 00:54:05.930 align:middle line:84% in high enough temperature so the threading arms could 00:54:05.930 --> 00:54:08.180 align:middle line:84% run to the edge of the wafer and how 00:54:08.180 --> 00:54:10.190 align:middle line:90% to grade the germanium content. 00:54:10.190 --> 00:54:13.100 align:middle line:84% So you start growing at silicon lattice. 00:54:13.100 --> 00:54:15.740 align:middle line:84% And then you gradually add germanium as you go up 00:54:15.740 --> 00:54:17.750 align:middle line:84% and you grade it over many microns. 00:54:17.750 --> 00:54:20.310 align:middle line:84% It turns out, when you do that, you can relax the silicon 00:54:20.310 --> 00:54:20.810 align:middle line:90% germanium. 00:54:20.810 --> 00:54:24.110 align:middle line:84% But you can reduce the number of arms that thread to the surface. 00:54:24.110 --> 00:54:26.235 align:middle line:84% All the arms run out to the edge of the wafer where 00:54:26.235 --> 00:54:27.735 align:middle line:84% you don't care about them you're not 00:54:27.735 --> 00:54:29.990 align:middle line:84% going make any devices at the edge of the wafer. 00:54:29.990 --> 00:54:32.150 align:middle line:84% So the threading dislocation density 00:54:32.150 --> 00:54:35.580 align:middle line:84% can go down from 10 to the 9th to less than 10 to the 5th 00:54:35.580 --> 00:54:37.430 align:middle line:90% if you do the grading right. 00:54:37.430 --> 00:54:39.620 align:middle line:84% So a nice thing about this is all of a sudden, 00:54:39.620 --> 00:54:42.200 align:middle line:84% on silicon substrates in the early '90s, people 00:54:42.200 --> 00:54:43.970 align:middle line:84% had a way to take a silicon substrate, 00:54:43.970 --> 00:54:46.040 align:middle line:84% grow a relaxed silicon germanium layer, 00:54:46.040 --> 00:54:48.800 align:middle line:84% and have it be pretty high quality and reasonable enough 00:54:48.800 --> 00:54:51.388 align:middle line:84% high quality that you could make devices on it that work. 00:54:51.388 --> 00:54:53.930 align:middle line:84% There's still a lot of threading dislocations, but not enough 00:54:53.930 --> 00:54:56.257 align:middle line:84% to screw up the devices too badly. 00:54:56.257 --> 00:54:58.590 align:middle line:84% So that was kind of the breakthrough that enabled people 00:54:58.590 --> 00:55:00.870 align:middle line:84% to make relaxed silicon germanium and then 00:55:00.870 --> 00:55:05.078 align:middle line:84% strained silicon MOSFETs on top of it. 00:55:05.078 --> 00:55:06.620 align:middle line:84% What are some of the materials issues 00:55:06.620 --> 00:55:09.290 align:middle line:84% if you're growing this relaxed silicon germanium? 00:55:09.290 --> 00:55:12.860 align:middle line:84% Well, I already mentioned the threading dislocation density. 00:55:12.860 --> 00:55:15.980 align:middle line:84% It actually decreases if you go to higher growth temperatures 00:55:15.980 --> 00:55:19.100 align:middle line:84% because the dislocations can move faster. 00:55:19.100 --> 00:55:21.650 align:middle line:84% The problem, though, is you also tend 00:55:21.650 --> 00:55:25.160 align:middle line:84% to get a very strange field that creates a crosshatch. 00:55:25.160 --> 00:55:29.240 align:middle line:84% In fact, if you look a plan view optical micrograph 00:55:29.240 --> 00:55:30.870 align:middle line:84% of a relaxed silicon germanium layer, 00:55:30.870 --> 00:55:32.540 align:middle line:90% it looks like a plaid shirt. 00:55:32.540 --> 00:55:36.470 align:middle line:84% This plaid is that buried strain field causes the surface 00:55:36.470 --> 00:55:41.240 align:middle line:84% to undulate ever so slightly, maybe 100 angstrom, high hills 00:55:41.240 --> 00:55:42.500 align:middle line:90% and valleys, very smooth. 00:55:42.500 --> 00:55:44.210 align:middle line:84% And the period of these undulations 00:55:44.210 --> 00:55:45.570 align:middle line:90% are the spacing is quite large. 00:55:45.570 --> 00:55:47.660 align:middle line:90% It can be microns. 00:55:47.660 --> 00:55:49.160 align:middle line:90% So these are smooth undulations. 00:55:49.160 --> 00:55:51.590 align:middle line:84% They don't really mess with the electron mobility. 00:55:51.590 --> 00:55:53.900 align:middle line:84% The electrons and holes don't care about them. 00:55:53.900 --> 00:55:56.120 align:middle line:84% It makes the wafers look ugly, however. 00:55:56.120 --> 00:55:58.430 align:middle line:84% And it makes lithography difficult 00:55:58.430 --> 00:56:01.070 align:middle line:84% because optical machines don't like looking at plaid. 00:56:01.070 --> 00:56:02.510 align:middle line:84% They like things that look smooth. 00:56:02.510 --> 00:56:04.332 align:middle line:90% So it is a practical problem. 00:56:04.332 --> 00:56:05.790 align:middle line:84% People have found a ways around it. 00:56:05.790 --> 00:56:09.085 align:middle line:84% You can CNP the layer, and you can remove all the undulations. 00:56:09.085 --> 00:56:10.710 align:middle line:84% And it's not necessarily a big problem. 00:56:10.710 --> 00:56:13.010 align:middle line:90% But it is an issue. 00:56:13.010 --> 00:56:14.990 align:middle line:84% If you grow the silicon germanium too hot, 00:56:14.990 --> 00:56:18.050 align:middle line:84% it turns out it doesn't even grow as a planar film. 00:56:18.050 --> 00:56:21.320 align:middle line:84% It grows as big little clumps, little mountains or islands. 00:56:21.320 --> 00:56:22.940 align:middle line:84% It doesn't wet the surface properly. 00:56:22.940 --> 00:56:26.660 align:middle line:84% And this type of growth mode is generally 00:56:26.660 --> 00:56:29.360 align:middle line:84% considered not very useful because it's not really 00:56:29.360 --> 00:56:30.590 align:middle line:90% a smooth film. 00:56:30.590 --> 00:56:33.410 align:middle line:84% So you cannot grow silicon germanium at too high 00:56:33.410 --> 00:56:35.285 align:middle line:84% a temperature for a given germanium fraction. 00:56:35.285 --> 00:56:38.700 align:middle line:90% 00:56:38.700 --> 00:56:41.880 align:middle line:84% What do these relaxed silicon germanium graded buffer layers 00:56:41.880 --> 00:56:42.390 align:middle line:90% look like? 00:56:42.390 --> 00:56:44.473 align:middle line:84% I said, well, they have very few defects that 00:56:44.473 --> 00:56:45.390 align:middle line:90% thread to the surface. 00:56:45.390 --> 00:56:47.100 align:middle line:90% That's indeed true. 00:56:47.100 --> 00:56:49.282 align:middle line:84% These are some cross section TEM micrographs 00:56:49.282 --> 00:56:50.490 align:middle line:90% of relaxed silicon germanium. 00:56:50.490 --> 00:56:53.678 align:middle line:84% See at 10%, this is the silicon substrate. 00:56:53.678 --> 00:56:54.720 align:middle line:90% This is the graded layer. 00:56:54.720 --> 00:56:56.880 align:middle line:84% There are lots of dislocations in the graded layer. 00:56:56.880 --> 00:56:58.540 align:middle line:90% That's what you want. 00:56:58.540 --> 00:57:00.580 align:middle line:84% You want dislocations down in the graded layer 00:57:00.580 --> 00:57:02.580 align:middle line:84% because you want to relax the lattice parameter. 00:57:02.580 --> 00:57:05.640 align:middle line:84% You want it to expand to be its equilibrium parameter. 00:57:05.640 --> 00:57:09.030 align:middle line:84% But what you see is as you get up to the surface, 00:57:09.030 --> 00:57:11.520 align:middle line:84% if the graded layer is thick enough into the box layer, 00:57:11.520 --> 00:57:15.157 align:middle line:84% there are too many dislocations of visible in cross-section PEM. 00:57:15.157 --> 00:57:16.740 align:middle line:84% Again, this is up here is where you're 00:57:16.740 --> 00:57:17.980 align:middle line:90% going to build your device. 00:57:17.980 --> 00:57:19.330 align:middle line:90% So that's what you care about. 00:57:19.330 --> 00:57:23.610 align:middle line:84% But if you compare a 10% to a 20% to a 30% germanium, 00:57:23.610 --> 00:57:25.140 align:middle line:84% you see a higher and higher density 00:57:25.140 --> 00:57:27.457 align:middle line:84% of these buried dislocations, of course, 00:57:27.457 --> 00:57:29.040 align:middle line:84% as you increase the germanium fraction 00:57:29.040 --> 00:57:33.110 align:middle line:84% because you have to relieve more lattice mismatch. 00:57:33.110 --> 00:57:36.040 align:middle line:90% So the layers get pretty ugly. 00:57:36.040 --> 00:57:37.650 align:middle line:84% Even though the device is up on top, 00:57:37.650 --> 00:57:39.730 align:middle line:84% there's still a lot of integration issues. 00:57:39.730 --> 00:57:43.600 align:middle line:84% You can imagine trying to make CMOS circuits on a substrate 00:57:43.600 --> 00:57:44.575 align:middle line:90% that looks like this. 00:57:44.575 --> 00:57:47.147 align:middle line:84% All these buried dislocations are going to worry 00:57:47.147 --> 00:57:48.730 align:middle line:84% and they're going to bother the device 00:57:48.730 --> 00:57:51.670 align:middle line:84% engineers because after all, we do have CMOS wells. 00:57:51.670 --> 00:57:56.890 align:middle line:84% We do have deep junctions in some of these structures. 00:57:56.890 --> 00:58:00.400 align:middle line:84% So this silicon germanium can affect our my shallow trench 00:58:00.400 --> 00:58:03.610 align:middle line:84% isolation fabrication, formation of silicides, 00:58:03.610 --> 00:58:05.320 align:middle line:90% and the way dopants diffuse. 00:58:05.320 --> 00:58:06.805 align:middle line:90% So it certainly is an issue. 00:58:06.805 --> 00:58:09.380 align:middle line:90% 00:58:09.380 --> 00:58:12.620 align:middle line:84% In fact, there's a whole field of research going on right now, 00:58:12.620 --> 00:58:15.040 align:middle line:84% which is the study of the diffusion of dopants 00:58:15.040 --> 00:58:17.510 align:middle line:90% in silicon germanium. 00:58:17.510 --> 00:58:21.830 align:middle line:84% And I'm showing here on slide 32, 00:58:21.830 --> 00:58:24.500 align:middle line:84% this is some data for silicon germanium alloys 00:58:24.500 --> 00:58:26.490 align:middle line:90% with 20% germanium in them. 00:58:26.490 --> 00:58:28.400 align:middle line:84% And what we're looking at is the diffusivity 00:58:28.400 --> 00:58:29.510 align:middle line:90% of different dopants. 00:58:29.510 --> 00:58:32.970 align:middle line:84% So this is the effective equilibrium diffusivity. 00:58:32.970 --> 00:58:34.760 align:middle line:84% So its diffusion coefficient in centimeter 00:58:34.760 --> 00:58:38.660 align:middle line:84% squared per second as a function of inverse temperature, well, 00:58:38.660 --> 00:58:40.370 align:middle line:84% the usual kind of Arrhenius plot. 00:58:40.370 --> 00:58:43.850 align:middle line:84% And if you look at boron, so the dashed line here 00:58:43.850 --> 00:58:45.290 align:middle line:90% is for diffusion in silicon. 00:58:45.290 --> 00:58:48.230 align:middle line:84% And the solid is for diffusion in silicon germanium. 00:58:48.230 --> 00:58:52.890 align:middle line:84% If you look at boron here, in fact, in 20% silicon germanium, 00:58:52.890 --> 00:58:54.060 align:middle line:90% the diffusivities reduced. 00:58:54.060 --> 00:58:55.852 align:middle line:84% I'd said before, a factor of 10, looks more 00:58:55.852 --> 00:58:58.410 align:middle line:84% like a factor of 6 or so, depending on the temperature. 00:58:58.410 --> 00:59:01.528 align:middle line:84% But so the boron diffusivity goes down 00:59:01.528 --> 00:59:03.570 align:middle line:84% in silicon germanium, which is nice because boron 00:59:03.570 --> 00:59:05.550 align:middle line:90% is a fast diffuser in silicon. 00:59:05.550 --> 00:59:07.770 align:middle line:84% But if you look at arsenic, this is arsenic 00:59:07.770 --> 00:59:09.000 align:middle line:90% in silicon, the dashed line. 00:59:09.000 --> 00:59:11.520 align:middle line:84% Arsenic in silicon germanium is the solid line here 00:59:11.520 --> 00:59:12.790 align:middle line:90% with the green arrow. 00:59:12.790 --> 00:59:14.250 align:middle line:84% So the arsenic in silicon germanium 00:59:14.250 --> 00:59:15.490 align:middle line:90% is diffusing a lot faster. 00:59:15.490 --> 00:59:17.700 align:middle line:84% In fact, arsenic in silicon germanium 00:59:17.700 --> 00:59:19.710 align:middle line:84% is diffusing at about the same rate as born 00:59:19.710 --> 00:59:22.270 align:middle line:90% in silicon germanium at 20%. 00:59:22.270 --> 00:59:24.450 align:middle line:84% So we've taken the boron diffusion coefficient, 00:59:24.450 --> 00:59:25.470 align:middle line:90% it goes down. 00:59:25.470 --> 00:59:27.970 align:middle line:84% Arsenic, which is usually very slow in silicon, now goes up. 00:59:27.970 --> 00:59:30.512 align:middle line:84% Now they diffuse about equally, which is kind of interesting. 00:59:30.512 --> 00:59:32.730 align:middle line:84% You might make some interesting device structures 00:59:32.730 --> 00:59:34.320 align:middle line:84% when you have symmetric diffusion 00:59:34.320 --> 00:59:37.110 align:middle line:84% coefficients between the N and the P type doping. 00:59:37.110 --> 00:59:39.720 align:middle line:84% Unfortunately, what it means is that in silicon germanium 00:59:39.720 --> 00:59:41.790 align:middle line:84% MOSFETs, we actually are much more 00:59:41.790 --> 00:59:44.370 align:middle line:84% worried about the n-type dopant diffusion 00:59:44.370 --> 00:59:45.750 align:middle line:84% because that's what's going fast. 00:59:45.750 --> 00:59:48.330 align:middle line:84% And we're less worried about the p-type doping diffusion 00:59:48.330 --> 00:59:49.890 align:middle line:84% because the boron is going slower. 00:59:49.890 --> 00:59:51.880 align:middle line:84% Phosphorus doesn't help the situation. 00:59:51.880 --> 00:59:55.810 align:middle line:84% Phosphorus is also enhanced, maybe about a factor of 2 or 3, 00:59:55.810 --> 00:59:58.630 align:middle line:90% depending upon the temperature. 00:59:58.630 --> 01:00:01.630 align:middle line:84% So and this line here gives you an idea roughly 01:00:01.630 --> 01:00:03.880 align:middle line:90% of the diffusivity of germanium. 01:00:03.880 --> 01:00:06.632 align:middle line:84% Well, this is at a very low germanium fraction 01:00:06.632 --> 01:00:07.840 align:middle line:90% as a function of temperature. 01:00:07.840 --> 01:00:10.855 align:middle line:84% It's diffusing slower than the dopants, at least 01:00:10.855 --> 01:00:11.980 align:middle line:90% at this germanium fraction. 01:00:11.980 --> 01:00:15.380 align:middle line:90% 01:00:15.380 --> 01:00:19.030 align:middle line:84% So I mentioned to you that arsenic moves faster 01:00:19.030 --> 01:00:20.000 align:middle line:90% in silicon germanium. 01:00:20.000 --> 01:00:21.667 align:middle line:84% This is actually a big problem if you're 01:00:21.667 --> 01:00:24.670 align:middle line:84% trying to make a strained silicon mMOSFET with silicon 01:00:24.670 --> 01:00:26.210 align:middle line:90% germanium in the substrate. 01:00:26.210 --> 01:00:29.770 align:middle line:84% Let's say you're trying to make a shallow source drains. 01:00:29.770 --> 01:00:34.720 align:middle line:84% So you implant 50 keV at 3 times 10 to 15, 01:00:34.720 --> 01:00:37.660 align:middle line:84% may be a typical source drain implant. 01:00:37.660 --> 01:00:42.110 align:middle line:84% And you anneal it at 1,000 degrees for various times. 01:00:42.110 --> 01:00:45.800 align:middle line:84% And the dashed lines here are for a silicon substrate 01:00:45.800 --> 01:00:48.090 align:middle line:84% and the solid are for silicon germanium. 01:00:48.090 --> 01:00:51.500 align:middle line:84% So if you just look at, say, this dashed line here, 01:00:51.500 --> 01:00:54.830 align:middle line:84% this red dashed line, for the silicon substrate 01:00:54.830 --> 01:00:57.947 align:middle line:84% with this 22nd anneal at 1,000, you 01:00:57.947 --> 01:00:59.780 align:middle line:84% have a junction depth of about 500 angstrom. 01:00:59.780 --> 01:01:00.950 align:middle line:90% That's perfect. 01:01:00.950 --> 01:01:02.640 align:middle line:84% For the silicon germanium substrate, 01:01:02.640 --> 01:01:04.850 align:middle line:84% which is the solid line, 22nd anneal, 01:01:04.850 --> 01:01:08.000 align:middle line:84% your junction depth is more than twice that. 01:01:08.000 --> 01:01:10.230 align:middle line:90% It's like a 1,200 angstroms. 01:01:10.230 --> 01:01:12.650 align:middle line:84% So the arsenic and the silicon germanium 01:01:12.650 --> 01:01:14.180 align:middle line:90% is going much more rapidly. 01:01:14.180 --> 01:01:16.430 align:middle line:90% You see this big, big shift. 01:01:16.430 --> 01:01:18.260 align:middle line:90% And that's a problem. 01:01:18.260 --> 01:01:20.030 align:middle line:84% If you use very short anneal times, 01:01:20.030 --> 01:01:21.830 align:middle line:84% you can minimize that difference. 01:01:21.830 --> 01:01:24.440 align:middle line:84% But 1,000 degrees, one second is not 01:01:24.440 --> 01:01:26.330 align:middle line:84% really enough to really repair all the damage 01:01:26.330 --> 01:01:27.630 align:middle line:90% and to activate all the dopant. 01:01:27.630 --> 01:01:30.410 align:middle line:84% So there's a very interesting issue here 01:01:30.410 --> 01:01:35.815 align:middle line:84% is, how do we optimize the annealing time for anti dopant? 01:01:35.815 --> 01:01:37.940 align:middle line:84% So we get good activation in the silicon germanium. 01:01:37.940 --> 01:01:41.760 align:middle line:84% But we don't introduce too much diffusion, otherwise, 01:01:41.760 --> 01:01:47.050 align:middle line:84% we can't control the device fabrication. 01:01:47.050 --> 01:01:48.707 align:middle line:84% Here's another issue on slide 34. 01:01:48.707 --> 01:01:51.040 align:middle line:84% I'm not trying to make silicon germanium look difficult. 01:01:51.040 --> 01:01:52.510 align:middle line:90% But it is a new material. 01:01:52.510 --> 01:01:55.120 align:middle line:90% And a lot of things change. 01:01:55.120 --> 01:01:59.470 align:middle line:84% Slide 34 shows you that indeed, the germanium itself 01:01:59.470 --> 01:02:01.010 align:middle line:90% can diffuse. 01:02:01.010 --> 01:02:04.750 align:middle line:84% Well, we're talking about taking a silicon germanium thick layer 01:02:04.750 --> 01:02:06.970 align:middle line:84% and putting a thin layer, say 100 angstroms, 01:02:06.970 --> 01:02:08.140 align:middle line:90% of strained silicon on it. 01:02:08.140 --> 01:02:09.640 align:middle line:90% That's my channel. 01:02:09.640 --> 01:02:10.930 align:middle line:90% OK, that sounds fine. 01:02:10.930 --> 01:02:12.550 align:middle line:90% We can grow that epitaxially. 01:02:12.550 --> 01:02:14.770 align:middle line:84% But now, I need to implant the source trays 01:02:14.770 --> 01:02:16.015 align:middle line:90% and anneal it at 1,000. 01:02:16.015 --> 01:02:18.640 align:middle line:84% What's going to happen to that silicon germanium, the germanium 01:02:18.640 --> 01:02:20.170 align:middle line:90% and silicon germanium? 01:02:20.170 --> 01:02:23.020 align:middle line:84% Well, it's going to diffuse up into the silicon. 01:02:23.020 --> 01:02:27.477 align:middle line:84% Once we get germanium into the channel, that's a no no. 01:02:27.477 --> 01:02:29.560 align:middle line:84% The electrons don't like germanium in the channel. 01:02:29.560 --> 01:02:31.560 align:middle line:84% The electrons can be scattered by the germanium. 01:02:31.560 --> 01:02:33.080 align:middle line:90% Mobility goes way down. 01:02:33.080 --> 01:02:34.990 align:middle line:90% And your device is ruined. 01:02:34.990 --> 01:02:36.592 align:middle line:84% You lose a lot of your strain too. 01:02:36.592 --> 01:02:38.800 align:middle line:84% So you have to keep the germanium out of the channel. 01:02:38.800 --> 01:02:40.092 align:middle line:90% Well, it doesn't sound too bad. 01:02:40.092 --> 01:02:42.970 align:middle line:84% But if you just look at some of the diffusion coefficients, 01:02:42.970 --> 01:02:46.030 align:middle line:84% in fact, these different lines, these are measured. 01:02:46.030 --> 01:02:47.980 align:middle line:84% This is the germanium self-diffusion 01:02:47.980 --> 01:02:51.380 align:middle line:84% in silicon germanium alloys for different germanium content. 01:02:51.380 --> 01:02:52.460 align:middle line:90% So this is for silicon. 01:02:52.460 --> 01:02:54.370 align:middle line:84% So this will be silicon self diffusion. 01:02:54.370 --> 01:03:00.830 align:middle line:84% This is 10% germanium, 20%, 30, 40 and 50. 01:03:00.830 --> 01:03:05.127 align:middle line:84% So if you just look at a given temperature, say at 950 degrees. 01:03:05.127 --> 01:03:07.460 align:middle line:84% Look it's what's happening to the diffusion coefficient. 01:03:07.460 --> 01:03:11.930 align:middle line:84% In pure silicon, germanium diffusing very slowly 01:03:11.930 --> 01:03:14.600 align:middle line:84% for a low germanium content, here, at around 10 01:03:14.600 --> 01:03:17.160 align:middle line:84% to the minus 17th centimeter squared per second. 01:03:17.160 --> 01:03:21.050 align:middle line:84% But if you go up to a 40% alloy, the diffusivity 01:03:21.050 --> 01:03:23.190 align:middle line:84% goes up by two orders of magnitude. 01:03:23.190 --> 01:03:25.670 align:middle line:84% So what this is saying is basically, 01:03:25.670 --> 01:03:28.700 align:middle line:84% you get an exponential increase, it 01:03:28.700 --> 01:03:32.240 align:middle line:84% looks like, in the germanium diffusion coefficient 01:03:32.240 --> 01:03:34.750 align:middle line:84% as you increase the amount of germanium in the alloy. 01:03:34.750 --> 01:03:36.560 align:middle line:90% So it's a big issue. 01:03:36.560 --> 01:03:39.412 align:middle line:84% If you want to do strained silicon on 20%, 01:03:39.412 --> 01:03:41.120 align:middle line:84% you're going to have a silicon germanium. 01:03:41.120 --> 01:03:42.828 align:middle line:84% You have a certain diffusion coefficient. 01:03:42.828 --> 01:03:44.780 align:middle line:84% If you do it on 30%, all of a sudden 01:03:44.780 --> 01:03:48.480 align:middle line:84% your diffusion coefficient, factor of 5 or so higher. 01:03:48.480 --> 01:03:49.970 align:middle line:84% So that's a big issue because you 01:03:49.970 --> 01:03:52.137 align:middle line:84% don't want the germanium diffusing into the channel. 01:03:52.137 --> 01:03:55.850 align:middle line:84% So this data is not exactly interdiffusion data. 01:03:55.850 --> 01:03:58.590 align:middle line:84% That whole problem needs to be worked out. 01:03:58.590 --> 01:04:02.070 align:middle line:84% But interdiffusion is definitely observed for like 1,000 degrees 01:04:02.070 --> 01:04:03.640 align:middle line:90% 10 seconds at 20%. 01:04:03.640 --> 01:04:05.473 align:middle line:90% So that that's a no no. 01:04:05.473 --> 01:04:07.140 align:middle line:84% So that's going to be a big issue in how 01:04:07.140 --> 01:04:08.723 align:middle line:84% we're going to activate source strains 01:04:08.723 --> 01:04:10.110 align:middle line:90% in those kind of structures. 01:04:10.110 --> 01:04:12.950 align:middle line:90% 01:04:12.950 --> 01:04:16.190 align:middle line:84% That's an interesting research area. 01:04:16.190 --> 01:04:19.640 align:middle line:84% You might be saying, well, what is the mechanism of diffusion 01:04:19.640 --> 01:04:21.350 align:middle line:90% of germanium in silicon anyway? 01:04:21.350 --> 01:04:22.790 align:middle line:90% Actually, it's not really known. 01:04:22.790 --> 01:04:26.990 align:middle line:84% On page 35, there's an example of people are just 01:04:26.990 --> 01:04:28.320 align:middle line:90% trying to get a handle on this. 01:04:28.320 --> 01:04:32.330 align:middle line:84% This was published about a year and a half ago. 01:04:32.330 --> 01:04:35.020 align:middle line:84% And this particular experiment, they 01:04:35.020 --> 01:04:37.850 align:middle line:84% did a process where they had strained silicon 01:04:37.850 --> 01:04:39.410 align:middle line:90% on relaxed silicon germanium. 01:04:39.410 --> 01:04:42.410 align:middle line:84% And they did an anneal in a rapid thermal annealer 01:04:42.410 --> 01:04:43.800 align:middle line:90% where they put in ammonia. 01:04:43.800 --> 01:04:46.160 align:middle line:84% So they're doing rapid thermal nitratation. 01:04:46.160 --> 01:04:49.310 align:middle line:84% So what do we know about nitratation? 01:04:49.310 --> 01:04:52.063 align:middle line:84% Well, nitratation, we believe injects vacancies, right? 01:04:52.063 --> 01:04:52.980 align:middle line:90% That's what we use it. 01:04:52.980 --> 01:04:55.400 align:middle line:84% So in the samples that were nitrided, 01:04:55.400 --> 01:04:57.785 align:middle line:84% the germanium diffused right through the strained silicon 01:04:57.785 --> 01:04:58.910 align:middle line:90% all the way to the surface. 01:04:58.910 --> 01:05:01.130 align:middle line:84% And the sims profile for the blue 01:05:01.130 --> 01:05:03.170 align:middle line:84% here shows a germanium went right through that 01:05:03.170 --> 01:05:04.170 align:middle line:90% set 100 angstrom layer. 01:05:04.170 --> 01:05:06.060 align:middle line:90% It was gone. 01:05:06.060 --> 01:05:07.707 align:middle line:84% The red line is for the germanium 01:05:07.707 --> 01:05:09.290 align:middle line:84% that was annealed in an inert ambient. 01:05:09.290 --> 01:05:13.040 align:middle line:84% It diffused a little bit during this rapid thermal anneal. 01:05:13.040 --> 01:05:15.680 align:middle line:84% But ignore the surface peak because that's 01:05:15.680 --> 01:05:17.150 align:middle line:90% sort of a sims artifact. 01:05:17.150 --> 01:05:19.230 align:middle line:84% But at least it didn't go all the way. 01:05:19.230 --> 01:05:22.197 align:middle line:84% So this is a very preliminary result. 01:05:22.197 --> 01:05:23.280 align:middle line:90% And it has to be repeated. 01:05:23.280 --> 01:05:25.550 align:middle line:84% But it's sort of indicates that processes 01:05:25.550 --> 01:05:27.537 align:middle line:84% that inject vacancies probably should 01:05:27.537 --> 01:05:29.120 align:middle line:84% be avoided if you're trying to control 01:05:29.120 --> 01:05:31.520 align:middle line:90% the diffusion of germanium. 01:05:31.520 --> 01:05:34.310 align:middle line:84% And so there are a lot of interest 01:05:34.310 --> 01:05:38.090 align:middle line:84% in doing processes to inject interstitials, inject vacancies. 01:05:38.090 --> 01:05:41.450 align:middle line:84% All the same kind of tests that have done in silicon 01:05:41.450 --> 01:05:44.360 align:middle line:84% can now be done in strained silicon on silicon germanium 01:05:44.360 --> 01:05:46.387 align:middle line:84% to see how the germanium diffuses 01:05:46.387 --> 01:05:47.720 align:middle line:90% or even how the dopants diffuse. 01:05:47.720 --> 01:05:52.160 align:middle line:90% 01:05:52.160 --> 01:05:54.953 align:middle line:84% I want to say a little bit now about these MOSFETs on insulator 01:05:54.953 --> 01:05:57.620 align:middle line:84% because I've talked a little bit about how the silicon germanium 01:05:57.620 --> 01:05:58.967 align:middle line:90% is becoming a problem. 01:05:58.967 --> 01:06:00.800 align:middle line:84% How can we get around some of these problems 01:06:00.800 --> 01:06:05.160 align:middle line:84% by doing strained silicon on insulator? 01:06:05.160 --> 01:06:08.730 align:middle line:84% Well, on page 37, I'm showing an example of a potential thing 01:06:08.730 --> 01:06:09.810 align:middle line:90% you might be thinking of. 01:06:09.810 --> 01:06:13.230 align:middle line:84% Well, let's say we could get rid of-- if we go back here, 01:06:13.230 --> 01:06:19.230 align:middle line:84% let me go back a few slides, and just look at this on slide 31, 01:06:19.230 --> 01:06:21.840 align:middle line:84% this ugly looking thing with all these dislocations 01:06:21.840 --> 01:06:23.970 align:middle line:84% and that thick silicon germanium substrate. 01:06:23.970 --> 01:06:25.810 align:middle line:84% I need them to relax the strain so I 01:06:25.810 --> 01:06:27.810 align:middle line:84% can grow a thin layer of strained silica on top. 01:06:27.810 --> 01:06:30.780 align:middle line:84% But I don't really need them once the strain is relaxed. 01:06:30.780 --> 01:06:33.390 align:middle line:84% So that idea was, what if you could take this layer? 01:06:33.390 --> 01:06:35.400 align:middle line:84% The top portion looks pretty good. 01:06:35.400 --> 01:06:37.590 align:middle line:84% If you could take this whole layer and bond it 01:06:37.590 --> 01:06:41.010 align:middle line:84% to another wafer and then get rid of, etch away, 01:06:41.010 --> 01:06:43.350 align:middle line:84% all the dislocated material, then you'd 01:06:43.350 --> 01:06:46.680 align:middle line:84% have relaxed silicon germanium on insulator, for example. 01:06:46.680 --> 01:06:48.420 align:middle line:84% And all this dislocation material 01:06:48.420 --> 01:06:50.670 align:middle line:84% would have been etched off or maybe removed 01:06:50.670 --> 01:06:52.410 align:middle line:90% by a smart cut like process. 01:06:52.410 --> 01:06:54.520 align:middle line:84% I think some of you are studying smart cut. 01:06:54.520 --> 01:06:58.030 align:middle line:84% We talked about a little bit we talked about SOI formation. 01:06:58.030 --> 01:07:01.920 align:middle line:84% So, in fact, that's what this cartoon on slide 37 01:07:01.920 --> 01:07:05.050 align:middle line:84% shows where you might have a thin layer of relaxed silicon 01:07:05.050 --> 01:07:07.150 align:middle line:84% germanium that has been generated 01:07:07.150 --> 01:07:09.572 align:middle line:84% by growing a thick layer and bonding it on. 01:07:09.572 --> 01:07:11.530 align:middle line:84% And then you could grow strained silicon on top 01:07:11.530 --> 01:07:15.200 align:middle line:84% and make your nMOS and your pMOS that way. 01:07:15.200 --> 01:07:16.930 align:middle line:84% So this is one idea that people have 01:07:16.930 --> 01:07:23.470 align:middle line:84% been pursuing to get around some of those dislocation issues. 01:07:23.470 --> 01:07:25.410 align:middle line:84% In fact, on slide 38, if you're interested, 01:07:25.410 --> 01:07:27.243 align:middle line:84% I've listed a number of different references 01:07:27.243 --> 01:07:30.510 align:middle line:84% here on how people are trying to form relaxed silicon 01:07:30.510 --> 01:07:31.800 align:middle line:90% germanium on insulator. 01:07:31.800 --> 01:07:34.440 align:middle line:84% It turns out, there are a lot of different processes, not just 01:07:34.440 --> 01:07:35.550 align:middle line:90% the bonding process. 01:07:35.550 --> 01:07:38.460 align:middle line:84% There is a SIMOX-like process where 01:07:38.460 --> 01:07:41.490 align:middle line:84% people grow silicon germanium on silicon epitaxily, 01:07:41.490 --> 01:07:43.410 align:middle line:84% and then they ion implant the oxygen. 01:07:43.410 --> 01:07:45.690 align:middle line:90% That's exactly how SOI is made. 01:07:45.690 --> 01:07:47.880 align:middle line:84% So SOI, they ion implant into pure silicon. 01:07:47.880 --> 01:07:49.680 align:middle line:84% Here, they ion implant the oxygen 01:07:49.680 --> 01:07:51.270 align:middle line:90% into a silicon germanium layer. 01:07:51.270 --> 01:07:55.140 align:middle line:84% So the SIMOX-like process, Toshiba has talked about that. 01:07:55.140 --> 01:07:56.790 align:middle line:90% There are some problems though. 01:07:56.790 --> 01:07:59.580 align:middle line:84% For high germanium contents of above about 15%, 01:07:59.580 --> 01:08:01.890 align:middle line:90% it doesn't work very well. 01:08:01.890 --> 01:08:04.290 align:middle line:84% There's another very interesting technique, 01:08:04.290 --> 01:08:08.340 align:middle line:84% also came out of Toshiba, called germanium condensation. 01:08:08.340 --> 01:08:10.575 align:middle line:84% They actually start with an SOI wafer. 01:08:10.575 --> 01:08:13.230 align:middle line:84% So they start with a wafer you buy from a commercial vendor. 01:08:13.230 --> 01:08:14.760 align:middle line:84% It would have silicon on insulator. 01:08:14.760 --> 01:08:17.652 align:middle line:84% They epitaxially grow a thin layer of silicon germanium 01:08:17.652 --> 01:08:19.319 align:middle line:84% with a relatively low germanium content. 01:08:19.319 --> 01:08:21.160 align:middle line:84% And then the oxidized the whole thing. 01:08:21.160 --> 01:08:24.100 align:middle line:84% It turns out, in the oxidation, the germanium 01:08:24.100 --> 01:08:26.979 align:middle line:84% is snow plowed forward into the layer. 01:08:26.979 --> 01:08:28.367 align:middle line:90% And the layer relaxes. 01:08:28.367 --> 01:08:29.950 align:middle line:84% So they end up with a higher germanium 01:08:29.950 --> 01:08:31.600 align:middle line:90% content after oxidation. 01:08:31.600 --> 01:08:33.130 align:middle line:84% And they have written in the layer-- 01:08:33.130 --> 01:08:34.922 align:middle line:84% originally strained silicon germanium layer 01:08:34.922 --> 01:08:37.810 align:middle line:84% ends up relaxing by the oxidation process 01:08:37.810 --> 01:08:39.710 align:middle line:90% at very high temperatures. 01:08:39.710 --> 01:08:41.569 align:middle line:90% So that seems to be interesting. 01:08:41.569 --> 01:08:45.069 align:middle line:84% There's a very similar technique called melt solidification 01:08:45.069 --> 01:08:47.410 align:middle line:90% by another group. 01:08:47.410 --> 01:08:50.050 align:middle line:84% And the bond and etch back is what I mentioned before. 01:08:50.050 --> 01:08:53.649 align:middle line:84% You grow the relaxed silicon germanium, bond it 01:08:53.649 --> 01:08:56.335 align:middle line:84% to an oxide wafer, and etch it back or smart cut it off. 01:08:56.335 --> 01:08:59.140 align:middle line:90% 01:08:59.140 --> 01:09:01.255 align:middle line:84% This is just that same process on slide 39. 01:09:01.255 --> 01:09:02.560 align:middle line:90% I won't go in great detail. 01:09:02.560 --> 01:09:06.250 align:middle line:84% But you can imagine, by bonding and selective etching, 01:09:06.250 --> 01:09:08.800 align:middle line:84% you grow your graded silicon germanium relax layer, 01:09:08.800 --> 01:09:12.850 align:middle line:84% you grow your relaxed cap, and then you 01:09:12.850 --> 01:09:16.330 align:middle line:84% flip it over and bond it to an oxidized handle wafer, 01:09:16.330 --> 01:09:19.210 align:middle line:84% just like you would in a bond an etch SOI. 01:09:19.210 --> 01:09:21.160 align:middle line:84% And then you etch everything back. 01:09:21.160 --> 01:09:23.410 align:middle line:84% And you end up with just relaxed silicon germanium 01:09:23.410 --> 01:09:28.010 align:middle line:84% after you've done CMP, relaxed silicon germanium on insulator. 01:09:28.010 --> 01:09:30.080 align:middle line:84% So that's a technique that's being 01:09:30.080 --> 01:09:32.899 align:middle line:84% pursued to form relaxed silicon germanium uninsulated. 01:09:32.899 --> 01:09:36.439 align:middle line:84% Then you can grow strained silicon afterwards. 01:09:36.439 --> 01:09:39.770 align:middle line:84% You can also do bonding and hydrogen-induced delamination 01:09:39.770 --> 01:09:41.750 align:middle line:90% or smart cut. 01:09:41.750 --> 01:09:43.609 align:middle line:84% If we study smart cut for silicon, 01:09:43.609 --> 01:09:46.130 align:middle line:84% here, we start with a relaxed silicon germanium epi layer. 01:09:46.130 --> 01:09:49.069 align:middle line:84% We ion implant with hydrogen. And at the peak of the hydrogen, 01:09:49.069 --> 01:09:51.899 align:middle line:90% bubbles form. 01:09:51.899 --> 01:09:55.190 align:middle line:84% So after we've done the bonding, we heat it up. 01:09:55.190 --> 01:09:57.920 align:middle line:84% And those bubbles end up cleaving it off. 01:09:57.920 --> 01:09:59.510 align:middle line:84% You get the delamination process. 01:09:59.510 --> 01:10:02.370 align:middle line:84% This is exactly analogous to what happens in SOI fab. 01:10:02.370 --> 01:10:04.370 align:middle line:84% And now you have silicon germanium on insulator. 01:10:04.370 --> 01:10:06.800 align:middle line:84% Of course, it has to be polished and smoothed. 01:10:06.800 --> 01:10:09.200 align:middle line:84% But it is a way to form that material. 01:10:09.200 --> 01:10:11.888 align:middle line:90% 01:10:11.888 --> 01:10:14.430 align:middle line:84% I'll skip through this at this point because we're basically, 01:10:14.430 --> 01:10:16.972 align:middle line:84% it shows, when you make that, and you make a strained silicon 01:10:16.972 --> 01:10:18.750 align:middle line:84% MOSFET, you get the same mobility 01:10:18.750 --> 01:10:22.380 align:middle line:84% that people have seen in the past. 01:10:22.380 --> 01:10:24.073 align:middle line:84% Again, I'll skip that, interest of time. 01:10:24.073 --> 01:10:26.490 align:middle line:84% And what I want to end up with is there's another approach 01:10:26.490 --> 01:10:28.890 align:middle line:84% people thought, well, why don't you 01:10:28.890 --> 01:10:31.890 align:middle line:84% just transfer the strained silicon layer itself? 01:10:31.890 --> 01:10:33.510 align:middle line:90% Why do we even have to transfer? 01:10:33.510 --> 01:10:35.790 align:middle line:84% Once you grow the relaxed silicon germanium, 01:10:35.790 --> 01:10:38.460 align:middle line:84% you can grow strained silicon on top of it. 01:10:38.460 --> 01:10:41.310 align:middle line:84% Bond that to oxide, and then remove everything 01:10:41.310 --> 01:10:43.090 align:middle line:90% except the strained silicon. 01:10:43.090 --> 01:10:45.600 align:middle line:84% And if the strain remains, then you 01:10:45.600 --> 01:10:47.970 align:middle line:84% have strained silicon directly on the insulator. 01:10:47.970 --> 01:10:49.860 align:middle line:84% You remove the silicon germanium completely. 01:10:49.860 --> 01:10:52.360 align:middle line:84% So you don't have to worry about silicon germanium diffusion 01:10:52.360 --> 01:10:54.640 align:middle line:84% and diffusion of dope into silicon germanium and all 01:10:54.640 --> 01:10:55.180 align:middle line:90% of that. 01:10:55.180 --> 01:10:57.840 align:middle line:84% So this is an ultra thin strained silicon directly 01:10:57.840 --> 01:10:59.520 align:middle line:90% on insulator sort of concept. 01:10:59.520 --> 01:11:01.530 align:middle line:90% This is just a cartoon. 01:11:01.530 --> 01:11:06.170 align:middle line:84% It's actually been demonstrated fairly recently here at MIT 01:11:06.170 --> 01:11:07.270 align:middle line:90% on page 44. 01:11:07.270 --> 01:11:09.200 align:middle line:84% In fact, this is the process that was used. 01:11:09.200 --> 01:11:12.660 align:middle line:84% A very complicated stack of epi layers was grown. 01:11:12.660 --> 01:11:14.210 align:middle line:84% But the important thing is the top 01:11:14.210 --> 01:11:16.100 align:middle line:84% of the stack was strained silicon. 01:11:16.100 --> 01:11:19.910 align:middle line:84% And that whole stack was bonded to an oxidized silicon wafer. 01:11:19.910 --> 01:11:22.350 align:middle line:84% And then everything was removed by selective etching. 01:11:22.350 --> 01:11:24.290 align:middle line:84% So when you're done, all you have left 01:11:24.290 --> 01:11:27.200 align:middle line:84% is strained silicon directly on the oxide. 01:11:27.200 --> 01:11:30.020 align:middle line:84% And you use etch top layers, so you 01:11:30.020 --> 01:11:32.120 align:middle line:84% can etch back the entire wafer and just 01:11:32.120 --> 01:11:35.230 align:middle line:90% leave strained silicon behind. 01:11:35.230 --> 01:11:37.360 align:middle line:84% And in fact, on page 45, what you see 01:11:37.360 --> 01:11:41.290 align:middle line:84% is that the strained silicon is bonded to the oxide. 01:11:41.290 --> 01:11:43.280 align:middle line:84% It's low defect density material. 01:11:43.280 --> 01:11:45.820 align:middle line:84% And if you do Raman scattering, the strain remains. 01:11:45.820 --> 01:11:48.160 align:middle line:84% Even after you remove the silicon germanium, 01:11:48.160 --> 01:11:50.230 align:middle line:84% you annealing at high temperatures, 01:11:50.230 --> 01:11:52.670 align:middle line:84% even up to 1,000 degrees, the strain doesn't go away. 01:11:52.670 --> 01:11:54.320 align:middle line:90% So the bond is very strong. 01:11:54.320 --> 01:11:57.040 align:middle line:84% So now, you have strained silicon directly on insulator. 01:11:57.040 --> 01:11:59.800 align:middle line:84% Looks just like SOI, you can process it just like SOI. 01:11:59.800 --> 01:12:01.310 align:middle line:84% But the strain is already built in. 01:12:01.310 --> 01:12:05.438 align:middle line:84% You don't have any germanium that you need to worry about. 01:12:05.438 --> 01:12:07.730 align:middle line:84% So I think that's going to be a pretty interesting area 01:12:07.730 --> 01:12:08.960 align:middle line:90% for the future. 01:12:08.960 --> 01:12:10.730 align:middle line:84% Just yesterday, I read in the paper 01:12:10.730 --> 01:12:14.030 align:middle line:84% that it had been announced that a company called Soitec is now 01:12:14.030 --> 01:12:16.700 align:middle line:84% manufacturing 300-millimeter strained silicon directly 01:12:16.700 --> 01:12:18.590 align:middle line:90% on insulator SSDOI. 01:12:18.590 --> 01:12:21.350 align:middle line:84% And once it becomes available 300 millimeter, 01:12:21.350 --> 01:12:25.350 align:middle line:84% a lot of companies will be able to use it for their production. 01:12:25.350 --> 01:12:27.680 align:middle line:84% So that's about all I have to say on that. 01:12:27.680 --> 01:12:29.930 align:middle line:90% I know I need to leave you time. 01:12:29.930 --> 01:12:31.623 align:middle line:90% I'm glad I see the surveys here. 01:12:31.623 --> 01:12:33.290 align:middle line:84% I hope you haven't been waiting outside. 01:12:33.290 --> 01:12:37.730 align:middle line:84% But please take 10 minutes or so and go ahead and fill out 01:12:37.730 --> 01:12:41.440 align:middle line:90% the course evaluation survey. 01:12:41.440 --> 01:12:43.000 align:middle line:90%