WEBVTT 00:00:00.000 --> 00:00:06.440 align:middle line:84% JUDY HOYT: Point and start with our actual lecture. 00:00:06.440 --> 00:00:12.170 align:middle line:84% So we are in the middle of talking about chapter 4. 00:00:12.170 --> 00:00:15.048 align:middle line:84% Again, I hope you're keeping up with the reading in the text. 00:00:15.048 --> 00:00:17.090 align:middle line:84% That's part of the reason I don't assign homework 00:00:17.090 --> 00:00:18.740 align:middle line:90% continuously. 00:00:18.740 --> 00:00:20.990 align:middle line:84% Between now and Thursday, you don't have any homework. 00:00:20.990 --> 00:00:24.860 align:middle line:84% You can have time to be reading the text. 00:00:24.860 --> 00:00:27.350 align:middle line:90% Chapter 4, we're talking-- 00:00:27.350 --> 00:00:29.990 align:middle line:84% discusses wafer cleaning and gettering. 00:00:29.990 --> 00:00:34.730 align:middle line:84% Last time, we talked about some important aspects. 00:00:34.730 --> 00:00:37.460 align:middle line:84% We talked about point defects in silicon, 00:00:37.460 --> 00:00:41.060 align:middle line:84% neutral and charged vacancies and interstitials. 00:00:41.060 --> 00:00:43.580 align:middle line:84% Other important properties of silicon wafers we discussed 00:00:43.580 --> 00:00:46.430 align:middle line:84% were the carbon content and the oxygen content. 00:00:46.430 --> 00:00:48.920 align:middle line:84% We're going to talk more today about oxygen and Czochralski 00:00:48.920 --> 00:00:49.670 align:middle line:90% silicon. 00:00:49.670 --> 00:00:52.400 align:middle line:84% And we got a little bit of a taste last time 00:00:52.400 --> 00:00:56.390 align:middle line:84% of some qualitative information on clean rooms and how-- 00:00:56.390 --> 00:00:59.970 align:middle line:84% and that's-- we call that the level 1 introduction 00:00:59.970 --> 00:01:03.180 align:middle line:84% to controlling contamination on wafers. 00:01:03.180 --> 00:01:04.943 align:middle line:84% And these notes, we're going to talk 00:01:04.943 --> 00:01:07.110 align:middle line:84% about the two other major techniques for controlling 00:01:07.110 --> 00:01:09.900 align:middle line:84% contamination, which are to clean the wafers 00:01:09.900 --> 00:01:11.160 align:middle line:90% throughout the process. 00:01:11.160 --> 00:01:13.350 align:middle line:84% Every few steps, we do a wafer cleaning. 00:01:13.350 --> 00:01:15.070 align:middle line:84% And we'll talk about how that works. 00:01:15.070 --> 00:01:17.760 align:middle line:84% And then finally, the principles and modeling of a process 00:01:17.760 --> 00:01:19.770 align:middle line:90% called impurity gettering. 00:01:19.770 --> 00:01:23.990 align:middle line:84% So that's what this handout is about. 00:01:23.990 --> 00:01:27.140 align:middle line:84% Let's go on to slide number 2 of this handout. 00:01:27.140 --> 00:01:30.980 align:middle line:84% And it shows schematically this level 2 approach 00:01:30.980 --> 00:01:32.480 align:middle line:90% of reducing contamination. 00:01:32.480 --> 00:01:33.440 align:middle line:90% And that's to do-- 00:01:33.440 --> 00:01:34.580 align:middle line:90% wafer cleaning. 00:01:34.580 --> 00:01:38.552 align:middle line:84% And there are two alternate paths here 00:01:38.552 --> 00:01:40.010 align:middle line:84% indicated on the left and the right 00:01:40.010 --> 00:01:43.160 align:middle line:84% that we might take a silicon wafer through. 00:01:43.160 --> 00:01:44.960 align:middle line:84% The path on the left is primarily 00:01:44.960 --> 00:01:46.940 align:middle line:84% what we'll be talking about in this course 00:01:46.940 --> 00:01:51.660 align:middle line:84% because it pertains to the front end of the process. 00:01:51.660 --> 00:01:53.220 align:middle line:84% In the front end, by definition, you 00:01:53.220 --> 00:01:55.470 align:middle line:84% have no-- you have not done any metallization 00:01:55.470 --> 00:01:57.510 align:middle line:84% or have not put any metal contacts on the wafer. 00:01:57.510 --> 00:01:58.885 align:middle line:84% That's what we mean by front end. 00:01:58.885 --> 00:02:00.480 align:middle line:84% And that's what this course is about. 00:02:00.480 --> 00:02:02.790 align:middle line:84% So typically, throughout a process, 00:02:02.790 --> 00:02:06.600 align:middle line:84% we talked about there might be 16 or 20 mask levels. 00:02:06.600 --> 00:02:09.990 align:middle line:84% Each time you do a masking step and you do photolithography, 00:02:09.990 --> 00:02:13.270 align:middle line:84% you have this organic gloop, this photoresist, on your wafer. 00:02:13.270 --> 00:02:14.250 align:middle line:90% It needs to be removed. 00:02:14.250 --> 00:02:17.010 align:middle line:84% So before you can go on to the next process step, 00:02:17.010 --> 00:02:19.650 align:middle line:84% you need to remove that organic material, that resist. 00:02:19.650 --> 00:02:22.470 align:middle line:84% And we do that in what's called a resist strip. 00:02:22.470 --> 00:02:23.520 align:middle line:90% That's if necessary. 00:02:23.520 --> 00:02:25.780 align:middle line:84% That is if there's photoresist on the wafer. 00:02:25.780 --> 00:02:30.270 align:middle line:84% And then after you remove this bulk gross macroscopic organic 00:02:30.270 --> 00:02:33.240 align:middle line:84% contamination, then trace organics and trace metals 00:02:33.240 --> 00:02:36.060 align:middle line:84% are removed in a step called the RCA clean. 00:02:36.060 --> 00:02:37.950 align:middle line:84% And that's named for where it was invented, 00:02:37.950 --> 00:02:41.820 align:middle line:84% the old RCA Sarnoff labs in New Jersey. 00:02:41.820 --> 00:02:43.680 align:middle line:84% That's where that comes from, the RCA clean. 00:02:43.680 --> 00:02:46.240 align:middle line:84% And then you execute the next process step. 00:02:46.240 --> 00:02:48.030 align:middle line:84% So typically, the RCA clean would 00:02:48.030 --> 00:02:51.400 align:middle line:84% precede any high-temperature processing step 00:02:51.400 --> 00:02:53.800 align:middle line:84% that we'll talk about in this course, such as a diffusion 00:02:53.800 --> 00:02:57.040 align:middle line:84% step in a furnace or an oxidation step. 00:02:57.040 --> 00:03:00.040 align:middle line:84% You would have to do this RCA cleaning. 00:03:00.040 --> 00:03:02.800 align:middle line:84% By contrast, if you're in the back end, 00:03:02.800 --> 00:03:04.690 align:middle line:90% something similar happens. 00:03:04.690 --> 00:03:07.060 align:middle line:84% You do have to strip the photoresist. 00:03:07.060 --> 00:03:10.030 align:middle line:84% And you do have to do some kind of post-metal cleaning, 00:03:10.030 --> 00:03:12.520 align:middle line:90% usually before another step. 00:03:12.520 --> 00:03:14.230 align:middle line:84% But the solutions that you would use 00:03:14.230 --> 00:03:16.450 align:middle line:84% are much less caustic than the ones we're 00:03:16.450 --> 00:03:19.925 align:middle line:84% going to talk about here in the acidic and basic solutions 00:03:19.925 --> 00:03:22.300 align:middle line:84% because metals can-- you don't want to etch all the metal 00:03:22.300 --> 00:03:24.083 align:middle line:90% wiring off your wafer. 00:03:24.083 --> 00:03:25.750 align:middle line:84% So the difference with backend cleaning, 00:03:25.750 --> 00:03:27.082 align:middle line:90% it's generally more gentle. 00:03:27.082 --> 00:03:29.290 align:middle line:84% And it's not designed to remove metals because that's 00:03:29.290 --> 00:03:31.880 align:middle line:90% what you have in the back end. 00:03:31.880 --> 00:03:34.660 align:middle line:90% You have metal wiring. 00:03:34.660 --> 00:03:37.120 align:middle line:84% So let's go on to slide number 3. 00:03:37.120 --> 00:03:40.310 align:middle line:84% And this, I've taken out of your text. 00:03:40.310 --> 00:03:42.610 align:middle line:84% It talks about these-- this front end cleaning 00:03:42.610 --> 00:03:45.310 align:middle line:90% process and the steps involved. 00:03:45.310 --> 00:03:48.750 align:middle line:84% So if we start at the top, this section up here, 00:03:48.750 --> 00:03:52.230 align:middle line:84% these top few steps illustrate photoresist stripping. 00:03:52.230 --> 00:03:54.330 align:middle line:84% Again, every time you do a masking level, 00:03:54.330 --> 00:03:56.340 align:middle line:84% you typically have this organic film 00:03:56.340 --> 00:04:01.140 align:middle line:84% that's somewhere in the order of several microns of this material 00:04:01.140 --> 00:04:02.250 align:middle line:90% that needs to be stripped. 00:04:02.250 --> 00:04:03.840 align:middle line:90% And it's fairly gross. 00:04:03.840 --> 00:04:05.310 align:middle line:90% It's quite thick. 00:04:05.310 --> 00:04:06.570 align:middle line:90% It's macroscopic material. 00:04:06.570 --> 00:04:09.053 align:middle line:84% So it can be stripped in a number of ways. 00:04:09.053 --> 00:04:10.470 align:middle line:84% This is a very common way of doing 00:04:10.470 --> 00:04:14.220 align:middle line:84% it is to take hot, very hot, mixture of sulfuric and hydrogen 00:04:14.220 --> 00:04:14.730 align:middle line:90% peroxide. 00:04:14.730 --> 00:04:16.769 align:middle line:84% That's a very, very strong oxidizer. 00:04:16.769 --> 00:04:22.140 align:middle line:84% It really attacks organics, especially photoresist. 00:04:22.140 --> 00:04:26.400 align:middle line:84% Sometimes in fabs today, prior to doing 00:04:26.400 --> 00:04:28.200 align:middle line:84% the actual wet chemistry to remove 00:04:28.200 --> 00:04:30.720 align:middle line:84% the photoresist, sometimes people do a dry process. 00:04:30.720 --> 00:04:33.960 align:middle line:84% And we'll talk later in this course on dry etching. 00:04:33.960 --> 00:04:36.930 align:middle line:84% But there are ways to create in the gas phase 00:04:36.930 --> 00:04:40.320 align:middle line:84% reactive species that will also attack the resist and remove it. 00:04:40.320 --> 00:04:44.950 align:middle line:84% And an oxygen plasma is often a way 00:04:44.950 --> 00:04:47.320 align:middle line:84% that people use to remove photoresist, sometimes 00:04:47.320 --> 00:04:50.740 align:middle line:84% prior to doing the wet chemical strip. 00:04:50.740 --> 00:04:54.490 align:middle line:84% So after you do this sulfuric peroxide step, 00:04:54.490 --> 00:04:57.730 align:middle line:84% there's typically a water rinse, which 00:04:57.730 --> 00:04:59.890 align:middle line:90% is shown in here as implicit. 00:04:59.890 --> 00:05:04.900 align:middle line:84% And then you dip the wafers in a mixture of water and HF. 00:05:04.900 --> 00:05:09.250 align:middle line:84% HF is an etchant that etches oxide, which is 00:05:09.250 --> 00:05:10.910 align:middle line:90% all forms of silicon dioxide. 00:05:10.910 --> 00:05:13.360 align:middle line:84% So in this step, there's a strong oxidizer. 00:05:13.360 --> 00:05:15.550 align:middle line:84% You end up forming a thin chemical oxide 00:05:15.550 --> 00:05:17.380 align:middle line:90% on the silicon surface. 00:05:17.380 --> 00:05:19.880 align:middle line:90% This strips it off. 00:05:19.880 --> 00:05:21.310 align:middle line:84% And then you do a final DI rinse. 00:05:21.310 --> 00:05:25.450 align:middle line:84% So at that point, you have all the gross macroscopic 00:05:25.450 --> 00:05:27.400 align:middle line:84% photoresist film removed from your wafer. 00:05:27.400 --> 00:05:30.820 align:middle line:84% You still have trace impurities, trace organics, 00:05:30.820 --> 00:05:33.160 align:middle line:84% and trace metals that have gotten onto the wafer 00:05:33.160 --> 00:05:37.670 align:middle line:84% during the process step that you just did prior to this. 00:05:37.670 --> 00:05:41.080 align:middle line:84% So in order to remove those trace impurities, 00:05:41.080 --> 00:05:43.175 align:middle line:84% we do what's called the RCA clean. 00:05:43.175 --> 00:05:45.550 align:middle line:84% There are a lot of different variations on the RCA clean. 00:05:45.550 --> 00:05:47.770 align:middle line:84% But this is-- the one I'm outlining here 00:05:47.770 --> 00:05:50.110 align:middle line:90% is the most classic RCA clean. 00:05:50.110 --> 00:05:53.470 align:middle line:84% It has two steps, typically, major steps, one called 00:05:53.470 --> 00:05:58.450 align:middle line:84% silicon clean 1, SC-1, and one called silicon clean 2. 00:05:58.450 --> 00:06:02.080 align:middle line:84% The major purpose of SC-1 is to strip organics. 00:06:02.080 --> 00:06:06.100 align:middle line:84% It does attack a certain amount of metals, and it also-- 00:06:06.100 --> 00:06:07.220 align:middle line:90% to remove particles. 00:06:07.220 --> 00:06:10.930 align:middle line:84% So it's good at etching off particles. 00:06:10.930 --> 00:06:14.860 align:middle line:84% And SC-1 itself consists of a solution of 511. 00:06:14.860 --> 00:06:20.770 align:middle line:84% So this 511 represents the concentration ratio of water 00:06:20.770 --> 00:06:24.370 align:middle line:84% to hydrogen peroxide and ammonium hydroxide. 00:06:24.370 --> 00:06:27.040 align:middle line:84% And it's usually heated to this temperature range, about 80 00:06:27.040 --> 00:06:29.890 align:middle line:90% degrees C for about 10 minutes. 00:06:29.890 --> 00:06:32.200 align:middle line:84% After that SC-1, after that treatment, 00:06:32.200 --> 00:06:34.465 align:middle line:84% then it goes through a deionized water rinse. 00:06:34.465 --> 00:06:35.590 align:middle line:90% Again, we talked last time. 00:06:35.590 --> 00:06:37.870 align:middle line:84% You have to use very pure chemicals, 00:06:37.870 --> 00:06:40.413 align:middle line:84% a water rinse at room temperature and then very often, 00:06:40.413 --> 00:06:42.580 align:middle line:84% not always, but very often, right in this step here, 00:06:42.580 --> 00:06:45.323 align:middle line:84% in between SC-1 and SC-2, there's an HF dip. 00:06:45.323 --> 00:06:47.740 align:middle line:84% Again, the purpose of that is to remove the chemical oxide 00:06:47.740 --> 00:06:50.650 align:middle line:84% that you grew in this step to create a fresh silicon 00:06:50.650 --> 00:06:53.020 align:middle line:90% surface and then a DI rinse. 00:06:53.020 --> 00:06:54.610 align:middle line:84% It wasn't indicated in your text. 00:06:54.610 --> 00:06:56.620 align:middle line:90% So I don't-- I put a box for it. 00:06:56.620 --> 00:07:00.460 align:middle line:84% But it is very common to have HF dip in the middle there. 00:07:00.460 --> 00:07:03.220 align:middle line:90% And then you go to SC-2. 00:07:03.220 --> 00:07:08.440 align:middle line:84% This is-- SC-1 was an alkaline solution, like a basic solution. 00:07:08.440 --> 00:07:10.480 align:middle line:90% SC-2 is low pH. 00:07:10.480 --> 00:07:11.350 align:middle line:90% It's acidic. 00:07:11.350 --> 00:07:13.480 align:middle line:84% And it consists of 611 or something 00:07:13.480 --> 00:07:16.610 align:middle line:84% in that ratio, water peroxide to hydrogen chloride. 00:07:16.610 --> 00:07:19.180 align:middle line:90% So it's very strong acid. 00:07:19.180 --> 00:07:21.400 align:middle line:84% And the key for SC-2, its main role 00:07:21.400 --> 00:07:25.450 align:middle line:84% is to remove any other remaining trace metals that were not 00:07:25.450 --> 00:07:26.920 align:middle line:90% effectively removed by SC-1. 00:07:26.920 --> 00:07:28.180 align:middle line:84% And there are some metals that don't come off 00:07:28.180 --> 00:07:29.230 align:middle line:90% very well in SC-1. 00:07:29.230 --> 00:07:31.090 align:middle line:90% And also alkali ions. 00:07:31.090 --> 00:07:32.270 align:middle line:90% We'll talk about those. 00:07:32.270 --> 00:07:35.350 align:middle line:84% But the mobile ions, sodium, potassium, calcium, 00:07:35.350 --> 00:07:38.360 align:middle line:84% are very effectively removed in the SC-2 00:07:38.360 --> 00:07:41.160 align:middle line:84% and not so effectively removed in SC-1. 00:07:41.160 --> 00:07:43.620 align:middle line:84% And then there's a DI water rinse. 00:07:43.620 --> 00:07:46.080 align:middle line:84% Sometimes there's a final HI dip right 00:07:46.080 --> 00:07:48.660 align:middle line:84% at the end, where you remove the Chemox. 00:07:48.660 --> 00:07:51.630 align:middle line:84% So you create a pure silicon surface. 00:07:51.630 --> 00:07:54.120 align:middle line:84% Sometimes there's not, especially 00:07:54.120 --> 00:07:56.190 align:middle line:84% if the next step-- if the wafer is going directly 00:07:56.190 --> 00:07:58.423 align:middle line:84% into an oxidation furnace, people 00:07:58.423 --> 00:08:00.840 align:middle line:84% leave the surface passivated with a little bit of chemical 00:08:00.840 --> 00:08:02.250 align:middle line:90% oxide on it anyway. 00:08:02.250 --> 00:08:06.570 align:middle line:84% So again, this is very much a rough estimate 00:08:06.570 --> 00:08:07.320 align:middle line:90% of what people do. 00:08:07.320 --> 00:08:11.570 align:middle line:84% There are a lot of variations on this theme. 00:08:11.570 --> 00:08:14.750 align:middle line:84% Let's move on to slide number 4 and talk a little bit more 00:08:14.750 --> 00:08:16.130 align:middle line:90% about the chemistry. 00:08:16.130 --> 00:08:18.180 align:middle line:84% There's more details about this in your text. 00:08:18.180 --> 00:08:21.470 align:middle line:84% So if you want to read about that, that will be helpful. 00:08:21.470 --> 00:08:23.990 align:middle line:84% An important thing we need to do in these solutions 00:08:23.990 --> 00:08:27.170 align:middle line:84% is to remove metal atoms from the surface of the wafer. 00:08:27.170 --> 00:08:29.165 align:middle line:84% Again, trace metals are very bad. 00:08:29.165 --> 00:08:30.290 align:middle line:90% We'll talk more about this. 00:08:30.290 --> 00:08:33.080 align:middle line:84% We saw last time they can reduce the lifetime. 00:08:33.080 --> 00:08:35.929 align:middle line:84% They can cause all kinds of problems for logic and memory 00:08:35.929 --> 00:08:37.380 align:middle line:90% devices. 00:08:37.380 --> 00:08:39.590 align:middle line:84% So the way that we remove them from the surface 00:08:39.590 --> 00:08:44.030 align:middle line:84% is to convert them into ions so that they're no longer neutral. 00:08:44.030 --> 00:08:46.660 align:middle line:84% And these ions end up being soluble in the cleaning 00:08:46.660 --> 00:08:47.160 align:middle line:90% solution. 00:08:47.160 --> 00:08:49.910 align:middle line:84% So they dissolve into the wet solution. 00:08:49.910 --> 00:08:53.960 align:middle line:84% And they stay in the solution instead of on your wafer. 00:08:53.960 --> 00:08:57.590 align:middle line:84% So to do this, we need to oxidize the metal atoms. 00:08:57.590 --> 00:08:59.090 align:middle line:84% That is, we're going to use the term 00:08:59.090 --> 00:09:02.030 align:middle line:84% oxidation as referring to removing electrons 00:09:02.030 --> 00:09:03.480 align:middle line:90% from the metal atoms. 00:09:03.480 --> 00:09:06.440 align:middle line:84% So this top equation here, shown on this line, 00:09:06.440 --> 00:09:09.590 align:middle line:84% shows that the oxidation process of silicon, 00:09:09.590 --> 00:09:12.000 align:middle line:84% taking silicon atoms at the surface, 00:09:12.000 --> 00:09:15.090 align:middle line:84% combining them with an oxidizer, such as water, 00:09:15.090 --> 00:09:18.990 align:middle line:84% and forming SiO2, plus evolving some hydrogen 00:09:18.990 --> 00:09:20.230 align:middle line:90% ions and some electrons. 00:09:20.230 --> 00:09:21.990 align:middle line:84% So this is an oxidation reaction. 00:09:21.990 --> 00:09:25.290 align:middle line:90% It ends up removing electrons. 00:09:25.290 --> 00:09:27.653 align:middle line:84% Similarly, this M here, the second equation, 00:09:27.653 --> 00:09:29.320 align:middle line:84% M is meant to represent a generic metal. 00:09:29.320 --> 00:09:30.028 align:middle line:90% It could be iron. 00:09:30.028 --> 00:09:31.360 align:middle line:90% It could be copper, whatever. 00:09:31.360 --> 00:09:33.150 align:middle line:90% Just say iron. 00:09:33.150 --> 00:09:35.310 align:middle line:84% By an oxidation reaction can have 00:09:35.310 --> 00:09:37.110 align:middle line:84% its electrons removed or stripped 00:09:37.110 --> 00:09:42.570 align:middle line:84% and form a positive ion, along with some electrons 00:09:42.570 --> 00:09:45.063 align:middle line:90% in the solution. 00:09:45.063 --> 00:09:46.480 align:middle line:84% And if you read through your text, 00:09:46.480 --> 00:09:49.120 align:middle line:84% this is a table I took directly from your text. 00:09:49.120 --> 00:09:52.980 align:middle line:84% This is a table of oxidation reduction reactions 00:09:52.980 --> 00:09:54.885 align:middle line:84% for various elements, say some metals. 00:09:54.885 --> 00:09:57.150 align:middle line:90% It has silicon dioxide itself. 00:09:57.150 --> 00:10:00.210 align:middle line:90% Here's iron and chromium. 00:10:00.210 --> 00:10:04.440 align:middle line:84% And what's shown here is the standard oxidation potential 00:10:04.440 --> 00:10:07.410 align:middle line:84% in volts and the oxidation reduction 00:10:07.410 --> 00:10:11.320 align:middle line:84% reaction, the actual chemical reaction, in the third column. 00:10:11.320 --> 00:10:15.360 align:middle line:84% So example, the second row here shows the oxidation of silicon 00:10:15.360 --> 00:10:17.280 align:middle line:90% to form SiO2. 00:10:17.280 --> 00:10:21.360 align:middle line:84% This row here with the iron shows the oxidation of iron 00:10:21.360 --> 00:10:26.610 align:middle line:84% to form Fe3 plus, a positive to remove electrons. 00:10:26.610 --> 00:10:29.610 align:middle line:84% In general, a stronger oxidant, something 00:10:29.610 --> 00:10:34.080 align:middle line:84% that's better at stripping off electrons from an atom, 00:10:34.080 --> 00:10:36.130 align:middle line:84% has a more negative oxidation potential. 00:10:36.130 --> 00:10:39.118 align:middle line:84% So as you go down in this table, you're 00:10:39.118 --> 00:10:41.410 align:middle line:84% getting to more and more negative oxidation potentials. 00:10:41.410 --> 00:10:44.910 align:middle line:84% These are very strong oxidants, oxidizers. 00:10:44.910 --> 00:10:50.010 align:middle line:84% So H2O2, hydrogen peroxide, is a very strong oxidizer. 00:10:50.010 --> 00:10:54.100 align:middle line:84% Ozone, O3, is a very strong oxidizer. 00:10:54.100 --> 00:10:56.640 align:middle line:84% So the way to interpret this table 00:10:56.640 --> 00:10:59.760 align:middle line:84% is we're going from fairly weak oxidizers 00:10:59.760 --> 00:11:02.430 align:middle line:84% to very strong oxidizers as we go to the bottom. 00:11:02.430 --> 00:11:06.540 align:middle line:84% In general, if you have a lot of these reactions taking place 00:11:06.540 --> 00:11:08.580 align:middle line:84% on the wafer or in the solution at once, 00:11:08.580 --> 00:11:11.430 align:middle line:84% the lowest reaction in the table is going to dominate. 00:11:11.430 --> 00:11:14.320 align:middle line:84% And it's going to tend to go towards the left. 00:11:14.320 --> 00:11:16.230 align:middle line:84% So for example, say I have a solution that 00:11:16.230 --> 00:11:19.110 align:middle line:90% contains hydrogen peroxide. 00:11:19.110 --> 00:11:21.240 align:middle line:84% It's quite down in the-- far on the table. 00:11:21.240 --> 00:11:22.830 align:middle line:84% It's going to tend to be pushed-- 00:11:22.830 --> 00:11:24.000 align:middle line:90% it's a strong oxidizer. 00:11:24.000 --> 00:11:26.710 align:middle line:84% It's going to tend to be pushed to the left. 00:11:26.710 --> 00:11:31.570 align:middle line:84% And that tends to drive all the reactions above it 00:11:31.570 --> 00:11:35.840 align:middle line:84% to the right, basically, just as a rule of thumb. 00:11:35.840 --> 00:11:38.090 align:middle line:84% So for example, hydrogen peroxide, 00:11:38.090 --> 00:11:43.210 align:middle line:84% which is well below iron, tend to take electrons from iron, 00:11:43.210 --> 00:11:48.520 align:middle line:84% forming this Fe3 plus ions, and create ions that 00:11:48.520 --> 00:11:49.790 align:middle line:90% are soluble in the solution. 00:11:49.790 --> 00:11:52.150 align:middle line:84% So it's an effective way of removing-- 00:11:52.150 --> 00:11:55.030 align:middle line:84% of ionizing iron and removing it from the wafer. 00:11:55.030 --> 00:11:58.870 align:middle line:84% At the same time as this happens, this is happening, 00:11:58.870 --> 00:12:01.960 align:middle line:84% the silicon itself will be oxidized by the H2O2, 00:12:01.960 --> 00:12:03.200 align:middle line:90% by the peroxide. 00:12:03.200 --> 00:12:07.840 align:middle line:84% So we're forming these ions on the wafer, these metallic ions 00:12:07.840 --> 00:12:09.040 align:middle line:90% that can then dissolve. 00:12:09.040 --> 00:12:10.498 align:middle line:84% And at the same time, we're growing 00:12:10.498 --> 00:12:13.330 align:middle line:84% a little bit of SiO2, a very thin layer of this silicon 00:12:13.330 --> 00:12:13.930 align:middle line:90% dioxide. 00:12:13.930 --> 00:12:15.820 align:middle line:84% And that gets stripped, that silicon dioxide, 00:12:15.820 --> 00:12:16.930 align:middle line:90% in the HF strip-- 00:12:16.930 --> 00:12:18.800 align:middle line:90% the HF dip, of a next step. 00:12:18.800 --> 00:12:21.220 align:middle line:84% So that's some idea of the chemistry. 00:12:21.220 --> 00:12:24.220 align:middle line:84% Again, there are-- your book has a reference to chemistry texts 00:12:24.220 --> 00:12:25.550 align:middle line:90% if you want to read that. 00:12:25.550 --> 00:12:31.600 align:middle line:84% That's involved in some of this cleaning, cleaning technology. 00:12:31.600 --> 00:12:32.920 align:middle line:90% Let's go on to page number 5. 00:12:32.920 --> 00:12:34.390 align:middle line:90% Now this is an actual-- 00:12:34.390 --> 00:12:36.670 align:middle line:84% some actual data I pulled out of-- 00:12:36.670 --> 00:12:40.060 align:middle line:90% some laboratory data on slide 5. 00:12:40.060 --> 00:12:41.680 align:middle line:84% And what it's showing, it's actually 00:12:41.680 --> 00:12:43.780 align:middle line:84% taken from a wafer mapping tool made 00:12:43.780 --> 00:12:45.490 align:middle line:90% by a company called KLA-Tencor. 00:12:45.490 --> 00:12:46.780 align:middle line:90% It's a picture. 00:12:46.780 --> 00:12:49.598 align:middle line:84% It's a face-on picture of a wafer. 00:12:49.598 --> 00:12:50.890 align:middle line:90% Up here, it's a six-inch wafer. 00:12:50.890 --> 00:12:52.430 align:middle line:90% Here's a flat on top. 00:12:52.430 --> 00:12:53.680 align:middle line:90% It didn't reproduce very well. 00:12:53.680 --> 00:12:54.010 align:middle line:90% I'm sorry. 00:12:54.010 --> 00:12:55.420 align:middle line:84% So this is the flat of the wafer. 00:12:55.420 --> 00:12:58.640 align:middle line:84% This is the six-inch diameter edge of the wafer. 00:12:58.640 --> 00:13:00.610 align:middle line:84% And this is a machine that takes a laser 00:13:00.610 --> 00:13:02.660 align:middle line:84% and scans it across the surface of the wafer. 00:13:02.660 --> 00:13:06.520 align:middle line:84% And it looks for spurious reflections. 00:13:06.520 --> 00:13:09.220 align:middle line:84% Every time you hit a particle, you no longer 00:13:09.220 --> 00:13:10.840 align:middle line:90% have specular reflection. 00:13:10.840 --> 00:13:11.940 align:middle line:90% And it notes that. 00:13:11.940 --> 00:13:13.690 align:middle line:84% And it says, aha, there must be a particle 00:13:13.690 --> 00:13:16.960 align:middle line:84% or what they call a light point defect, LPD. 00:13:16.960 --> 00:13:19.242 align:middle line:84% And this is to prevent having to have people 00:13:19.242 --> 00:13:20.950 align:middle line:84% sit in the clean room and literally count 00:13:20.950 --> 00:13:21.730 align:middle line:90% the particles on a wafer. 00:13:21.730 --> 00:13:23.438 align:middle line:84% That would be too tedious and inaccurate. 00:13:23.438 --> 00:13:25.750 align:middle line:84% So it's all automated these days. 00:13:25.750 --> 00:13:28.480 align:middle line:84% And it even plots out, the computer plots for you, 00:13:28.480 --> 00:13:29.630 align:middle line:90% the point on the wafer. 00:13:29.630 --> 00:13:31.390 align:middle line:84% These little black points are places 00:13:31.390 --> 00:13:33.190 align:middle line:84% where it found light point defects, where 00:13:33.190 --> 00:13:35.500 align:middle line:84% it found nonspecular reflection, something 00:13:35.500 --> 00:13:37.030 align:middle line:84% that looks like as if there's dirt 00:13:37.030 --> 00:13:40.220 align:middle line:84% or some kind of macroscopic little particle on the wafer. 00:13:40.220 --> 00:13:42.340 align:middle line:84% And in fact, it not only does that, it actually 00:13:42.340 --> 00:13:44.330 align:middle line:84% tells you roughly the size of the particle. 00:13:44.330 --> 00:13:47.890 align:middle line:84% So if you look at these bins, there's bin 1 through 8 here. 00:13:47.890 --> 00:13:49.773 align:middle line:90% So it bins them for you. 00:13:49.773 --> 00:13:50.815 align:middle line:90% And these are in microns. 00:13:50.815 --> 00:13:51.610 align:middle line:90% So I apologize. 00:13:51.610 --> 00:13:54.193 align:middle line:84% It wasn't written down, if you want to put that on your notes. 00:13:54.193 --> 00:13:56.260 align:middle line:84% So bin 1 goes-- is any particle that 00:13:56.260 --> 00:14:00.200 align:middle line:84% is between something like 0.2 microns to 0.3 microns. 00:14:00.200 --> 00:14:03.790 align:middle line:84% So in that size range, it found 16 particles. 00:14:03.790 --> 00:14:06.687 align:middle line:84% In fact, the system only found 21 on the entire wafer. 00:14:06.687 --> 00:14:08.770 align:middle line:84% So there aren't very many particles on this wafer. 00:14:08.770 --> 00:14:10.900 align:middle line:90% It's pretty clean. 00:14:10.900 --> 00:14:13.300 align:middle line:84% 16 of them are in this range, reasonably small, 00:14:13.300 --> 00:14:15.363 align:middle line:90% 0.2 to 0.3 microns. 00:14:15.363 --> 00:14:17.530 align:middle line:84% I don't know which-- where they are exactly on here. 00:14:17.530 --> 00:14:21.370 align:middle line:84% But these 21 points are the particles shown on the right. 00:14:21.370 --> 00:14:24.640 align:middle line:84% And it found five of them in the range of 0.3, 0.4. 00:14:24.640 --> 00:14:27.100 align:middle line:84% And it found zero in all the other bins. 00:14:27.100 --> 00:14:29.530 align:middle line:84% Bin 3 is 0.4 to 0.5 all the way up 00:14:29.530 --> 00:14:32.740 align:middle line:84% to, say, bin 8, which is very large, 00:14:32.740 --> 00:14:35.278 align:middle line:90% a micron or 1 to 7 microns. 00:14:35.278 --> 00:14:37.070 align:middle line:84% So it didn't find any particles that large. 00:14:37.070 --> 00:14:39.970 align:middle line:84% So this is a pretty darn clean wafer. 00:14:39.970 --> 00:14:43.510 align:middle line:84% This actually happens to be a wafer map from a wafer that's 00:14:43.510 --> 00:14:44.330 align:middle line:90% fresh from a box. 00:14:44.330 --> 00:14:47.470 align:middle line:84% So this is a prime grade wafer taken right out of the wafer 00:14:47.470 --> 00:14:49.210 align:middle line:90% manufacturer's box. 00:14:49.210 --> 00:14:52.480 align:middle line:84% It didn't spend more than a few seconds in the clean room 00:14:52.480 --> 00:14:54.220 align:middle line:84% before it got inserted into this machine. 00:14:54.220 --> 00:14:56.650 align:middle line:84% And it counted what the wafers looked like. 00:14:56.650 --> 00:14:58.960 align:middle line:84% So that just gives you an idea of how sophisticated 00:14:58.960 --> 00:15:00.920 align:middle line:90% these systems can be these days. 00:15:00.920 --> 00:15:02.380 align:middle line:84% And they're getting ones that can 00:15:02.380 --> 00:15:05.150 align:middle line:84% go to shorter and shorter wavelengths. 00:15:05.150 --> 00:15:06.982 align:middle line:84% So that's a way of mapping particles. 00:15:06.982 --> 00:15:08.440 align:middle line:84% Again, particles are only one thing 00:15:08.440 --> 00:15:09.880 align:middle line:90% we're trying to remove, though. 00:15:09.880 --> 00:15:10.900 align:middle line:84% Remember we talked about, they're 00:15:10.900 --> 00:15:12.130 align:middle line:90% just randomly distributed. 00:15:12.130 --> 00:15:13.840 align:middle line:84% A particle is a macroscopic object. 00:15:13.840 --> 00:15:17.740 align:middle line:84% It has some diameter or size, 0.2 to 0.3 microns. 00:15:17.740 --> 00:15:22.448 align:middle line:84% There may be at the atomic scale atoms of iron, atoms of copper 00:15:22.448 --> 00:15:23.740 align:middle line:90% distributed across the surface. 00:15:23.740 --> 00:15:25.073 align:middle line:90% They're not in particulate form. 00:15:25.073 --> 00:15:27.320 align:middle line:84% But they're just atomic level contaminants. 00:15:27.320 --> 00:15:29.670 align:middle line:84% Those are the things we need to remove as well. 00:15:29.670 --> 00:15:32.750 align:middle line:84% In fact, those are more of a problem than these particles. 00:15:32.750 --> 00:15:34.310 align:middle line:84% The particles are fairly localized. 00:15:34.310 --> 00:15:37.550 align:middle line:84% And they don't hurt you in-- for the purpose 00:15:37.550 --> 00:15:39.480 align:middle line:90% of doing a lot of experiments. 00:15:39.480 --> 00:15:42.200 align:middle line:84% Let's look at the next slide, though, slide number 6. 00:15:42.200 --> 00:15:44.390 align:middle line:84% Here's another example of a wafer map. 00:15:44.390 --> 00:15:45.920 align:middle line:90% This is actually the same wafer. 00:15:45.920 --> 00:15:48.200 align:middle line:84% Actually, ironically, I'm just trying 00:15:48.200 --> 00:15:52.250 align:middle line:84% to give you a little bit of practical education here. 00:15:52.250 --> 00:15:54.620 align:middle line:90% This is after an RCA clean. 00:15:54.620 --> 00:15:57.183 align:middle line:84% But it was a non-optimal process. 00:15:57.183 --> 00:15:58.600 align:middle line:84% Now, obviously, it was non-optimal 00:15:58.600 --> 00:16:00.800 align:middle line:84% because it looks like there's a lot more dots on the wafer 00:16:00.800 --> 00:16:02.530 align:middle line:84% now than when it was fresh out of the box. 00:16:02.530 --> 00:16:04.697 align:middle line:84% And this can happen if you're in a clean room that's 00:16:04.697 --> 00:16:08.780 align:middle line:84% not that clean or if you have water-- particles in your DI 00:16:08.780 --> 00:16:09.280 align:middle line:90% water. 00:16:09.280 --> 00:16:11.230 align:middle line:84% The last step is a DI water rinse. 00:16:11.230 --> 00:16:14.500 align:middle line:84% Or if you put it through a dryer, which heats up the wafer 00:16:14.500 --> 00:16:15.490 align:middle line:90% and spins it around. 00:16:15.490 --> 00:16:18.347 align:middle line:84% And if that dryer, the spin dryer, is very dirty, 00:16:18.347 --> 00:16:19.930 align:middle line:84% it can throw particles at the very end 00:16:19.930 --> 00:16:21.097 align:middle line:90% throughout your whole wafer. 00:16:21.097 --> 00:16:24.103 align:middle line:84% So just to make sure we understand the RCA clean, 00:16:24.103 --> 00:16:25.270 align:middle line:90% there's chemistry behind it. 00:16:25.270 --> 00:16:27.353 align:middle line:84% But there also has to be a lot of practical things 00:16:27.353 --> 00:16:28.760 align:middle line:90% that have to be done properly. 00:16:28.760 --> 00:16:31.840 align:middle line:84% And so as it turns out, now instead of 21 particles, 00:16:31.840 --> 00:16:33.190 align:middle line:90% this wafer now has 1,020-- 00:16:33.190 --> 00:16:36.370 align:middle line:84% 1,012 or so, quite a few, most of them 00:16:36.370 --> 00:16:37.580 align:middle line:90% added in a very small range. 00:16:37.580 --> 00:16:39.790 align:middle line:84% You'll notice they're quite small particles. 00:16:39.790 --> 00:16:42.430 align:middle line:90% Most of these are below 0.2-- 00:16:42.430 --> 00:16:44.960 align:middle line:90% 0.4 microns or so in diameter. 00:16:44.960 --> 00:16:47.680 align:middle line:84% They are perhaps the most difficult to remove of all 00:16:47.680 --> 00:16:48.250 align:middle line:90% things. 00:16:48.250 --> 00:16:51.430 align:middle line:84% At the atomic level, removing copper, iron, sodium, 00:16:51.430 --> 00:16:54.400 align:middle line:84% we can do that with the-- the particles have a certain mass 00:16:54.400 --> 00:16:55.060 align:middle line:90% to them. 00:16:55.060 --> 00:16:56.530 align:middle line:90% And they also tend to get stuck. 00:16:56.530 --> 00:16:58.113 align:middle line:84% They're electrostatically attracted 00:16:58.113 --> 00:16:59.280 align:middle line:90% because they can be charged. 00:16:59.280 --> 00:17:00.830 align:middle line:84% So they're attracted to the wafer. 00:17:00.830 --> 00:17:03.620 align:middle line:84% So it's a little bit trickier to get particles off. 00:17:03.620 --> 00:17:07.099 align:middle line:84% And people are always working on new particle reduction 00:17:07.099 --> 00:17:10.069 align:middle line:90% technologies. 00:17:10.069 --> 00:17:14.119 align:middle line:84% This is some other experimental evidence on slide 7, 00:17:14.119 --> 00:17:16.339 align:middle line:84% just to show you that the RCA clean does work 00:17:16.339 --> 00:17:18.530 align:middle line:84% and that it can work quite effectively. 00:17:18.530 --> 00:17:20.089 align:middle line:90% This is not tracking particles. 00:17:20.089 --> 00:17:22.849 align:middle line:84% But what it's looking at is how well 00:17:22.849 --> 00:17:27.619 align:middle line:84% it can remove certain mobile ions and metal contaminants. 00:17:27.619 --> 00:17:32.120 align:middle line:84% And this is a picture of a wafer and a process 00:17:32.120 --> 00:17:34.820 align:middle line:84% or a measurement technique called TXRF. 00:17:34.820 --> 00:17:39.260 align:middle line:84% And next lecture, we'll talk about total X-ray fluorescence. 00:17:39.260 --> 00:17:43.380 align:middle line:84% It's a metrology technique that can scan across the wafer. 00:17:43.380 --> 00:17:46.310 align:middle line:84% And in certain areas and one-centimeter spots, 00:17:46.310 --> 00:17:49.520 align:middle line:84% it can measure by X-ray fluorescence 00:17:49.520 --> 00:17:52.850 align:middle line:84% the atomic density of certain contaminants on the wafer. 00:17:52.850 --> 00:17:56.990 align:middle line:84% And these five points, center, top, right, left, and bottom, 00:17:56.990 --> 00:17:57.920 align:middle line:90% were taken. 00:17:57.920 --> 00:18:01.530 align:middle line:84% And they correspond to these xy coordinates here. 00:18:01.530 --> 00:18:05.060 align:middle line:84% So TXRF was performed at five points on the wafer. 00:18:05.060 --> 00:18:08.718 align:middle line:84% And these elements on the left column 00:18:08.718 --> 00:18:10.260 align:middle line:84% here are the ones that were measured, 00:18:10.260 --> 00:18:12.660 align:middle line:90% sulfur through bromine. 00:18:12.660 --> 00:18:14.092 align:middle line:84% And there are two different types 00:18:14.092 --> 00:18:15.300 align:middle line:90% of wafers that were measured. 00:18:15.300 --> 00:18:17.790 align:middle line:84% Here on the right in red, these are 00:18:17.790 --> 00:18:20.640 align:middle line:84% wafers that were subject to a fairly dirty process called 00:18:20.640 --> 00:18:22.410 align:middle line:90% chemical mechanical polishing. 00:18:22.410 --> 00:18:24.523 align:middle line:84% That's actually tends to be part of back end. 00:18:24.523 --> 00:18:26.940 align:middle line:84% But it can actually be used in front end in shallow trench 00:18:26.940 --> 00:18:29.550 align:middle line:90% isolation, STI, process. 00:18:29.550 --> 00:18:34.260 align:middle line:84% CMP uses a slurry of particles and all kinds 00:18:34.260 --> 00:18:37.020 align:middle line:84% of various solutions, one of which 00:18:37.020 --> 00:18:40.180 align:middle line:84% contains a lot of potassium and perhaps even some calcium. 00:18:40.180 --> 00:18:41.530 align:middle line:90% So it's a fairly dirty process. 00:18:41.530 --> 00:18:44.460 align:middle line:84% So after CMP, you really have to get the wafer clean. 00:18:44.460 --> 00:18:47.220 align:middle line:84% So this was a test on the right in the red, 00:18:47.220 --> 00:18:50.370 align:middle line:84% CMP plus a dilute sulfuric clean only. 00:18:50.370 --> 00:18:51.930 align:middle line:90% So it didn't do the RCA clean. 00:18:51.930 --> 00:18:55.020 align:middle line:84% And you can see, look at the amount of calcium 00:18:55.020 --> 00:18:56.640 align:middle line:84% here, reasonably high quantities. 00:18:56.640 --> 00:18:59.880 align:middle line:84% These are all measured in units of 10 00:18:59.880 --> 00:19:02.827 align:middle line:84% to the 10th atoms per square centimeter. 00:19:02.827 --> 00:19:03.660 align:middle line:90% So this is looking-- 00:19:03.660 --> 00:19:05.867 align:middle line:84% if you looked in a square centimeter, 00:19:05.867 --> 00:19:07.575 align:middle line:84% so imagine going into a square centimeter 00:19:07.575 --> 00:19:11.040 align:middle line:84% and counting the atoms, you'd find 10 to the tenth of them, 00:19:11.040 --> 00:19:12.240 align:middle line:90% roughly. 00:19:12.240 --> 00:19:13.770 align:middle line:84% Well, actually, in this position, 00:19:13.770 --> 00:19:17.393 align:middle line:84% you'd find 17 times 10 to the 10th, or roughly 1.7 times 10 00:19:17.393 --> 00:19:18.810 align:middle line:84% to the 11th per square centimeter, 00:19:18.810 --> 00:19:20.830 align:middle line:90% sitting on that surface. 00:19:20.830 --> 00:19:23.170 align:middle line:84% So it's a way of quantifying that. 00:19:23.170 --> 00:19:25.190 align:middle line:84% So here it's in the tens, twenties 00:19:25.190 --> 00:19:28.750 align:middle line:84% to hundreds on the right-hand side in the solution that 00:19:28.750 --> 00:19:32.130 align:middle line:90% used sulfuric only. 00:19:32.130 --> 00:19:33.750 align:middle line:84% On the other hand, when the sulfuric 00:19:33.750 --> 00:19:36.570 align:middle line:84% was followed by an RCA clean, all the calcium 00:19:36.570 --> 00:19:40.950 align:middle line:84% went down below the detection limit of the measurement. 00:19:40.950 --> 00:19:43.980 align:middle line:84% The measurement technique can't measure below about 2 times 10 00:19:43.980 --> 00:19:46.560 align:middle line:84% to the 10th atoms per square centimeter. 00:19:46.560 --> 00:19:49.440 align:middle line:84% If you compare these numbers to what's 00:19:49.440 --> 00:19:54.335 align:middle line:84% required in the ITRS roadmap, in fact, 00:19:54.335 --> 00:19:55.710 align:middle line:84% I think some of your homework was 00:19:55.710 --> 00:19:57.990 align:middle line:84% about what's required in the ITRS roadmap, 00:19:57.990 --> 00:20:00.990 align:middle line:84% you know that on the right-hand side, 17, 50, whatever, 00:20:00.990 --> 00:20:03.090 align:middle line:84% a hundred and some times 10 to the 10th, 00:20:03.090 --> 00:20:06.550 align:middle line:84% that's well above what's allowed to be 00:20:06.550 --> 00:20:11.090 align:middle line:84% processed on a wafer for the starting material. 00:20:11.090 --> 00:20:13.270 align:middle line:84% So we really need to get this down into this range. 00:20:13.270 --> 00:20:15.770 align:middle line:84% And the RCA clean does that here for calcium. 00:20:15.770 --> 00:20:16.990 align:middle line:90% Another example is iron. 00:20:16.990 --> 00:20:18.132 align:middle line:90% It's not quite as obvious. 00:20:18.132 --> 00:20:20.590 align:middle line:84% But on the right-hand side, there's a little bit more iron. 00:20:20.590 --> 00:20:21.610 align:middle line:90% It's measured here. 00:20:21.610 --> 00:20:24.550 align:middle line:84% Again, it varies from point to point because there may be 00:20:24.550 --> 00:20:26.920 align:middle line:90% some particles on this wafer. 00:20:26.920 --> 00:20:29.110 align:middle line:84% And it's picking up particles, perhaps spuriously, 00:20:29.110 --> 00:20:30.585 align:middle line:90% which contain iron. 00:20:30.585 --> 00:20:32.710 align:middle line:84% This particular spot on the wafer looks very dirty. 00:20:32.710 --> 00:20:36.400 align:middle line:84% It's got a high count, about 12 times 10 to the 10th. 00:20:36.400 --> 00:20:38.080 align:middle line:90% And zinc comes down uniformly. 00:20:38.080 --> 00:20:39.610 align:middle line:84% Look at the amount of zinc that's 00:20:39.610 --> 00:20:41.950 align:middle line:84% introduced by the CMP process that is not 00:20:41.950 --> 00:20:44.110 align:middle line:90% removed by sulfuric acid. 00:20:44.110 --> 00:20:45.970 align:middle line:84% It's not until it goes to the RCA clean 00:20:45.970 --> 00:20:49.220 align:middle line:84% that you get that down below the background. 00:20:49.220 --> 00:20:52.540 align:middle line:84% So the RCA clean can be very effective in removing 00:20:52.540 --> 00:20:55.455 align:middle line:84% contaminants, even ones that come from a very dirty process, 00:20:55.455 --> 00:20:57.580 align:middle line:84% which is basically like a sandpaper type of process 00:20:57.580 --> 00:21:00.340 align:middle line:90% almost, the CMP slurry. 00:21:00.340 --> 00:21:04.050 align:middle line:90% So it's certainly required. 00:21:04.050 --> 00:21:06.105 align:middle line:90% So let's go on to page number 8. 00:21:06.105 --> 00:21:08.480 align:middle line:84% There are actually-- the RCA clean is very old-fashioned. 00:21:08.480 --> 00:21:12.570 align:middle line:84% It's been around for, I don't know, 30 years, a long time. 00:21:12.570 --> 00:21:14.720 align:middle line:84% And since its invention, a lot of people 00:21:14.720 --> 00:21:16.860 align:middle line:90% have tried to improve upon it. 00:21:16.860 --> 00:21:18.410 align:middle line:90% And there are some improvements. 00:21:18.410 --> 00:21:20.600 align:middle line:84% There are a lot of advanced cleaning processes 00:21:20.600 --> 00:21:22.950 align:middle line:84% that are being developed over the years. 00:21:22.950 --> 00:21:24.920 align:middle line:84% And in fact, this is one that's-- 00:21:24.920 --> 00:21:27.718 align:middle line:84% I took from your text by a professor 00:21:27.718 --> 00:21:30.260 align:middle line:84% named Professor Omhi in Japan, who has proposed this cleaning 00:21:30.260 --> 00:21:31.070 align:middle line:90% process. 00:21:31.070 --> 00:21:36.290 align:middle line:84% And his goals were to try to get a process that would operate 00:21:36.290 --> 00:21:39.050 align:middle line:84% at lower temperature, at room temperature, 00:21:39.050 --> 00:21:42.020 align:middle line:84% and that would not remove, perhaps, so much silicon. 00:21:42.020 --> 00:21:44.158 align:middle line:84% And it also wouldn't use so many chemicals. 00:21:44.158 --> 00:21:45.200 align:middle line:90% This is an amazing thing. 00:21:45.200 --> 00:21:49.490 align:middle line:84% But there's a real concern about environmental factors. 00:21:49.490 --> 00:21:51.350 align:middle line:84% There's a lot of research today trying 00:21:51.350 --> 00:21:54.950 align:middle line:84% to improve the environmental friendliness of silicon 00:21:54.950 --> 00:21:57.530 align:middle line:84% IC processing because the chemicals I just mentioned, 00:21:57.530 --> 00:21:59.600 align:middle line:84% all of that concentrated sulfuric acid, 00:21:59.600 --> 00:22:01.400 align:middle line:84% hydrochloric acid, it's very concentrated. 00:22:01.400 --> 00:22:05.540 align:middle line:84% All that has to go down, be diluted somehow 00:22:05.540 --> 00:22:09.860 align:middle line:84% and treated before it goes out into our wastewater stream. 00:22:09.860 --> 00:22:13.070 align:middle line:84% So people are trying to minimize the amount of these chemicals 00:22:13.070 --> 00:22:15.380 align:middle line:84% that are dumped into the environment 00:22:15.380 --> 00:22:17.420 align:middle line:84% by inventing new cleaning methods that are more 00:22:17.420 --> 00:22:18.620 align:middle line:90% environmentally friendly. 00:22:18.620 --> 00:22:21.200 align:middle line:84% So he feels this is more friendly. 00:22:21.200 --> 00:22:23.510 align:middle line:84% But it has a lot of the similar principles. 00:22:23.510 --> 00:22:27.860 align:middle line:84% Notice the first step is to put it 00:22:27.860 --> 00:22:31.260 align:middle line:84% in ozonized water, so water that has ozone in it. 00:22:31.260 --> 00:22:34.500 align:middle line:84% So if we go back a couple of slides, 00:22:34.500 --> 00:22:37.190 align:middle line:84% let's see, if I go back to slide number 4 here, 00:22:37.190 --> 00:22:38.720 align:middle line:90% look at ozone, O3. 00:22:38.720 --> 00:22:41.900 align:middle line:84% It's the strongest oxidizer on the list, so even stronger 00:22:41.900 --> 00:22:43.920 align:middle line:90% than peroxide. 00:22:43.920 --> 00:22:47.510 align:middle line:84% So he creates ozone and puts it in the water. 00:22:47.510 --> 00:22:49.910 align:middle line:84% And a very strong oxidizer to strip 00:22:49.910 --> 00:22:53.730 align:middle line:84% organics that photoresist and even some metals. 00:22:53.730 --> 00:22:56.075 align:middle line:84% The next step-- again, and these are 00:22:56.075 --> 00:22:57.700 align:middle line:84% all at room temperature, which is nice. 00:22:57.700 --> 00:23:02.015 align:middle line:84% So you don't have to heat things up because, when you boil acids, 00:23:02.015 --> 00:23:04.390 align:middle line:84% they tend to vaporize and they go up into the atmosphere. 00:23:04.390 --> 00:23:06.598 align:middle line:84% And that also creates a lot of atmospheric pollution. 00:23:06.598 --> 00:23:09.060 align:middle line:90% So not having to boil is nice. 00:23:09.060 --> 00:23:11.940 align:middle line:84% Then he has a little bit of HF, peroxide, and water 00:23:11.940 --> 00:23:13.870 align:middle line:90% in a surfactant. 00:23:13.870 --> 00:23:18.460 align:middle line:84% Surfactant helps remove particles from the surface. 00:23:18.460 --> 00:23:19.710 align:middle line:90% And he puts it in a megasonic. 00:23:19.710 --> 00:23:23.760 align:middle line:84% This is a very high-frequency vibrational technique 00:23:23.760 --> 00:23:26.070 align:middle line:84% that actually tries to vibrate the particles off. 00:23:26.070 --> 00:23:28.320 align:middle line:84% So he puts a soap-like substance, 00:23:28.320 --> 00:23:30.780 align:middle line:84% which is a surfactant, a very high purity, 00:23:30.780 --> 00:23:34.980 align:middle line:84% to remove particles and more metals. 00:23:34.980 --> 00:23:36.840 align:middle line:84% He then goes back to this strong-- oxidizes 00:23:36.840 --> 00:23:42.300 align:middle line:84% the ozonized water to strip any chemicals 00:23:42.300 --> 00:23:44.730 align:middle line:84% that adhere to the wafer and then 00:23:44.730 --> 00:23:47.520 align:middle line:84% does an HF dip to get rid of the chemical oxide, 00:23:47.520 --> 00:23:50.890 align:middle line:90% produce a passive surface. 00:23:50.890 --> 00:23:53.490 align:middle line:84% So there's a lot of variations on the RCA theme. 00:23:53.490 --> 00:23:56.040 align:middle line:84% I've included here one reference from 1999 that I-- 00:23:56.040 --> 00:23:58.260 align:middle line:84% from Mark Hines, I think is a nice reference, 00:23:58.260 --> 00:24:02.370 align:middle line:84% it's a little old now, but on alternate cleaning processes 00:24:02.370 --> 00:24:04.710 align:middle line:84% and the efficiency of cleaning processes with respect 00:24:04.710 --> 00:24:07.568 align:middle line:84% to how well they clean the surface prior to gate oxidation. 00:24:07.568 --> 00:24:09.360 align:middle line:84% Gate oxidation is one of the most sensitive 00:24:09.360 --> 00:24:11.580 align:middle line:84% high-temperature processes you have to do. 00:24:11.580 --> 00:24:14.860 align:middle line:84% We don't want to incorporate any mobile ions, or iron, 00:24:14.860 --> 00:24:16.110 align:middle line:90% or anything in our gate oxide. 00:24:16.110 --> 00:24:17.700 align:middle line:90% So he's got some-- 00:24:17.700 --> 00:24:19.218 align:middle line:84% it's an interesting article if you 00:24:19.218 --> 00:24:21.510 align:middle line:84% want to know more about details about surface cleaning. 00:24:21.510 --> 00:24:24.610 align:middle line:90% 00:24:24.610 --> 00:24:27.490 align:middle line:90% So that's basically cleaning. 00:24:27.490 --> 00:24:29.110 align:middle line:84% Let's go now-- now I'm going to talk 00:24:29.110 --> 00:24:33.038 align:middle line:84% a little bit, the last topic, on contamination control, which 00:24:33.038 --> 00:24:33.580 align:middle line:90% is gettering. 00:24:33.580 --> 00:24:36.400 align:middle line:84% But before we go to gettering, I've 00:24:36.400 --> 00:24:39.400 align:middle line:84% just been indicating certain elements are bad 00:24:39.400 --> 00:24:42.122 align:middle line:90% or whatever, have properties. 00:24:42.122 --> 00:24:44.080 align:middle line:84% Let's say a few words about the elements, which 00:24:44.080 --> 00:24:46.990 align:middle line:90% we call deep levels in silicon. 00:24:46.990 --> 00:24:50.590 align:middle line:84% And this is a chart that I took from Simon Sze's famous book 00:24:50.590 --> 00:24:52.630 align:middle line:84% on the physics of semiconductor devices. 00:24:52.630 --> 00:24:54.280 align:middle line:84% I took it right out of his chart-- 00:24:54.280 --> 00:24:55.030 align:middle line:90% his book. 00:24:55.030 --> 00:24:57.550 align:middle line:84% And what it shows is the silicon bandgap here 00:24:57.550 --> 00:24:59.260 align:middle line:84% is being-- at the top, the conduction 00:24:59.260 --> 00:25:01.780 align:middle line:84% band, the bottom of this-- straight line 00:25:01.780 --> 00:25:03.400 align:middle line:90% at the bottom, the valence band. 00:25:03.400 --> 00:25:07.720 align:middle line:84% And the dashed line is the mid-gap point or the gap center. 00:25:07.720 --> 00:25:11.230 align:middle line:84% And what he has is a series of elements that, when you put them 00:25:11.230 --> 00:25:14.460 align:middle line:84% into silicon, what kind of states 00:25:14.460 --> 00:25:16.890 align:middle line:90% do they create in the bandgap? 00:25:16.890 --> 00:25:19.380 align:middle line:84% There are some shallow levels here. 00:25:19.380 --> 00:25:21.210 align:middle line:84% For instance, if we put boron in silicon, 00:25:21.210 --> 00:25:22.668 align:middle line:84% we all know that's a p-type doping. 00:25:22.668 --> 00:25:23.430 align:middle line:90% It's very shallow. 00:25:23.430 --> 00:25:25.860 align:middle line:84% It lies very close to the valence band. 00:25:25.860 --> 00:25:30.360 align:middle line:84% And it can accept electrons, and create holes, 00:25:30.360 --> 00:25:31.890 align:middle line:90% and dope your p-type. 00:25:31.890 --> 00:25:36.280 align:middle line:84% So boron is not too much of a problem. 00:25:36.280 --> 00:25:39.910 align:middle line:84% But the things that I have put a box around here, 00:25:39.910 --> 00:25:43.000 align:middle line:84% for instance, zinc and gold, have deep levels 00:25:43.000 --> 00:25:44.770 align:middle line:84% very close to the middle of the band gap 00:25:44.770 --> 00:25:46.990 align:middle line:90% here right near the dashed line. 00:25:46.990 --> 00:25:49.180 align:middle line:84% Copper as well has a mid-gap level. 00:25:49.180 --> 00:25:51.400 align:middle line:84% Iron has a level right next to mid-gap. 00:25:51.400 --> 00:25:56.080 align:middle line:84% It turns out when you have this deep level, this state, 00:25:56.080 --> 00:25:58.180 align:middle line:84% this electronic state that's close to mid-gap, 00:25:58.180 --> 00:26:01.375 align:middle line:90% it creates a generation and-- 00:26:01.375 --> 00:26:04.000 align:middle line:84% it creates a center where a lot of generation and recombination 00:26:04.000 --> 00:26:04.820 align:middle line:90% can happen. 00:26:04.820 --> 00:26:09.050 align:middle line:84% So it can trap carriers and reduce carrier lifetime. 00:26:09.050 --> 00:26:11.800 align:middle line:84% So the worst place, the worst types of elements then, 00:26:11.800 --> 00:26:13.690 align:middle line:84% the bad actors, are the ones that have 00:26:13.690 --> 00:26:15.143 align:middle line:90% deep levels right near mid-gap. 00:26:15.143 --> 00:26:16.810 align:middle line:84% And you see there are some of the things 00:26:16.810 --> 00:26:19.120 align:middle line:84% that people commonly talk about in clean rooms 00:26:19.120 --> 00:26:20.590 align:middle line:90% as you need to avoid. 00:26:20.590 --> 00:26:22.300 align:middle line:84% For example, copper, right here, it's 00:26:22.300 --> 00:26:26.060 align:middle line:84% got a number of deep levels, one at mid-gap, iron, zinc, 00:26:26.060 --> 00:26:27.480 align:middle line:90% and gold. 00:26:27.480 --> 00:26:30.660 align:middle line:84% So they're very-- deep levels are very undesirable. 00:26:30.660 --> 00:26:32.970 align:middle line:84% You should take this, though, summary 00:26:32.970 --> 00:26:35.640 align:middle line:84% chart with a grain of salt. It gives you a rough idea. 00:26:35.640 --> 00:26:37.950 align:middle line:90% Don't take it too literally. 00:26:37.950 --> 00:26:40.430 align:middle line:90% It's representative, however. 00:26:40.430 --> 00:26:42.180 align:middle line:84% If you want to look at it a different way, 00:26:42.180 --> 00:26:44.460 align:middle line:90% you move to the next slide. 00:26:44.460 --> 00:26:46.740 align:middle line:90% And I took this chart. 00:26:46.740 --> 00:26:50.430 align:middle line:84% This is actually a colored version of the periodic chart 00:26:50.430 --> 00:26:52.470 align:middle line:90% that I took from your textbook. 00:26:52.470 --> 00:26:55.607 align:middle line:84% And it characterizes the elements 00:26:55.607 --> 00:26:57.690 align:middle line:84% according to their position in the periodic table. 00:26:57.690 --> 00:27:02.370 align:middle line:84% And here in column 4 is where silicon is, of course. 00:27:02.370 --> 00:27:03.720 align:middle line:90% That's our semiconductor. 00:27:03.720 --> 00:27:07.020 align:middle line:84% The shallow acceptors, which make the semiconductor 00:27:07.020 --> 00:27:09.330 align:middle line:84% p-type which we need to add in, are typically boron. 00:27:09.330 --> 00:27:11.940 align:middle line:90% They're shown here in column 3. 00:27:11.940 --> 00:27:13.890 align:middle line:84% The shallow donors are in column 5. 00:27:13.890 --> 00:27:17.190 align:middle line:84% Of course, they have one extra valence electron 00:27:17.190 --> 00:27:18.510 align:middle line:90% compared to silicon. 00:27:18.510 --> 00:27:20.310 align:middle line:90% So those are fine elements. 00:27:20.310 --> 00:27:21.750 align:middle line:90% The problems are here. 00:27:21.750 --> 00:27:23.430 align:middle line:84% The problem elements, as you notice, 00:27:23.430 --> 00:27:25.290 align:middle line:84% tend to be in the transition metals. 00:27:25.290 --> 00:27:28.720 align:middle line:84% These form these deep levels, these Shockley-Read-Hall 00:27:28.720 --> 00:27:30.100 align:middle line:90% recombination centers. 00:27:30.100 --> 00:27:33.610 align:middle line:84% And a lot of them are in the transition elements. 00:27:33.610 --> 00:27:35.920 align:middle line:84% The other problems are shown over here on the left, 00:27:35.920 --> 00:27:40.960 align:middle line:84% the alkali ions, lithium, sodium, potassium, as well as 00:27:40.960 --> 00:27:44.240 align:middle line:84% calcium is also included in that category. 00:27:44.240 --> 00:27:46.733 align:middle line:84% These are a problem not so much because they 00:27:46.733 --> 00:27:48.400 align:middle line:84% create deep, deep levels in the silicon, 00:27:48.400 --> 00:27:51.190 align:middle line:84% but because they impact the quality, the electrical quality, 00:27:51.190 --> 00:27:54.940 align:middle line:84% of the gate oxide, its ability to act as a good insulator. 00:27:54.940 --> 00:27:56.560 align:middle line:84% They can cause shifts in the threshold 00:27:56.560 --> 00:27:57.580 align:middle line:90% voltage, their presence. 00:27:57.580 --> 00:28:01.660 align:middle line:84% They tend to be mobile in oxide under bias. 00:28:01.660 --> 00:28:02.930 align:middle line:90% So that's a problem. 00:28:02.930 --> 00:28:06.250 align:middle line:84% So we certainly want to get rid of these. 00:28:06.250 --> 00:28:08.458 align:middle line:84% So the idea here in gettering, which 00:28:08.458 --> 00:28:10.000 align:middle line:84% is a process we're going to talk now, 00:28:10.000 --> 00:28:12.130 align:middle line:84% is we're going to take all these unwanted elements, 00:28:12.130 --> 00:28:15.610 align:middle line:84% the transition metals, some of the heavy metals, 00:28:15.610 --> 00:28:16.900 align:middle line:90% the alkali ions. 00:28:16.900 --> 00:28:19.552 align:middle line:84% And we want to collect them in a region of the chip. 00:28:19.552 --> 00:28:21.260 align:middle line:84% Either we want to avoid introducing them, 00:28:21.260 --> 00:28:23.542 align:middle line:84% which we'll try to do, or if we can't totally, 00:28:23.542 --> 00:28:25.750 align:middle line:84% we're going to collect them in the region of the chip 00:28:25.750 --> 00:28:28.110 align:middle line:90% where they won't be harmful. 00:28:28.110 --> 00:28:30.170 align:middle line:84% In order to do gettering, and gettering 00:28:30.170 --> 00:28:32.750 align:middle line:84% is most effective for the transition metals, 00:28:32.750 --> 00:28:34.460 align:middle line:84% we need to take the elements wherever 00:28:34.460 --> 00:28:37.190 align:middle line:84% they are during the process and make them mobile. 00:28:37.190 --> 00:28:40.310 align:middle line:84% So they have to get away from their position near the device 00:28:40.310 --> 00:28:43.070 align:middle line:84% and be attracted to another part of the wafer. 00:28:43.070 --> 00:28:44.478 align:middle line:90% So they have to then diffuse. 00:28:44.478 --> 00:28:45.770 align:middle line:90% So they have to be made mobile. 00:28:45.770 --> 00:28:49.670 align:middle line:84% They have to then mobilize and diffuse to the trapping site. 00:28:49.670 --> 00:28:51.510 align:middle line:84% And then they have to stick there. 00:28:51.510 --> 00:28:53.180 align:middle line:84% So there's three processes that you 00:28:53.180 --> 00:28:55.970 align:middle line:84% have to do in order to getter impurities. 00:28:55.970 --> 00:28:58.860 align:middle line:90% 00:28:58.860 --> 00:29:00.350 align:middle line:84% So let's go on to slide number 11. 00:29:00.350 --> 00:29:03.470 align:middle line:84% And this is a schematic illustration and discussion 00:29:03.470 --> 00:29:08.340 align:middle line:84% of the methods of gettering for alkali ions. 00:29:08.340 --> 00:29:10.018 align:middle line:84% And this is a cross-section, meant 00:29:10.018 --> 00:29:12.060 align:middle line:84% to be a cross-section of your wafer starting here 00:29:12.060 --> 00:29:14.640 align:middle line:84% at the bottom in red, the back side of the wafer. 00:29:14.640 --> 00:29:18.280 align:middle line:84% And up at top is the front side of the wafer. 00:29:18.280 --> 00:29:19.770 align:middle line:84% There are some devices here shown. 00:29:19.770 --> 00:29:21.990 align:middle line:84% These little regions are supposed 00:29:21.990 --> 00:29:24.220 align:middle line:90% to be p-n junctions, perhaps. 00:29:24.220 --> 00:29:26.910 align:middle line:84% And in pink at the very top is what's called-- 00:29:26.910 --> 00:29:31.570 align:middle line:84% someone has put on the wafer a layer of PSG, 00:29:31.570 --> 00:29:33.850 align:middle line:84% also called phosphosilicate glass. 00:29:33.850 --> 00:29:36.885 align:middle line:84% It's actually an SiO2 layer, which 00:29:36.885 --> 00:29:39.650 align:middle line:84% it has a lot of phosphorous in it, up to 5%. 00:29:39.650 --> 00:29:41.590 align:middle line:90% So it's an alloy. 00:29:41.590 --> 00:29:43.750 align:middle line:84% This is an old-fashioned technique. 00:29:43.750 --> 00:29:46.270 align:middle line:84% People used to use it because alkali ions tend 00:29:46.270 --> 00:29:49.900 align:middle line:90% to be trapped in that glass. 00:29:49.900 --> 00:29:53.650 align:middle line:84% It's not so much in use anymore because it tends to absorb water 00:29:53.650 --> 00:29:56.090 align:middle line:90% later on in the process. 00:29:56.090 --> 00:29:58.030 align:middle line:84% And if you are using aluminum metallization, 00:29:58.030 --> 00:30:00.800 align:middle line:90% it can cause corrosion. 00:30:00.800 --> 00:30:02.895 align:middle line:84% So the approach to alkali ions these days 00:30:02.895 --> 00:30:04.270 align:middle line:84% as opposed to trying to gettering 00:30:04.270 --> 00:30:05.740 align:middle line:90% them is a little different. 00:30:05.740 --> 00:30:09.200 align:middle line:84% People just try not to get them in during the process. 00:30:09.200 --> 00:30:11.048 align:middle line:84% And then once the chip is fabbed, 00:30:11.048 --> 00:30:13.090 align:middle line:84% they don't want them to get in after fabrication. 00:30:13.090 --> 00:30:14.570 align:middle line:90% They put a protective layer. 00:30:14.570 --> 00:30:16.810 align:middle line:84% So the alternative of putting PSG in here 00:30:16.810 --> 00:30:19.660 align:middle line:90% is to put silicon nitride. 00:30:19.660 --> 00:30:21.752 align:middle line:90% It's a relatively tough layer. 00:30:21.752 --> 00:30:22.960 align:middle line:90% It doesn't scratch very well. 00:30:22.960 --> 00:30:26.260 align:middle line:84% And it protects the chip from alkali ion contamination 00:30:26.260 --> 00:30:27.160 align:middle line:90% after processing. 00:30:27.160 --> 00:30:28.900 align:middle line:84% So once the chip is finished, there's 00:30:28.900 --> 00:30:31.880 align:middle line:84% a scratch mask put down of silicon nitride. 00:30:31.880 --> 00:30:34.910 align:middle line:84% And then alkali ions that would come on the chip surface 00:30:34.910 --> 00:30:40.903 align:middle line:84% later after fabrication can't get through it. 00:30:40.903 --> 00:30:43.070 align:middle line:84% Now, so this isn't really so much gettering anymore. 00:30:43.070 --> 00:30:44.870 align:middle line:84% It's more like preventing it from happening. 00:30:44.870 --> 00:30:47.120 align:middle line:84% And this assumes, of course, that the processing steps 00:30:47.120 --> 00:30:48.700 align:middle line:84% are free of alkali ion contamination. 00:30:48.700 --> 00:30:52.150 align:middle line:84% That's why it's so important in processing, 00:30:52.150 --> 00:30:53.950 align:middle line:84% if we go back to this periodic chart, 00:30:53.950 --> 00:30:58.390 align:middle line:84% that we try to keep people out of the lab because all of us 00:30:58.390 --> 00:31:00.010 align:middle line:84% are full of sodium and potassium. 00:31:00.010 --> 00:31:01.930 align:middle line:84% Without sodium and potassium, we'd all die immediately. 00:31:01.930 --> 00:31:03.180 align:middle line:90% Our hearts would stop beating. 00:31:03.180 --> 00:31:08.507 align:middle line:84% So we're just exuding sodium, potassium, calcium, our bones, 00:31:08.507 --> 00:31:09.340 align:middle line:90% calcium, everything. 00:31:09.340 --> 00:31:12.787 align:middle line:84% So people are one of the worst, worst things in the lab. 00:31:12.787 --> 00:31:14.620 align:middle line:84% And that's why you would never touch a wafer 00:31:14.620 --> 00:31:16.840 align:middle line:84% or do anything where anything from your body 00:31:16.840 --> 00:31:18.700 align:middle line:84% can get onto the wafer because we have 00:31:18.700 --> 00:31:21.770 align:middle line:90% a lot of alkali ions in us. 00:31:21.770 --> 00:31:25.040 align:middle line:84% So trace metals are a little different. 00:31:25.040 --> 00:31:28.670 align:middle line:84% There are some pretty effective ways of gettering trace metals. 00:31:28.670 --> 00:31:30.120 align:middle line:90% In fact, there are two methods. 00:31:30.120 --> 00:31:35.460 align:middle line:84% One, they're both shown here schematically on slide 12. 00:31:35.460 --> 00:31:37.298 align:middle line:84% The first one is called extrinsic gettering. 00:31:37.298 --> 00:31:38.340 align:middle line:90% What does extrinsic mean? 00:31:38.340 --> 00:31:39.900 align:middle line:90% Well, extra. 00:31:39.900 --> 00:31:44.700 align:middle line:84% You externally put something-- do something to the wafer. 00:31:44.700 --> 00:31:46.320 align:middle line:84% Usually, on the back side, you often 00:31:46.320 --> 00:31:52.290 align:middle line:84% put a layer of something, which is created to trap ions or trap 00:31:52.290 --> 00:31:54.600 align:middle line:90% metal ions down here. 00:31:54.600 --> 00:31:56.820 align:middle line:84% So it's usually extrinsic to the wafer. 00:31:56.820 --> 00:31:57.900 align:middle line:90% That is, it's extra. 00:31:57.900 --> 00:32:00.660 align:middle line:90% It's been added on. 00:32:00.660 --> 00:32:04.680 align:middle line:84% The second way is to do what they call intrinsic gettering. 00:32:04.680 --> 00:32:06.720 align:middle line:84% And that's to, in the center of the wafer, 00:32:06.720 --> 00:32:08.850 align:middle line:84% in the middle region here that's shaded, 00:32:08.850 --> 00:32:11.220 align:middle line:84% is to intentionally form a region 00:32:11.220 --> 00:32:13.520 align:middle line:90% full of oxygen precipitates. 00:32:13.520 --> 00:32:15.740 align:middle line:84% And these precipitates create little traps 00:32:15.740 --> 00:32:19.432 align:middle line:84% where the metal atoms can diffuse to and get stuck. 00:32:19.432 --> 00:32:21.140 align:middle line:84% And they're within the bulk of the wafer. 00:32:21.140 --> 00:32:23.665 align:middle line:84% Generally, for most ICs, the bulk of the wafer 00:32:23.665 --> 00:32:25.040 align:middle line:84% is not part of the active region. 00:32:25.040 --> 00:32:28.070 align:middle line:84% It's just a handle just holding the thing. 00:32:28.070 --> 00:32:29.990 align:middle line:84% So it's OK to have the metals there. 00:32:29.990 --> 00:32:32.755 align:middle line:84% Now, of course, if you're making a power device, where 00:32:32.755 --> 00:32:34.880 align:middle line:84% the currents flow through from the top of the wafer 00:32:34.880 --> 00:32:37.047 align:middle line:84% to the bottom of the wafer, this would be a problem. 00:32:37.047 --> 00:32:43.270 align:middle line:84% But for most microprocessor memory, this would be effective. 00:32:43.270 --> 00:32:44.980 align:middle line:84% Intrinsic gettering is more popular 00:32:44.980 --> 00:32:46.840 align:middle line:90% these days than extrinsic. 00:32:46.840 --> 00:32:49.030 align:middle line:84% For these reasons, it's a little bit-- 00:32:49.030 --> 00:32:50.800 align:middle line:90% has a little better control. 00:32:50.800 --> 00:32:54.580 align:middle line:84% The gettering region here is actually closer to the devices. 00:32:54.580 --> 00:32:56.360 align:middle line:84% It can be within 10 or 20 microns. 00:32:56.360 --> 00:32:59.980 align:middle line:84% So it's not so hard to get the bad elements out of the devices 00:32:59.980 --> 00:33:02.200 align:middle line:90% and into the gettering region. 00:33:02.200 --> 00:33:06.160 align:middle line:84% And the other thing about it is that these SiO2 precipitates 00:33:06.160 --> 00:33:08.330 align:middle line:84% tend to be thermally stable once they're formed. 00:33:08.330 --> 00:33:12.550 align:middle line:84% So it's usually works throughout the entire process. 00:33:12.550 --> 00:33:16.960 align:middle line:84% Not always the case with the backside gettering. 00:33:16.960 --> 00:33:20.417 align:middle line:84% Let's go on to slide number 13 and think about, 00:33:20.417 --> 00:33:21.500 align:middle line:90% what are we talking about? 00:33:21.500 --> 00:33:23.920 align:middle line:84% We're taking elements away from-- 00:33:23.920 --> 00:33:26.950 align:middle line:84% impurities out of the device regions during the processing. 00:33:26.950 --> 00:33:29.613 align:middle line:84% And we somehow have to get them to diffuse to these trap sites. 00:33:29.613 --> 00:33:31.030 align:middle line:84% So the first thing we need to know 00:33:31.030 --> 00:33:33.670 align:middle line:84% is, how rapidly at a given temperature 00:33:33.670 --> 00:33:35.290 align:middle line:84% are these elements going to diffuse? 00:33:35.290 --> 00:33:36.860 align:middle line:84% If they don't diffuse fast enough, 00:33:36.860 --> 00:33:39.277 align:middle line:84% there's no way you're ever going to get them to the traps. 00:33:39.277 --> 00:33:40.970 align:middle line:90% 00:33:40.970 --> 00:33:43.770 align:middle line:84% So there's some good news and some bad news. 00:33:43.770 --> 00:33:47.330 align:middle line:84% The good news is that most metals diffuse as interstitials, 00:33:47.330 --> 00:33:51.090 align:middle line:84% so in the interstitial spaces in silicon very rapidly, 00:33:51.090 --> 00:33:53.180 align:middle line:84% which means you have a chance of grabbing them, 00:33:53.180 --> 00:33:55.280 align:middle line:84% of getting them to go from the device region 00:33:55.280 --> 00:33:57.990 align:middle line:84% to the back of the wafer or into the bulk of the wafer. 00:33:57.990 --> 00:33:59.930 align:middle line:84% And in fact, here's some numbers for you 00:33:59.930 --> 00:34:02.330 align:middle line:90% quantitatively on this plot. 00:34:02.330 --> 00:34:04.400 align:middle line:84% The left axis shows the diffusivity. 00:34:04.400 --> 00:34:07.820 align:middle line:84% And we measure that in units of a length squared per second. 00:34:07.820 --> 00:34:10.580 align:middle line:84% We're going to talk about the process or per unit time. 00:34:10.580 --> 00:34:14.017 align:middle line:84% We'll talk about the process of diffusion later in the course. 00:34:14.017 --> 00:34:16.100 align:middle line:84% But the bigger this number, let's put it that way, 00:34:16.100 --> 00:34:19.489 align:middle line:84% the faster the elements will move at a given temperature 00:34:19.489 --> 00:34:21.076 align:middle line:90% throughout the wafer. 00:34:21.076 --> 00:34:22.909 align:middle line:84% And this is what's called an Arrhenius plot. 00:34:22.909 --> 00:34:27.380 align:middle line:84% So it's plotted versus 1,000 or 1 over temperature 00:34:27.380 --> 00:34:28.250 align:middle line:90% on the bottom axis. 00:34:28.250 --> 00:34:31.100 align:middle line:84% If you want to read temperatures in actual units of centigrade, 00:34:31.100 --> 00:34:33.290 align:middle line:84% you can read it up at the top axis from here, 00:34:33.290 --> 00:34:35.960 align:middle line:84% from 800, all the way up to 1,200. 00:34:35.960 --> 00:34:38.900 align:middle line:84% And the different curves are for different elements. 00:34:38.900 --> 00:34:41.630 align:middle line:84% And all the metals here are indicated in red. 00:34:41.630 --> 00:34:44.239 align:middle line:84% So here's some common metals, things 00:34:44.239 --> 00:34:46.820 align:middle line:84% that we mentioned as being very bad for lifetime killers. 00:34:46.820 --> 00:34:47.630 align:middle line:90% Here's copper. 00:34:47.630 --> 00:34:50.065 align:middle line:84% Look at copper, almost irrespective 00:34:50.065 --> 00:34:51.440 align:middle line:84% of the temperature in this range. 00:34:51.440 --> 00:34:53.570 align:middle line:84% Copper is zipping through the wafer faster 00:34:53.570 --> 00:34:56.060 align:middle line:90% than any of these others. 00:34:56.060 --> 00:34:59.450 align:middle line:84% Here's gold, Au, as an interstitial form. 00:34:59.450 --> 00:35:01.790 align:middle line:90% It diffuses quite rapidly. 00:35:01.790 --> 00:35:04.130 align:middle line:84% Gold here, when it's substitutional on the lattice, 00:35:04.130 --> 00:35:06.330 align:middle line:84% so the gold atom ends up being substitutional, 00:35:06.330 --> 00:35:07.760 align:middle line:90% doesn't diffuse quite so fast. 00:35:07.760 --> 00:35:11.142 align:middle line:90% It's orders of magnitudes down. 00:35:11.142 --> 00:35:13.600 align:middle line:84% Looking here at these, these are the diffusion coefficients 00:35:13.600 --> 00:35:17.697 align:middle line:84% in blue here for the shallow impurities, 00:35:17.697 --> 00:35:19.030 align:middle line:90% like arsenic, boron, phosphorus. 00:35:19.030 --> 00:35:20.810 align:middle line:90% They diffuse much slower. 00:35:20.810 --> 00:35:27.370 align:middle line:84% And this dark, I guess, stippled region that's colored in tan 00:35:27.370 --> 00:35:31.300 align:middle line:84% is roughly the region where people believe the silicon 00:35:31.300 --> 00:35:32.650 align:middle line:90% interstitial diffuses. 00:35:32.650 --> 00:35:35.260 align:middle line:84% Again, no one's been able to actually measure exactly 00:35:35.260 --> 00:35:37.150 align:middle line:84% because you can't see a silicon interstitial. 00:35:37.150 --> 00:35:39.340 align:middle line:84% So you can't profile it, how rapidly it diffuses. 00:35:39.340 --> 00:35:41.380 align:middle line:84% But there's indirect ways of inferring it. 00:35:41.380 --> 00:35:43.960 align:middle line:84% So somewhere in this band, this brown band, 00:35:43.960 --> 00:35:46.450 align:middle line:84% is where the silicon interstitial diffusivity lies. 00:35:46.450 --> 00:35:47.630 align:middle line:90% It's not bad. 00:35:47.630 --> 00:35:49.855 align:middle line:84% It's not as fast as the fastest metals. 00:35:49.855 --> 00:35:53.890 align:middle line:84% It's not as fast as copper or sub-- or interstitial gold 00:35:53.890 --> 00:35:54.400 align:middle line:90% or iron. 00:35:54.400 --> 00:35:58.682 align:middle line:84% But it's in the range of some of the metals. 00:35:58.682 --> 00:36:01.140 align:middle line:84% Just to give you an example here, this bottom bullet shows, 00:36:01.140 --> 00:36:03.780 align:middle line:84% let's say take copper at 900 degrees. 00:36:03.780 --> 00:36:07.560 align:middle line:84% You can read off this chart that diffusivity, D, is about 10 00:36:07.560 --> 00:36:10.080 align:middle line:84% to the minus 4 centimeters squared per second. 00:36:10.080 --> 00:36:12.090 align:middle line:84% We'll learn that a rough estimate 00:36:12.090 --> 00:36:15.600 align:middle line:84% of how far a profile diffuses can be the square root of D 00:36:15.600 --> 00:36:16.690 align:middle line:90% times the time. 00:36:16.690 --> 00:36:20.670 align:middle line:84% So if I put this wafer that accidentally has some copper 00:36:20.670 --> 00:36:24.330 align:middle line:84% by accident on the surface, in the furnace at 900 00:36:24.330 --> 00:36:26.580 align:middle line:84% degrees in one minute, the square root of dt 00:36:26.580 --> 00:36:28.470 align:middle line:90% is 780 microns. 00:36:28.470 --> 00:36:31.840 align:middle line:84% And the wafer is only typically 600 microns thick. 00:36:31.840 --> 00:36:35.760 align:middle line:84% So basically, that copper can go right through the wafer 00:36:35.760 --> 00:36:37.277 align:middle line:90% and be anywhere in the wafer. 00:36:37.277 --> 00:36:39.360 align:middle line:84% That's bad because, if you accidentally get copper 00:36:39.360 --> 00:36:41.620 align:middle line:84% on the back of the wafer, let's say where you say, 00:36:41.620 --> 00:36:43.440 align:middle line:84% oh, it won't hurt anything, you put it in the furnace, 00:36:43.440 --> 00:36:44.160 align:middle line:90% well, it sure will. 00:36:44.160 --> 00:36:45.450 align:middle line:84% It'll be in the front real quickly. 00:36:45.450 --> 00:36:46.740 align:middle line:84% It's good if you're trying to getter 00:36:46.740 --> 00:36:49.050 align:middle line:84% because, if the copper happens to be in your device region, 00:36:49.050 --> 00:36:51.030 align:middle line:84% you can try to get it to the back of the wafer. 00:36:51.030 --> 00:36:52.687 align:middle line:90% And the key is to hold it there. 00:36:52.687 --> 00:36:53.770 align:middle line:90% So it's both good and bad. 00:36:53.770 --> 00:36:57.015 align:middle line:84% But metals are very fast diffusers. 00:36:57.015 --> 00:37:01.500 align:middle line:84% Well, and that's-- copper is an extreme example. 00:37:01.500 --> 00:37:05.330 align:middle line:84% So let's go on to slide 14, which talks about metal 00:37:05.330 --> 00:37:06.750 align:middle line:90% gettering a little bit more. 00:37:06.750 --> 00:37:10.460 align:middle line:84% And this plot on the vertical axis is temperature. 00:37:10.460 --> 00:37:13.730 align:middle line:84% And the bottom, the horizontal axis, 00:37:13.730 --> 00:37:15.060 align:middle line:90% is atoms per cubic centimeter. 00:37:15.060 --> 00:37:19.340 align:middle line:84% It represents the solubilities of the fast-diffusing metal 00:37:19.340 --> 00:37:20.760 align:middle line:90% as a function of temperature. 00:37:20.760 --> 00:37:22.890 align:middle line:84% So if you look at any point on this curve, 00:37:22.890 --> 00:37:24.800 align:middle line:84% if you have a concentration higher than that 00:37:24.800 --> 00:37:28.550 align:middle line:84% at that temperature, then the metal will tend to precipitate 00:37:28.550 --> 00:37:33.370 align:middle line:90% out, will not be in solution. 00:37:33.370 --> 00:37:35.850 align:middle line:84% So metals preferentially like to reside 00:37:35.850 --> 00:37:38.560 align:middle line:84% in sites on the silicon lattice where there's imperfections. 00:37:38.560 --> 00:37:39.310 align:middle line:90% So this is key. 00:37:39.310 --> 00:37:41.880 align:middle line:90% We can use this, this idea. 00:37:41.880 --> 00:37:42.970 align:middle line:90% Why might this be? 00:37:42.970 --> 00:37:45.962 align:middle line:84% Well, the metals don't fit because of their atomic size 00:37:45.962 --> 00:37:46.920 align:middle line:90% different from silicon. 00:37:46.920 --> 00:37:49.740 align:middle line:84% They don't fit into the silicon lattice. 00:37:49.740 --> 00:37:51.720 align:middle line:84% The other thing is that if you have 00:37:51.720 --> 00:37:54.450 align:middle line:84% a fault or a disordered region in the crystal, 00:37:54.450 --> 00:37:56.850 align:middle line:84% it might be able to accommodate this size difference 00:37:56.850 --> 00:37:59.440 align:middle line:90% and trap the metal atom there. 00:37:59.440 --> 00:38:01.980 align:middle line:84% And it turns out that dislocations and stacking 00:38:01.980 --> 00:38:04.410 align:middle line:84% faults, these are crystal imperfections that 00:38:04.410 --> 00:38:08.250 align:middle line:84% can exist in the wafer, sometimes, they are well 00:38:08.250 --> 00:38:11.760 align:middle line:90% known as trapping sites for-- 00:38:11.760 --> 00:38:15.210 align:middle line:84% or sites that where copper, gold, and iron tend 00:38:15.210 --> 00:38:17.350 align:middle line:90% to accumulate and precipitate. 00:38:17.350 --> 00:38:20.715 align:middle line:84% So if you have dislocations, depending on where they are, 00:38:20.715 --> 00:38:21.840 align:middle line:90% that could be a good thing. 00:38:21.840 --> 00:38:23.610 align:middle line:84% You could try to create dislocations. 00:38:23.610 --> 00:38:25.890 align:middle line:84% And these, they tend to be decorated, 00:38:25.890 --> 00:38:29.550 align:middle line:84% so to speak, with these copper, gold, and iron. 00:38:29.550 --> 00:38:31.410 align:middle line:84% So the trick, the name of the game, 00:38:31.410 --> 00:38:34.810 align:middle line:84% is to form these kind of defects intentionally, not 00:38:34.810 --> 00:38:36.910 align:middle line:84% in your device region, because then you'd 00:38:36.910 --> 00:38:39.430 align:middle line:84% have copper, and gold, and iron there, but away 00:38:39.430 --> 00:38:40.670 align:middle line:90% from the active region. 00:38:40.670 --> 00:38:43.000 align:middle line:84% And so the idea of metal gettering 00:38:43.000 --> 00:38:46.930 align:middle line:84% is how to form these types of things. 00:38:46.930 --> 00:38:49.290 align:middle line:84% So on slide 16, we talk a little bit 00:38:49.290 --> 00:38:53.100 align:middle line:84% about how extrinsic gathering can take place. 00:38:53.100 --> 00:38:55.287 align:middle line:84% Usually, in the extrinsic case, they 00:38:55.287 --> 00:38:57.120 align:middle line:84% try to form, people try to form, these sites 00:38:57.120 --> 00:38:58.200 align:middle line:90% on the back side of the wafer. 00:38:58.200 --> 00:38:59.992 align:middle line:84% And there are a number of ways of doing it, 00:38:59.992 --> 00:39:03.217 align:middle line:84% some very crude, grinding and sandpaper. 00:39:03.217 --> 00:39:05.550 align:middle line:84% Sometimes you'll look at the back of a wafer, especially 00:39:05.550 --> 00:39:06.420 align:middle line:90% an older wafer. 00:39:06.420 --> 00:39:09.840 align:middle line:84% And you look at it and there's this ground pattern, almost look 00:39:09.840 --> 00:39:11.220 align:middle line:90% like it was put in a lathe. 00:39:11.220 --> 00:39:13.980 align:middle line:84% And it goes around, this pattern that goes on the back side. 00:39:13.980 --> 00:39:16.530 align:middle line:84% That was intentionally ground into the back of the wafer 00:39:16.530 --> 00:39:20.040 align:middle line:84% to make a lot of these gettering sites. 00:39:20.040 --> 00:39:21.780 align:middle line:84% They used to use sandpaper abrasion. 00:39:21.780 --> 00:39:23.950 align:middle line:90% That's not so popular anymore. 00:39:23.950 --> 00:39:27.540 align:middle line:84% People these days use the cleaner processes or the cleaner 00:39:27.540 --> 00:39:28.800 align:middle line:90% methods. 00:39:28.800 --> 00:39:31.230 align:middle line:84% Ion implantation can be used to damage 00:39:31.230 --> 00:39:33.690 align:middle line:84% the back of the wafer, a deposition 00:39:33.690 --> 00:39:35.040 align:middle line:90% of a polycrystalline film. 00:39:35.040 --> 00:39:38.760 align:middle line:84% It's not that unusual to find polysilicon 00:39:38.760 --> 00:39:41.040 align:middle line:84% on the back of the wafer or a region 00:39:41.040 --> 00:39:42.630 align:middle line:84% very highly doped with phosphorus 00:39:42.630 --> 00:39:44.022 align:middle line:90% on the back of the wafer. 00:39:44.022 --> 00:39:45.480 align:middle line:84% These are all regions where metals, 00:39:45.480 --> 00:39:49.160 align:middle line:84% if they diffuse back there, will get trapped and stick. 00:39:49.160 --> 00:39:51.580 align:middle line:84% So the whole idea is, let's make extended defects 00:39:51.580 --> 00:39:54.190 align:middle line:84% that are stable throughout the high-temperature processing. 00:39:54.190 --> 00:39:56.320 align:middle line:84% The problem is if they tend-- these extended defects tend 00:39:56.320 --> 00:39:58.060 align:middle line:84% to heal, the metals will then be released 00:39:58.060 --> 00:39:59.977 align:middle line:84% and they'll go back to the front of the wafer. 00:39:59.977 --> 00:40:04.110 align:middle line:90% 00:40:04.110 --> 00:40:05.820 align:middle line:90% So that's extrinsic gettering. 00:40:05.820 --> 00:40:09.440 align:middle line:84% So if we go on to page or slide number 16, 00:40:09.440 --> 00:40:12.060 align:middle line:84% we have a schematic of this intrinsic gathering, 00:40:12.060 --> 00:40:14.920 align:middle line:90% which is a little more common. 00:40:14.920 --> 00:40:16.650 align:middle line:84% Remember the last time we talked, 00:40:16.650 --> 00:40:19.725 align:middle line:84% we said that oxygen is present in all Czochralski silicon. 00:40:19.725 --> 00:40:20.850 align:middle line:90% It comes from the crucible. 00:40:20.850 --> 00:40:23.650 align:middle line:84% There's nothing we can do about it. 00:40:23.650 --> 00:40:26.700 align:middle line:84% And it can be introduced in this supersaturated form. 00:40:26.700 --> 00:40:30.120 align:middle line:84% That is, oxygen can be dissolved in there 00:40:30.120 --> 00:40:33.060 align:middle line:84% because the wafer was cooled at a relatively rapid weight-- 00:40:33.060 --> 00:40:34.230 align:middle line:90% rate. 00:40:34.230 --> 00:40:36.930 align:middle line:84% But there's more oxygen there than the crystal 00:40:36.930 --> 00:40:40.330 align:middle line:84% wants to hold at that temperature in equilibrium. 00:40:40.330 --> 00:40:43.815 align:middle line:84% So if you were to heat that crystal up and give it 00:40:43.815 --> 00:40:46.620 align:middle line:84% a long enough time, the oxygen could then tend to form-- 00:40:46.620 --> 00:40:49.800 align:middle line:90% precipitate out and form SiO2. 00:40:49.800 --> 00:40:54.060 align:middle line:84% So oxygen forms these SiO2 precipitates in the crystal. 00:40:54.060 --> 00:40:56.970 align:middle line:84% And the interesting thing is that these precipitates 00:40:56.970 --> 00:40:58.950 align:middle line:84% are usually accompanied by some kind of defect, 00:40:58.950 --> 00:41:01.140 align:middle line:84% mechanical defect, like a stacking fault 00:41:01.140 --> 00:41:04.310 align:middle line:84% or an extended defect, like dislocations. 00:41:04.310 --> 00:41:06.060 align:middle line:84% And that's because of the volume mismatch. 00:41:06.060 --> 00:41:07.860 align:middle line:90% SiO2 wants to expand. 00:41:07.860 --> 00:41:09.852 align:middle line:84% These little SiO2 precipitates want to expand. 00:41:09.852 --> 00:41:11.560 align:middle line:84% They compress the silicon lattice around. 00:41:11.560 --> 00:41:13.680 align:middle line:84% If they compress it enough, it can pop out 00:41:13.680 --> 00:41:16.540 align:middle line:90% these extended defects. 00:41:16.540 --> 00:41:20.430 align:middle line:84% So the key point of gettering is to get this precipitation 00:41:20.430 --> 00:41:24.335 align:middle line:84% to occur in the bulk, in the bulk of the crystal, 00:41:24.335 --> 00:41:25.710 align:middle line:84% throughout the bulk of the wafer, 00:41:25.710 --> 00:41:27.968 align:middle line:84% but not in the near surface region. 00:41:27.968 --> 00:41:29.260 align:middle line:90% So how are we going to do that? 00:41:29.260 --> 00:41:30.635 align:middle line:84% Well, have to have a near surface 00:41:30.635 --> 00:41:34.630 align:middle line:84% region that doesn't have high oxygen so it can't precipitate. 00:41:34.630 --> 00:41:36.810 align:middle line:84% So what we do is we create something 00:41:36.810 --> 00:41:39.927 align:middle line:84% called a denuded zone, which is low in oxygen. 00:41:39.927 --> 00:41:41.760 align:middle line:84% And there's a couple of methods of doing it. 00:41:41.760 --> 00:41:44.640 align:middle line:84% This denuded zone, which is pictured schematically here, 00:41:44.640 --> 00:41:47.190 align:middle line:84% typically might be 10 to 20 microns thick. 00:41:47.190 --> 00:41:48.900 align:middle line:84% It sits on top of the wafer, the wafer, 00:41:48.900 --> 00:41:52.390 align:middle line:90% again, being full of oxygen. 00:41:52.390 --> 00:41:55.320 align:middle line:84% So we can do that-- there's a process we can use 00:41:55.320 --> 00:41:57.060 align:middle line:84% called epitaxial growth, which we're 00:41:57.060 --> 00:42:00.180 align:middle line:84% going to talk about later in this term, where we can grow 00:42:00.180 --> 00:42:02.820 align:middle line:84% material on top that's very low in oxygen. 00:42:02.820 --> 00:42:05.370 align:middle line:84% Or if you don't want to do epitaxial growth, 00:42:05.370 --> 00:42:08.820 align:middle line:84% because it's fairly expensive, you can just take the wafer. 00:42:08.820 --> 00:42:10.740 align:middle line:84% And the very first step you can do 00:42:10.740 --> 00:42:13.050 align:middle line:84% is go to a very high temperature and cause the oxygen 00:42:13.050 --> 00:42:16.090 align:middle line:84% to outdiffuse, out the surface, out the front side of the wafer. 00:42:16.090 --> 00:42:20.110 align:middle line:84% So whatever oxygen was in the top 10 microns can go out. 00:42:20.110 --> 00:42:22.853 align:middle line:84% Oxygen buried deeper below that at this high temperature 00:42:22.853 --> 00:42:24.520 align:middle line:84% can't make it because it can't diffuse-- 00:42:24.520 --> 00:42:28.670 align:middle line:84% it doesn't have enough time to diffuse that amount of distance. 00:42:28.670 --> 00:42:31.180 align:middle line:84% So one of the first steps people use 00:42:31.180 --> 00:42:33.160 align:middle line:84% for doing intrinsic gettering is sometimes 00:42:33.160 --> 00:42:36.940 align:middle line:84% a high-temperature step to create this denuded zone 00:42:36.940 --> 00:42:39.685 align:middle line:84% or to grow an epitaxial silicon layer. 00:42:39.685 --> 00:42:42.460 align:middle line:90% 00:42:42.460 --> 00:42:45.250 align:middle line:84% So let's go to the next slide, slide 17. 00:42:45.250 --> 00:42:46.960 align:middle line:84% And the next couple of slides are 00:42:46.960 --> 00:42:48.940 align:middle line:84% going to outline the thermal processing. 00:42:48.940 --> 00:42:51.040 align:middle line:84% So just by doing these thermal processing steps, 00:42:51.040 --> 00:42:53.430 align:middle line:90% you can do intrinsic gettering. 00:42:53.430 --> 00:42:55.440 align:middle line:84% And this plot shows schematically 00:42:55.440 --> 00:42:58.290 align:middle line:84% on the vertical axis the wafer temperature and time. 00:42:58.290 --> 00:43:00.870 align:middle line:84% Again, time isn't really indicated here quantitatively, 00:43:00.870 --> 00:43:02.520 align:middle line:90% just to give you a rough idea. 00:43:02.520 --> 00:43:04.180 align:middle line:84% But the first step is to take the wafer 00:43:04.180 --> 00:43:06.390 align:middle line:84% at a very high temperature, say 1,100 degrees, 00:43:06.390 --> 00:43:07.975 align:middle line:90% and do the outdiffusion step. 00:43:07.975 --> 00:43:10.350 align:middle line:84% So we're going to take that wafer, put it in the furnace, 00:43:10.350 --> 00:43:13.620 align:middle line:84% and let the oxygen go out, not all of it, but whatever 00:43:13.620 --> 00:43:16.020 align:middle line:90% is in the near surface region. 00:43:16.020 --> 00:43:20.430 align:middle line:84% And this expression in the middle of the slide on slide 17 00:43:20.430 --> 00:43:23.600 align:middle line:84% shows the diffusivity as a function of temperature. 00:43:23.600 --> 00:43:24.780 align:middle line:90% It gives you a rough idea. 00:43:24.780 --> 00:43:27.840 align:middle line:84% And the nice thing is you don't have to remove all the oxygen. 00:43:27.840 --> 00:43:30.790 align:middle line:84% You just have to get the level below a certain level. 00:43:30.790 --> 00:43:32.760 align:middle line:84% So for instance, you need to get it 00:43:32.760 --> 00:43:36.190 align:middle line:84% from about 20 part per million down to less than 10. 00:43:36.190 --> 00:43:38.377 align:middle line:84% If it gets low enough, the precipitation 00:43:38.377 --> 00:43:39.960 align:middle line:84% won't take place because, again, it'll 00:43:39.960 --> 00:43:41.370 align:middle line:90% be below the solubility limit. 00:43:41.370 --> 00:43:43.440 align:middle line:84% So you don't have to worry about precipitation. 00:43:43.440 --> 00:43:46.410 align:middle line:84% So you're just trying to reduce it down. 00:43:46.410 --> 00:43:49.470 align:middle line:84% And usually, 1,100 to 1,200 is sufficient to create 00:43:49.470 --> 00:43:50.160 align:middle line:90% a denuded zone. 00:43:50.160 --> 00:43:53.350 align:middle line:84% And that's in the order of 10 to 20 microns deep. 00:43:53.350 --> 00:43:55.650 align:middle line:84% It has-- the lower oxygen concentration 00:43:55.650 --> 00:43:56.770 align:middle line:90% then will precipitate. 00:43:56.770 --> 00:44:00.000 align:middle line:84% So that's the first step is a very high-temperature step. 00:44:00.000 --> 00:44:03.090 align:middle line:84% Next step in the process, shown on slide 18, 00:44:03.090 --> 00:44:04.890 align:middle line:90% is called nucleation. 00:44:04.890 --> 00:44:07.020 align:middle line:84% So here we go from this high temperature, 00:44:07.020 --> 00:44:09.640 align:middle line:84% where we got rid of the oxygen in the surface region. 00:44:09.640 --> 00:44:12.240 align:middle line:84% And now we want to nucleate the precipitates in the bulk. 00:44:12.240 --> 00:44:16.330 align:middle line:84% So we go down to a lower temperature, say 700 degrees. 00:44:16.330 --> 00:44:19.140 align:middle line:84% And we want to cause very, very small precipitates 00:44:19.140 --> 00:44:20.100 align:middle line:90% to take place. 00:44:20.100 --> 00:44:24.160 align:middle line:84% And the optimum temperature for this is somewhere around 700. 00:44:24.160 --> 00:44:28.050 align:middle line:84% And this is discussed in chapter 3 of your text. 00:44:28.050 --> 00:44:32.100 align:middle line:84% It's optimum because you need these nuclei to grow 00:44:32.100 --> 00:44:34.470 align:middle line:84% to a minimum critical size to prevent them 00:44:34.470 --> 00:44:39.180 align:middle line:84% from shrinking later when you start to ramp up the wafer 00:44:39.180 --> 00:44:40.210 align:middle line:90% temperature. 00:44:40.210 --> 00:44:43.050 align:middle line:84% So the critical size is about two nanometers, 00:44:43.050 --> 00:44:44.840 align:middle line:84% say, one to three nanometers in diameter, 00:44:44.840 --> 00:44:46.650 align:middle line:90% this precipitate of SiO2. 00:44:46.650 --> 00:44:49.530 align:middle line:84% And you need to-- you typically shoot for a density about 10 00:44:49.530 --> 00:44:51.203 align:middle line:84% to the 11th per cubic centimeter. 00:44:51.203 --> 00:44:52.620 align:middle line:84% So there aren't too many of these. 00:44:52.620 --> 00:44:55.500 align:middle line:84% But you definitely have a density of them. 00:44:55.500 --> 00:44:57.300 align:middle line:90% So we create these nuclei. 00:44:57.300 --> 00:44:58.890 align:middle line:84% And then we actually do-- we grow 00:44:58.890 --> 00:45:02.310 align:middle line:84% them to be a little bigger so they become stable. 00:45:02.310 --> 00:45:04.950 align:middle line:84% So once you've created this nucleation step, you then 00:45:04.950 --> 00:45:07.320 align:middle line:84% take it to higher temperature, say 900 degrees. 00:45:07.320 --> 00:45:10.743 align:middle line:84% And you grow these nuclei to make them a little bit larger. 00:45:10.743 --> 00:45:11.910 align:middle line:90% So you're not creating more. 00:45:11.910 --> 00:45:14.220 align:middle line:90% You're just making them larger. 00:45:14.220 --> 00:45:17.010 align:middle line:84% And you typically want a size, a minimum size, 00:45:17.010 --> 00:45:20.748 align:middle line:84% say, in the range of 50 to 100 nanometers. 00:45:20.748 --> 00:45:22.540 align:middle line:84% You have to be careful how you ramp this up 00:45:22.540 --> 00:45:24.610 align:middle line:84% so you don't cause them to all shrink. 00:45:24.610 --> 00:45:26.890 align:middle line:84% The basic idea is then you've enabled 00:45:26.890 --> 00:45:28.728 align:middle line:84% yourself to create these SiO2 precipitates 00:45:28.728 --> 00:45:29.770 align:middle line:90% in the bulk of the wafer. 00:45:29.770 --> 00:45:32.720 align:middle line:90% 00:45:32.720 --> 00:45:35.140 align:middle line:84% So we go on to the next slide, slide 20. 00:45:35.140 --> 00:45:36.460 align:middle line:90% This, I took from your text. 00:45:36.460 --> 00:45:39.250 align:middle line:84% These are some actual cross-section 00:45:39.250 --> 00:45:42.190 align:middle line:90% scanning electron micrographs. 00:45:42.190 --> 00:45:44.340 align:middle line:84% And you're actually looking at the cross-section 00:45:44.340 --> 00:45:46.080 align:middle line:84% of a full wafer here in each picture. 00:45:46.080 --> 00:45:47.770 align:middle line:90% Each picture is a full wafer. 00:45:47.770 --> 00:45:51.150 align:middle line:84% So this top here at the very top is the top of the wafer. 00:45:51.150 --> 00:45:52.830 align:middle line:84% This bottom surface here at the bottom 00:45:52.830 --> 00:45:57.153 align:middle line:84% of one little rectangular region is the bottom of the wafer. 00:45:57.153 --> 00:45:59.820 align:middle line:84% And you'll notice, what you see are all these little black dots, 00:45:59.820 --> 00:46:03.180 align:middle line:90% are these SiO2 precipitates. 00:46:03.180 --> 00:46:05.640 align:middle line:84% And there's a bunch of snapshots here shown, 00:46:05.640 --> 00:46:07.225 align:middle line:90% starting from the top. 00:46:07.225 --> 00:46:08.850 align:middle line:84% And there are two different directions. 00:46:08.850 --> 00:46:12.610 align:middle line:84% If you go down vertically, we're reading the nucleation time. 00:46:12.610 --> 00:46:14.760 align:middle line:84% So remember, the nucleation time is the time 00:46:14.760 --> 00:46:16.620 align:middle line:90% that we nucleate these. 00:46:16.620 --> 00:46:18.170 align:middle line:90% And so if you-- 00:46:18.170 --> 00:46:19.920 align:middle line:84% starting at the top here in the upper left 00:46:19.920 --> 00:46:21.720 align:middle line:84% and moving down the first column, 00:46:21.720 --> 00:46:23.850 align:middle line:84% we're increasing the pre-anneal time. 00:46:23.850 --> 00:46:25.380 align:middle line:90% And what you're seeing is-- 00:46:25.380 --> 00:46:26.848 align:middle line:84% what you're doing is you're getting 00:46:26.848 --> 00:46:28.890 align:middle line:84% a higher and higher density of these precipitates 00:46:28.890 --> 00:46:33.390 align:middle line:84% as I move from the upper left down to the lower left. 00:46:33.390 --> 00:46:36.510 align:middle line:84% And then as I move right on this chart, 00:46:36.510 --> 00:46:39.210 align:middle line:90% I'm increasing the growth time. 00:46:39.210 --> 00:46:41.800 align:middle line:84% That was called the precipitation time or the amount 00:46:41.800 --> 00:46:43.450 align:middle line:90% of time we're growing these. 00:46:43.450 --> 00:46:46.790 align:middle line:84% Now these are grown at about 1,000 degrees. 00:46:46.790 --> 00:46:48.350 align:middle line:84% And you can see them growing here. 00:46:48.350 --> 00:46:51.050 align:middle line:84% And so by the time I get to this upper right, 00:46:51.050 --> 00:46:53.410 align:middle line:84% we have a certain density of these. 00:46:53.410 --> 00:46:57.550 align:middle line:84% But in general, if I move to longer nucleation 00:46:57.550 --> 00:47:01.530 align:middle line:84% time, so down here, and longer growth times, 00:47:01.530 --> 00:47:04.330 align:middle line:84% so I'm moving to the right, I get the highest density 00:47:04.330 --> 00:47:08.680 align:middle line:84% and a reasonable size of these precipitates in the wafer bulk. 00:47:08.680 --> 00:47:11.680 align:middle line:84% And you can see all of them here in the lower-- the bottom right 00:47:11.680 --> 00:47:12.947 align:middle line:90% corner. 00:47:12.947 --> 00:47:14.030 align:middle line:90% So these are precipitates. 00:47:14.030 --> 00:47:16.830 align:middle line:84% And look at the very surface of the wafer, the top 10 microns. 00:47:16.830 --> 00:47:17.690 align:middle line:90% You don't see any. 00:47:17.690 --> 00:47:18.860 align:middle line:90% That's the denuded zone. 00:47:18.860 --> 00:47:21.800 align:middle line:84% And that denuded zone was formed, of course, 00:47:21.800 --> 00:47:24.380 align:middle line:90% at the top and bottom. 00:47:24.380 --> 00:47:27.920 align:middle line:84% Denuded zone was created by doing an initial 1,000 degree C 00:47:27.920 --> 00:47:30.960 align:middle line:84% 10-hour anneal before any of this pre-annealing. 00:47:30.960 --> 00:47:33.950 align:middle line:84% So during that initial annealing, 00:47:33.950 --> 00:47:37.010 align:middle line:84% the oxygen diffused out at the top and bottom of the wafer 00:47:37.010 --> 00:47:39.560 align:middle line:90% for about 10 microns distance. 00:47:39.560 --> 00:47:41.370 align:middle line:84% And then it didn't diffuse out in here. 00:47:41.370 --> 00:47:43.843 align:middle line:84% And that's why when you go through this processing step, 00:47:43.843 --> 00:47:45.260 align:middle line:84% you can create these precipitates. 00:47:45.260 --> 00:47:47.420 align:middle line:84% And these are the precipitates that we 00:47:47.420 --> 00:47:50.000 align:middle line:84% want there so that metal ions can then diffuse 00:47:50.000 --> 00:47:52.850 align:middle line:90% from the devices and get stuck. 00:47:52.850 --> 00:47:55.260 align:middle line:84% And then keep them away from your devices. 00:47:55.260 --> 00:47:57.500 align:middle line:84% So that's the pictorial schematic of the process. 00:47:57.500 --> 00:48:01.120 align:middle line:90% 00:48:01.120 --> 00:48:07.330 align:middle line:84% So then let's go on to slide number 21 and now talk about way 00:48:07.330 --> 00:48:08.915 align:middle line:90% people model this process. 00:48:08.915 --> 00:48:11.290 align:middle line:84% And there are typically three steps people need to model. 00:48:11.290 --> 00:48:15.350 align:middle line:84% The first step here, shown number one, happens to be-- 00:48:15.350 --> 00:48:19.640 align:middle line:84% I'm showing a specific case for the gettering of gold 00:48:19.640 --> 00:48:21.473 align:middle line:84% because gold is a very, very common element. 00:48:21.473 --> 00:48:22.515 align:middle line:90% It needs to be gettering. 00:48:22.515 --> 00:48:24.100 align:middle line:84% And it has some unique properties. 00:48:24.100 --> 00:48:26.560 align:middle line:84% Step number 1, we need to mobilize the atom. 00:48:26.560 --> 00:48:29.060 align:middle line:84% So remember we talked about gold when it was substitutional. 00:48:29.060 --> 00:48:30.470 align:middle line:90% It diffused very slowly. 00:48:30.470 --> 00:48:32.750 align:middle line:84% If it's interstitial, it diffuses fast. 00:48:32.750 --> 00:48:35.540 align:middle line:84% So to make it mobilized, I need to somehow get it 00:48:35.540 --> 00:48:38.840 align:middle line:84% off substitutional sites and make it interstitial. 00:48:38.840 --> 00:48:41.180 align:middle line:84% The second step is shown schematically, step number 00:48:41.180 --> 00:48:43.850 align:middle line:84% 2 here, which is the gold has to actually diffuse away 00:48:43.850 --> 00:48:47.690 align:middle line:84% from the device region down to the trapping site, wherever 00:48:47.690 --> 00:48:49.110 align:middle line:90% it might be. 00:48:49.110 --> 00:48:54.790 align:middle line:84% And the third is, of course, the trapping mechanism, number 3. 00:48:54.790 --> 00:48:57.740 align:middle line:84% So people have tried to model these different aspects, 00:48:57.740 --> 00:49:00.670 align:middle line:84% the making mobile, the diffusion, the trapping, 00:49:00.670 --> 00:49:02.140 align:middle line:90% with mixed success. 00:49:02.140 --> 00:49:04.570 align:middle line:84% The models are actually not universally accepted. 00:49:04.570 --> 00:49:08.290 align:middle line:84% There is no really good program you can sit there and run 00:49:08.290 --> 00:49:10.648 align:middle line:84% that describes gettering, and very quantitatively, 00:49:10.648 --> 00:49:12.190 align:middle line:84% in the way that you can for oxidation 00:49:12.190 --> 00:49:13.520 align:middle line:90% that we'll see in this course. 00:49:13.520 --> 00:49:16.450 align:middle line:84% But they give you a physical mechanistic picture 00:49:16.450 --> 00:49:17.980 align:middle line:90% of how the getting operates. 00:49:17.980 --> 00:49:21.332 align:middle line:84% But they're not very quantitative. 00:49:21.332 --> 00:49:22.790 align:middle line:84% So let's look at those three steps. 00:49:22.790 --> 00:49:25.750 align:middle line:84% The first step was to make the atoms mobile. 00:49:25.750 --> 00:49:27.820 align:middle line:84% And remember, gold and a number of atoms 00:49:27.820 --> 00:49:31.810 align:middle line:84% can actually exist either on the lattice or in interstitial form. 00:49:31.810 --> 00:49:33.377 align:middle line:84% The diffusion is-- the diffusivity 00:49:33.377 --> 00:49:34.960 align:middle line:84% is much higher when it's interstitial. 00:49:34.960 --> 00:49:37.510 align:middle line:84% So to make the mobile really means 00:49:37.510 --> 00:49:40.342 align:middle line:84% to get them off the lattice into the interstitial spaces where 00:49:40.342 --> 00:49:41.050 align:middle line:90% they can diffuse. 00:49:41.050 --> 00:49:43.990 align:middle line:90% 00:49:43.990 --> 00:49:48.580 align:middle line:84% So then these-- just to show you some examples here 00:49:48.580 --> 00:49:51.010 align:middle line:84% of common metals and the different types of solubilities 00:49:51.010 --> 00:49:54.520 align:middle line:84% they have in interstitial or substitutional form. 00:49:54.520 --> 00:49:57.640 align:middle line:84% So copper and nickel have higher solubilities 00:49:57.640 --> 00:49:59.110 align:middle line:90% in interstitial form. 00:49:59.110 --> 00:50:01.030 align:middle line:84% So they tend to exist already interstitial. 00:50:01.030 --> 00:50:02.740 align:middle line:90% So mobilizing them is easy. 00:50:02.740 --> 00:50:05.020 align:middle line:84% They're already in the interstitial space. 00:50:05.020 --> 00:50:07.450 align:middle line:84% Gold and platinum actually have higher solubilities 00:50:07.450 --> 00:50:09.700 align:middle line:90% in substitutional form. 00:50:09.700 --> 00:50:11.340 align:middle line:84% But they can diffuse very rapidly 00:50:11.340 --> 00:50:12.840 align:middle line:90% once they become interstitial. 00:50:12.840 --> 00:50:15.540 align:middle line:90% So these have to be mobilized. 00:50:15.540 --> 00:50:20.070 align:middle line:84% Titanium and moly, molybdenum, are actually issue-- 00:50:20.070 --> 00:50:20.730 align:middle line:90% problems. 00:50:20.730 --> 00:50:23.190 align:middle line:84% They're primarily substitutional, 00:50:23.190 --> 00:50:26.190 align:middle line:84% but they have relatively slow diffusion rates, 00:50:26.190 --> 00:50:28.170 align:middle line:84% whether they're substitutional or interstitial. 00:50:28.170 --> 00:50:30.780 align:middle line:84% So Ti and moly are going to be hard to getter 00:50:30.780 --> 00:50:33.065 align:middle line:84% because their diffusion rates are not very large. 00:50:33.065 --> 00:50:35.190 align:middle line:84% So to get them all the way to the back of the wafer 00:50:35.190 --> 00:50:37.330 align:middle line:90% is going to take a long time. 00:50:37.330 --> 00:50:38.710 align:middle line:90% So it's just not very effective. 00:50:38.710 --> 00:50:43.170 align:middle line:84% So to getter Ti and moly in these intermediate or 00:50:43.170 --> 00:50:46.650 align:middle line:84% slow-diffusing metals, we prefer to use intrinsic gettering. 00:50:46.650 --> 00:50:49.770 align:middle line:84% So instead of going 500 microns, all it has to getter through 00:50:49.770 --> 00:50:50.670 align:middle line:90% is the denuded zone. 00:50:50.670 --> 00:50:53.080 align:middle line:84% All it has to diffuse through is about 10 microns. 00:50:53.080 --> 00:50:55.883 align:middle line:84% So that's another reason why intrinsic gettering 00:50:55.883 --> 00:50:56.550 align:middle line:90% is more popular. 00:50:56.550 --> 00:50:57.925 align:middle line:84% You don't have to diffuse as far. 00:50:57.925 --> 00:51:01.030 align:middle line:90% 00:51:01.030 --> 00:51:02.550 align:middle line:90% So let's go on to slide 23. 00:51:02.550 --> 00:51:04.830 align:middle line:84% And we'll just do an example of how 00:51:04.830 --> 00:51:09.140 align:middle line:84% people try to model the gettering of gold and silicon. 00:51:09.140 --> 00:51:11.850 align:middle line:84% Or this also applies to platinum. 00:51:11.850 --> 00:51:16.950 align:middle line:84% So the first thing, we said that gold can react with a silicon 00:51:16.950 --> 00:51:17.910 align:middle line:90% interstitial. 00:51:17.910 --> 00:51:22.350 align:middle line:84% And this Si sub i is meant to represent a silicon interstitial 00:51:22.350 --> 00:51:24.990 align:middle line:90% and form an interstitial gold. 00:51:24.990 --> 00:51:27.240 align:middle line:84% So an interstitial silicon atom can come along, 00:51:27.240 --> 00:51:29.280 align:middle line:84% take the gold which is on the silicon lattice, 00:51:29.280 --> 00:51:31.240 align:middle line:84% and knock it off, and take its place. 00:51:31.240 --> 00:51:33.250 align:middle line:84% And that's called the kickout mechanism. 00:51:33.250 --> 00:51:36.390 align:middle line:84% So that's one way of getting the gold off of lattice site 00:51:36.390 --> 00:51:38.340 align:middle line:84% because there are other mechanisms. 00:51:38.340 --> 00:51:41.610 align:middle line:84% An interstitial can react with a vacancy 00:51:41.610 --> 00:51:44.250 align:middle line:90% and form substitutional gold. 00:51:44.250 --> 00:51:47.520 align:middle line:84% So basically, the idea, if you look at this first reaction, 00:51:47.520 --> 00:51:50.100 align:middle line:84% this first equation, any process that 00:51:50.100 --> 00:51:53.542 align:middle line:84% creates excess interstitials should be helpful in gettering. 00:51:53.542 --> 00:51:55.500 align:middle line:84% If I have a lot of excess silicon interstitials 00:51:55.500 --> 00:51:57.503 align:middle line:84% around whenever there's a gold, it 00:51:57.503 --> 00:51:59.670 align:middle line:84% can just knock it off the lattice and get it moving. 00:51:59.670 --> 00:52:02.180 align:middle line:90% 00:52:02.180 --> 00:52:04.652 align:middle line:84% So process that-- any process that 00:52:04.652 --> 00:52:06.110 align:middle line:84% creates excess interstitials should 00:52:06.110 --> 00:52:10.856 align:middle line:84% be helpful in this gettering, this mobilization process. 00:52:10.856 --> 00:52:13.898 align:middle line:84% Processes that create excess vacancies will tend to hinder 00:52:13.898 --> 00:52:16.190 align:middle line:84% gettering because they're-- if you have a lot of excess 00:52:16.190 --> 00:52:18.660 align:middle line:84% vacancies, it's going to drive this back to the left, 00:52:18.660 --> 00:52:21.320 align:middle line:84% the second reaction, and cause gold to go back 00:52:21.320 --> 00:52:23.110 align:middle line:90% to substitutional. 00:52:23.110 --> 00:52:27.550 align:middle line:84% So interestingly, just from an empirical point of view, 00:52:27.550 --> 00:52:29.440 align:middle line:84% a lot of the things people do for gettering 00:52:29.440 --> 00:52:33.370 align:middle line:84% happen to also be known to create excess interstitials. 00:52:33.370 --> 00:52:36.110 align:middle line:84% So for example, backside phosphorus diffusion, 00:52:36.110 --> 00:52:38.280 align:middle line:84% when we talk about phosphorus diffusion, 00:52:38.280 --> 00:52:41.650 align:middle line:84% we'll see how it tends to inject a lot of excess interstitials. 00:52:41.650 --> 00:52:45.820 align:middle line:84% Ion implantation, in which we shoot ions into the crystal, 00:52:45.820 --> 00:52:48.160 align:middle line:84% tend to knock silicon off lattice sites 00:52:48.160 --> 00:52:50.500 align:middle line:84% and create a lot of interstitials. 00:52:50.500 --> 00:52:52.840 align:middle line:84% And internal gathering or intrinsic gettering 00:52:52.840 --> 00:52:56.050 align:middle line:84% involves oxidizing silicon, which we will find out. 00:52:56.050 --> 00:52:59.980 align:middle line:84% Creating SiO2 precipitates tends to punch out 00:52:59.980 --> 00:53:02.410 align:middle line:84% a lot of excess silicon interstitials. 00:53:02.410 --> 00:53:07.720 align:middle line:84% So qualitatively can we say that the things that we know work 00:53:07.720 --> 00:53:10.570 align:middle line:84% for gettering also tend to create a lot of interstitials. 00:53:10.570 --> 00:53:12.880 align:middle line:84% And that would be consistent with this type of model, 00:53:12.880 --> 00:53:16.090 align:middle line:84% the fact that some of the metals need to be able to be mobilized. 00:53:16.090 --> 00:53:18.620 align:middle line:90% 00:53:18.620 --> 00:53:20.330 align:middle line:90% So then go on to slide 24. 00:53:20.330 --> 00:53:22.280 align:middle line:84% And the second step is, once its mobile, 00:53:22.280 --> 00:53:26.260 align:middle line:84% is to get the metal to diffuse to the gettered site 00:53:26.260 --> 00:53:28.160 align:middle line:90% in this interstitial form. 00:53:28.160 --> 00:53:30.310 align:middle line:84% And so it has to diffuse to the back side 00:53:30.310 --> 00:53:33.550 align:middle line:90% or into the intrinsic region. 00:53:33.550 --> 00:53:37.340 align:middle line:84% Let's say we're doing diffusion to the back side. 00:53:37.340 --> 00:53:42.730 align:middle line:84% This plot on slide number 24 is a plot of the concentration 00:53:42.730 --> 00:53:44.320 align:middle line:90% of that gold. 00:53:44.320 --> 00:53:50.350 align:middle line:84% It has a function of depth in the wafer, where the 0,0 00:53:50.350 --> 00:53:52.900 align:middle line:84% point, or this point right here, is supposed to represent 00:53:52.900 --> 00:53:54.850 align:middle line:90% the wafer backside. 00:53:54.850 --> 00:53:56.350 align:middle line:84% So that's the backside of the wafer. 00:53:56.350 --> 00:53:59.168 align:middle line:84% As you move to the right, you're going through the thickness 00:53:59.168 --> 00:53:59.710 align:middle line:90% of the wafer. 00:53:59.710 --> 00:54:01.510 align:middle line:84% So you're going into the center of the wafer. 00:54:01.510 --> 00:54:03.010 align:middle line:84% And I move all the way to the window, 00:54:03.010 --> 00:54:04.593 align:middle line:84% I'd be at the front side of the wafer. 00:54:04.593 --> 00:54:07.590 align:middle line:84% So imagine the wafer is standing up sideways. 00:54:07.590 --> 00:54:10.520 align:middle line:84% And this is a plot of different time 00:54:10.520 --> 00:54:14.000 align:middle line:84% contours of what the gold concentration would look like. 00:54:14.000 --> 00:54:17.260 align:middle line:84% Initially, at time t equals 0, it's a flat line. 00:54:17.260 --> 00:54:18.960 align:middle line:90% It's just at 10 to the 15. 00:54:18.960 --> 00:54:20.750 align:middle line:84% So let's say, somehow, the wafer got 00:54:20.750 --> 00:54:23.570 align:middle line:84% 10 to the 15th per cubic atoms of gold in it 00:54:23.570 --> 00:54:25.110 align:middle line:90% uniformly distributed. 00:54:25.110 --> 00:54:27.630 align:middle line:84% Now you heat it up and you increase the amount of time. 00:54:27.630 --> 00:54:32.660 align:middle line:84% And as I heat it, basically, the gold 00:54:32.660 --> 00:54:34.310 align:middle line:90% is diffusing out of the wafer. 00:54:34.310 --> 00:54:35.870 align:middle line:84% It's lowering here in the center. 00:54:35.870 --> 00:54:37.830 align:middle line:84% And at the edge, it's diffusing down. 00:54:37.830 --> 00:54:40.160 align:middle line:84% So this is what we expect the gold profiles 00:54:40.160 --> 00:54:45.490 align:middle line:84% to look like over time, just in a rough sense. 00:54:45.490 --> 00:54:48.460 align:middle line:84% If we go on to slide 25, actually, there 00:54:48.460 --> 00:54:49.930 align:middle line:90% are some plots which-- 00:54:49.930 --> 00:54:52.870 align:middle line:84% there's a plot which shows, as a function of time, 00:54:52.870 --> 00:54:55.780 align:middle line:84% at 1,000 degrees, what happens to the gold concentration 00:54:55.780 --> 00:54:58.570 align:middle line:84% profiles, actually, experimentally. 00:54:58.570 --> 00:55:01.120 align:middle line:84% And this represents the observed profiles. 00:55:01.120 --> 00:55:03.790 align:middle line:84% They're actually not quite like what was shown in slide 24. 00:55:03.790 --> 00:55:06.620 align:middle line:84% If I go back here, you see this is a much more-- in slide 24, 00:55:06.620 --> 00:55:09.580 align:middle line:84% it's a much more gradual drop off on this plot. 00:55:09.580 --> 00:55:12.310 align:middle line:84% In slide 25, in fact, it's very abrupt. 00:55:12.310 --> 00:55:18.010 align:middle line:84% It's very sudden than predicted by this simple model. 00:55:18.010 --> 00:55:21.100 align:middle line:84% Well, it turns out that people explain the sudden drop off 00:55:21.100 --> 00:55:24.910 align:middle line:84% by the fact that the silicon interstitial diffusion is fast. 00:55:24.910 --> 00:55:28.260 align:middle line:84% But it's actually slower than the metal atom diffusion. 00:55:28.260 --> 00:55:30.380 align:middle line:84% So the rate-limiting process in this step 00:55:30.380 --> 00:55:32.450 align:middle line:84% is for the silicon interstitials to diffuse in 00:55:32.450 --> 00:55:37.250 align:middle line:84% the back or inside into the wafer and mobilize the gold. 00:55:37.250 --> 00:55:39.050 align:middle line:84% So as soon as the silicon interstitials 00:55:39.050 --> 00:55:41.053 align:middle line:84% arrive at a particular location, the gold 00:55:41.053 --> 00:55:42.470 align:middle line:84% is converted to interstitial form. 00:55:42.470 --> 00:55:43.880 align:middle line:90% And then it's got-- 00:55:43.880 --> 00:55:48.050 align:middle line:84% it's diffusing so fast, it can rapidly diffuse out the wafer. 00:55:48.050 --> 00:55:51.920 align:middle line:84% So what's rate limiting here is the indiffusion of the silicon 00:55:51.920 --> 00:55:53.090 align:middle line:90% interstitials. 00:55:53.090 --> 00:55:56.660 align:middle line:84% So this sharp drop off point here, you see at 12 seconds-- 00:55:56.660 --> 00:56:00.500 align:middle line:84% 12 minutes, it's dropping off here, 20, in 25, 30 minutes. 00:56:00.500 --> 00:56:03.530 align:middle line:84% That sharp drop point corresponds to the depth people 00:56:03.530 --> 00:56:05.900 align:middle line:84% believe which the silicon interstitials must 00:56:05.900 --> 00:56:07.440 align:middle line:90% have diffused in a given time. 00:56:07.440 --> 00:56:09.860 align:middle line:84% In fact, in the old days, people used this 00:56:09.860 --> 00:56:13.640 align:middle line:84% as a marker for how to measure or estimate silicon interstitial 00:56:13.640 --> 00:56:14.510 align:middle line:90% diffusion. 00:56:14.510 --> 00:56:16.310 align:middle line:84% They couldn't see the interstitials, 00:56:16.310 --> 00:56:19.250 align:middle line:84% but they hypothesized they were having this effect on the gold. 00:56:19.250 --> 00:56:22.580 align:middle line:84% So they would model the gold diffusion and the silicon 00:56:22.580 --> 00:56:25.440 align:middle line:84% interstitial in diffusion that's associated with that. 00:56:25.440 --> 00:56:30.870 align:middle line:84% And so that was one method of estimating silicon in diffusion 00:56:30.870 --> 00:56:34.297 align:middle line:84% So anyway, that's just a practical issue with the way 00:56:34.297 --> 00:56:35.505 align:middle line:90% gold diffuses through wafers. 00:56:35.505 --> 00:56:39.140 align:middle line:90% 00:56:39.140 --> 00:56:41.960 align:middle line:84% And in fact, on slide 26, this is just 00:56:41.960 --> 00:56:44.360 align:middle line:84% to remind you of how the relationship 00:56:44.360 --> 00:56:47.750 align:middle line:84% or the relative magnitude of the gold diffusion, 00:56:47.750 --> 00:56:49.790 align:middle line:84% look at the gold interstitial diffusivity 00:56:49.790 --> 00:56:51.800 align:middle line:84% up here, compared to the silicon-- 00:56:51.800 --> 00:56:56.220 align:middle line:84% the diffusion of silicon self-interstitials. 00:56:56.220 --> 00:57:00.420 align:middle line:84% If you have an element like titanium, which diffuses slower 00:57:00.420 --> 00:57:02.220 align:middle line:84% than silicon interstitials, it's going 00:57:02.220 --> 00:57:04.140 align:middle line:84% to have a more classical type of profile, 00:57:04.140 --> 00:57:07.530 align:middle line:84% as shown here on the lower right, 00:57:07.530 --> 00:57:10.200 align:middle line:84% for its diffusion profile of titanium 00:57:10.200 --> 00:57:13.740 align:middle line:84% outdiffusing or diffusing to the back of the wafer. 00:57:13.740 --> 00:57:16.758 align:middle line:90% 00:57:16.758 --> 00:57:17.800 align:middle line:90% So we've made the mobile. 00:57:17.800 --> 00:57:19.220 align:middle line:84% We got them diffused to the site. 00:57:19.220 --> 00:57:20.290 align:middle line:90% The final thing is trap them. 00:57:20.290 --> 00:57:21.820 align:middle line:84% It doesn't do any good to make them diffuse there 00:57:21.820 --> 00:57:23.695 align:middle line:84% and then have them come right back out again. 00:57:23.695 --> 00:57:26.320 align:middle line:90% So they have to be trapped. 00:57:26.320 --> 00:57:28.750 align:middle line:84% And this tends to be very, very empirical. 00:57:28.750 --> 00:57:30.400 align:middle line:84% But people have experimentally observed 00:57:30.400 --> 00:57:33.310 align:middle line:84% that certain types of backside damage trap metal atoms. 00:57:33.310 --> 00:57:36.100 align:middle line:84% If you measure the concentration in that backside region 00:57:36.100 --> 00:57:38.350 align:middle line:84% after you've damaged it, it has a higher concentration 00:57:38.350 --> 00:57:40.840 align:middle line:90% of metals in it. 00:57:40.840 --> 00:57:42.310 align:middle line:90% It's hard to model this. 00:57:42.310 --> 00:57:44.650 align:middle line:84% But one approach is to model it using 00:57:44.650 --> 00:57:50.740 align:middle line:84% the mathematics of segregation and to talk about the gold 00:57:50.740 --> 00:57:52.330 align:middle line:84% as it exists in the bulk and the gold 00:57:52.330 --> 00:57:54.455 align:middle line:84% as it exists in the trap site and compare those two 00:57:54.455 --> 00:57:56.830 align:middle line:90% concentrations. 00:57:56.830 --> 00:57:59.770 align:middle line:84% Remember, in chapter 3, we talked about segregation, 00:57:59.770 --> 00:58:02.890 align:middle line:84% described the doping behavior as it segregated 00:58:02.890 --> 00:58:05.620 align:middle line:84% between a liquid phase and the solid phase 00:58:05.620 --> 00:58:07.280 align:middle line:90% during Czochralski growth. 00:58:07.280 --> 00:58:10.720 align:middle line:84% Well, the same concept here really applies here. 00:58:10.720 --> 00:58:12.910 align:middle line:84% We have the segregation of these gold 00:58:12.910 --> 00:58:14.590 align:middle line:84% or this metal atom between the silicon 00:58:14.590 --> 00:58:17.870 align:middle line:84% bulk and the trapped region, or the backside trapped region. 00:58:17.870 --> 00:58:22.780 align:middle line:84% So we can use similar mathematics. 00:58:22.780 --> 00:58:24.280 align:middle line:84% And that's what's shown on slide 28. 00:58:24.280 --> 00:58:28.180 align:middle line:84% This is a relatively simple way of looking at a segregation 00:58:28.180 --> 00:58:31.690 align:middle line:84% model for metal atom trapping during gettering. 00:58:31.690 --> 00:58:34.210 align:middle line:84% And what we do is we write down the solubility 00:58:34.210 --> 00:58:35.620 align:middle line:90% of gold in two states. 00:58:35.620 --> 00:58:37.900 align:middle line:84% On the upper equation is the solubility 00:58:37.900 --> 00:58:40.310 align:middle line:90% of gold in bulk silicon. 00:58:40.310 --> 00:58:42.100 align:middle line:84% So this is not in the gettered region, 00:58:42.100 --> 00:58:43.810 align:middle line:90% just existing in the silicon. 00:58:43.810 --> 00:58:46.660 align:middle line:84% And it has some exponential dependence on an activation 00:58:46.660 --> 00:58:50.320 align:middle line:90% energy, Ea1 over kT. 00:58:50.320 --> 00:58:53.530 align:middle line:84% And the second equation down is the solubility 00:58:53.530 --> 00:58:59.230 align:middle line:84% of gold in the gettered site, G. So it's got this G next to it 00:58:59.230 --> 00:59:00.280 align:middle line:90% in the gettered region. 00:59:00.280 --> 00:59:04.540 align:middle line:84% And here Ng, N sub g, is the density of gathering sites. 00:59:04.540 --> 00:59:08.725 align:middle line:84% N sub silicon here is the density of silicon atomic sites. 00:59:08.725 --> 00:59:11.420 align:middle line:90% 00:59:11.420 --> 00:59:14.140 align:middle line:84% And what we define as the segregation coefficient, 00:59:14.140 --> 00:59:16.360 align:middle line:84% just as we did before in liquid solid segregation, 00:59:16.360 --> 00:59:23.200 align:middle line:84% K0, as to be the ratio of the concentration of the total gold 00:59:23.200 --> 00:59:27.050 align:middle line:84% in the wafer divided by that in-- that's in the silicon. 00:59:27.050 --> 00:59:29.680 align:middle line:84% So that's our segregation coefficient. 00:59:29.680 --> 00:59:31.870 align:middle line:90% And we can just simply ratio-- 00:59:31.870 --> 00:59:34.300 align:middle line:84% plug-in these exponentials and ratio them, as 00:59:34.300 --> 00:59:36.900 align:middle line:90% shown in the equation. 00:59:36.900 --> 00:59:41.900 align:middle line:84% So if Ea1, the activation energy a1, minus Ea2 00:59:41.900 --> 00:59:44.180 align:middle line:84% is a positive quantity, then K0 is 00:59:44.180 --> 00:59:46.500 align:middle line:84% going to decrease as temperature increases. 00:59:46.500 --> 00:59:51.400 align:middle line:84% So this tells us something about where the metal wants to reside. 00:59:51.400 --> 00:59:55.900 align:middle line:84% In fact, empirically for phosphorus backside gettering, 00:59:55.900 --> 00:59:58.210 align:middle line:84% it's been found that the segregation coefficient can 00:59:58.210 --> 01:00:01.690 align:middle line:84% be written like this, a simple equation, where it has 01:00:01.690 --> 01:00:03.890 align:middle line:90% a positive activation energy. 01:00:03.890 --> 01:00:06.640 align:middle line:84% So what this says is we need to increase the number of gettering 01:00:06.640 --> 01:00:08.860 align:middle line:84% sites, Ng, which, of course, just 01:00:08.860 --> 01:00:10.510 align:middle line:84% means keep the amount of phosphorus 01:00:10.510 --> 01:00:12.130 align:middle line:90% in the backside region high. 01:00:12.130 --> 01:00:15.493 align:middle line:84% And we keep the temperature low, trap it there 01:00:15.493 --> 01:00:16.910 align:middle line:84% with a relatively low temperature. 01:00:16.910 --> 01:00:19.270 align:middle line:84% So we don't want to go too high to get the highest 01:00:19.270 --> 01:00:20.780 align:middle line:90% amount of segregation. 01:00:20.780 --> 01:00:22.420 align:middle line:84% So again, it's fairly qualitative. 01:00:22.420 --> 01:00:24.700 align:middle line:84% But it gives some estimate of what-- 01:00:24.700 --> 01:00:28.690 align:middle line:84% of how gettering works as a function of temperature 01:00:28.690 --> 01:00:31.210 align:middle line:90% trapping. 01:00:31.210 --> 01:00:34.310 align:middle line:90% That's one example of a model. 01:00:34.310 --> 01:00:36.640 align:middle line:84% The second example is shown on page 29. 01:00:36.640 --> 01:00:40.450 align:middle line:84% And this actually goes back to work 01:00:40.450 --> 01:00:45.480 align:middle line:84% that Shockley and Mole did very long time ago back in the 1960s. 01:00:45.480 --> 01:00:50.262 align:middle line:84% And this work was to understand the enhanced solubility 01:00:50.262 --> 01:00:51.720 align:middle line:84% of metals in heavily-doped silicon. 01:00:51.720 --> 01:00:55.500 align:middle line:84% It was observed that certain metals are more soluble. 01:00:55.500 --> 01:00:57.690 align:middle line:84% They tend to want to be in heavily-doped silicon. 01:00:57.690 --> 01:01:00.092 align:middle line:84% They have a higher concentration in heavily-doped silicon 01:01:00.092 --> 01:01:01.800 align:middle line:84% than they would in lightly-doped silicon. 01:01:01.800 --> 01:01:04.410 align:middle line:90% 01:01:04.410 --> 01:01:07.380 align:middle line:84% And this is a model that helps to explain that. 01:01:07.380 --> 01:01:10.410 align:middle line:84% So in substitutional form, we know 01:01:10.410 --> 01:01:13.850 align:middle line:84% that gold introduces deep levels in the bandgap 01:01:13.850 --> 01:01:15.100 align:middle line:90% and near the mid-gap position. 01:01:15.100 --> 01:01:17.520 align:middle line:84% In fact, there's one here, an acceptor level marked here. 01:01:17.520 --> 01:01:20.175 align:middle line:84% And there's one here, a donor level, marked there. 01:01:20.175 --> 01:01:23.240 align:middle line:90% 01:01:23.240 --> 01:01:26.950 align:middle line:84% So this Au minus level is going to be 01:01:26.950 --> 01:01:30.500 align:middle line:84% created whenever the gold can capture a free electron. 01:01:30.500 --> 01:01:32.620 align:middle line:84% So if there's a lot of free electrons, 01:01:32.620 --> 01:01:37.430 align:middle line:84% we can create this Au minus state, so this acceptor level. 01:01:37.430 --> 01:01:40.360 align:middle line:84% So I can just write down this simple chemical equation, gold, 01:01:40.360 --> 01:01:45.245 align:middle line:84% one gold atom, plus an electron goes to an Au minus ion. 01:01:45.245 --> 01:01:46.870 align:middle line:84% Now, if you're familiar with chemistry, 01:01:46.870 --> 01:01:48.520 align:middle line:84% you know you can write any reaction. 01:01:48.520 --> 01:01:50.830 align:middle line:84% You can write an equilibrium constant 01:01:50.830 --> 01:01:52.750 align:middle line:84% associated with that reaction rate that 01:01:52.750 --> 01:01:55.770 align:middle line:90% depends only on temperature. 01:01:55.770 --> 01:01:58.260 align:middle line:84% So we write this number, K sub equilibrium. 01:01:58.260 --> 01:02:02.000 align:middle line:90% It depends only on temperature. 01:02:02.000 --> 01:02:05.900 align:middle line:84% It says that the reactants, the concentration of reactants, 01:02:05.900 --> 01:02:10.040 align:middle line:84% Au minus, divided by the concentration of the product 01:02:10.040 --> 01:02:12.470 align:middle line:90% of the concentration of the-- 01:02:12.470 --> 01:02:15.080 align:middle line:84% or this is the products divided by-- 01:02:15.080 --> 01:02:17.960 align:middle line:84% I multiply the concentration of the reactants. 01:02:17.960 --> 01:02:20.145 align:middle line:90% That has to be a constant. 01:02:20.145 --> 01:02:22.520 align:middle line:84% So the right-hand side of the equation, the concentration 01:02:22.520 --> 01:02:26.610 align:middle line:84% of Au minus divided by the gold concentration times the electron 01:02:26.610 --> 01:02:28.110 align:middle line:84% concentration, has to be a constant. 01:02:28.110 --> 01:02:30.540 align:middle line:90% It depends only on temperature. 01:02:30.540 --> 01:02:35.150 align:middle line:84% So that's an important equilibrium relationship. 01:02:35.150 --> 01:02:37.220 align:middle line:84% If we go on to slide number 30, here 01:02:37.220 --> 01:02:39.490 align:middle line:84% just at the upper left-hand side, 01:02:39.490 --> 01:02:42.090 align:middle line:90% just repeated that equation. 01:02:42.090 --> 01:02:47.260 align:middle line:84% This equation has to hold both when we have intrinsic silicon-- 01:02:47.260 --> 01:02:49.450 align:middle line:84% an intrinsic silicon, the electron concentration 01:02:49.450 --> 01:02:50.470 align:middle line:90% is just ni-- 01:02:50.470 --> 01:02:52.720 align:middle line:90% or when it's extrinsic. 01:02:52.720 --> 01:02:54.640 align:middle line:84% In extrinsic, the electron concentration 01:02:54.640 --> 01:03:00.740 align:middle line:84% is just n, whatever you've doped it to, the donor density level. 01:03:00.740 --> 01:03:03.010 align:middle line:84% So if that's the case, I can write this equation 01:03:03.010 --> 01:03:05.530 align:middle line:90% for those two cases. 01:03:05.530 --> 01:03:06.880 align:middle line:90% And it has to hold. 01:03:06.880 --> 01:03:11.770 align:middle line:84% So basically, I can write down, as it turns out, 01:03:11.770 --> 01:03:17.740 align:middle line:84% that the ratio of the gold, Au minus, 01:03:17.740 --> 01:03:21.130 align:middle line:84% in N-type material to its concentration 01:03:21.130 --> 01:03:25.370 align:middle line:84% in intrinsic material just has to go like n over ni. 01:03:25.370 --> 01:03:28.640 align:middle line:84% So then what that means is the solubility of gold 01:03:28.640 --> 01:03:31.860 align:middle line:84% as an acceptor is going to be higher in N-type material. 01:03:31.860 --> 01:03:34.850 align:middle line:84% So as I raise n over ni, let's say I dope-- 01:03:34.850 --> 01:03:36.320 align:middle line:90% I'm at 1,000 degrees. 01:03:36.320 --> 01:03:38.210 align:middle line:90% ni is 7 times 10 to the 18th. 01:03:38.210 --> 01:03:42.100 align:middle line:84% And I create a heavily doped region, 10 to the 21. 01:03:42.100 --> 01:03:47.470 align:middle line:84% Then just looking at n over ni is a factor of 100, roughly. 01:03:47.470 --> 01:03:49.240 align:middle line:84% So that means the gold solubility 01:03:49.240 --> 01:03:52.450 align:middle line:84% is going to be 100 times higher in this heavily doped 01:03:52.450 --> 01:03:55.960 align:middle line:84% 10 to the 21 region compared to the intrinsic region 01:03:55.960 --> 01:03:57.700 align:middle line:90% of the wafer. 01:03:57.700 --> 01:04:01.870 align:middle line:84% So again, this is looking at it from the electrochemical point 01:04:01.870 --> 01:04:02.470 align:middle line:90% of view. 01:04:02.470 --> 01:04:04.090 align:middle line:84% And it's a little bit hand-wavey. 01:04:04.090 --> 01:04:05.650 align:middle line:84% But it's consistent with the fact 01:04:05.650 --> 01:04:08.440 align:middle line:84% that when people do heavy n plus diffusion regions 01:04:08.440 --> 01:04:10.360 align:middle line:84% on the back of the wafer, like phosphorus, 01:04:10.360 --> 01:04:12.820 align:middle line:84% they can very effectively hold metals 01:04:12.820 --> 01:04:15.820 align:middle line:84% because it has a higher solubility, just because 01:04:15.820 --> 01:04:18.505 align:middle line:84% of the electrochemistry involved. 01:04:18.505 --> 01:04:21.400 align:middle line:90% 01:04:21.400 --> 01:04:25.570 align:middle line:84% So on slide 21, it's qualitatively 01:04:25.570 --> 01:04:27.610 align:middle line:84% listed as a series of explanations 01:04:27.610 --> 01:04:30.040 align:middle line:84% of why n plus silicon is a good getter. 01:04:30.040 --> 01:04:32.530 align:middle line:90% It's a good trap, trapping area. 01:04:32.530 --> 01:04:35.050 align:middle line:84% We just saw the statistics of the acceptor and donor levels 01:04:35.050 --> 01:04:37.120 align:middle line:84% versus doping from Shockley's paper that showed-- 01:04:37.120 --> 01:04:39.760 align:middle line:90% that is consistent with that. 01:04:39.760 --> 01:04:41.650 align:middle line:84% There's also people that have an ion pairing 01:04:41.650 --> 01:04:45.260 align:middle line:84% model, that a large atom like gold, well, gold is very large, 01:04:45.260 --> 01:04:47.650 align:middle line:84% it might want to pair with a small atom like phosphorus 01:04:47.650 --> 01:04:50.890 align:middle line:84% and form some kind of AuP complex 01:04:50.890 --> 01:04:54.550 align:middle line:90% because this minimizes strain. 01:04:54.550 --> 01:04:57.530 align:middle line:84% The other issue is that point defect concentrations 01:04:57.530 --> 01:04:59.530 align:middle line:84% are much higher in doped silicon than intrinsic. 01:04:59.530 --> 01:05:01.030 align:middle line:84% Remember, last time we talked about, 01:05:01.030 --> 01:05:02.740 align:middle line:84% as I move the Fermi level around, 01:05:02.740 --> 01:05:07.730 align:middle line:84% I can create more interstitials or vacancies. 01:05:07.730 --> 01:05:10.730 align:middle line:84% Gold diffuses primarily by an interstitial mechanism. 01:05:10.730 --> 01:05:12.380 align:middle line:84% Once it arrives in the getter region, 01:05:12.380 --> 01:05:15.350 align:middle line:84% it needs to find a lattice site to become substitutional. 01:05:15.350 --> 01:05:18.290 align:middle line:84% So you can imagine an interstitial gold might react 01:05:18.290 --> 01:05:20.088 align:middle line:90% with a vacancy in the lattice. 01:05:20.088 --> 01:05:22.130 align:middle line:84% And it gets stuck there to make it substitutional 01:05:22.130 --> 01:05:24.047 align:middle line:84% because, remember, substitutional gold doesn't 01:05:24.047 --> 01:05:25.440 align:middle line:90% diffuse very fast. 01:05:25.440 --> 01:05:27.860 align:middle line:84% So in n plus silicon, we saw last time, 01:05:27.860 --> 01:05:30.140 align:middle line:84% it has a much higher population of vacancies 01:05:30.140 --> 01:05:31.268 align:middle line:90% than in intrinsic silicon. 01:05:31.268 --> 01:05:33.560 align:middle line:84% And so that's going to tend to drive this to the right. 01:05:33.560 --> 01:05:35.840 align:middle line:84% So once the gold gets-- it diffuses 01:05:35.840 --> 01:05:38.030 align:middle line:84% by interstitial mechanism, it gets to the back side 01:05:38.030 --> 01:05:40.310 align:middle line:84% to the n plus, there's plenty of vacancies 01:05:40.310 --> 01:05:43.960 align:middle line:84% around to give it a lattice site to get stuck onto. 01:05:43.960 --> 01:05:46.570 align:middle line:84% So that's another, again, somewhat qualitative 01:05:46.570 --> 01:05:49.600 align:middle line:90% explanation. 01:05:49.600 --> 01:05:51.810 align:middle line:84% So that's for backside gettering with an n 01:05:51.810 --> 01:05:55.980 align:middle line:84% plus region, such as phosphorus, which was the classic method. 01:05:55.980 --> 01:05:58.940 align:middle line:90% How about intrinsic gettering? 01:05:58.940 --> 01:06:00.860 align:middle line:84% What kind of models do people have 01:06:00.860 --> 01:06:05.720 align:middle line:84% for gettering atoms or-- metal atoms near SiO2 precipitates? 01:06:05.720 --> 01:06:10.100 align:middle line:84% Well, actually, this tends to be a lot more qualitative. 01:06:10.100 --> 01:06:14.740 align:middle line:84% But if you go back to chapter 3, chapter 3 01:06:14.740 --> 01:06:19.070 align:middle line:84% talks about how SiO2 precipitation takes place. 01:06:19.070 --> 01:06:21.620 align:middle line:84% And in fact, it involves a net-- volume expansion. 01:06:21.620 --> 01:06:23.530 align:middle line:90% The SiO2, when you add-- 01:06:23.530 --> 01:06:27.910 align:middle line:84% when you form it, it tends to increase the lattice locally, 01:06:27.910 --> 01:06:30.190 align:middle line:90% wherever that is. 01:06:30.190 --> 01:06:32.710 align:middle line:84% And it actually compresses the lattice around it. 01:06:32.710 --> 01:06:35.680 align:middle line:84% So people have written this-- it looks fairly complex. 01:06:35.680 --> 01:06:37.540 align:middle line:90% But each term has a meaning. 01:06:37.540 --> 01:06:41.380 align:middle line:84% This equation, which describes the formation internally 01:06:41.380 --> 01:06:45.160 align:middle line:84% of silicon reacting with oxygen interstitials. 01:06:45.160 --> 01:06:47.475 align:middle line:84% So here's a silicon lattice site here. 01:06:47.475 --> 01:06:48.850 align:middle line:84% A certain number of lattice sites 01:06:48.850 --> 01:06:53.770 align:middle line:84% are involved, reacting with oxygen interstitial, O sub i, 01:06:53.770 --> 01:06:55.870 align:middle line:90% and forming SiO2. 01:06:55.870 --> 01:06:57.110 align:middle line:90% It could be at the surface. 01:06:57.110 --> 01:06:58.090 align:middle line:90% It could be internally. 01:06:58.090 --> 01:07:02.230 align:middle line:84% But you're forming a small SiO2 region, along 01:07:02.230 --> 01:07:05.938 align:middle line:84% with some stress that gets involved. 01:07:05.938 --> 01:07:07.480 align:middle line:84% Look at this fairly complex equation. 01:07:07.480 --> 01:07:10.780 align:middle line:84% But some of the elements of it are identified. 01:07:10.780 --> 01:07:12.980 align:middle line:84% Gamma here is the number of interstitials 01:07:12.980 --> 01:07:16.400 align:middle line:84% that contribute to the precipitation process per oxygen 01:07:16.400 --> 01:07:21.030 align:middle line:84% atom that join into the SiO2 precipitate. 01:07:21.030 --> 01:07:24.557 align:middle line:90% So a certain number of silicon-- 01:07:24.557 --> 01:07:25.640 align:middle line:90% silicon is a lattice site. 01:07:25.640 --> 01:07:28.010 align:middle line:84% A certain number of silicon lattice sites 01:07:28.010 --> 01:07:30.530 align:middle line:90% participate in this reaction. 01:07:30.530 --> 01:07:33.350 align:middle line:84% Interestingly, look at the process. 01:07:33.350 --> 01:07:35.070 align:middle line:90% It consumes vacancies. 01:07:35.070 --> 01:07:39.140 align:middle line:84% So vacancies on the left have to be consumed. 01:07:39.140 --> 01:07:41.150 align:middle line:84% And it injects on the right interstitials. 01:07:41.150 --> 01:07:43.288 align:middle line:84% So the formation of SiO2, and we're 01:07:43.288 --> 01:07:45.080 align:middle line:84% going to see this in a lot more detail when 01:07:45.080 --> 01:07:49.070 align:middle line:84% we talk about oxidation, planar oxidation, 01:07:49.070 --> 01:07:52.382 align:middle line:84% involves the injection of excess interstitials, which 01:07:52.382 --> 01:07:54.590 align:middle line:84% is interesting because excess interstitials, we know, 01:07:54.590 --> 01:07:57.840 align:middle line:84% are important for mobilizing certain metal atoms. 01:07:57.840 --> 01:08:03.840 align:middle line:84% So that could be one reason why SiO2 precipitation is important. 01:08:03.840 --> 01:08:07.173 align:middle line:84% We know that the stress term is generated because it compresses 01:08:07.173 --> 01:08:08.090 align:middle line:90% the crystal around it. 01:08:08.090 --> 01:08:10.490 align:middle line:84% This can cause macroscopic defects. 01:08:10.490 --> 01:08:13.010 align:middle line:84% And the metal atoms tend to be located 01:08:13.010 --> 01:08:17.120 align:middle line:84% around these stacking faults or dislocations. 01:08:17.120 --> 01:08:19.430 align:middle line:84% If you want to do it in a quantitative or 01:08:19.430 --> 01:08:23.540 align:middle line:84% semi-quantitative manner, you can use the same model. 01:08:23.540 --> 01:08:25.130 align:middle line:84% Same mathematical segregation model 01:08:25.130 --> 01:08:29.479 align:middle line:84% was used as was used for previously. 01:08:29.479 --> 01:08:32.060 align:middle line:84% And again, this reaction creates interstitials, 01:08:32.060 --> 01:08:36.350 align:middle line:84% which will help free up substitutional gold 01:08:36.350 --> 01:08:40.410 align:middle line:84% to make it ready, interstitial for diffusion. 01:08:40.410 --> 01:08:42.323 align:middle line:84% So on page 20 or slide 23, I just 01:08:42.323 --> 01:08:43.490 align:middle line:90% want to summarize gettering. 01:08:43.490 --> 01:08:44.779 align:middle line:90% It's still fairly qualitative. 01:08:44.779 --> 01:08:45.920 align:middle line:90% I think you can tell. 01:08:45.920 --> 01:08:48.649 align:middle line:84% These gettering methods really haven't 01:08:48.649 --> 01:08:52.100 align:middle line:84% been changed much in the last 10 to 15 years. 01:08:52.100 --> 01:08:55.490 align:middle line:84% There is a better understanding of how it works. 01:08:55.490 --> 01:08:58.609 align:middle line:84% But we really still need better models. 01:08:58.609 --> 01:09:00.380 align:middle line:84% In the future, intrinsic gettering, 01:09:00.380 --> 01:09:03.050 align:middle line:84% this process of creating the denuded zone either by epi 01:09:03.050 --> 01:09:04.670 align:middle line:84% or whatever, is going to dominate 01:09:04.670 --> 01:09:09.200 align:middle line:84% because the thermal budgets for IC processing will decrease. 01:09:09.200 --> 01:09:11.180 align:middle line:84% And trying to diffuse those atoms all the way 01:09:11.180 --> 01:09:13.923 align:middle line:84% to the back of the wafer, especially titanium and moly, 01:09:13.923 --> 01:09:15.590 align:middle line:84% which have lower diffusion coefficients, 01:09:15.590 --> 01:09:18.510 align:middle line:90% is not very practical. 01:09:18.510 --> 01:09:21.649 align:middle line:84% So having the gettering site close is better. 01:09:21.649 --> 01:09:24.260 align:middle line:84% Clearly, the amount of oxygen, the oxygen concentration 01:09:24.260 --> 01:09:26.212 align:middle line:84% in the wafer, is going to be critical for 01:09:26.212 --> 01:09:27.420 align:middle line:90% this whole gettering process. 01:09:27.420 --> 01:09:30.680 align:middle line:84% So we're going to need tighter controls so that we can really 01:09:30.680 --> 01:09:32.660 align:middle line:84% control the gettering process, and also 01:09:32.660 --> 01:09:35.029 align:middle line:84% to control other things like wafer warpage. 01:09:35.029 --> 01:09:37.310 align:middle line:84% We talked earlier about how oxygen actually 01:09:37.310 --> 01:09:39.290 align:middle line:84% increases the mechanical strength of silicon. 01:09:39.290 --> 01:09:44.960 align:middle line:84% So newer techniques are needed to do gettering 01:09:44.960 --> 01:09:45.960 align:middle line:90% at lower temperatures. 01:09:45.960 --> 01:09:48.290 align:middle line:90% So that's an area for work. 01:09:48.290 --> 01:09:50.660 align:middle line:84% And there is really no accurate simulation tool. 01:09:50.660 --> 01:09:52.158 align:middle line:84% In this class, for homework, you're 01:09:52.158 --> 01:09:54.200 align:middle line:84% going to learn how to use simulation tools called 01:09:54.200 --> 01:09:57.690 align:middle line:84% SUPREM-IV to model processes like diffusion, oxidation, 01:09:57.690 --> 01:09:58.190 align:middle line:90% whatever. 01:09:58.190 --> 01:10:00.560 align:middle line:84% There is no such simple tool for gettering. 01:10:00.560 --> 01:10:04.220 align:middle line:84% It's still very hand-waving qualitative arguments, 01:10:04.220 --> 01:10:07.650 align:middle line:90% like we've given today. 01:10:07.650 --> 01:10:10.440 align:middle line:84% And I'll finish up by summarizing chapter 4. 01:10:10.440 --> 01:10:13.500 align:middle line:84% The last couple of lectures, we talked about particle control, 01:10:13.500 --> 01:10:15.180 align:middle line:84% wafer cleaning, and gettering processes. 01:10:15.180 --> 01:10:17.550 align:middle line:84% These three things, all of these are 01:10:17.550 --> 01:10:20.400 align:middle line:84% crucial for successful manufacturing 01:10:20.400 --> 01:10:23.880 align:middle line:84% and even for doing the research that you do here in MTL. 01:10:23.880 --> 01:10:25.850 align:middle line:84% There's a three-term tiered approach. 01:10:25.850 --> 01:10:28.740 align:middle line:84% It was to keep the air and the water clean, 01:10:28.740 --> 01:10:32.460 align:middle line:84% continuously clean the wafers in these chemicals, 01:10:32.460 --> 01:10:36.672 align:middle line:84% and use the RCA clean, and finally, to do gettering. 01:10:36.672 --> 01:10:39.360 align:middle line:84% Gettering involves releasing the impurities, 01:10:39.360 --> 01:10:42.510 align:middle line:84% diffusing them to the trapping site, and getting them trapped. 01:10:42.510 --> 01:10:46.170 align:middle line:84% The good thing is that most metals diffuse relatively fast. 01:10:46.170 --> 01:10:48.300 align:middle line:90% That's both good and bad. 01:10:48.300 --> 01:10:49.620 align:middle line:90% There are a lot of means. 01:10:49.620 --> 01:10:51.037 align:middle line:84% And next time, we're going to talk 01:10:51.037 --> 01:10:54.480 align:middle line:84% about the physical, electrical means for characterizing 01:10:54.480 --> 01:10:55.830 align:middle line:90% the wafer contamination. 01:10:55.830 --> 01:10:57.408 align:middle line:84% And how effective is your gettering? 01:10:57.408 --> 01:10:58.950 align:middle line:84% And how contaminated are your wafers? 01:10:58.950 --> 01:11:02.010 align:middle line:84% And I showed you some data today without explaining 01:11:02.010 --> 01:11:03.845 align:middle line:90% TXRF and things like that. 01:11:03.845 --> 01:11:05.220 align:middle line:84% We'll talk next time a little bit 01:11:05.220 --> 01:11:08.220 align:middle line:84% about how some of those contamination measurement 01:11:08.220 --> 01:11:09.850 align:middle line:90% techniques work. 01:11:09.850 --> 01:11:12.190 align:middle line:84% So that's all I have for this lecture. 01:11:12.190 --> 01:11:15.570 align:middle line:84% Please come up and hand in your homework in the orange folder 01:11:15.570 --> 01:11:18.210 align:middle line:90% here, homework number 1. 01:11:18.210 --> 01:11:20.150 align:middle line:90% Thanks. 01:11:20.150 --> 01:11:23.660 align:middle line:90% [SIDE CONVERSATIONS] 01:11:23.660 --> 01:11:29.000 align:middle line:90%