Electron-Enhanced Atomic Layer Deposition (EE-ALD)

Dr. Steven George – University of Colorado at Boulder


This talk on electron-enhanced atomic layer deposition (EE-ALD) will highlight our recent work on BN EE-ALD and Co EE-ALD. EE-ALD uses electrons as a reactant in an ALD process. Low energy electrons (e.g. 100-200 eV) are able to desorb surface species by electron stimulated desorption (ESD). This desorption opens up empty sites and facilitates precursor adsorption. Subsequently, the adsorbed precursor is illuminated with more electrons to continue the EE-ALD process. EE-ALD has many advantages such as low temperature ALD, rapid nucleation of ALD films and topographically selective ALD. This talk will also highlight our recent work using a hollow cathode plasma electron source (HC-PES) that provides much higher electron currents, illuminates a larger surface area and is much more chemically robust than our earlier electron gun. The talk concludes with TEM results for topographically selective Co EE-ALD in vias using the HC-PES.


0:00:12   Hello, this is steve George from the University of Colorado. I am happy to be part of this advanced
0:00:19   Surface engineering summit.
0:00:22   I'm going to talk about electron enhanced atomic layer deposition or E. L. D. Yeah. This electron
0:00:30   enhanced process we can think of as using the electrons in an L. D. Process where the electrons are
0:00:37   going to replace one of the reactions that we typically use in an A. B process. This work has been
0:00:44   supported by the semiconductor research Corporation through the jump program as well as Darkwa.
0:00:53   So why do we want to do electronic enhanced L. D. Mm. Well part of the reason is that electron
0:01:00   enhancement gives us non thermal excitation and this enables the E. L. D. To be done at much lower
0:01:08   temperatures to give you a little bit of an idea of the temperatures required. Some films really
0:01:14   have quite high temperatures for both CVD and L. D. For example, gallium nitride temperatures around
0:01:22   1000. See silicon also nine 11,100 Boron Nitride even higher. Now we can pull the temperatures down
0:01:30   a little bit in A. L. D. Processes but they're still quite high for gallium nitride and silicon as
0:01:35   well as the end. On the other hand, when we use a non thermal excitation process as we as we can
0:01:43   with the electrons, we can get too much lower temperatures and this is really taking advantage of
0:01:48   the process of electron stimulated disorder option that can occur in the presence of these electrons.
0:01:56   So let's say a little bit about electron stimulated description. Electron stimulated deception
0:02:02   brings in electrons to the surface and these are electrons. Low energy electrons typically may be in
0:02:08   the range of 100, 150 e. V. something in that range. Or or or maybe higher. And those electrons have
0:02:15   the ability to knock off Liggins from the surface. So what I show here an example is hydrogen. So we
0:02:21   can come in with electrons Say 150 e. V. and kick off the hydrogen on the surface. That might be
0:02:28   that self limiting ligand that we would have in a typical A. L. D. Reaction. The removal of the
0:02:35   hydrogen then leaves an open site or if this was silicon, maybe a dangling bond. And those open
0:02:42   sites then are very receptive then for absorption uh two other precursors. So the idea is then we're
0:02:49   going to produce reactive sites as a result of the electron stimulated disruption process. And this
0:02:55   electron stimulated disruption process then is essentially going to act like a reaction in an L. D.
0:03:00   Process. So instead of having a self limiting reaction will remove the Liggins and then come in with
0:03:07   a precursor and go back and forth between electrons and precursors. And that's what's shown here in
0:03:14   the cartoon. So this is using electron stimulated disruption in the sequential a lady process where
0:03:20   the first, the first process the reaction, the electrons would come in and say remove hydrogen in
0:03:25   this case and produce these reactive sites or dangling bonds. And then in the second reaction the
0:03:33   process we could come in with our precursor and the precursor now is going to absorb onto these
0:03:38   reactive sites to deposit some species of interest. And then maybe if the precursor also has
0:03:43   hydrogen in it, the hydrogen would be essentially like the terminating ligand. So we could come back
0:03:49   to the A reaction and again apply the electrons to sort of the hydrogen make the reactive sites and
0:03:54   then repeat the process and then basically go back and forth A B a B just like we would in an in an
0:04:00   L. D. Process except now we're using the electrons as essentially one of the precursors. Now, a good
0:04:07   example of this type of L. D. Growth is the growth of silicon using electrons. It's in the range 100,
0:04:17   e. V. And I styling. And so in this process, the electrons come in remove the hydrogen to produce
0:04:25   reactive sites on the surface and then in the second reaction, the dice island cannot sword onto
0:04:31   those reactive sites to deposit silicon. That's the species that were growing. And then in the next
0:04:39   cycle, the electrons could come in and then remove the hydrogen that was on the silicon that came in
0:04:44   with the dye styling to make more reactive sites and then we can repeat the process. So we found in
0:04:50   this is This is one of our earlier studies uh that we get a growth rate of silicon of about .3
0:04:57   angstrom cycle in this process with the electrons at 100 250 E. B. So that's that's the basic idea
0:05:05   of the electron enhanced LD. So what I'd like to talk about today is more recent work. And I'll
0:05:12   start out with looking at results from boron nitride E. L. D. And these are going to be results that
0:05:19   were done with an electron gun. And then we'll also take a look at results for cobalt electron
0:05:25   enhanced LD. That were also done with the electron gun. And then I'll move on to the the topic glass
0:05:33   topic here, which is the cobalt E. L. D. With a new source, a new hollow cathode plasma electron
0:05:38   source. And this has really been a big innovation for us to go to the hollow cathode as you'll see,
0:05:45   it can produce much more current and it's much more chemically robust than the electron gun. Okay.
0:05:51   But first let's just take a look at the electron uh L. D. Reactor and the high vacuum chamber that
0:05:58   we use for the E. L. D. Uh So this this these experiments that I'm going to show have been done in
0:06:04   this high vacuum chamber. And the reason really to go to the high vacuum is just to avoid any kind
0:06:10   of surface contamination. Uh We're making reactive species on the surface with electron stimulated
0:06:17   deception. And so any background gas could also compete with our precursor for those reactive sites.
0:06:22   And so to make things simple, we run in a high vacuum chamber. We also on this chamber have a sample
0:06:28   load lock. We can transfer the change the samples in and out of the vacuum chamber. Uh We have
0:06:35   electron gun. Uh We'll have a hollow cathode also uh in subsequent experiments and then we'll be
0:06:41   measuring the thickness of the film uh films using a lip symmetry. So we have a lip symmetry ports
0:06:47   so that we can reflect the beam off the sample and do a lip symmetry for thin film thickness
0:06:53   measurements. Okay, so the first system I like to look at is boron nitride. More on nitride. E. L. D.
0:06:59   Can be done with a single source precursor. And this is Boris scene were seen is a very interesting
0:07:06   molecule. It has equal amounts of boron and nitrogen. And that's very convenient. So in one
0:07:14   precursor we can come in and if we can remove the hydrogen with electron stimulated disruption
0:07:20   process, then we can potentially deposit the boron nitride. And so here's a picture of boron nitrate.
0:07:27   It basically looks like benzene except the nitrogen and boron are replacing what would be carbon in
0:07:32   benzene. So in the course of the boron E. L. D. We're going to absorb the borough seen on the
0:07:38   surface and then we'll come in with the electrons to remove the hydrogen. Using the electron
0:07:43   stimulated deception. And then we'll just repeat that process. So we'll go back and forth between
0:07:47   absorption of Berezin and then removing the hydrogen with electron stimulated distortion. What we
0:07:55   find. And in this work and again, this is a little geometry measurements of the film thickness of
0:08:01   the boron nitride as a function of number of the E. L. D. Cycles is we can get very linear growth.
0:08:08   The growth rates are a little different depending on what electron energy uh electrons have. And we
0:08:16   see here we can get 3 3 Angstrom. The cycle at 120 V. Uh we can go down a little bit in growth rate
0:08:24   when we go to 300 DV 2.5 angstrom. And then at 400 E. V. We're getting to angstrom a cycle. Now this
0:08:30   is work done with the electron gun For fairly long electron exposures, 240 seconds and electronic
0:08:37   currents. In this case we're about 300 micrograms, but we can see very linear growth. So it's very
0:08:43   much like an L. D. Process where we get very linear growth versus number of cycles. We can also tell
0:08:51   that the electron beam is rather limited in its spatial extent on the sample. So we can bring the
0:08:57   electron beam in and it turns out that it focuses down to a spot that's on the order of about a
0:09:02   centimeter squared. And we can easily see it. This is after 500 cycles. We can easily see where the
0:09:08   boron nitride has formed on the silicon wafer. And this is also the sample, a Platen that where we
0:09:14   put the silicon wafer on and then we can grow the film on the silicon. And these were with rather
0:09:19   modest for isn't exposures. And again, electron energy about 100 TV. And note the temperature were
0:09:26   at room temperature or only slightly above room temperature because of the electron gun electron
0:09:32   current heating of the sample. So we can we can see the deposit. So there's no doubt that we're
0:09:37   growing the film. When we take the films out, we can do analysis of those films. And so when we do X
0:09:45   ray diffraction of the boron nitride film, we can see that we have a crystalline sample. We have a
0:09:50   very strong 00 to feature in the X. R. D. Which is which is consistent with hexagonal boron nitride.
0:09:59   But we do have very broad peaks in the X ray diffraction. And when we use the sherrer formula, we're
0:10:06   getting crystal lights that are on the order of 1 to 2 nanometers. So these are kind of a poly
0:10:11   crystalline and with fairly small crystals forming this sample. Now we can also take the sample out
0:10:20   and cross section it and do high resolution transmission electron microscopy. And when we do that we
0:10:28   can see really the beautiful silicon wafer the silicon surface underneath the boron nitride film. We
0:10:35   have a fairly amorphous region right at the interface. But then notice these kind of lines and way
0:10:42   venous. These are what we're looking at here are the basil planes of the boron nitride and they're
0:10:47   approximately parallel to the surface. There are little wavy, they're not perfectly parallel and
0:10:53   these kind of wavy planes that orient themselves approximately parallel to the surface is consistent
0:10:59   with a turbo strategy boron nitride which is a hexagonal boron nitride where the individual planes
0:11:06   that are stacked kind of like cards in a deck of cards have just tilted a little bit in terms of
0:11:12   their rotation angle relative to each other. So it's not perfect alignment, but a little bit of
0:11:16   rotation between the planes to give rise to turbo strategy or on nitride. But still this is a
0:11:22   crystalline material and you can see probably why we have the the slightly wider X ray diffraction
0:11:29   patterns is because the the individual domains are not that large. But you can see that overall
0:11:34   there is a nice parallel alignment of these lines relative to the silicon sample. So what do we
0:11:41   think is happening with the boron nitride? E L. D. We're getting growth rates of around three
0:11:47   angstrom. In fact, the maximum growth rate was about 3.2 Angstrom. And that is very close to the
0:11:54   hexagonal C axis spacing in boron nitride. Which is shown here. So here here's one basil plane,
0:12:00   another basil plane, another basil plane. And there's about 3.3 Angstrom is per basil plain. And so
0:12:07   our growth rate is almost exactly what we would expect if we were growing crystalline boron nitride
0:12:13   where we were actually depositing a mono layer of the hexagonal boron nitride every cycle. And so
0:12:20   what we think is happening is that we're absorbing bora seen down on the surface. And we can imagine
0:12:26   that the Brazilians are lying flat parallel to the surface. We come in with the electrons and
0:12:33   electron stimulated deception, is able to remove the hydrogen. And so now we take away the hydrogen
0:12:38   which are shown here in green. And then after the hydrogen are gone, these molecules can essentially
0:12:44   knit together to form a basil plane of boron nitride. And then once we have that basal plane of
0:12:51   boron nitride that we've we've just deposited, then we can absorb Oracene on top and it will form
0:12:56   again on top of that Basil plane and then we can repeat the process. So this is our speculation as
0:13:01   to how the films are growing to give us that nice 3.2 Angstrom recycle, which is very close to the
0:13:08   the distance between the basil planes. Okay, so that's the story for boron nitride. So now let's
0:13:15   move on and take a look at results for cobalt electron enhanced L. D. And these are also be done
0:13:21   with the electron gun for cobalt E A L. D. We use a precursor for cobalt that's called cobalt. Try
0:13:28   carbon, Neil nitrous ill. So it's a rather simple cobalt precursor with The cobalt, the Middle three
0:13:33   carbon. He'll groups here and then a nitrous seal on top. And the model for the growth is based on
0:13:41   molecular absorption of this cobalt try carbon and nitrous seal on the surface and then electron
0:13:47   stimulated absorption with which we're now the electrons come in and now, instead of disturbing
0:13:52   hydrogen, like we saw before, the electrons are disturbing either carbon monoxide or no. And so
0:13:59   those would go off into the gas phase and leave behind cobalt and so we go back and forth to do
0:14:04   cobalt E. L. D between the precursor absorption and then the electron stimulated disruption process.
0:14:11   So that's that's the same idea. Except now we're disturbing the Ceo and the N. O. As opposed to
0:14:16   hydrogen. Now, in this case we have the spectroscopic ellipse on it or on in fact it's on all the
0:14:23   time. So we're doing a continuous monitoring of the film thickness. And so what we see here is
0:14:29   really something quite extraordinary that we can we can actually see the absorption of the cobalt
0:14:35   precursor. And we see that as an equivalent thickness increase. And then after the dose, the film
0:14:42   thickness stays largely constant. And then when we come in with electrons, which in this case was
0:14:47   electrons at 150 E V. We see the thickness essentially effectively decrease. But that's the
0:14:54   distortion of the Liggins off of the molecular precursor on the surface. And then after the electron
0:14:59   dose again, the thickness stays constant until the next precursor dose. So we see a nice jump up
0:15:06   with the cobalt try Carbonneau nitrous seal and then leveling off as it absorbs and is constant on
0:15:12   the surface. And then electrons drive it drive the Liggins off and then again it's constant before
0:15:17   we repeat that process. So we're getting a nice uh Increase of the thickness. We get about 1.3
0:15:24   angstrom cycle for this cobalt growth process. And it's very reproducible and going cycle after
0:15:31   cycle after cycle to deposit the cobalt. Now, one thing I should add is that we're using electron
0:15:39   gun to do these experiments and the electron gun, even though it's, it's working quite well, it does
0:15:44   have its problems. And so if we look at how long each cycle time each cycle takes, uh it turns out
0:15:51   it takes a long time because the electron gun takes a long time to warm up about it 120 seconds. The
0:15:57   electron exposures for this cobalt case, we're about 60 seconds. But then the electron gun, we also
0:16:03   have to let it cool down because if we come in with precursor with the hot electron gun filament, we
0:16:09   actually get a reaction on the filament which then is corrosive to the filament. So we have to wait
0:16:14   for the gun to cool. And then we do the precursor exposure which is negligible in terms of the total
0:16:18   time. And then we course purge again because we don't want the any precursors in the chamber when we
0:16:24   also warm up the electron gun. So all told then we really have a quite long cycle time using the
0:16:30   electron gun. So to get away with uh to get away from the electron gun and to obtain a shorter cycle
0:16:38   time and potentially even a better source maybe with higher flux and also more chemically robust
0:16:45   electron source. We have developed a new hollow cathode plasma electron source, which I think is
0:16:50   really a great enabler for electron enhanced L. D. Now just let me say a few words about the hollow
0:16:57   cathode plasma. The hollow cathode is a very simple plasma, basically what you do is you create a
0:17:04   cavity. The walls of this cavity are at a negative voltage. And so what will happen is that
0:17:11   basically the ions will get attracted to the walls there negatively uh they're they're set up in a
0:17:18   negative voltage. So like if you make an eye on like argon for example, shown here, it will get
0:17:22   accelerated the walls when that argon gets accelerated to the walls. Uh it will hit the wall with
0:17:29   the reasonable velocity. And then also what will happen is secondary electrons will get omitted. And
0:17:35   so those electrons will come in to the cavity. And because the walls are negatively charged, they'll
0:17:40   essentially oscillate in a in a somewhat pendulum way back and forth. And that process then will
0:17:45   store quite a lot of electrons in the hollow Catherine. Also, the electrons can can disassociate
0:17:51   more argon because we're having argon gas is the is the gas that defines this plasma. And then those
0:17:57   argon ions that are formed of course where they're going to get accelerated to the walls again, hit
0:18:01   the walls, produce more electrons. And you can see that this process just kind of keeps cascading
0:18:06   and keeps building up electrons in the hollow cathode. Now, um the cathode is essentially the walls
0:18:14   but there is a release of the electrons and there you can set up a little grid bias and a little
0:18:20   aperture at the end and allow the electrons to escape. And that's what we do. And this becomes then
0:18:25   our source for electrons for the electron enhanced L. D. So here's a picture. Just a schematic of
0:18:32   the Oregon plasma that defines the hollow cathode. There's a grid bias that will pull the electrons
0:18:38   out and establish the energy of the electrons and then we can take those electrons in and hit our
0:18:43   substrate. Now this design was developed by scott Walton at US naval Research Lab and they helped us
0:18:50   build our hell catholic plasma source. In fact, here here is a picture of the hollow cathode on the
0:18:58   back end. The other thing I wanted to say is that the electrons, this is a great source of electrons.
0:19:03   But because of the sputtering process that goes on when the argon ions hit the walls, there also is
0:19:09   some sputter flux that comes out of the hollow cathode. So what we have to do is we have to add
0:19:14   magnets um at the at the at the at the output of the hollow cathode to bend the electron beam so
0:19:21   that the sputter flux just comes forward and hits the wall without going to our substrate. Whereas
0:19:26   the electrons get bent around this bend and then go into the chamber and hit the hit the substrate.
0:19:33   So that's what we do in order to clean up the beam and uh and steer the electrons to the substrate.
0:19:40   But this hollow cathode is really a fantastic source of electrons. If we compare like the sample
0:19:45   current of the hollow cathode versus the electron gun, we find that we have almost three orders of
0:19:52   magnitude more electron current coming from the hollow cathode than from the electron gun. So that's
0:19:57   quite an improvement. The other huge improvement is that the hollow catholic can be turned on and
0:20:03   off just in the course of On the order of 10 milliseconds or less. So we can go on and off with the,
0:20:09   with the source under really incredibly short times. Whereas the electron gun was Requiring, you
0:20:15   know, 10 to hundreds of seconds to fully get up to the proper temperature for the electron gun to
0:20:21   operate. And so that's that's been a huge improvement in reducing our cycle time and also reducing
0:20:27   our cycle time because we have so much more current coming from the hollow cathode. And also when we
0:20:34   have this higher current, it turns out we have really excellent new creation of these films. So this
0:20:40   is a comparison of the cobalt film thickness that we get. And this is just on a native oxide on
0:20:45   silicon. And you can see what the electron gun. Uh it was taking us on the order of 50 60 cycles to
0:20:52   really get into the linear growth regime for cobalt E. L. D at at that 1.3 and streams that we saw
0:20:59   few sides ago. On the other hand, with the hollow cathode, we can get higher growth rates and we
0:21:06   also can get new creation that's almost immediately within the first couple cycles of the of the
0:21:11   process. Uh notice the times here to the electron gun was really requiring multiple hours to grow
0:21:18   the film because the cycle time was so long. Uh and the electron flux was so small that required
0:21:24   that really stretched out the cycle time. On the other, other hand, the hollow cathode we're now
0:21:30   we're looking at instead of ours, we're looking at just minutes uh to to grow a film that's even
0:21:35   thicker than than we show on the left here for the electron gun. So they're really tremendous
0:21:39   improvement with the hollow cathode. The other thing the hollow cathode can do it is because the
0:21:44   electron flux is so much higher, we can illuminate a much larger area of the sample and still get
0:21:51   saturation of the electron stimulated destruction process. So what I show here these are results
0:21:56   from the electron gun where we're seeing a fairly flat top on top of a, the cobalt profile is about
0:22:04   one centimeter squared deposition and it's flat on top because it's flat when the chemistry
0:22:10   saturates. And so when we get that really self limiting process where we have saturated the amount
0:22:15   of VSD the electron stimulated description, we get a flat, we get really reach a constant amount of
0:22:21   material that can be deposited. Whereas when we when we roll off the edge of the electric focused
0:22:26   electron beam of course, then we fall below saturation. And then we really end up with Not much at
0:22:32   all outside of that one cm square. On the other hand, for the hollow cathode, we are able to
0:22:39   illuminate really our whole silicon sample on the sample stage. And so we're able to get a very
0:22:45   uniform thickness of the cobalt across the entire sample because the process is going to saturation
0:22:52   over the whole sample. And so when we go to saturation, we end up with a very constant growth at any
0:22:58   location. And so that's this is essentially a flat top, but it just goes on indefinitely because it
0:23:04   extends beyond what the flat top region would basically be extending beyond the size of our sample.
0:23:11   Now, when we look at the the actual cobalt growth with the hollow cathode, what we see is something
0:23:17   very similar. We see a kind of a stepwise growth. Every time we come in with the precursor, we get
0:23:23   the growth of the of the cobalt film. Uh and then we wait in it, flattens out in that waiting
0:23:28   process. And then we come in with the electrons and we see a drop off of the thickness just like we
0:23:33   did before the electron gun. And then when we wait until the next precursor dose and then we dose
0:23:38   again. So we see again a very digital process. This looks like a very uh very A L. D. Like process
0:23:45   where this could be essentially very similar to the results that you would get with the court's
0:23:50   crystal micro balance monitoring the L. D. Process. And again with the hollow cathode we get a
0:23:55   Higher growth rate to a 2.4 Angstrom recycle. And so these results also we're done with
0:24:01   spectroscopic lips symmetry to measure the cobalt thickness. And uh yeah, we're so we're very very
0:24:09   pleased to see this rapid nuclear nation on again on a native oxide on on silicon. Now you might say,
0:24:17   well what are what are the advantages of the E. A. L. D. Well for sure we've seen that we can get
0:24:23   much shorter cycle times because we can get higher electron currents. Uh We can also get uh we have
0:24:31   a much more chemically robust source so that we don't have to worry about how long it takes to purge.
0:24:37   And we don't have to wait for the electron gun to warm up and cool down. Because the it turns out
0:24:42   the hello cathode is very immune to whatever gas is present in the chamber up to about in the
0:24:48   military range. And so what that means is that we can use E. L. D. For a number of things. Like one
0:24:54   thing that we believe that E. L. D. Will be very useful for is Nuclear Nation enhancement. Because a
0:24:59   lot of it turns out a lot of uh L. D. Processes especially things like the thermal A. L. D. Of
0:25:05   metals and oxides. They really have very poor new creation. And it might take tens, hundreds even
0:25:10   multiple hundreds of cycles to get a film to nuclear and grow. And then by the time you have a
0:25:16   continuous film maybe the film is already 40 50 nanometer stick. Where as we've seen with electron
0:25:22   enhanced LD we can nuclear it in just as few as a few cycles, 3 4 cycles in the case of cobalt. And
0:25:29   this will actually be very important for applications that require very ultra thin films, especially
0:25:35   as a barrier liners for example, in the back and interconnect for example. Um And and the other
0:25:44   thing that the electron enhanced L. D. Has is also it has electrons which are directional. And so I
0:25:49   haven't really said much about that. But the electrons actually can come in and they come in usually
0:25:55   the configuration as they come in normal to the surface. And what that means is that then there's a
0:25:59   lot more electron flux along the horizontal surfaces then along side walls. So you end up then with
0:26:06   a lot more electrons hitting the horizontal surfaces here. And that means then that you're going to
0:26:11   get more deposition of your material. This is this is a cartoon showing more cobalt deposited on the
0:26:17   horizontal surfaces then on the side walls. And in this case it was a slope side walls. So there is
0:26:24   some but there's certainly more at the bottom and the top because that's where more the electrons
0:26:29   are going to hit. So to uh to demonstrate that and to show that that is indeed the case, we did uh
0:26:37   electron enhanced um L. D. Of cobalt on some on some vehicles that had fairly straight vertical
0:26:43   sidewalls. And so here's the picture then. So we had a sample Where we had a via going through s. i.
0:26:50   0. 2 and then we coated the cobalt primarily on the horizontal surfaces up at the top and at the
0:26:58   bottom. And we didn't see very much cobalt on the vertical sidewalls. And then of course this is T.
0:27:04   E. M. So there's another layer to visualize the T. E. M. That was deposited on top. So this is the
0:27:10   cartoon. The sputtered platinum is shown in the gray. And then here we're showing the actual tm
0:27:16   image that again shows the vertical sidewalls. You can kind of see the cobalt film down here.
0:27:23   There's not much cobalt on the side wall, so it's more difficult to see. But if we blow it up, you
0:27:27   can again see the cobalt at the bottom and again, not that much cobalt showing up on the side walls.
0:27:32   So we're clearly getting a lot more cobalt deposited at the bottom of this via uh than we are on the
0:27:39   side wall. And these results were done with electrons Energies of 400, e. V. And this was just 45
0:27:46   cycles for this cobalt e. l. d. processing. So that was the cartoon. And then we can actually uh go
0:27:56   in and scan across the side wall and across the bottom just to get an idea of what the thicknesses
0:28:01   are. So when we scan with this, now this is this is with the DS scanning across the side wall. What
0:28:10   we find is that we really don't have that much cobalt on the side wall. We can, the cobalt is shown
0:28:16   here in the kind of reddish salmon color. And you can't you really can't see much of that color on
0:28:23   the side wall, but you can see a lot at the bottom. So we're getting, it turns out when you scan
0:28:27   across, you're getting about 55 nanometers, all there isn't that Much that much signal. So it's a
0:28:32   rather noisy signal, but about five nm on the side wall. However, when you look at the scan across
0:28:39   the bottom of the horizontal, the horizontal surface at the bottom of the veal. Now, what you're
0:28:44   seeing is you're seeing a lot more well, it's also it's a lot more obvious that you have more cobalt
0:28:50   at the bottom. And when you scan across that Deposition at the bottom, it turns out it has a
0:28:56   thickness of about 20 nm. So if you look at the side wall thickness versus the bottom thickness,
0:29:02   you're getting about 20 nanometer over a five nanometer or horizontal. The vertical of about a 4 to
0:29:08   1 topographical selectivity. Which means that it might be possible then to use this process of
0:29:14   electron enhanced L. D. To essentially do a bottom up Phil of the via. Since there's a lot more
0:29:20   deposited the bottom then on the side wall and that would be very useful for a back end inter
0:29:26   connects. Okay so I'm just about done. I just wanted to say a little bit about where the research is
0:29:32   going. It's turning out that right now we've been able to actually introduce now background gases in
0:29:40   with the E. L. D. Process. So because the because the hollow catholic plasma electron source can
0:29:48   tolerate pressures in the in the chamber up to in the military range we can put background gases in
0:29:54   the chamber and then run the E. L. D. Process. So we can come in with a precursor purge electron
0:29:59   purge precursor purge electron purge and all the while have a background gas present. So this is
0:30:07   really great because now we can take advantage then of the background gas to help clean the film to
0:30:13   give us higher purity film and also depending on the composition of the background gas, we can
0:30:19   actually tune the composition also of the film that we're growing. And so we're very excited about
0:30:24   the ability now to do the E. L. D. Using this hollow cathode that can tolerate these higher
0:30:29   pressures in the chamber and then use a background gas to be able to manipulate the film that we're
0:30:35   depositing much more with basically another knob to turn in the E. L. D. Process. Okay so I'm let me
0:30:44   finish up that now with some conclusions. So what we've seen is that we can do E. L. D. And we can
0:30:50   do it at very low temperatures at room temperature using this non thermal electron stimulated
0:30:55   disruption process. And I showed you results for the E. L. D. For boron nitride and cobalt. And then
0:31:02   also I showed you a new electron source that we've developed for E. L. D. Based upon the hollow
0:31:08   cathode plasma electron source which just has orders of magnitude more electron current and is much
0:31:14   more chemically robust and can be turned on and off much faster than the electron gun. So we're very
0:31:19   excited about this as an electron source. And then also I showed results which which which indicated
0:31:27   that we can use E. L. D. For new creation enhancement. So a very rapid new creation of cobalt uh
0:31:34   with E. L. D. On the native oxide of silicon and then also some T E. M. Results. That illustrated
0:31:40   that we could do bottom up Phil potentially with the E. E. L. D. Process. Because we were able to
0:31:45   get this 4-1 topographical selectivity for the cobalt E. L. D. So before I finish, let me just
0:31:52   acknowledge all the people that did this work. The boron nitride electron enhanced L. D. Was done by
0:31:57   Jacqueline Springer. In fact she she started our work in electron enhanced L. D. And now she's at
0:32:03   intel uh the cobalt E. L. D. Work was done by Zach civil and he's still a graduate student in the
0:32:10   group and still going strong. He worked with the electron gun and then also worked with Andrew
0:32:15   Cavanagh on the construction of the hollow cathode plasma electron source. Which is really sped sped
0:32:21   things up in many ways. And then also I wanted to thank and acknowledge scott Walton and David Boris
0:32:27   at the Naval Research Lab for their design of this hollow catholic plasma electron source, which I
0:32:32   really think is going to innovate all the variety of things that can be done with electron enhanced
0:32:39   L. D. Okay, so that's all for the talk. How so? Thank you for listening.