Particle Atomic Layer Deposition at Argonne National Laboratory

Applications and Equipment

Session Notes:


Dr. Jeffrey W. Elam

Senior Chemist Applied Materials Division Argonne National Laboratory 




0:00:01   Hi, everyone. This is Jeff Elam from Argonne I'm gonna be talking to today about particle atomic
0:00:07   layer deposition at Argonne National Laboratory. I wanted to start out by thanking forged Nano for
0:00:13   inviting me to speak at their summit. And also, I wanted to thank all of you for dialing in and
0:00:19   listening to my presentation. I will begin with a brief introduction. I will then discuss three
0:00:28   different applications for particle ale de Catalunya sis energy storage and clean water. Next, I'll
0:00:36   talk a little bit about equipment for particle Ale de and finally I will conclude.
0:00:45   So let's start with the introduction. I know that we're all familiar with atomic layer deposition,
0:00:51   but it's always good toe briefly review the salient aspect. Uh, so it is, of course, a way to growth
0:00:59   in films using chemical vapors and self limiting reactions between those papers and a solid surface.
0:01:08   There are many different materials that we can grow by a topic layer deposition, including oxides,
0:01:13   nitrites, metals and other materials, and that allows us toe use these materials to pursue a broad
0:01:21   variety of applications.
0:01:25   And then finally on. Perhaps most importantly, for today's summit, we can coat non player services.
0:01:32   So they'll deacon precisely code three D objects, including particles.
0:01:41   I wanted to talk briefly about a related technology called sequential infiltration synthesis, or S s.
0:01:47   And that's most easily done by comparing it with atomic layer Deposition. So shown here,
0:01:53   schematically is a pulse sequence for a hypothetical ale de material, a B where were growing that 80
0:01:59   material on top of a polymer film. So the polymer is a substrate, and we're growing a film on top of
0:02:04   that substrate if you extend the exposure conditions. So you use longer exposures and larger partial
0:02:12   pressures than those A and B precursors can diffuse into that polymer film. And if there are
0:02:19   reactive sites for those precursors, they will react and they will stick. And if you repeat that
0:02:26   process, you'll end up with a hybrid material where you've infused it with the Inter Vanik baby. And
0:02:32   this is typically done at lower temperatures compared to LD so that you don't melt or degrade the
0:02:36   Palmer. So I'll refer to S I s later in the in the presentation.
0:02:42   So let's talk now about some applications for particle ale de. So the applications for particle ale
0:02:48   de are many, and their diverse I list here some of the applications that were pursuing at Argonne.
0:02:56   These include catalysts licking my and batteries and sore bits and the other items on that list. Of
0:03:03   course, the if you include the research going on it other institutions. That list is much longer and
0:03:09   includes things like pharmaceuticals and explosives and propellants and so on. So a lot of different
0:03:15   applications for this technology and that, of course, is what motivates this summit. So let's start
0:03:22   out by talking about atomic lawyer deposition for catalysts. If you wanted to engineer the ideal
0:03:28   catalyst, you'd want to be able to start with any high surface area substrate so that shown by this
0:03:36   yellow circle here, you then want to have the ability to put down a support on top of that substrate
0:03:42   That's signified by this by this red shell, and that support would be perhaps a metal oxide that is
0:03:51   not the active catalyst, but it participates in the catalytic reactions. It's maybe a co catalyst.
0:03:57   On top of that, you want to be able to have small, modern, dispersed nanoparticles shown by these
0:04:02   blue dots, so they should all be exactly the same size and all just the right size to catalyze the
0:04:09   reaction most effectively. And then finally, you'd want to have some means for preventing century.
0:04:14   So stopping these particles from a glom aerating ast the catalyst is undergoing chemical reactions.
0:04:22   So we can do this by atomic layer. Deposition weaken, Of course. Start out with any substrate. This
0:04:27   could be a piece of silicon where this could be a high surface area powder or this could be a
0:04:31   membrane. We can then grow the support layer by atomic layer deposition because there are many metal
0:04:37   oxides, for instance that we could grow by a lady. We can then grow nanoparticles on top of that
0:04:43   support. So there are a lot of ale, the metal processes that naturally form nanoparticles rather
0:04:50   than a continuous film. And then finally, we can add another layer on top of that to stabilize these
0:04:55   nanoparticles. So let's talk about some of these different aspect.
0:05:02   So palladium atomic layer deposition could be accomplished using palladium, h back and formaldehyde,
0:05:10   and we're gonna look at that reaction on top of the aluminum oxide service. So our support here is
0:05:14   gonna be aluminum oxide. And what we notice is that the thickness as a function of the number of
0:05:23   palladium ale de cycles, is not linear. It starts out very low. And then after some period of time,
0:05:29   it becomes linear. And so we can divide the growth into the nuclear ation phase, and then the growth
0:05:35   face. And then this happens at about 100 Hail disciples. This transition thing, you might ask, Why
0:05:44   is this so why doesn't it just grow a steady growth rate right from the very first Michael? And of
0:05:50   course, the answer is that it's not growing as uniformed layers a scoring his particles. So shown
0:05:56   here on the left is a transmission electron Micrografx, and these white dots are the palladium
0:06:03   nanoparticles. We have a zoomed in higher resolution TM that shows that the palladium nanoparticles
0:06:13   are about two nanometers inside. It's 1 to 2 nanometers, and they are crystal in the history. Graham
0:06:21   here on the right hand side, extracted from the TM and shows that they are again between about one
0:06:26   and two nanometers and size with a fairly tight particle size distribution. So we do palladium male
0:06:32   deal. We end up with small model dispersed related nanoparticles. And if we test these using
0:06:38   catalytic reactions, we find that, for instance, and methanol decomposition, they are about twice as
0:06:43   active as palladium nanoparticles made by conventional means made by incipient witness. Let's talk
0:06:49   now about over quoting the palladium nanoparticle catalysts using a lady aluminum oxide. We're gonna
0:06:56   probe these catalysts using this methanol decomposition reaction. That is gonna be our measure of
0:07:02   the catalytic activity. We prepared a series of samples where we over coated the Palladian
0:07:09   nanoparticles with different numbers of aluminum oxide, hailed the cycles. What we found was that
0:07:15   after between 30 and 35 over coating cycles, the conversion had essentially dropped to zero. And
0:07:22   that's not surprising because at that point we had completely buried the palladium in aluminum oxide.
0:07:28   However, for smaller numbers of over quoting cycles, the catalytic activity was maintained and it
0:07:34   was actually maintained in a special way. We found that we had turned off a Coke forming catalytic
0:07:42   reaction which made the catalyst more selected towards this methanol decomposition which we wanted.
0:07:49   So we want. So what we accomplish was to create a catalyst which was more selective. But we did
0:07:57   something else We also greatly improved the thermal stability and that was the whole motivation for
0:08:04   doing this over coating on that improve stability has shown in this pair of TM images on the left we
0:08:11   see the palladium nanoparticle catalyst without the overcoat after it's been heated to 500
0:08:17   centigrade for six hours and these white spots that you can see that a very obvious those air 10 to
0:08:23   15 nanometer particles. So those have have centered tremendously compared to their initial size and
0:08:30   contrast to that. On the right hand side, you can barely see the nanoparticles all because of the
0:08:36   scale for this TM. And so they had maintained their wanted to nanometer size. So we find that the
0:08:41   ailed aluminum oxide not only inhibits, uh not only reduces the coke formation but also inhibits
0:08:49   this injury. Let's talk next about a l. D for lithium ion battery cathodes, so the cathodes and
0:08:57   lithium ion batteries are made out of powders of material like lithium cobalt oxide. And as everyone
0:09:04   is well aware, if you charge and discharge of lithium ion battery many times, the capacity goes down
0:09:11   and one of the mechanisms responsible for this is degradation. of the Catholic powders. It turns out
0:09:18   that if you coat thes powders with thin films by Atomic Layer deposition, you can improve their
0:09:26   cycling stability so you can prevent this loss of capacity with cycling. And in some cases you can
0:09:32   also achieve higher capacity and even higher rates. Some of the mechanisms that are responsible for
0:09:39   this improvement are for since preventing contact with the electrolyte and scavenging HF that would
0:09:45   otherwise degree of the cathode. So here are some examples from Argonne of using a Lady Thin films.
0:09:54   The top two graphs are ailed the thin films of a tungsten carbide aluminum fluoride, nano composite
0:10:01   on lithium cobalt oxide powders. And in these graphs, the red data sets are for the coated materials,
0:10:09   and the black data sets are for the uncoated materials and you could see as a function of cycles. We
0:10:14   have better stability with the ale, decoded materials and as a function of rate. We can also get
0:10:19   higher charge discharge rates. The lower graph is for a different system. It's ailed de lithium,
0:10:26   aluminum oxide, Hyundai lick, lithium, nickel, manganese oxide, cathode and again, if we look at the
0:10:35   blue data sets with lithium aluminum oxide. If you compare that to the uncoated powder, the black
0:10:41   data said, you see that we have higher? Uh huh. Capacity retention versus cycles. So this is not
0:10:47   true for all coatings on all capitals. But there are many examples where they'll d films can be very
0:10:52   beneficial. The final application that I wanted to talk about his hail dese orbits. This is probably
0:10:59   unusual in a particle ale de summit, but you'll see later why I included this. So if we start with
0:11:05   polyurethane foam and we used the S I s sequential infiltration synthesis of aluminum oxide, we can
0:11:12   put a thin inorganic layer on the ligaments of the polyurethane foam. If we follow that with a
0:11:19   chemical salinization to put down an olio Philip model year, then we imparts, um, unique properties
0:11:26   to the to the polyurethane. It becomes simultaneously holy Oh Filic. So it will ads or boil
0:11:33   hydrophobic reusable. And it's of course, an inexpensive, so this could be used as assortment.
0:11:42   So this looping video shows that properties of the resulting treated foam eso the blue liquid is
0:11:50   vegetable oil which has been dyed blue, so you can see it and it is floating on top of water. And
0:11:56   each time that we apply this this'll olio sponge this treated polyurethane, it rapidly soaks up the
0:12:02   oil. Okay, so what? What can we use this for? Well, this is a still photograph taken underwater at
0:12:11   the home set facility. This is a facility for testing and training for oil spill response. In this
0:12:21   photograph, there are plumes of crude oil underneath the saltwater and their impinge ing on this
0:12:27   wall of holiest bunch phone. And in the experiments that we ran, we found that if the oldest bunch
0:12:32   was ableto adds orb, these sub surface oil droplets very effectively. We imagine that this could be
0:12:39   used in ah scenario where, uh, an underwater oil plumes needs to be cleaned up, and fishing trawlers
0:12:48   contract holiest bunch material in a net. It will ads or the oil. It can be squeezed out and reused,
0:12:56   So this could be a highly cost effective and also efficient way of cleaning up oil. But of course,
0:13:03   we need to make a lot of it in order to do that. And so there's this question about scale up, one of
0:13:09   the difficulties that needs to be overcome to scale up. This folios bunch is shown in this slide. In
0:13:15   the experiments that we did that produced the small cubes of film that we use for our tests. We
0:13:20   place the polyurethane cubes inside of ah tube. The two was part of our viscous flow ale, the
0:13:26   reactor where the precursors in the carrier gas were flowing in this direction. And what we found
0:13:31   was that the thickness of the coding on the probably your thing depended a lot on the position
0:13:36   inside the tube, and in particular there were thinner coatings downstream. So, you know, the
0:13:41   question is, how can we do this effectively such that all of these cubes received the same dose and
0:13:49   we'll talk about that later. So the next part of my presentation is about particle ale, the
0:13:53   equipment I'm going to start by describing how at Argonne we've been doing particle ale de in a very
0:14:01   straightforward way, using our existing tubular reactors. So this is a photograph of a nail, the
0:14:08   reactor that we use for lots of different activities as a general purpose research system. This is a
0:14:15   A wrapped up in heating tape is a two inch diameter tube. There are many projects that use this with
0:14:21   many different building materials. So, uh, however we cook, the powders needs to be easily
0:14:27   integrated with all the other activities where we're coding other things like silicon and memories
0:14:31   and so on. And so the simplest way to achieve that that we found was simply that Take the particles
0:14:37   of the power that we want a coat and to spread them out in a in a thin layer in a tray. And so, in
0:14:43   this photograph here, you can see the white powder is spread out in that stainless steel tray, and
0:14:48   then above the tray is a cover that is a porous stainless steel mesh that can keep the powder from
0:14:55   from from exiting the trade on. Then this schematic over here shows in cross section that the
0:15:02   precursors can diffuse through this wire mesh, and then they can then diffuse into this particle bed
0:15:08   so coding powder in a static bed in a metal tray works fairly well shown. Here On the left is a
0:15:15   graph where we've taken one gram samples of silica gel of powder with AH 100 square meters programme
0:15:25   surface area, and we have coded that with aluminum oxide by atomic layer deposition using different
0:15:32   increasing TM exposure times. And we find that the weight gain saturates after about 60 seconds
0:15:38   under these conditions, and we calculate that we use about only 30% of the precursor. The remaining
0:15:44   70% flows over the sample and is not reacted, so it's not terribly efficient, but it is simple and
0:15:52   effective. This photograph of the bottom shows two of these powder trays, one after another on a
0:15:59   metal carrier that we could slide into our tube reactor. And it's what we do fairly often is toe
0:16:05   place pieces of silicon around these powder carries, so the weekend afterwards measured the
0:16:11   thickness of the film using ill. It's on a tree,
0:16:16   so one way to coat larger quantities that powder is to use a larger train. So, yeah, we have a panic
0:16:22   TFS 500 a lady system, which is also biscuits flow across flow tight arrangements similar to our two
0:16:30   reactors. So if we insert ah larger train to this system, we can coat it in just the same way. And
0:16:36   in this example, there's about 20 grams of Catholic powder on this for inch by eight inch trey. So
0:16:44   one of the disadvantages of coding powder and a static bed by Atomic Layer deposition is shown in
0:16:49   this photograph. In this example, we were coating silica gel, which is white with a lady tungsten,
0:16:56   which, if it coats the silica gel uniformly should appear black. On the top surface of the silica
0:17:02   gel is coated and appears black. But once you disturbed that powder and you start toe, reveal some
0:17:08   of the powder underneath, you can see that it is not coded. So under these conditions, there's a non
0:17:14   uniform thickness of the coding as a function of depth into the bed on. That's a result of the non
0:17:21   uniform precursor exposure you get, because the precursor needs to defuse from the top down your of
0:17:27   necessity going to get larger exposures on the top.
0:17:32   So one way to alleviate the problem that I just described is, of course, that you can agitate the
0:17:38   powder while you're coating it. So to accomplish that, we have this rotating drum fixture that we
0:17:45   designed for our ale, the tubular reactor. So it's Ah, long skinny rotating drum, and the outside of
0:17:53   it is made of a 325 stainless steel mesh so it can contain particles that are greater than about 40
0:18:01   microns in diameter. Uh, we use a rotary feed through to couple this rotating drum to a motor that's
0:18:11   on the outside of the chamber, which can go between zero and 120 rotations per minute, and this can
0:18:17   hold about 100 grams of power.
0:18:21   So the graph here on the left shows the effect of using the rotating drum powder coating fixture to
0:18:29   perform aluminum oxide. Atomic lawyer Deposition on silica gel powder. And in this case, we're using
0:18:35   about 1500 square meters of powder in the in the powder drum, and we measure the saturation time.
0:18:44   That's the time to saturate the tri methyl aluminum reaction as measured by weight. Gain the change
0:18:51   in weight after coating the powder, and we find that in the absence of rotation, the saturation time
0:18:57   is about 160 seconds. But if we wrote it, the drama even just a few rpm that saturation time drops
0:19:03   down to about 100 seconds. We next performed Tungsten Atomic Layer deposition on the same Silica Joe,
0:19:11   and we found that when we removed the powder. When we use the pattern drum, that power came out
0:19:18   black and uniform, indicating a uniform coating. On the contrast of that, When we took the powder
0:19:25   from the fixed bed, we found that it was, uh, both black and white. So has described earlier.
0:19:33   There's a lot of the silica gel particles that just don't get coated it all those particles that are
0:19:37   below the surface of the layer so the powder room provides faster SAT saturation improved Cody
0:19:46   uniformity on the powder, and we achieve about 80% precursor utilization. I should mention that in
0:19:54   this conference there is a poster that is presented by Matthew Coil that describes the powder drum
0:19:59   in greater detail and also mentioned one final note, which is that even though we get better
0:20:05   uniformity by by rotating the powder drum, there still is. There is the possibility of having non
0:20:13   uniformity in the direction of flow because even when you rotate the powder, there's no way to get
0:20:17   around. The fact that the precursor is going first is gonna first encounter the powder upstream, and
0:20:25   then it's gonna flow downstream, and so doing it will become depleted and so downstream you will
0:20:29   always have smaller pickers exposures.
0:20:35   So one way to overcome that problem that I just described about the precursor exposure non
0:20:43   uniformity in the direction of float is to use a different ale de system. This is our powder coating
0:20:49   reactor. The rotating drum fixture I just described is something that fits into our existing lt
0:20:55   systems, and so it's very convenient on because it the ale, the systems don't need to be dedicated
0:21:01   to a particular purpose. This PCR this has dedicated Justin coating powers, and so some of the
0:21:08   features or some of the specifications are listed over here. It holds about a kilogram of powder. It
0:21:13   was designed to cut about 10,000 square meters of surface area. It uses high pressure, static dozing.
0:21:19   It has ways of handling temperature transients. Those is the heat that is generated as a result of
0:21:28   the exo thermic ale de reactions because if that isn't done properly, you can power can get quite
0:21:34   hot. We also use direct liquid injection of pre Christer, and there is an easy way to get the power
0:21:40   in and out of this system.
0:21:44   So this next live distrust schematically how this PCR operates. It's essentially Ah, hot wall
0:21:53   chamber with a large static volume. It's large that you can fit a large number of precursor
0:21:58   molecules in there at any one time, and in addition to that, so it's got it's got a rotating drum
0:22:05   inside, and in addition to that, it has a method of conviction. For were circulating the precursor
0:22:12   during a static dose. And it's this conviction that allows us to achieve more homogeneous precursor
0:22:20   exposure because there's not a single direction afloat the way that there is in the biscuits flow
0:22:27   reactor. So we're able to achieve about 100% Christie utilization with this system and, as I
0:22:32   described it, accomplishes more uniformed precursor exposures compared to the two reactor,
0:22:40   all of the particle ale. The equipment that I've described so far is equipment that we built
0:22:44   ourselves so that we could do bench scale research in our labs where we were primarily interested in
0:22:52   the properties of the materials that we're making. We're trying to make catalysts and cathode
0:22:57   materials that had unique properties. However, it's also important for us to try toe, translate the
0:23:07   materials and the processes that we come up with to industry and in order to do that, we need to
0:23:12   operate under conditions that are closer to those used commercially and so. One example of that is
0:23:22   that a particular implementation of particle ale de is the semi continuous process, it forging nano
0:23:27   and to try to be more close in the pressure flow conditions. We have recently purchased a fluid eyes
0:23:34   bed ale de system,
0:23:39   so I thought I would leave you with one crazy idea of what we might do with a food I spent system.
0:23:45   And it requires that we expand our definition of particles somewhat to include polyurethane cubes.
0:23:53   So I mentioned we have this difficulty coding polyurethane in a cross flowing uld system because we
0:24:00   were getting non uniform precursor exposure, and that was leading toe non uniformed coatings. If we
0:24:06   could tumble these cubes and a fluid eyes bed, that might be a way to try toe homogenize the
0:24:10   precursor exposure, and we might end up with a better product, and ultimately, this would then be a
0:24:16   way that we could clean up oil slicks from below the ocean Service. I'll let you decide if you think
0:24:22   that's a practical thing to do,
0:24:26   so let me just conclude by saying there are many, many applications for particle ale de I just told
0:24:32   you about a few and I'm sure that you're gonna hear many more in this summit. Also, that bench scale
0:24:39   research is not that difficult to accomplish in particle ale de, especially if you're not so worried
0:24:47   about optimizing the process. But you're are concentrating just on coding particles and doing
0:24:53   research with those particles. And then finally, that scale up on commercialisation will will
0:24:59   require specialized equipment that goes beyond things like fix beds and even maybe rotating drums.
0:25:07   As you might imagine, many people contributed to the work that I described in this presentation. I
0:25:12   will not read out the list of names, but I'll let you read them yourselves. In addition to that, we
0:25:17   had funding from a number of different sources. I will just point out funding from the high act
0:25:23   Canalys, ISI, FRC, as well as the seas energy storage TFRC. Both of these fr seas are funded by the
0:25:30   Department of Energy. Thanks very much for your attention