Engineering Challenges and Solutions with Particle ALD 

Session Notes:


James Ragonesi

Director of Engineering at Forge Nano 



ALD research has presented many promising avenues for the enhancement of powder materials.  At Forge Nano, we have engineered unique solutions to both enable and commercialize PALD research.  In this presentation we will investigate some of the controls, features, and unique flexibility of our research equipment.  Then we will dive into the commercialization approaches that Forge Nano has undertaken for the scaling of ALD research.  The focus will be on bulk handling of powders, vaporization of large precursor quantities, and throughput enhancements for economical PALD.   

0:00:01   Hi, everyone. My name's James, Ragonesi. I'm the director of engineering at Fortune Nano. And
0:00:06   today I'd like to talk with you about some of the engineering challenges associated with particle
0:00:10   atomic layer deposition and some of the scaling challenges and solutions that we've come up with.
0:00:17   Thanks for stopping by at the Fortune Nano engineering presentation for the P A. L D. Summit. As you
0:00:23   just saw in the video, I'm gonna be taking you through some of the challenges and some of the
0:00:27   solutions for engineering LD systems on particles.
0:00:35   All right, To start out for today, I'm gonna kind of split the agenda between our lab systems and
0:00:40   our commercial systems for lab systems. This is what tends to get the bulk of our scientists time.
0:00:46   They use it as a platform system. There's a lot of different capabilities and were frequently
0:00:50   creating or adding custom is ations. And when we say lab system were usually referring to a research
0:00:57   system that can process materials on the scale of one millimeter to one leader of substrate, her
0:01:03   powder or material per day on these systems will discuss some of the general design considerations.
0:01:09   When you're building them methods for controlling the system and reactor shapes and sizes, as well
0:01:16   as integration into some other lab systems that will show you where commercial systems they tend to
0:01:22   get the bulk of our large tolling orders and provide a great needs for us to scale up a lot of the
0:01:27   research that we're doing it for Jay Nano And when I say commercial systems are really referring to
0:01:32   systems that might process anywhere from one leader to 1000 leaders of substrate per day on these
0:01:39   systems all discuss obtaining economic throughput, handling bulk materials. So how do you handle
0:01:46   them coming in and out of the door vaporization of large quantities of precursor and, of course, the
0:01:51   actual commercial processing of those materials. And I'm sure as we go along, you'll probably
0:01:58   develop a few questions of your own that you might want to ask. Feel free to shoot me an email
0:02:03   connect with me on linked in whatever you're comfortable doing. Happy to discuss it with you further.
0:02:08   Okay, Most of the research systems at 14 0 use a fluid ized bed reactors set up to perform a L. D.
0:02:16   Most of you have probably seen something similar set up before. But the general premise is that
0:02:20   you're gonna be taking and two or on a nerve gas. You're going to flow that through a manifold at a
0:02:26   fixed speed. It's going to enter a reactor that's housing your powdered materials or your target
0:02:31   substrate and those materials. We're going to begin to fluid eyes or become suspended within the
0:02:37   travelling and to gas. Once that's done, you can start to release vapor from your precursor into the
0:02:44   same manifold, and that's gonna become entrained within the end to It's going to make its way to
0:02:49   these particles, and you're gonna have the opportunity to begin performing a nail. The reaction
0:02:54   typical setups will have you know another precursor on here so that you can start switching back and
0:02:58   forth, and you can create lots of different configurations and lots of different ways of performing
0:03:04   L. D. Toe add more granularity to the process. You would attach mass flow controllers like the one
0:03:10   you see here, maybe a needle valve to kind of choke this precursor flow and control that pressure
0:03:16   transducers, a residual gas analyzer, things that can give you insight into the process. That's our
0:03:22   general design philosophy for the systems that we have here for Conan O.
0:03:29   So minimum fluid ization is gonna be your ideal operating regime for patters. But you'll probably
0:03:35   find that most powders aren't going to do this easily. Many powders are gonna be too small and
0:03:40   cohesive to fluid eyes easily. So what our engineering team does is fabricate these clear glass
0:03:45   reactors for our scientists and that allows them to kind of fully vet out and characterize the fluid
0:03:51   ization properties of a challenging powder set. And when you have more consistent fluid ization,
0:03:57   you'll find that your reactions happen faster and more repeatedly. You also find that your precursor
0:04:02   used to just probably gonna be a lot more efficient when I go to an example of that right now, Waken
0:04:09   see that kind of very light boiling action that these powders air exhibiting when they're out a nice
0:04:14   fluid ization regime.
0:04:17   So here you can see the research system that the engineering team designs and builds for the
0:04:22   research team. We call this our Prometheus system. It comes with a lot of different features and
0:04:28   modules that you can both add or remove to the system that really makes it a highly adaptable system
0:04:34   we also have. It is a commercially available system for customers who are looking for, say, a
0:04:39   turnkey ready ale de system for particles. And I stress adaptability because we don't want a
0:04:46   situation where, like you can see in the photo on the right, where we've had to move a lot of valve
0:04:51   ing around and the system is constantly in flux or evolution and be starts to become very difficult
0:04:57   to troubleshoot, and you start to have issues getting repeatable results, and we find that
0:05:03   adaptability really starts with the control system. So that's what I'll take you to. Now, from an
0:05:09   engineering perspective is to give our scientists a lot of flexibility in Our system is set up
0:05:15   specific. So you get kind of a flexible sandbox to work in where they can add various controls,
0:05:23   temperature monitoring and a lot of sensors and things of that nature sort of provide, possibly
0:05:29   involving system, but something that making keep under control and keep saying way try to keep this
0:05:35   is clean. It's possible, even though it's an ever changing.
0:05:40   So as you just saw with See Rio, we also have a lab, you be I. And it's a gooey that allows us to
0:05:47   generate recipes. And our goal with these recipes is to minimize the physical operation to just
0:05:55   loading and unloading of powers. We strive to make everything else automated by the user. So on this
0:06:02   screen we have our recipe building interface, and users can control the state of the valves. The
0:06:08   temperature is the NFC is all in this interface here, and then they can save that state as a single
0:06:14   step in the recipe. They can name that step, and they can choose triggers or events or timers that
0:06:21   begin and end that step. So say, for five seconds, I want to hold two valves open, and then my next
0:06:28   step is to hold a different valve open until the pressure reaches 100 tour. They can generate all
0:06:35   these recipes right here on their laptop. They could then execute that as a CS V and send it over to
0:06:43   the system at their leisure and import that and begin the run that way. So although ale de is kind
0:06:50   of the obvious thing that you're going to generate these recipes for, we also have recipes for
0:06:56   figuring out minimum fluid ization, ramping up heating or ramping down heating, a kneeling and other
0:07:03   auxiliary processes that could benefit from automation.
0:07:08   In continuing with the theme of adaptability, we also equip our systems with the pneumatic relay
0:07:14   bank, and this gives really granular or fine control of the valve system to the end user. The user
0:07:21   can add thematic lines for valves, pistons, imp, actors, motors. They can build you manifolds. Put
0:07:28   those together, add them to the system kind of whenever they need them. And then they just use the
0:07:33   recipe builder to decide which ports are active when they want the valves of the relays, in this
0:07:38   case to be held open. And then we also provide them with an interlocking interface so that they can
0:07:43   prevent, let's say, unsafe conditions. So if there's two valves that they really want to make sure
0:07:48   can't be opened at the same time, even by accident, you can program in those interlocks very easily
0:07:53   as well. So we like Teoh have This feature is sort of a midway point between adding lots of system
0:08:00   adaptability but also being able to maintain a safe system environment,
0:08:07   okay and Now, of course, we have our fluid as bed reactors. Our typical design rule of thumb is to
0:08:14   not let the total bed height. So just kind of how far the resting bed of the powder is going to be.
0:08:20   We don't want that to exceed 2/3 of this straight section here called the Bed Zone. We try to
0:08:26   prevent that so that as you start to fluid as the powders, they don't get up into this expansion
0:08:32   zone, where the gas velocity is starting to slow down. We find that this just lends itself to
0:08:37   inconsistent fluid ization. When you feel it that far, we also have this disengagement zone, where
0:08:43   you really want the gas velocities slow down and all the powders to begin falling down because
0:08:48   otherwise they're going to catch on these filters here, which we're going to need in order to pull
0:08:52   an aggressive vacuum without pulling a bunch of powder into the vacuum system itself. The system is
0:08:59   usually built using conflict flanges conflict plan just lend themselves to having good vacuum
0:09:04   capability, but also a lot of good high temperature capability, which you'll find that you'll
0:09:09   probably are gonna want to have as an option for your ale de process. If you're really interested in
0:09:16   learning more about fluid ization of different powders like this, I recommend reading up on gelled
0:09:21   our classes, and that will give you more information about what powders are gonna flu dies easily
0:09:26   and how to design your reactors.
0:09:31   So for our research in labs systems, we find that when we use a common CF land size here, we're
0:09:39   gonna have a lot more usability than if all over the reactor types are unique. The reason for that
0:09:45   is that the mounting plate doesn't have to be exchanged every time you want to disengage the reactor
0:09:51   and swap it in for something else. So although this reactor on the left here looks a little goofy,
0:09:56   it allows us to take one system and have a range of Let's say this is this is about a five million a
0:10:02   reactor all the way up to a one leader reactor, all on the same system, all within the same
0:10:08   environment. And I'll show you a video of that now breeze and using our tools. One of the things
0:10:16   that the engineering team is providing ranging possible our reactors. Scientists take a 10 million
0:10:25   reactor such as this one. Quickly attention to system, do a nail and then remove this tractor
0:10:35   something much larger sailing
0:10:40   and it quickly attached to the same system with minimal amount. Richie. Never recently.
0:10:49   So the last thing I'd like to discuss about our research systems is integration into glove boxes and
0:10:56   both these scenarios. We've flanges a reactor onto the exterior of the glove box so that the user is
0:11:02   allowed to do is perform some sort of pre or post treatment of their substrate within the glove box,
0:11:09   loaded into the ale, the reactor system. Do the process and then unloaded from the ale the reactor
0:11:17   system into the glove box and maintain an inert environment the entire time. And then we always
0:11:25   provide a leak tight lid that could be attached to the reactor here. And then you can remove this
0:11:30   reactor if you need some kind of service or some other treatment. That has to happen with the glove
0:11:34   box. Okay, that concludes our talk about lab systems. You probably noticed that the focus for those
0:11:41   lab systems was mostly on adaptability and ease of use now that removing on the commercial systems.
0:11:49   The focus will really change more towards safety, cost efficiency and throughput efficiency. So when
0:11:56   we're talking about commercial scaling and things like that were usually looking to do something
0:12:01   called Spatial Ale de, this is different than your typical batch. LD where you're going tohave
0:12:09   pumping purge of precursor through a single chamber and your substrates just going to remain in that
0:12:14   chamber throughout the process with special ale de, you're actually going to allow the substrate to
0:12:20   move through different precursors owns. This technique actually removes vacuum purging from your
0:12:27   throughput calculation and allows you to operate more of, ah, call it continuous ale de or
0:12:33   continuous processing mode. So I've sort of mocked up what a spatial ale de process could look like.
0:12:40   What the great represents here is powder that's getting ready to be processed, and then in each
0:12:46   chamber reactor. Whatever you wanna call it, we've got different precursors pre charged into each
0:12:53   reactor, so this could be your precursor, a precursor. Be back to a back to be, and these will be
0:13:01   isolated either through some sort of inner gas or through calving. When you're ready to begin the
0:13:07   process, you would allow this substrate to move through these different zones and undergo the ale,
0:13:13   the reaction as it's traveling through. Once this powder is moving through the system, you can
0:13:20   reload this plot IQ end and begin processing more power, and then you're going to end up with sort
0:13:26   of a more of a cyclical or continuous system here. And that allows us to really reduce the overall
0:13:36   cost of operations as well as increase the throughput. So some new challenges that we're going to
0:13:43   face now is how do we deal with these large quantities of material? We're talking about tonnage of
0:13:48   powder now instead of a few grams of powder. And then how do we deal with the large quantities of
0:13:54   precursor that we're going to be using? Because, as many of you probably know, most ale de
0:14:00   precursors have a lot of hazards associated with them, and now we're adding a substantial amount of
0:14:07   precursor to our everyday use.
0:14:12   So first we'll talk about the handling of these large quantities of powder. So when you're receiving,
0:14:19   let's say tonnage of powder, you're usually going to be receiving them in a super sack or a bulk bag.
0:14:25   That's kind of the industry term for it. I have some images of them below here. Usually these are
0:14:31   about one cubic yard, and they have these loops here, which could be lived into a hoist or a
0:14:36   forklift. This material is usually like a a woven polypropylene. It feels kind of like a heavy duty
0:14:43   tarp that you might buy at a big box store. And typically there's an inner liner here that's
0:14:49   actually sealed, and that's what's protecting the raw material from the shipping environment. So
0:14:56   especially if you're shipping freight or a shipping on a boat or something like that, you want to
0:15:00   make sure that that materialise, sealed and not exposed to that type of atmosphere. So what
0:15:06   Fortunato actually does is in a great what you would maybe refer to a standard bulk handling methods
0:15:13   into their ale de process. I'll show you that next.
0:15:19   So at this point we have some images of Forge Nano's, I guess call it shipping and receiving type of
0:15:28   process on the left Here. This would be sort of how we receivable bag, so either the boat bag is
0:15:33   already sitting on a storage show. Or maybe we just received it off of a truck fork lift would then
0:15:40   bring that bulk bag over to this framework here. And this yellow arm here is the end of a hoist. You
0:15:48   wouldn't then lower that hoist and attach the loops to these different arms. You would then raise
0:15:54   the bulk bag up, and you can attach that bulk bag. You can't see it very well here, but there's a
0:15:59   sort of a cone attachment, and you're able to attach the inner liner and seal it against this code,
0:16:06   and that allows us to in Erdely handle whatever material it is that we're receiving. At that point,
0:16:14   we can empty the bulk bag into this cone, and it's going to be new magically conveyed. So when I say
0:16:20   dramatically conveyed, I'm talking about almost a giant vacuum cleaner that has a long hose here
0:16:26   that's attached to the bottom of the cone and that will then move the material up and into this
0:16:32   giant hopper. Here. This hopper is under an end to blanket, so now we can store this material in a
0:16:39   safe operating environment that's not going to be contaminated by the atmosphere, anything like that.
0:16:46   At this point, we can actually batch out small quantities of material. We have a little rotary
0:16:51   mechanism on here, and then we can dramatically convey that to different areas of the facility if we
0:16:57   want to operate in small quantities or we can unload the entire bulk bag of material into one of our
0:17:04   larger commercial processes. So this can actually allow us to distribute materials throughout the
0:17:08   facility when materials air finished. We once again can dramatically convey them. And we have a
0:17:15   sieve or a commercial screening system here and that allows us to break down any if you had a
0:17:22   cooperation or if there was any contaminants in this in the material that would allow us to capture
0:17:27   it here and then that material is going to fall through the system here, the finished goods. And in
0:17:33   this scenario, we have a glove box, and what we're going to be doing here is actually matching this
0:17:40   out into small bags so that we can distribute the material to multiple destinations. Another
0:17:47   scenario would be that you could actually attach your book bag. Here. You can attach the inner liner
0:17:52   to this outlet cone, and you can fill an entire bulk bags seal it and then ship it off as a bulk.
0:17:58   Good. So we have either option that we've developed here it for Jim, and that's kind of our way of
0:18:04   dealing with large quantities of material.
0:18:09   Okay, let's discuss a bit on handling large quantities of precursor on lab systems. It's fairly
0:18:17   typical to directly heat the bubbler or container that your precursors in an induced vaporization
0:18:23   that way. But when you get to commercial systems, that would mean heating a much larger quantity of
0:18:29   precursors, say, like 50 leaders and their safety implications to doing that. You might also degrade
0:18:36   that precursor, depending on what temperature your heating it to and for how long. So, for that
0:18:41   reason, you don't really want to heat those large quantities of precursor. Now you might be able to
0:18:47   incorporate a sort of a commercial vaporizer. Call it off the shelf that is typically used in
0:18:53   semiconductor ale, de or semiconductor CVD. The only issue that we run into with those is they tend
0:18:59   to run very, very dilute, so you have a small quantity of precursor mixed in with your carrier. Gas
0:19:05   powders have a much higher surface area than flats that you might encounter in the semiconductor
0:19:10   industry. So we're going to spend a lot longer running through a nail D process because it's so
0:19:14   deluded and that really affects the throughput. So what we do unfortunate are what we would
0:19:20   recommend here is to incorporate, say, a liquid flow meter or liquid flow controller to dispense
0:19:28   really, really precise quantities of precursor and then run that through your vaporization process
0:19:34   so we might refer to that as like a direct injection process. We also recommend attaching a nozzle
0:19:41   at the outlet of your direct injection so that you can Aarhus allies the liquid as best you can. And
0:19:48   that will really put you in a better position to do vaporization. So try to show you an example of
0:19:54   that right now we have this mocked up. This is just the outlet of our dispenser with a nozzle on it,
0:20:00   and we're just dispensing water right now.
0:20:06   Looks like it's gonna play in the frame. Okay, so the water just got dispensed through the nozzle
0:20:13   and you can see what looks like steam. It's really just crystallized water. But because the nozzle
0:20:18   is so fine, we're able to get really, really small micro droplets, and that puts us in a much better
0:20:23   position to Vape arise that precursor. Thank you. Okay, let's finally get into the commercial
0:20:32   processing of these powders. So we've managed to unload the powders. We know how to vaporizer.
0:20:39   Precursors are actually ready to perform the ale de. So how do we do that in a way that's fast and
0:20:45   cost effective. So here I'll be showing you an atmospheric application of spatial. They'll be. This
0:20:52   is for a lady processes that don't really require a vacuum, and that just needed a nerd environment
0:20:59   in order to undergo deposition. So what you'll notice here is that we're going to have a bed of
0:21:05   powder that's going to be traveling along. Call it a conveyor and you're going tohave gas that's
0:21:12   coming out from the bottom of this conveyor into the different zones. So the first zone being at two
0:21:20   and will be running our precursor A and to blanket again and then Precursor B, and we can just
0:21:27   continue that trend throughout the entire process. So an example of what that might look like here,
0:21:33   where we're going to load this material into a hopper and then unloaded into our conveyor here for
0:21:39   the actual ale de process.
0:21:44   Okay, so I have a quick video here so you can see how this process actually looks so that you can
0:21:49   visualize the powder bed traveling through the different zones. In this scenario, we have a powder
0:21:57   bed of cathode materials, and we're gonna dress drop some white talc powder on there so that you can
0:22:03   kind of visualize the bed moving. Start that So that talc is gonna drop in just a second here, there
0:22:10   it goes. And now you can actually see the bed traveling and it's gonna hit the first precursor zone
0:22:16   here in just a second.
0:22:22   There goes. That was the first zone right there. And you can kind of see the little bit of mixing
0:22:26   that took place. It's about to hit the second zone, and you can see that this is moving at a really
0:22:33   decent clip. We've already gone through one full LD zone, so our throughput is really regulated by
0:22:40   the bed speed. So how fast this bed is moving through the different systems and then how wide this
0:22:47   bet is, and of course, how high this bet is. So this bed is actually fairly shallow here. It's
0:22:52   probably not much more than 1/4 inch to 1/2 inch high, But as we increase those numbers weaken
0:22:58   directly, increase the throughput, and then the number of ale the cycles that you can perform is
0:23:03   really just limited by the length of this conveyor system. You could also consider doing a recycle
0:23:09   loop. If you want to just double the amount of cycles, you could do it that way or if your footprint
0:23:15   is limited. Perhaps you would just combine two of these in parallel.
0:23:21   All right, well, now that you've seen the powder traveling through the system and kind of have a
0:23:26   fuel for how the system operates, I want to sort of zoom in here and focus on the reactive zone or
0:23:33   the deposition zone. So okay, good started. And over here we can see the actual reactive zone, and
0:23:41   you can see the gases penetrating through the powder bed. It almost looks like a fluid ization is
0:23:46   occurring here. So in order to accomplish this, we have to sort of optimize for each powder that
0:23:53   we're going to use, so we might have to adjust the gas velocity here we might have to adjust the bed
0:23:59   height. It just depends on the powder properties, and once we've optimized it, we can kind of plug
0:24:07   and go and raw materials very, very quickly. Additionally, we can take other measures to deal with.
0:24:14   Precursors are sorry substrates that are less reactive or precursors that are less reactive. So in
0:24:21   order to deal with that, we will slow down the bed height and we will allow the residents time to be
0:24:27   increased for the powers in these reactive zones. And that will just maximize the amount of time
0:24:32   that the powder has to react with the precursor. Maximize your chances of having 100% L D coverage.
0:24:41   Additionally, weaken even lengthen physically lengthen the zones. If we need to really increase that
0:24:46   duration, we can also adjust things like gas velocity already mentioned bed, height, bed with things
0:24:54   of that nature in order to play with the throughput and optimize even further if there's really a
0:25:01   powder that we're trying to hone in on. So that kind of concludes our commercial processing section
0:25:09   that does conclude the content of the presentation. Thank you all for listening, and I hope that you
0:25:13   found a presentation useful once again. If you do have any questions, comments or you just wanna
0:25:19   chat, shoot me an email. It was a pleasure presenting thanks again and enjoy the rest of the Senate.