Below is the text version for the "Development of Real-Time Characterization Tools and Associated Efforts to Assist Membrane Electrode Assembly Manufacturing Scale-Up" webinar, held on August 22, 2018.

Eric Parker, Fuel Cell Technologies Office

Hello, everyone, and welcome to the U.S. Department of Energy's Fuel Cell Technologies Office webinar. We've got a great presentation this month from Michael Ulsh on the Development of Real-Time Characterization Tools and Associated Efforts to Assist Membrane Electrode Assembly Manufacturing Scale-Up.

My name is Eric Parker. I provide program support within the Fuel Cell Technologies Office, and I'm the organizer for the meeting. We'll begin shortly, but first, I have a few housekeeping items to tell you about. Today's webinar is being recorded, and the recording, along with all the slides you'll see today, will be posted eventually, and we'll be sure to let you know.

All attendees will be on mute during the webinar, but if you have a question, you can submit it at any time to the host in the WebEx chat box you see on your screen.

With that, I would like to introduce today's DOE webinar host, Nancy Garland, who is joining us here at headquarters.

Nancy Garland, U.S. Department of Energy

Okay. Thank you, Eric. I just want to say a few words about Mike—Mike Ulsh—who's our presenter today. He's developed new techniques for identifying manufacturing defects in MEAs. His work defines sensitivity requirements for diagnostics that will lead to improved production tolerances and lower cost MEAs.

His approach, to learn from industry partners what quality control is needed, used modeling from Lawrence Berkeley National Lab to guide the development of diagnostics in the lab, validate the diagnostics in-line, use in situ testing to understand the effects of defects on fuel cell performance, and finally, transfer the technology to industry.

Mike came to NREL from 3M, and we're glad he's here. Mike, take it away.

Michael Ulsh, National Renewable Energy Laboratory

Okay. Thank you very much, Eric and Nancy, and good morning or good afternoon to everyone. Let me share the file here. All right. It doesn't seem to want to let me do that. There we go. Okay. Does that look good, Eric?

Eric Parker

Yep. Go ahead, Mike.

Michael Ulsh

Okay. Great. So again, thank you to FCTO for this opportunity, and appreciate everyone's time who is listening in. Sorry. Let me quickly get rid of my email. Okay.

So, as Eric said, and as Nancy alluded to, I'm going to talk a little bit about a suite of activities that we have going on here at NREL related to assisting in the scale-up of membrane electrode assembly manufacturing. Let me see. Sorry. I'm not advancing here. Just—hopefully it'll catch up. Huh.

Eric Parker

Mike, try hitting the arrow keys.

Michael Ulsh

Yeah, I am. Okay. All right. So—okay, so Eric already talked about, I think the chatting and question and answer.

So, moving forward—again, apologize for that—want to make sure to mention right at the top here that the activities that I'm going to describe here have been supported for several years by FCTO, and in addition, support has been received from other areas of DOE as indicated here, as well as in industry-funded projects. I'm having troubles again. I'm sorry.

Eric Parker

Mike, if you're having issues, I can take over.

Michael Ulsh

Let me try this. Okay. So here's what I'm going to talk about this morning. I'm going to start out with going over a little bit of the context within which we think about MEA manufacturing, kind of the reason that we're doing what we're doing. Then most of the time I'll spend, again, talking about the development of these real-time in-line characterization techniques. So this is quality inspection for MEA materials as they're being fabricated in, for example, roll-to-roll processes.

Then at the end, I'll give a little bit of an overview, so shorter sections, but just to give you a flavor of kind of the two other parts of the project, one being the development and use of specialized in situ diagnostics to help us understand how variations in MEA materials can affect the cell performance and lifetime. And I want to be sure to differentiate here. We're not talking about defects that are created in cells as they're being operated, and how they impact durability. That's different activities supported under FCTO. What we're looking at specifically is as manufactured defects.

And then finally, work to understand the relationships between electrode materials, processing methods, and performance, so looking at the processes themselves.

Okay, so I'm going to start with a little bit of context of the bigger picture here. And so what I show here, the two graphics in the middle were shown by Sunita Satyapal, FCTO Office Director, in her plenary talk at the recent AMR. And basically, the synthesis says markets for multiple applications right now are expanding and have been expanding. In terms of units, in terms of power output, in terms of revenue, these are all increasing. And in addition to the extensive efforts supported by FCTO and other areas of DOE, there's an increasing amount of state activity beyond just California, for sure in the Northeast, and other states.

So markets are increasing, and due to that, we need to think about how to transition to high volume techniques for manufacturing MEAs.

So in that vein, I'm going to—I show a graph here that a lot of you are probably familiar with. This is from the strategic analysis cost model for automotive fuel cell systems. The plot shows fuel cell systems cost as a function of annual production rate. And there's three lines here that are actually showing cost projections for three different technology levels of the automotive system. But what I want to talk about here, how I want to use this graph, is to say that I think a lot of us are familiar with these cost analyses, but the point is that the system and stacked cost analysis assumes the use of high volume manufacturing methods for MEA materials, and yet the truth is that a lot of these modeled manufacturing technologies are not wholly proven out at scale, and so there's still learnings to be done here about scaling up these technologies.

In particular here, I show some examples. So these are three cartoons that were taken from strategic analysis annual reports, the most recent one. And I'm certainly not going to go through these in detail, but basically, they show graphically depictions of three different roll-to-roll manufacturing processes, one for membranes, one for electrodes, and one for integration of multilayer assemblies, including gaskets. And so, again, the point is just to say that these models assume these kinds of processes, and yet they're not always—all the information in terms of scaling these up are not always known.

So the premise for our activity here is that MEAs need to be fabricated using scalable processes to enable high volume and low cost, number one. Number two, the processes used are typically atmospheric pressure and solution-based processes, not always, but in many cases, because of the nature of the materials that we want to use, including polymers and particles.

And sort of the punchline is that when these materials, when processed in a variety of different methods, tend to have a variety of macro and microscale defects—several examples of those in a variety of the different layers of the MEA are shown there at the right—which may affect the performance and lifetime of the cell. And if they do, we call them a defect.

So based on that, here's the challenges that we're trying to address. Number one, how can we detect these defects in MEA materials in ways that are amenable to the fabrication process? In other words, during high volume roll-to-roll manufacturing of these materials? Second, how do we understand how the defects that can be formed during either the fabrication or subsequent handling of these materials can affect performance and lifetime? And again, I differentiate that from specific durability testing. And finally, how do we understand how the parameters of the processes themselves can affect the morphology and the performance of the fabricated layers?

And certainly as Nancy mentioned, our approach is very clearly and explicitly to work with industry to try to address these challenges. And you can see here a number of industry members that have—that are either currently working with us on the project, or that we've worked with in the past.

Okay, so I'm going to start now into, again, some detailed discussion of the development of these real-time in-line characterization techniques, but I'll start answering a couple of questions. So I think it's a valid question to ask why worry about quality in the first place.

So I show two analyses here. The top one is one that I worked through with the strategic analysis team, again using their automotive fuel cell cost model. And the plot then at the top right shows stack process yield, against on the Y axis, the increase in stack cost.

And so what you can see here is, if you run the model assuming 100 percent yield for all the processes used in fabricating the stack, and then run it again at only 90 percent yield in all those processes, the stack—the resulting change in the stack cost is an increase of 60 percent, so that's a significant impact. And you can see about the same thing in the example at the bottom, then. This is an analysis done by Lawrence Berkeley National Lab on a ten 10-kilowatt PEM backup power system. And you can see, again, if you look at the stacked areas sort of from the purple down, which is the stack part, you can see a high impact on system cost shown on the Y axis as a function of stack module yield. So quality is important, and can have a significant impact on cost.

And so this importance is then reflected in the multi-year R&D plan. What you see there are two tables from the manufacturing section of the MYR&D plan. On the left is a table in task five area, which is quality control itself. And then on the right is task one area, which is MEAs, but also embodies some aspects of quality. So it's these tasks and associated milestones, which are informed by industry that drive the work that we're doing.

So at a very sort of cartoon level, what we're trying to do is assist in the scale-up of these technologies, in particular by helping people transition from manual inspection of MEAs to in-line inspection that is amenable to these kinds of processes. And so that means that these techniques need to be rapid in terms of both the measurement itself, as well as the data processing, preferably. They need to be implementable in an in-line fashion. And so they need to be able to be set up and be operated in an environment, kind of what you see on the right there. They certainly need to be non-destructive, and in particular, what we focus on is techniques that can be aerial in nature, instead of point measurements.

And that's really important because we know that discrete defects in these different materials can have an impact, and so we want to be able to have imaging kinds of inspection that can detect discrete defects.

So this is an overview of the different techniques that we've developed. I won't talk about this line by line, but just to say that if you look along the left hand side there, you can see that we've focused on techniques that address different kinds of defects in each of the different layers within the MEA, as well as, if you look at the bottom, different kinds of sub-assemblies of the MEA, including the full MEA itself, where you might want to look at through-plane types of defects or abnormalities, such as shorting and failure of membrane integrity.

So I'll give some examples of all these different kinds of techniques, but I wanted to include kind of an overview that once you have this deck, if you'd like to download it, you can look at this in more detail.

Okay, so I'm going to start out now with some discussion of the techniques themselves, and I'm going to start talking about some infrared-based techniques. In particular, here to start with, some that are relevant to catalyst-coded membrane kinds of structures, both in terms of the electrode uniformity or defects, and then looking at shorting. And I want to say relative to using infrared, you know, unless we look at just differences in emissivity, which isn't necessarily all that useful, we—to use infrared, we need to create an excitation of the material that generates a thermal response that then we can measure with the infrared detector.

And so what I'll talk about as I go through this first part is different kinds of excitations that we can use on these materials to get a response that tells us something about the uniformity or a material property that we're interested in measuring.

So this first case we call IRDC, and it basically uses a direct current excitation, as you can kind of see in the little cartoon at the top right, to generate—so basically, if we contact two portions of the electrodes on a membrane, we can generate a voltage across that electrode layer. That results in a current, which causes some resistive heating, and of course, we can modulate that to get the kind of response that we want in terms of magnitude. So this gives us a rapid and aerial measurement that is—that tells us something about the thickness uniformity of the electrode.

And across the bottom, then, you can see various examples. On the left is a pristine electrode, where basically by the all-pink response you can see that we get a nice uniform response there, whereas the other three examples are cases where we've got this square defect shown in the cartoon in the middle of the electrode of various intensities, if you will, and you can see by the resulting thermal response that we can tell something about the uniformity of that electrode by the generated thermal response.

But of course, we want to do this in a fashion that is amenable to the web-line. So here, you can see a couple of pictures of setting this technique up on one of the web-lines that we have at NREL. In particular, on the right you can see this system with brass rollers that the CCM is running through. And using rotating electrical contacts, we can basically create that voltage difference and run a small current through the electrode as it's passing in between that series of rollers to generate the thermal response, and then you can see the IR camera at the top.

So now I'll give a couple of examples. You can see several here of different kinds of electrode defects. Just to talk about the one on the left as an example, in this case, we had a gas diffusion electrode where there was some carbon debris on that electrode prior to lamination of a membrane, and you can see the result of that debris in the dark red spots in the thermal image that's sort of underlaid underneath the optical image, and similar kinds of results from these other kinds of defects.

So in addition to seeing that we can detect defects to begin with, we want to understand the impact of processing conditions on the technique itself. And so, for example, we had this case where we had multiple different defects in this particular area of a sample, and we wanted to see could we equally detect these defects under different processing conditions, in particular under changing the line speed at which the sample was running by a factor of six. And as you can see, we were able to modulate the excitation conditions to give equivalent detection over a range of line speeds.

And I want to point out here that this particular technique had been demonstrated in a previous activity on an industry partner's roll-to-roll CCM manufacturing line.

Okay, so sticking with that same excitation idea, but now thinking about this through—more of a through-plane case, where we're interested in understanding, okay, I have a multilayer structure, maybe a full MEA, can I detect shorting pathways in that multilayer structure that might be caused by some things like the two examples that are shown at the top of the slide.

So here now, again, we kind of use the same idea, but now we apply the voltage to opposed rollers on either side of the multilayer structure, and using that, shorting pathways can allow a local current to pass, causing resistive heating at that location, which gives us a thermal response. And so again, where the previous technique was looking at more of an in-plane electrode uniformity measurement, this is now giving us an aerial identification of the location of shorts.

So again, several examples that you can look at here. For time, I'll just mention the one in the top left. This was a case where we had cell materials where the gas diffusion media had some rough fibers on the surface, and when that was hot pressed, those fibers penetrated into the electrode, enabling shorting pathways, which obviously, you can see here by the bright red and orange circles. So again, just one example here where we can see shorting pathways in a technique that's amenable to in-line measurement.

Okay, so I have been talking about CCM type constructions. Now I'm going to talk about some techniques that are perhaps a bit more applicable to gas diffusion electrodes, looking at uniformity in identifying defects. Also, kind of in the same way that we implemented an in-plane technique in a through-plane way, we're going to look at this different kind of excitation to look at membrane integrity, as well as measuring membranes in multilayer construction.

Okay, so the example here is showing a different excitation technique that we call reactive impinging flow. And what we're doing here is using a non-flammable reactive gas mixture, so containing a little bit of hydrogen, and we're using an array of jets to impinge that flow on to the GDE surface to generate a catalytic reaction of a controlled temperature in the GDE in such a way that non-uniformities in the electrode then would result in differences of the thermal response.

Now you can see an example of that at the bottom, where the right—the left-hand side of the image shows the GDE before the excitation, then that gas knife is in the middle, and then the material coming out to the right-hand side of the image has gotten this excitation and obviously has an increased temperature resulting from that catalytic reaction.

And so this gives us, again, a rapid aerial measurement of the GDE electrode in terms of thickness uniformity, and as we'll see, we can even get information about loading. So again, we want to implement this on the web-line. The image at the bottom left shows the entire web-line threaded up and running a GDE web material, and then the larger image—and let me see if I can get the pointer here—okay. So in the middle here, what I want to show, this metallic box is the plenum for the reactive gas, and on the bottom of the box that you can't see in this image is that array of openings that impinge the reacting flow onto the GDE web and create the excitation.

So again, some examples. In this case, this is a GDE web that we coated on our coating line here at NREL, and we used the technique to create a uniformity map of the entire coated sample, because we had specifically been running at different processing conditions, and we wanted to understand how that related to the uniformity of the coated electrode.

So as you can kind of see in the middle, then, the infrared image shows the results from different sections of that coated web, and on the left-hand side, you can see in the little inset optical image that we've got an electrode under whatever processing conditions were that is quite non-uniform and has some uncoated sections, and you can see that in the thermal image, because of the blue spots, or cool spots, so they indicate those uncoated or poorly coated regions, whereas as you move to the right, in that sample and in the infrared imaging, you can see that the imaging gets more and more uniform as the sample gets more uniform, as indicated by the second optical image farther to the right.

So we can see relatively—we can identify defects and understand uniformity, but at the bottom, then, you can see that we used XRF to quantitatively measure the loading in each of those GDE sections, and you can see that the average temperature generated during the excitation in that blue line plot just above the loading numbers indicates that the average temperature is in fact proportional to the loading.

Okay, so again, then, we took this same idea of reactive excitation, and now we're implementing it in more of a through-plane fashion. Here, we're exposing a membrane-containing assembly to a hydrogen-containing gas, again, as you can see in the cartoon at the top. In this case, the idea is that the hydrogen will head back through the pinhole or whatever fissure in the membrane, again, react on the catalyst to give us a thermal response.

And so here, again, we have a rapid aerial detection that now is indicative of the failure of membrane integrity. So again, several examples of this. As you could probably see on the previous slide, this current version of the technique is a static technique, but if we just look at the example on the right, we can see this case where we had a membrane with a 90 micron diameter pinhole, and using this technique with the reactive flow excitation, you can see then the hot spot in the middle of the measured region of the infrared image shows the—or indicates the location of that pinhole, so basically indicates that we have gas crossing over through the membrane as a result of this defect.

Okay, so one more technique here relative to infrared imaging. In this case, we're using the same kind of detector, but we're basically just putting heat into the multilayer structure that we're interested in here, either by using light or radiative sources.

We then measure the peak and decay of that input of heat with the detector, and we relate then that measurement real time to a thermal model of the material itself, so that we can back out physical properties that we're interested in, for example, thickness or porosity of a particular layer.

And so at the bottom then I give an example of a result here where we had a series of 10 different half cells—in other words, membranes laminated onto a GDE—and each of the membranes in those 10 different samples had a different thickness. And you can see then where we plot the thermal response across the bottom as a function of the thickness of the membranes in those different samples. Then we've got a nice, linear response.

Okay, so I've talked about infrared techniques so far, but we've also done a lot of work over the life of this project looking at different kinds of optical techniques for these materials in particular to detect defects in membranes and electrodes. And then I'll give an example of a newer technique that we're still really actively working on to use an optical technique to basically image or map in real-time the thickness of the membrane.

In general, the methodology here is that we can use either transmission or reflective imaging, depending on what we want to see, and we can do that in specular or diffuse modes. And so the idea there is it's important to understand and to control the various angles of the light source as well as the detector relative to the path or the measurement point. And so we can do that by using this flexible inspection apparatus that we have on our web-line that you can see in the middle image there, so you can see the—in this case, the line camera at the top, and a bank of light sources kind of at the lower left. I think that you can see that there's a membrane web that's running over that roller kind of in the middle. And so again, this apparatus allows us to change the angles of the light source and the detector in very controllable and repeatable ways to allow us to look at these different modes of inspection.

In addition, we have this filtered hood, which is shown in the image at the top right, that we affectionately refer to as Darth Vader's helmet. Nancy wanted to make sure that I worked that in. And the hood then basically goes down over top of this apparatus to let us eliminate external light, which gives us a better signal, as well as minimize contamination, which is pretty important, especially for looking at membranes, where a lot of the membrane materials, as many of you probably know, tend to like to grab fibers and dust in the air pretty quickly.

So in addition to developing the techniques themselves, though, we have spent a good amount of effort, especially in recent years, developing automated defect detection algorithms and classification algorithms at—which enable us to then—to provide full width and full length high resolution product roll mapping, and I'll show an example there.

Here on the right is just some examples of different kinds of defects and membrane materials. I imagine that none of these surprise many of you too much to see. Certainly, we've looked at PFSA membranes as well as other kinds of chemistries of membranes in terms of implementing these techniques, with a wide range of thickness. We've looked at both reinforced and not reinforced membranes. And as you can see in the examples, we can demonstrate detection of both discrete and aerial kinds of defects.

So then at the bottom, and this is simulated data, but to give kind of an idea of this automated full-roll metrics, so for example, if you're manufacturing a membrane, you might think that it's useful to understand all of the membranes that were detected on a particular roll in terms of how big they were and how dark they were, dark in terms of the resulting essentially detection in the optical imaging. And you can imagine that you might think, as the membrane manufacturer, that a dark defect might say one thing to you about the source of the defect, and a lighter defect might say something else to you. So you might be interested in understanding this.

And so this image is just kind of one way that we can output this data where we plot out all the defects—again, this is simulated data—all the defects on a roll as a function of their area and their intensity or darkness. So just one example of how we could plot these things out, to basically give you a map of an entire roll.

So here, I want to pause and just mention an activity that we've had for a few years now that is a great example of how we can transfer technology from the lab to industry. Here, we've had a collaboration with mainstream engineering under a series of SBIR projects that they've had, where they're taking some optical inspection technology that we develop here, they're adding some of their own technology, and the idea is that they're developing commercializable inspection equipment that we can then bring out to the industry to basically bring these different techniques into play.

And so, again, I'm not going to go over the details here, but just to give you this example of how we can work with industry, and in particular, that there's an activity that FCTO is supporting where hopefully we're bringing these kind of techniques to market with an industrial manufacturer that it can actually put these into relevant forms that can be implemented onto roll-to-roll lines.

Okay, so back to the techniques. Here's another technique, and I mentioned this specifically, where we're interested in here in actually mapping real-time the thickness of the membrane. And so here, we're using—the methodology is to use interference fringes and reflectance spectra where we then perform Fourier transform to find the thickness of the membrane in each pixel of the detector.

And importantly, we've seen that we can use this technique with membranes without and also with reinforcements, and critically, we can do this while the membrane is even still attached to, for example, a casting liner or other protective liners. And so you can see some examples of that on the right. And as I mentioned, this is still in development, so this is active, ongoing research here, but some very interesting initial results.

So just one more set of examples, then. Here, we're looking at identifying surface defects in electrodes. You can see both GDE and CCM examples on the right. And then on the left, this is, again, a GDE with different defects. But another example where we've applied automated defect detection, as you can see by the numeric indices on the image, identifying each of the defects.

Okay, so that was the presentation on the defect detection techniques, kind of giving you a broad overview and some examples. Now I want to spend the rest of the time, last 10 minutes or so, talking about—just giving a little bit of a flavor of the other two parts of the suite of projects here.

And the first one is looking at the development and use of specialized in situ diagnostics, again, to help us understand how these irregularities in the MEA materials could affect the performance and lifetime of the cell. In other words, are they a defect or are they not a defect?

And so I'm going to do that—excuse me—starting out with an example here, to kind of show you how we think through this.

So in this case, we made an MEA, and the membrane in that MEA had a pinhole that you can see characterized in the bottom left here. So we make an MEA, then, and if we do total cell initial performance and we compare that to the exact same materials and MEA structure but without a pinhole, you can see in that kind of top middle image of the polarization curve, there's really no difference between those two cells. But we want to understand this in more detail, because frankly, we don't believe that just looking at total performance tells us the whole picture.

So for example, we can take one more step, and we're still looking at initial performance, but if we use a segmented cell, and those results are then beside that in the top right, we can see that in fact if we look at the local performance in the cell, we can see reduced performance in the regions local to the location of the pinhole as indicated by the red spots there. So they indicate lower performance relative to the pristine cell.

So we get a sense of, okay, something is going on there, but that still doesn't tell us whether this defect or this irregularity is going to impact the operation of the cell over time. So then we can do the next step, and run a drive cycle, and that's shown in the middle graphic, and you can see in fact that the MEA with the pinhole in fact degrades faster than the pristine MEA over the duration of this drive cycle test.

So now we see that not only are there local impacts of that defect, but we see impact in degradation of performance over time. Finally, we can run an accelerated stress test, and the results there shown in the bottom center, and we can see that in fact that defected MEA fails faster than the pristine MEA. And by using some specialized cell hardware that we developed here, we can basically do infrared imaging of the cell during a hydrogen crossover test. And as you can see there, you can essentially see the crossover effect of that pinhole in the cell from the infrared imaging.

So the idea here is we can't look at this, we can't understand if these defects have an impact just by looking at total cell performance. We have to look at this spatially and temporally.

So let me see here. That initial example obviously was a membrane defect, so I just want to show this table—I've showed this for a couple of years in my AMR presentation. On the left, you see a variety of different electrode defect studies that we've done, so looking at different types and sizes and locations of defects. And again, this table basically shows the same thing, where if we just look at total cell initial performance, we really don't see much in terms of the impact of the irregularity, but if we look at local performance with the segmented cell and then look at these more temporally focused tests, it's really there that we can understand the impact of these kind of defects.

So a couple more examples, then. Here we're looking, again, at—or we're creating electrode defects. It's a bare spot in the center of the cathode. But what we're really looking at here is, what is the impact of the thickness of the membrane within which or under which, if you will, this electrode defect is sitting?

And so if we looked at that with the segmented cell, as you can see at the top right, and in fact, we're looking at two different sizes of that bare spot, and the top two images are on a 25-micron membrane, and the bottom two are on a 50-micron membrane. So, again, you can see locally we can see the impact of all of those defects, but there's really not much of a difference between the thicker membrane and the thinner membrane.

However, when we do these—when we go farther, we can see that these differences emerge. And in particular, if you look at the bottom left, we can see that when we run the drive cycle, there's a small impact of the defect on the thicker membrane, but there's a catastrophic impact of the same defect that's on the thinner membrane, so that's the upside-down triangle with the pink line. So there, we actually see a total loss of performance due to that defect in the drive cycle testing.

Similarly, if we look at those same defects then using the AST, we can see again in the upside-down triangle with the pink line that the defect on the thinner membrane fails much, much sooner than any of the other cases.

One more example, then, quickly, where here, we're now comparing, instead of—well, we're now comparing thin spots in the electrode to those bare spots that we looked at in the previous examples, and a thin spot is in fact 50 percent thickness reduction of the electrode instead of 100 percent thickness reduction. And again, we're looking at this on two different thicknesses of membranes, but first and foremost, what we see in both of these data sets at the top and bottom on the right is that the thin spots actually have about the same effect as the bare spot. So that's the—to us, that not only do we have to detect bare spots in electrodes, but we have to detect thin spots.

And in particular, then, if you look at the top graph, we can see that both defects on the thicker membrane have somewhat of a reduction—cause somewhat of a reduction in performance over time, but on the thinner membrane, they both create a catastrophic loss of performance, as you can see in the bottom graph. So this is the kind of work that we're doing to try to understand what's the impact of these different kind of defects that can emerge in these materials as a result of the manufacturing process.

Okay, finally, a little bit on this last part of the project, where we're trying to understand the impact of the process itself. And I like this graphic because it kind of shows pictorially what we're trying to do. We're really trying to help understand the transition from lab scale to scalable electrode production, in this case. And so on the left, then, you see the ultrasonic spray system that we have in our lab, which is kind of the lab standard at this point for research MEAs. And then obviously on the right you've got a roll-to-roll coating line.

And so the idea, though, is that each of these different kinds of processes have different conditions associated with them, both in terms of the ink, so you can see dilute versus much more concentrated, and the different kind of mixing that we might use for these different processes, as well as the processes themselves. And in particular, when you go from spray to roll-to-roll, you've got basically either a sequential build-up of multiple layers to get the loading that you want in a spray process, whereas a roll-to-roll process is generally going to coat a single layer all at once.

So all these things in terms of the ink and the process, we need to understand what those changes are.

And so to start here, we're looking at the rheology of the inks, and I'll describe these graphics a little bit, but I want to start by saying we want to understand the rheology, because number one, different roll-to-roll processes desire, if you will, different rheological behaviors in the inks that you use, and so we need to understand the rheology just from the perspective of getting a good ink for the process.

Secondly, the rheology tells us something about the level of agglomeration and interaction within the particles in the inks. And so obviously, that could impact the morphology of the coated layer, which then could impact the performance.

Okay, so all four of these graphs are showing—these are standard rheology graphs showing viscosity as a function of sheering, and then each of the data lines in the four graphs is showing the data as a function of the ionomer-to-carbon ratio, or addition of ionomer, in these inks.

And so at the top you can see carbon-only inks. You can see we've got that for both Vulcan and HSC supports. And basically, this says that addition of the ionomers stabilizes the carbon particles in these inks, and basically transitions them from sheer thinning at low ionomer concentrations to more Newtonian behavior as you get up towards what we would think to be more optimal [inaudible] from a performance perspective.

At the bottom, then, now we're looking at platinum on carbon, again, for the two support cases, and on the left, we can see for platinum on Vulcan that we get about the same behavior as just the carbon ink. But very interestingly and importantly, when we looked at platinum HSC, we do not see that same behavior. In particular, as we add more ionomer, this does not stabilize the particles, and we get a—maintaining that sheer thinning performance at higher and higher viscosities. So, again, kind of the punchline here is that the influence of platinum appears to be dependent on the supports, or said another way, the location of the platinum on the carbon support.

I won't go—the next slide is really just showing that we're doing these same kinds of things now with unsupported low-temperature electrolysis kinds of catalysts, so we're looking here at iridium oxide, and just quickly, we can see that in fact we see kind of the same rheological behavior of the iridium oxide as platinum on Vulcan, but at a much different concentration or wave fraction. So again, we have to take these things into account in terms of formulating the ink for a particular process.

Okay, so I talked about inks, but obviously, we're interested in understanding process parameters themselves. Here, we're looking at one possible roll-to-roll coating process, which is gravure coating. So if you're not familiar, basically, gravure works by spinning a patterned cylinder that you can see at the top through a trough of the ink, picking up the ink, and then wiping it against or frictioning it off against the web as it moves past that roller. And so at the bottom left, we're looking at different process parameters relative to the gravure coating process. In particular, we're looking at the volume factor of the cylinder, which is to say essentially the shape of the pattern, as a function of the roller speed ratio, which is the ratio between the rotational speed at which the cylinder is turning versus the linear speed that the substrate is moving.

And you can see clearly that both of these factors or parameters have an impact on the coat-ability and the defect formation in the coated electrode layer, so that's very interesting to understand. We also see in the image at the top right that these different parameters impact the achievable loading that we can reach in the electrode also.

And just to give you a sense that, yes, we can in fact make good electrodes with these processes, at the bottom right then is a comparison between our roll-to-roll coated electrode and the ultrasonically sprayed electrode, and you can see there that both in oxygen and in air we can get similar polarization performance.

Okay, so that's my presentation. I hope that was useful to you as an overview of the project here. I certainly want to acknowledge the contributions of many folks. I won't name them by name, but certainly, a lot of my colleagues here at NREL, as well as our partners at Lawrence Berkeley National Lab, who I didn't show any examples, but as Nancy mentioned, have been doing modeling for us on this project, both in terms of modeling of the diagnostic techniques, as well as doing in situ modeling to help us understand the impact of defects. In addition, partners at UC Irvine, collaborators at Argonne, who have done both USAX measurements and X-ray computed tomography measurements, all in the APS, and other partners at Georgia Tech and Colorado School of Mines.

With that, I think I will hand it back to Eric and Nancy, and thank you very much.

Eric Parker

Thank you, Mike. Very informative presentation. We'll begin our Q&A portion of the webinar now. I'll remind everyone to please submit your questions via the WebEx chat box on the slide you see here. And with that, I think Nancy's going to get us started with the first question.

Nancy Garland

Okay. Thanks, Mike, for a great presentation. With the roll-to-roll process, how fast can you go, and does it matter what kind of process you have? The RIF versus IRDC versus the through-plane reactive excitation?

Michael Ulsh

Great. Thank you, Nancy. Good question relative to the implementation of these techniques. So as I showed in the one example, we—on the web-line that we demonstrate these different techniques, we can go up to close to 100 foot per minute, which is, as far as I understand, at least at this point in time, that's a pretty relevant speed range for the processes that are being used for these fuel cell and electrolysis materials.

We ultimately want to look at that speed for each of the different techniques. We have done that to some extent, as shown in the one example for IRDC, although I would say that the differentiation between when we use one technique or another is really more of a function at this point of the material structure, where, for example, the IRDC is really more suited to a CCM structure, because you've got the single conductive layer on top of a non-conductive layer, and that can kind of drive the current through the electrode itself.

And so we're getting information specifically about the electrode, whereas we can't really do that in a GDE structure, because the gas diffusion layer is actually more electrically conductive than the electrode. So that's where we've developed a reactive excitation, which is probably better suited to a GDE structure.

Nancy Garland

Okay. Thank you. Can you elaborate on quality assurance methods of MEA framing and sub-gasketing in the roll-to-roll process?

Michael Ulsh

Yeah, so I—that's a great question, and I'll talk about needs here, because in general, we have not worked much at this point on implementation of QC for the incorporation of sub-gaskets. But certainly, that's a known need.

And the things that we think about there are really associated with alignment and registration, which is to say we're bringing together these multiple layers, and we need to make sure that all of the layers are correctly aligned. In particular, for example, we don't—if we're incorporating a sub-gasket, we don't want to have that sub-gasket skewed relative to the active area of the electrode, because where the sub-gasket and the electrode might overlap would create a thick spot within the MEA, and when you then compressed the MEA, that can lead to different degradation mechanisms that I think other people have elucidated.

So I think it's really—the thing that we have to focus on going forward relative to sort of looking at roll-to-roll processes for multilayer assembly is this idea of doing a good job of making sure that we have alignment and registration of the different layers.

Nancy Garland

Okay. The next question has to do with the testing process. Have you checked out to make sure that your inspection process does not affect fuel cell performance? Have you compared fuel cell performance say for example with the pristine sample, and one that has undergone inspection?

Michael Ulsh

Right. That's a great question, and I would say in some cases, yes, in other cases, not yet. But for example, for the IRDC and reactive impinging flow methods, to image the uniformity in defects in electrode layers, whether they be CCMs or GDEs, we have done that excitation, and we have then taken the cells that were sort of exposed to that excitation—I'm sorry, electrodes that were exposed that excitation, made cells out of them, and tested them, and compared them to pristine cells. And in fact, when we have appropriate excitation conditions—in other words, the conditions that we're really targeting, to get a detection, but not have overmuch excitation of the material, we see that we do not in fact see any impact of those excitations in the in situ performance.

Nancy Garland

Okay. Thanks. Next question wants to know are there any direct measurements for surface roughness in the gas diffusion electrode or the catalyst layer?

Michael Ulsh

Yeah, so I assume by direct—well, so I'll put it this way. In terms of ex situ measurement, so for example, an offline measurement, not something that we've looked at in-line, we have used things like both probe and optical profilometry to understand the roughness of—for GDLs or gas diffusion electrodes, with several of our partners, including Colorado School of Mines.

So those tend to be, at least in our experiences, good ways to characterize the roughnesses of those surfaces. We have not looked at any in-line methodologies to do that at this point. However, one of the things that we have looked at a little bit in, for example, some of our industry projects, some of which that we've reported on in our AMR presentations, is that we want to understand the impact on membrane surface roughness of different kinds of defects or other matter that you can imagine being incorporated in a membrane.

And so one of the things that we can see, for example, is that, again, we can change the configuration of the optical inspection technique to tell us something about the surface of the membrane and how certain kinds of materials might stand above the surface of the membrane because of how they're incorporated, and as you might expect, you might be interested in understanding that kind of information.

Nancy Garland

Okay. Thanks. The next one has to do with pinholes. Has the effect of pinholes been studied quantitatively? And then the follow-on question is, how small does a pinhole have to be to not really effect fuel cell performance?

Michael Ulsh

So again, great questions. To answer the first one, I would say that in fact, we have a paper right now that is in the submission process or publication process on exactly this topic, or at least starting to address this topic. I would say that there are quite a few studies out there in—that look at the in situ impact of pinholes in various ways, and of different sizes. So I think that there are a decent number of papers out there on that topic.

In terms of how large matters, is the question, for example, we presented some data at our AMR presentation just a couple of months ago where we were looking at a certain size of a pinhole that was let's say many tens of microns in diameter, and using this variety of different testing techniques, in fact, some of what I showed today was from that presentation.

So clearly, we can see the defects on that size scale can impact the performance and lifetime of the cell. We're in the process I would say of looking at smaller defects. We have looked at, for example, let's say MEAs with 10-micron or around 10-micron pinholes. We've only looked so far at initial performance, but I can say that we see no impact in initial performance of that size of pinhole, whereas we certainly see an impact in initial performance of that larger pinhole that I talked about today.

So, that's really an open question, and I can say that in the next year, certainly in our group here, we'll be continuing to explore that question.

Nancy Garland

Okay. And unfortunately, the last question that we've got, which AST did you use to assess the effect of pinholes?

Michael Ulsh

Yeah, so good question. Let's see. We did not use a specific, for example, DOE AST. We—the AST that we used is very similar to the combined mechanical and chemical AST sort of promulgated through Los Alamos and other activities, and now under FC-PAD. So it's basically that same combined chemical and mechanical AST. We just tweaked it a little bit because of some of the particulars of the operation of our test station. But it is essentially a combined chemical and mechanical AST that's very similar to the DOE AST.

Eric Parker

Okay. Thank you, Mike, for answering those great questions. Unfortunately, we are out of time. If your question did not get addressed, please feel free to contact Michael. If you could advance to the next slide, our contact information is available here, if you have further questions or want to reach out about specific slides you saw. But that does conclude our webinar for the day. I'd like to thank everyone for joining and remind you that the webinar and slides and the recording will be available online soon, so please be sure to check back on the website.

I also encourage everyone to sign up for our monthly newsletter, which includes information and registration for upcoming webinar topics. And other than that, have a great rest of your week, everyone, and goodbye.

[End of Audio]