Atomic Layer Deposition for Improved Biomass Conversion Catalysts
Wilson McNeary – National Renewable Energy Laboratory
Heterogeneous catalysts are a key enabler of the transition towards a sustainable, bio-based economy for fuels and chemicals. However, the harsh conditions in many biomass conversion processes lead to nanoparticle sintering, support collapse, and metal leaching in conventional PGM catalysts. This presentation will discuss ongoing work between the Catalytic Carbon Transformation and Scale-Up Center at NREL and various industrial partners to develop scalable and cost-effective atomic layer deposition (ALD) coatings for improving the performance of biomass conversion catalysts. Highlights include application of TiO2 ALD to dramatically increase the activity of a Pd catalyst towards aromatic hydrogenation, as well as preliminary results using ALD coatings to enhance catalyst selectivity in the production of sustainable aviation fuel from wet waste and reduce catalyst leaching in aqueous oxidation reactions for biobased chemicals.
0:00:01 Hello everyone. My name is Wilson McNeary and I'm a postdoc at NREL, in the catalytic carbon 0:00:06 transformation and scallops center. I'm going to talk today about some of our work using atomic 0:00:11 layer deposition to improve biomass conversion catalysts. 0:00:18 So as the world utilizes greater amounts of renewable energy, there is increasing discussion about 0:00:25 the hard to de carbonized sectors of the economy. These are areas such as heavy industry and 0:00:32 transport, which require large amounts of oil and oil products. There's a shift underway in 0:00:39 transportation for light duty transport towards electric vehicles. However, other methods of 0:00:47 transport, such as air travel and heavy shipping are going to be really hard to electrify. So 0:00:53 there's a need for research into carbon neutral or carbon negative fuels. one way in which this need 0:01:01 could be met is through the conversion of biomass into biofuels and bioproducts. This typically 0:01:07 involves a synthesis and upgrading step, which uses heterogeneous catalysts to convert the chemical 0:01:13 intermediates into the final bioproducts. In this step, the catalysts are often exposed to a number 0:01:20 of different stressors such as high temperature a condensed phase environment which may promote 0:01:26 leaching of the catalyst and impurities such as sulfur, which can deactivate the catalysts. So 0:01:33 there's great interest in developing a tailored catalyst solution that can mitigate all of these 0:01:38 different degradation pathways 0:01:44 here at in real we've been working to use atomic layer deposition or L. D. To protect catalyst in 0:01:51 these type of reactions. If you're watching this talk, I'm sure you're probably familiar with the 0:01:55 concept of a L. D. And how it works. But just as a general overview, this is a coding process in 0:02:01 which you can coat a starting surface using the four step process outlined below where you dose in a 0:02:10 metal organic precursor followed by a purge followed by a counter reactant, followed by another 0:02:15 purge that constitutes a single a lady cycle And you can repeat that as many times as you want. Some 0:02:22 of the unique aspects of a lady or that it's self limiting, which allows you to deposit formal films 0:02:28 on a range of different substrates. In particular, the ability to coat High surface area three 0:02:34 dimensional bulk powders has enabled a wide space of application around using L. D. Coatings to 0:02:40 protect catalytic nanoparticles. 0:02:46 one of the first examples of using led to protect catalysts was from a group at Argonne National 0:02:52 Laboratory. They showed that alumina L. D. Could help to preserve the small size of palladium 0:02:58 nanoparticles even when exposed to high temperatures for long amounts of time. They also found that 0:03:07 they had a lot of flexibility and how many cycles of A. L. D. They could put down on the catalyst 0:03:11 without degrading the catalytic activity. So up to 15 or so cycles, they found that the L. D. 0:03:18 Preferentially deposited on low coordination sites of the palladium nanoparticle, which were not 0:03:23 themselves active for the reaction they were looking at so they could get the protective effect of 0:03:30 the L. D. Film without negatively impacting reactivity. When they tested these catalysts in their 0:03:39 desire reaction, they found that they maintain their activity for a much longer time than uncoated 0:03:45 catalyst under really high temperatures and long reaction times. 0:03:53 This laid the groundwork for the work that we've been doing it in real trying to use a lady to 0:03:58 improve catalytic performance in reactions that are relevant to biomass conversion. Our first foray 0:04:05 into this was in collaboration with forged nano and johnson Matthey and we were looking at using L. 0:04:10 D. Coatings to improve performance in the Konik acid hydrogenation reaction. We found that aluminum 0:04:18 L. D. Could help to preserve the activity of these catalyst over long reaction times. It also helped 0:04:25 to impede the phase transition of the T. O. To support at high temperatures. It helped to preserve 0:04:32 the small size of the palladium nanoparticles and it had the added benefit of mitigating leaching of 0:04:38 the palladium in the acidic reaction environment. 0:04:44 The work I'm going to discuss today is a continuation of this, where we were looking at L. D. 0:04:48 Catalyst for use and aromatic hydrogenation reactions. Aromatic hydrogenation is an important step 0:04:55 in the production of biofuel. It's required to decrease aromatics in the final fuel, which in turn 0:05:03 allows for lower emissions when that fuel is burned. 0:05:09 The catalyst that we were looking at were done with an over coating procedure in which T 02 L. D. 0:05:14 Was deposited on a palladium supported on alumina catalyst. After an initial catalyst screening, we 0:05:21 found that the best performer of that bunch was one with 10 cycles of T. 02 which I'll be calling 10 0:05:28 c t 02 in this talk, that's the one I'm going to primarily focus on day today as we did a lot of in 0:05:35 depth characterization, reaction testing and computational modeling with this catalyst. 0:05:43 So when we examine this 10 C. T. 02 catalyst after the L. D. Procedure, we found that there was 0:05:49 about nine weight percent of T. I on the surface. Looking at the microscopy and E. DS maps, we can 0:05:56 see that there's a can formal layer of T 02 across the surface and the aluminum support was pretty 0:06:01 evenly coated. We did notice some, some significant changes in the kidneys option uptake both with 0:06:09 hydrogen and carbon monoxide. This indicates that the plane sites were at least partially covered by 0:06:16 the L. D. Process, although from just the kidneys option in the my Crosby images, we, the extent is 0:06:24 somewhat unclear, but this kind of brought us to our first big question in this work was how does 0:06:31 this coverage of the palladium by the T. O. To impact the reactivity. 0:06:41 We set out to examine these catalysts in batch reactions. Initially using a couple of different 0:06:48 aromatic molecules benzene, styrene and naphthalene. We found that in every case the 10 C. T. 02 0:06:55 actually had considerably higher conversion at the same time points than did the uncoated palladium 0:07:02 aluminum. We dug into this a little bit further with the napoli hydrogenation reaction. I'm showing 0:07:09 on the right the tetra land productivity of each catalyst, the 10 C. T. O. To the uncoated palladium 0:07:17 alumina and a palladium titania control catalyst. And you can see that the 10 C. T. 02 has higher 0:07:25 performance than those two uncoated materials. When we look at turnover frequency, which is 0:07:32 normalizing by the ceo kim absorption numbers that I showed on the previous slide. The difference is 0:07:38 even more stark with the 10 cto to drastically outperforming the other two catalysts. This is 0:07:44 because as we saw, there are fewer available palladium sites on this 10 C. T. 02 catalyst, but it's 0:07:50 doing more reaction with those fewer sites. 0:07:57 So this was an interesting finding that we kind of didn't expect, given that we knew we were going 0:08:02 to be covering some of the palladium sites. We wanted to make sure that we could replicate this in 0:08:07 flow reactions. So we tested napoli hydrogenation in a trickle bed reactor with those same three 0:08:13 catalysts and we found very similar trends where the 10 C. T. 02 outperformed both the uncoated 0:08:20 palladium alumina and the palladium titanium control case. 0:08:26 The 10 C. T. 02 ended up having about a 1.7 X. Higher activity than its uncoated base material. So 0:08:33 the application of the L. D. Layer alone is enough to increase the activity by that. Much. 0:08:39 Additionally, looking at the difference between the 10 C. T. 02 and the palladium supported on 0:08:46 titania, we can tell that there's there's a difference between the the L. D. Layer and its effect on 0:08:53 the catalytic performance versus just A T. 02 crystal and support. So there's something special 0:08:59 about having the close contact between the L. D. Layer and the palladium. So this brought us to our 0:09:07 next big question, which is why does the T. 02 L. D. Layer boost the activity? 0:09:17 We first aimed to address this question by characterizing the electronic structure of the catalyst 0:09:22 with XPS. And you can see here for the palladium uh segment of the XPS spectra, the binding energy 0:09:33 didn't really change when the 10 CT 02 L. D layer was added. We additionally got some X A. S done on 0:09:43 these catalysts and we found that the coordination number between palladium was exactly the same for 0:09:49 10 C. T. 02 versus the uncoated catalyst. So in both of these cases, application of the L. D layer 0:09:56 doesn't appear to really be changing the electronic structure of palladium itself. We also got some 0:10:03 I. S. S analysis done and we found for the aluminum and titanium peaks. Kind of the expected 0:10:11 behavior where for the 10 C t 02 and blew the alumina is almost entirely covered and there is the 0:10:18 emergence of a. T. I peak relative to the uncoated palladium alumina, additionally, zooming in on 0:10:25 the palladium peak in the spectrum. We can see evidence more evidence of the coating of palladium, 0:10:32 given that the intensity of the peak is slightly less than half for the 10 c t 02 catalyst. 0:10:43 So, since our experimental evidence pointed to no significant changes in the electronic structure, 0:10:49 but definite coverage of the palladium, we decided to look at this from a different angle. Using 0:10:55 some computational modeling. Our collaborators here at in real, developed a few different model 0:11:00 surfaces to test the absorption of different reaction intermediates on The one I'm going to focus on 0:11:08 is what we call the truncated T. 0 2 layer on palladium. Uh we expect that this is closest to the 0:11:16 experimental system we were working with, given that there's still available available platinum, but 0:11:21 it is partially covered by T. 02 We calculated the absorption energies of hydrogen, naphthalene and 0:11:29 petrol in on this truncated surface, which is shown in the right as the blue bars compared to bear 0:11:36 palladium. We saw some interesting shifts in the binding energy for these intermediates. The 0:11:44 hydrogen didn't really change. Um if anything, it got slightly more stable. However, the nazi lean 0:11:50 and petrol in got considerably less stable with the addition of that truncated T. 0. 2 Layer. We 0:11:58 believe this is because these molecules prefer to absorb in a flat configuration and they can't do 0:12:04 this is easily with the T. 02 blocking some of the large palladium domains on the surface. So we 0:12:15 believe that this combination of hydrogen binding in the hydrogenation reaction not really changing, 0:12:22 and a substantial decrease in the strongly bound aromatics, particularly the tetra lin product, have 0:12:31 the combined effect of increasing the rate of hydrogenation and giving us the phenomenon we're 0:12:37 seeing in our experiments, 0:12:43 as I mentioned, we were also interested in studying the stability of these catalysts with the L. D. 0:12:48 Layer. So we exposed them to a few different stressors and then retested them for naphthalene 0:12:54 hydrogenation activity. We took a subset of catalyst and exposed them to sulfur during the reaction 0:13:02 using dimethyl sulfide additive, we took some catalyst and expose them to a couple different thermal 0:13:10 treatments under dry air and we also exposed some catalyst to a hydrothermal treatment in liquid 0:13:17 water. Here, I'm showing the tetra in productivity of the catalyst after each of these different 0:13:24 treatments for the sulfide catalyst, we saw that they all lost some of their activity. However, the 0:13:32 ale decoded catalyst lost the least amount. 0:13:36 There were some interesting trends with The 450 c in the 750 c thermal treatment uh but one of the 0:13:44 biggest takeaways was that the 10 C. T. 02 catalyst not only retained its initial activity but 0:13:50 actually increased in activity over these different treatment steps. The hydrothermal treatment was 0:13:57 pretty detrimental to all of the catalysts shown in yellow, including the L. D. Catalyst, and we 0:14:03 believe a lot of that is due to evidence of support collapse and palladium centering in the 0:14:11 hydrothermal treatment. However, there were some interesting trends with the 10 C. T. 02 catalyst in 0:14:20 the ceo to I'm sorry ceo uptake, we found that it increased for each of these different treatments, 0:14:28 including the hydrothermal treatment. So we wanted to investigate that a little bit further 0:14:35 in order to dig into this further, we got some microscopy of the catalyst after the 7 50 C. 0:14:43 treatment and the 200 C hydrothermal treatment, you can see for the 10 C. T. O. Two, there is not a 0:14:51 significant change in the particle size which is for the 7 50 C. Thermal treatment, which is 0:14:56 promising, indicating that the L. D. Layer helped to preserve the palladium particles. But we do see 0:15:04 this really substantial increase in SEO uptake despite no real change in the particle size, We 0:15:10 believe this may be due to the formation of pores in the T. 02 layer. This calculation could open up 0:15:16 pores in the layer and expose more of the palladium, thereby increasing both the activity and the 0:15:22 ceo uptake. Mhm. The 200 C. Hydrothermal treatment as you saw, was really Detrimental to all the 0:15:30 catalysts, including the 10 c. t. 0. 2. We saw evidence of support collapse as well as in X. Rd 0:15:36 evidence of the alumina phase transition to bow My, which is known to be really detrimental to 0:15:41 catalytic performance. Interestingly though, in the 10 C. T. 02 we didn't see that much 0:15:48 agglomeration of the palladium as we did in the uncoated material. But there was a collaboration of 0:15:53 the A. L. D. Layer. You see those large chunks of T. I. On the surface indicate that the L. D. Layer 0:16:00 was unstable in that hot liquid water environment and a glom aerated together. So that's an area of 0:16:07 research. We're going to be following up on what the limitations are of these T. 02 L. D. Layers in 0:16:13 an Aquarius reaction environment and how much we can affect the stability and and keep those layers 0:16:21 from from showing the behavior that they had here. 0:16:27 Finally, we wanted to put some thought into the value proposition of ale de Forca Tallis is 0:16:32 specifically whether L. D. Would be worth it in the application of this hydrogenation catalyst. Our 0:16:40 collaborators at forge nano came up with a few different cost estimates for a 10 c t 02 L decoding 0:16:47 depending on the price of the L. D. Precursor and the capacity of the equipment used to do the 0:16:52 coding. We used a cost estimate of 43 84 per kilogram kind of as a worst case scenario, combine that 0:17:03 with a catalyst cost estimate from Israel's cat cost program and fed that into a simple techno 0:17:10 economic model of an aromatic hydrogenation process. 0:17:15 Were used this model to run a number of different scenarios in which we change the way to really 0:17:21 space velocity and the catalyst lifetime parameter. And these were meant to represent the different 0:17:29 performance improvements that might result from an ale decoding with waiter really. Space velocity 0:17:33 being an increased activity and catalyst lifetime being an increase in durability savings were found 0:17:43 in the positive direction for both of these parameters. In the case of weight over the space 0:17:46 velocity due to decreased operational and capital expenses and the case of catalyst lifetime due to 0:17:52 decreased catalyst purchase price, we found relative to the baseline that the cost of an L decoding 0:18:01 could be neutralized by just a two X increase in either of these parameters. So if the weight of the 0:18:08 space velocity or catalyst lifetime are doubled, the addition of an ale decoding more than pays for 0:18:15 itself, we think that a two X increase particularly weight early space velocity is achievable based 0:18:21 on the one point seven x activity increase that we saw experimentally in this work and of course Any 0:18:29 any increase in both parameters or beyond two x. Just compounds the benefit of the L. D. Coding even 0:18:37 further. So we think this is a really promising analysis that suggests LD coatings can be a feasible 0:18:44 option for catalytic processes if they increase the performance By at least two X or more. 0:18:59 So in conclusion, I've shown you that we can applying nail decoding to change the service absorption 0:19:05 energetic of aromatic molecules which can lead to improve process economics and we hope will 0:19:11 eventually lead to decreased harmful emissions by making it easier to remove aromatics from both bio 0:19:18 and fossil fuels. We're taking this work a step further right now by trying to use L. D. Catalyst in 0:19:28 the production of sustainable aviation fuels. And we're focused on a pathway to bio jet fuel that 0:19:35 involves making volatile fatty acids or V. F. A. S. From arrested anaerobic digestion and then 0:19:44 converting those into al canes for jet fuel through hydro de oxygenation or H. D. O. Given that HBO 0:19:53 has many similar mechanistic steps as hydrogenation. We felt that LD catalysts would also be 0:20:02 effective in that reaction. So far. We've done some preliminary where confound that with an L 0:20:08 decoded catalyst, we can dramatically shift the product distribution towards al canes in this multi 0:20:14 step reaction. Which is really promising and uh further indicates that there's utility of L. D. 0:20:21 Catalyst in the production of sustainable aviation fuel. Additionally, we've been working to use L. 0:20:29 D. Catalyst in oxidation reactions for making bio based chemicals such as glue connick acid. These 0:20:37 are challenging reaction environments given the oxidative atmosphere and the Aquarius Phase. So 0:20:45 we've been focused on reducing catalyst leaching. We found that in 15 hour leaching tests, LD 0:20:52 coatings can help to dramatically reduce leaching of the active metal species. Which is also 0:20:58 promising for this application as well. 0:21:05 I'd like to acknowledge funding for this work from the U. S. Department of Energy's bioenergy 0:21:09 Technologies office as well as our collaborators at forge nano and johnson Matthey. My group at in 0:21:16 real has led by Derek Vardon. And I'd also like to thank all of the collaborators in our group that 0:21:21 helped bring this work together. 0:21:25 I'd also like to thank you for your attention. I hope you enjoy the rest of the A. I. C. Summit and 0:21:30 I'd love to hear from you if you have any questions.