Atomic Layer Deposition for Enhanced Reactivity, Stability, and Sulfur Tolerance of Hydrogenation Catalysts

Nanoscale Changes and Macroscale Effects

Wilson McNeary – NREL


This work advances the application of ALD in heterogeneous catalysis by studying on the effects of a TiO2 coating on an aromatic hydrogenation catalyst. Scaleup of the ALD synthesis and industrial viability of the ALD-coated material were also considered.


0:00:01   Hello, everyone. My name is Wilson McNary. I'm a post doc it in RL on. I'm gonna be talking today
0:00:07   about some of our work using atomic layer deposition to enhance the reactivity, stability and sulfur
0:00:13   tolerance of hydrogenation catalysts. I'll begin with a little bit of background on the work that
0:00:21   we're doing. So some of you may know that Israel is located in the state of Colorado, which is well
0:00:28   loved for its outdoor scenery. We do have kind of a dirty secret here in Colorado, though, and
0:00:35   that's that. The air quality, particularly in Denver, is some of the worst in the country. It's
0:00:40   gotten so bad that in December of last year, the EPA downgraded the Denver area from moderate to
0:00:47   serious nonattainment in terms of its ozone pollution. So we clearly have a lot of work to do here
0:00:54   in Denver and around the world to address the problem of urban air pollution. One of the primary
0:01:03   contributors to urban air pollution is the combustion of diesel fuel. This can produce a range of
0:01:08   pollutants such as nitrogen and sulfur oxides, particular matter and polycyclic aromatic
0:01:14   hydrocarbons, or Ph is pH is air especially pernicious as they've been linked to increased incidence
0:01:21   of asthma, pulmonary disease as well as long and skin cancer. The amount of pH is produced during
0:01:28   combustion of a fuel is directly related to the amount of aromatics president in the fuel that's
0:01:33   being burned. Additionally, higher aromatic content has a negative impact on the quality of the fuel,
0:01:40   so it's in the interest of both the producer and the consumer to remove aromatics from diesel fuels.
0:01:48   This is typically done in industry through a catalytic hydrogenation process where in a two stage in
0:01:55   the second stage of a two stage process, AH, platinum group metal catalyst is used to saturate the
0:02:01   aromatic molecules and get rid of all the double bonds.
0:02:08   Even in conventional industrial processes, it's only typical to get down to around 20 to 30 weight
0:02:14   percent aromatics in the final fuel. And if we're really going to address the problem of urban air
0:02:20   pollution that many governments air pushing to lower the amount of aromatics that are allowed to be
0:02:27   in a fuel down to 15% or even less so, the Canada, the catalysis community, has been working on ways
0:02:34   to increase the activity of these hydrogenation catalysts to meet thes increasingly stringent
0:02:38   requirements in addition to the problem of activity catalysts, particularly those in hydrogenation
0:02:48   processes air prone to degradation. Uh, this is due to exposure to high temperatures, oxidizing and
0:02:55   reducing environments, as well as the condensed phase and coking from the carbonation species that
0:03:02   were trying to convert. Catalyst can degrade in these processes through a number of different
0:03:08   pathways, such as metal leaching, metal centering and the collapse of the support material.
0:03:17   Problems of catalyst, durability or especially pronounced in the biomass conversion space in which
0:03:22   biomass derived chemical intermediates are converted into biofuels and bio products. Catalytic
0:03:29   hydrogenation plays a role here as well, in processes such as the Valerie ization of lignin and
0:03:36   catalyst in these reactions were exposed to a lot of the same stressors as they are in fossil
0:03:42   derived aromatic hydrogenation. They may also be exposed to acidic and alkaline environment, such as
0:03:48   in the conversion of bio drive taconic acid into a depict acid. There's also the problem of sulfur
0:03:55   poisoning and other hetero atoms that air brought up into the biomass by the nutrients and the
0:04:02   plants that air processed, and the presence of these atoms. Such a sulfur can poison the platinum
0:04:09   group metal catalysts and degrade the catalytic activity. So taken together, there is a need to not
0:04:16   only increase the activity of hydrogenation catalysts but improve their durability as well. And our
0:04:24   hope is that we can find some kind of tailored catalyst solution that can address both of these
0:04:30   things. So in our work, we're investigating atomic layer deposition or L D. As that potential
0:04:41   tailored solution. I'm sure many of you watching this air familiar with how the L. D process works.
0:04:47   So I won't go into detail there. But the unique aspect of ale de that it can coat everything down
0:04:55   from the nano scale up to the bulk. Three D powder scale with a conform all film has opened up a
0:05:04   pretty wide application space and catalysis in using L. D. To protect catalytic nanoparticles.
0:05:15   This work was pioneered by the Lamb Group at Argonne National Lab. They found that illuminate L D
0:05:21   could protect palladium nanoparticles from centering and high temperatures. They also found that
0:05:28   they had ah, lot of flexibility with how much of a coding they applied. They could coat up. Thio 15
0:05:35   a. L D cycles or so before they started to degrade the activity of the catalysts by covering too
0:05:42   many sites. And that's because L. D selectively grows initially on the low coordination sites of
0:05:49   most PGM nanoparticles and those air sites that don't typically participate in reactions. So you can
0:05:57   get protective coatings without degrading the catalyst activity Azaz An extension of that. The
0:06:04   catalyst that they tested in this work. They found they could run them for really long times and get
0:06:12   great performance out of them, while an uncoated catalyst, which is degrade really quickly. So
0:06:19   there's a lot of potential here to use these for harsh reaction environments. So following in the
0:06:29   footsteps of this work, we've been using a lady codings for catalysts in biomass conversion. Here it
0:06:37   in RL. Our first foray into this was using aluminum coated palladium catalysts in the mechanic acid
0:06:44   hydrogenation. We found that these coatings enabled the catalyst to run for long reaction times
0:06:51   without losing significant amounts of activity. Three catalysts helped to impede the base
0:06:58   transformation of the Titanius support at high temperatures. They also helped preserve the plenum
0:07:05   nanoparticles and prevent centering, and we saw promising results as well with the lt coding helping
0:07:13   to mitigate leaching of the palladium in the acidic environment. The work I'm going to talk about
0:07:19   today is focused on aromatic hydrogenation with l decoded Catalyst in this project has been a
0:07:26   collaboration with Forge, Nano and Johnson Matthey. They initially gave us a whole suite of over
0:07:33   coated catalysts with different types of ale de coatings, and we screen those and found the top
0:07:39   performer was a palladium catalyst supported on aluminum with 10 cycles of T 02 over coated with a
0:07:48   lady. That's the one I'm going to spend the most time talking about today, given that we've done
0:07:53   some in depth characterization, reaction, testing and scale up analysis of that.
0:08:01   So this ale decoded catalyst with 10 cycles of t 02 I'm gonna be referring to as 10 c t i 02 When we
0:08:09   first analyze that catalyst after L. D. We found it had about nine weight percent of Titania
0:08:15   deposited on the service. Looking at Cem microscopy images, we concede that there's a conform a
0:08:21   layer of t I across the service, and it looks like the A luminous support was evenly coated by the
0:08:28   process. We did see some changes in the hydrogen and carbon monoxide uptake indicating that the
0:08:35   palladium was at least partially covered by the A L D layer. It's worth pointing out here that the
0:08:42   hydrogen uptake was less impacted than the CEO uptake. Waas on. This may point to some kind of
0:08:50   hydrogen spillover effect with the L. D Layer, where, since we're depositing a reducible oxide like
0:08:56   T I 02 this is creating high sites that hydrogen can absorb on at the same time that the palladium
0:09:07   sites are getting covered. So that may mitigate some of the losses in hydrogen uptake. But this
0:09:13   brought up kind of the first big question of this research, which, which was, How will the catalyst
0:09:20   perform when the palladium is partially covered by the L D layer? So in order to test this question,
0:09:29   we ran a Siris of aromatic hydrogenation reactions in batch. We initially looked at a couple of
0:09:38   different aromatic molecules such as benzene, styrene and nap. The late we found that the 10 cto to
0:09:45   catalyst doubled the conversion of the uncoated catalysts for each of these molecules.
0:09:54   Looking specifically Attn. Appling hydrogenation. We found that the productivity's of Tetra lengthy
0:10:03   single ring hydrogenated product was higher for 10 C t 02 than the uncoated catalyst, as well as a
0:10:11   control palladium on Titania Catalyst. When we normalized these results by the CEO Kim research
0:10:18   numbers to calculated turn over frequency, we saw even more starkly that the tin CTO to uh was much
0:10:26   more active than the other catalysts. And this is because the tin CTO, too, we know, has fewer
0:10:35   palladium sites accessible. So it's doing more reaction with fever sites.
0:10:47   These were very interesting results and we wanted to make sure that they could be replicated in flow.
0:10:51   So we tested those three catalysts again with Napoli hydrogenation in a trickle bed reactor, and we
0:10:58   found very much the same trend of the 10 CTO to outperforming the other materials. The 10 CTO to had
0:11:08   about a 1.7 times higher activity than the uncoated palladium on aluminum.
0:11:16   Additionally, we can see from these results that the mere presence of T 02 is not enough to get this
0:11:23   activity boost because the palladium supported on Titania does not have the same performance
0:11:29   enhancement. So there's something unique about the ultra thin ale de layer on top of the palladium
0:11:36   that's causing this activity, uh, boost. So that brought up the next question in this study, Which
0:11:44   was Why does the t o to ale dealer do this?
0:11:52   We first approached this question through the use of XPS here. I'm showing the palladium peaks for,
0:11:59   uh, the three catalyst and you can see that for palladium, alumina and black and 10 c t 02 in blue.
0:12:07   There doesn't appear to be any significant change in the binding energy of palladium, even with the
0:12:13   A L D layer, we also got some X A s analysis done. And from that we saw that there was no change in
0:12:21   the coordination number of palladium even when the A L D layer was deposited and when this analysis
0:12:29   was conducted in an institute reducing environment. So both of these results suggest that the L. D
0:12:37   layer does not really have much of an effect on the palladium Elektronik structure. We also got some
0:12:44   I s s done on these catalysts, which is a technique that looks at the first atomic layer and gives
0:12:50   compositional information. From that, we can confirm our our hypothesis that the alumina support is
0:12:58   mostly covered. You can see that The 10 c t 02 and blue. The aluminum peak is almost entirely
0:13:04   suppressed. Uh, for the palladium peak zoomed in on the right. You can see that the palladium
0:13:10   surface was partially covered, but as expected, we still got some palladium signal coming through
0:13:15   there. So that confirms what we were seeing with Kim absorption analysis that some of the palladium
0:13:21   is still accessible. So overall, we can see through this that the palladium in the Elektronik
0:13:29   structure, uh, is not changed by the t 02 So that is not the cause of the activity enhancement that
0:13:36   we're seeing in our reaction testing. We then approached this question from another direction.
0:13:44   Enlisting the help of a few collaborators for some dft studies. They set up a few computational
0:13:51   surfaces with bare palladium, a palladium covered by a full titania over layer and palladium covered
0:13:59   by a truncated titania over layer, which is shown here on the bottom. We expect this truncated over
0:14:05   layer is probably closest to the reality of our ale decoded catalyst. We then use these surfaces to
0:14:15   calculate the absorption energy of different intermediates on the surface is to see what affects the
0:14:23   L D layer might have on those intermediates. So on the right, I'm showing the binding energy of
0:14:30   hydrogen nap Selene and tetra lint on these three surfaces at the reaction temperature of 200 C. In
0:14:37   this plot, we could see a couple interesting trends. The hydrogen molecule appears to become more
0:14:43   stable with a with a more negative binding energy. In the case of the truncated over layer shown in
0:14:50   blue, however, Napoli and Tatchell in both become considerably less stable when the truncated over
0:14:56   layers added. This is because thes at aromatic molecules prefer to absorb in a flat configuration,
0:15:05   and when the T 02 layers there on the palladium surface, it interferes with the absorption of these
0:15:11   molecules and therefore increases their binding energy. We believe that this is actually the cause
0:15:21   of the increased activity that we see in our experiments because the surface is becoming more
0:15:27   favorable. The hydrogen binding, which is naturally good for a hydrogenation reaction, and we're
0:15:33   decreasing the binding of our strongly bound tetra and product, which clears the surface off faster
0:15:43   and helps speed up the reaction rate
0:15:49   I previously mentioned that we were interested as well in the stability of these catalysts. So we
0:15:54   did a series of experiments to examine how the A L. D layer impacted stability after exposure to a
0:16:01   few different stressors. The first was we exposed the catalyst to a sulfur containing molecule as a
0:16:09   contaminant. We also did a series of thermal treatments to simulate a regeneration cycle, and we did
0:16:20   a hydrothermal treatment to see how the catalyst held up when exposed to liquid water at high
0:16:25   temperature. Here I'm showing the results of these tests in terms of tetra and productivity. After
0:16:32   the exposure to the stressor, we can see for the soul flighted catalysts that they all lost some of
0:16:40   their initial activity. But the A L D catalysts held slightly more then the uncoated materials.
0:16:47   There were a number of interesting findings with the thermal treatments, but the big take away was
0:16:51   that the 10 cto to not only held its initial activity but actually gained activity at both the 450 c
0:16:59   and the 7 50 c thermal treatments. The hydrothermal treatment was pretty detrimental toe all of the
0:17:06   catalysts, but the A L. D catalyst did retain slightly more of its initial activity than the
0:17:12   uncoated materials waken see evidence of support, collapse and palladium centering in both the P D t
0:17:20   02 and the p D. A. L two a three catalysts. But interestingly, the 10 c t 02 appears to gain CEO
0:17:27   uptake as it's exposed both to the thermal and the hydrothermal treatments, which was a surprising
0:17:34   result.
0:17:39   We investigated this a little bit further with Cem postmortem microscopy. So for the 7 50 c thermal
0:17:46   treatment, that was one where the 10 c t. 02 gained a lot of, uh, CEO uptake, and we can see from
0:17:54   the microscope images that the T 02 layers still there, uh, after the 7. 50 c thermal treatment. So
0:18:01   what we believe happened is that the L. D layer may have formed pours during the callous a nation,
0:18:08   and this may have exposed the palladium nanoparticles and enabled MAWR accessibility for CEO and the
0:18:15   Kimmy's option experiments. This has been shown to happen previously with alumina over layers where
0:18:22   the thehyperfix chur oxidation could crack the films and crystallized them. Um, and they retain
0:18:29   their protective properties, But do allow slightly more, uh, reactant to get to the palladium site
0:18:39   with the hydrothermal treatment. There was, as I said, evidence of palladium centering in the
0:18:44   uncoated catalyst. But interestingly, in the coded catalyst, it looked like the palladium did not
0:18:51   center. But the L D layer actually did. So you can see those big blobs of titanium there on the
0:18:58   surface, uh, which are from the layer, crystallizing and retracting as it's exposed. That liquid
0:19:06   water environment. This was kind of an unexpected finding. And, uh, we also saw that the alumina
0:19:12   layer for both the coated and uncoated was transformed to the bow might phase, which had a negative
0:19:19   impact on the catalytic activity. We are gonna be doing some further work to understand the
0:19:26   mechanics of this ale de layer of collaboration in the liquid water environment, as this has
0:19:31   implications for using these catalysts in acquis chemistry's.
0:19:39   One of the most important aspects of this project was determining whether the catalyst that we made
0:19:45   could be scaled up in a reproducible way. So our collaborators at 40 Nano used there flu dies bed
0:19:52   technology to make us a couple additional batches of the 10 c. T I do catalyst. Our first batch was
0:19:59   made of the three grand scale, and these additional batches were made at the 100 grand scale. We
0:20:06   characterize these 100 grand batches and found that the activity enhancement was preserved. Even
0:20:12   though the catalyst synthesis has been scaled up by two orders of magnitude, you can see on the left
0:20:17   that thes 200 grand batches were actually slightly more active than the initially synthesized 10 c T
0:20:25   02 There was some deviation and the hydrogen uptake and B T surface areas you can see on the table
0:20:32   on the right. But once we got to the 100 grand scale, it appeared that the catalyst properties were
0:20:37   very repeatable. So these air positive results that indicate not only weaken make an effective
0:20:45   catalyst with L. D. But we can scale up the recipe to make this catalyst that industrially relevant
0:20:53   amounts.
0:20:57   Finally, we wanted to try to address the question of whether L D is actually worth it for
0:21:03   hydrogenation catalysts, because this comes up a lot. If we can demonstrate a lady in the lab,
0:21:10   that's all well and good. But is it prohibitively expensive? Thio Use in an industrial catalytic
0:21:17   process. So we approach this first by working with our collaborators that forged Nano to develop
0:21:25   some cost estimates for L D coatings at scale. Using the chemistry that we've been talking about in
0:21:32   this work, we found that, as expected, the coding cost is highly dependent on the assumption you
0:21:38   make about the price of the precursor that you're going to use. So we wanted to kind of evaluate a
0:21:45   worst case scenario here. So we took the maximum assumed coating cost of $43 a kilogram. Combine
0:21:55   that with an estimate from in rials, cat cost software, and put that into a simple model of an
0:22:01   industrial hydrogenation process. Just that second step where we're hydrogenated, uh, to a fixed
0:22:10   amount of aromatics in the final fuel.
0:22:14   We use this model to develop a sensitivity analysis where we change the weight hourly space velocity
0:22:21   of the catalyst as a stand in for catalyst activity, as well as the catalyst Lifetime to see what
0:22:28   kind of savings were generated in the in the product price as these parameters were changed. What
0:22:39   I'm showing here is the product price in dollars per metric ton relative to a baseline uh, selling
0:22:48   price as we improved the catalyst by changing the lifetime worthy activity
0:22:56   so you can see that in the lower left corner for the baseline. Just adding the L D coding does
0:23:04   increase the selling price quite a bit. But as we move in the positive direction in terms of
0:23:12   lifetime or wait hourly space velocity, we already generate savings beyond that level. That was,
0:23:22   that was added by the decoding cost and these savings air generated by in the case of weight, our
0:23:29   split wait hourly space velocity. You're decreasing the operational expense in the capital
0:23:34   expenditure by using a more active catalyst. And in the case of Lifetime, you're decreasing the
0:23:40   catalyst purchase costs because you don't have to spend as much replacing it as often, so we can see
0:23:49   that the L. D. The additional ale de cost is neutralized even in the case of improving the lifetime
0:23:56   or the weight. Early space velocity by two x, And given our experimental data showing that we had a
0:24:03   1.7 x increase in activity two X seems like a pretty achievable target. These savings will compound
0:24:11   with further improvements in either direction. So this points to a pretty accessible path for a lady
0:24:21   catalyst in aromatic hydrogenation. If process if improvements in the activity or lifetime, uh, can
0:24:28   be realized on the order of two X or greater. So I've shown you that we can use ailed de of T I 02
0:24:38   to make nano scale changes on the surface of a palladium catalyst. Thes changes significantly
0:24:46   improved the activity of the catalyst due to interfering with the absorption of aromatic molecules
0:24:53   on the surface. We believe that these nano scale changes can have macro scale effects because the
0:25:01   economics of the hydrogenation process can be significantly improved by the lifetime and activity
0:25:10   benefits that a nail decoding may in part. And this may enable greater dear of monetization of fuels,
0:25:18   thereby allowing for decreased harmful emissions. When those fuels air burned
0:25:27   for the future, Work at in Real were planning to continue working on a lady catalysts, but in the
0:25:34   capacity of oxidative chemical processes to produce different uh, bio derived molecules such as glue
0:25:42   connick acid. And we'll be focusing ah, lot on the stability of these catalysts and acquis
0:25:49   environments and preventing leaching and other phase transformations. With LD Layers so I'd like to
0:25:59   thank all of my collaborators. It forge Nano and Johnson Matthey as well as the P I on this project,
0:26:05   Derek Vardon and all of my teammates. It in RL and I would be happy. Thio. Take any questions you
0:26:15   have either through the live Q and A or or via email. Thank you for your attention.