Linking Catalyst and Process Development with Techno-Economic Analysis in the Conversion of Biomass to High-Octane Gasoline Webinar—Video Text Version
This is the text version for the Linking Catalyst and Process Development with Techno-Economic Analysis in the Conversion of Biomass to High-Octane Gasoline webinar.
>>Dan Ruddy: Thank you, everyone, for joining us today. My name's Dan Ruddy. I'm a senior scientist here at NREL, and it's a pleasure to discuss with you today how we've linked catalyst and process development with aspects of techno-economic analysis and process modeling in a research project that we've been exploring around biomass conversion to high-octane gasoline.
So, most of us are quite familiar with this approach to catalyst development and catalyst science, where we can couple theory, synthesis, and characterization of new materials with catalytic testing, and the results from each one of these aspects can feed back to one another to help improve computational models, design new materials to access your synthesis and characterization, and new reactions to run to understand catalytic performance. But, what I want to talk about today is how we've linked a second cycle around techno-economic analysis that can actually link back to this first more traditional catalyst development cycle.
By understanding some of the more high-cost aspects of the process, we can understand and gain information that can lead toward catalyst scaling and pilot-scale testing, but also, the results from the techno-economic analyses can feed back to that traditional loop in all aspects of catalyst testing, synthesis of new materials, and theoretical models that need to be developed. And that's what I'm gonna focus on today.
And all of our research and the conversion of biomass really comes back to this grand challenge, that stems from the complex functionality of biomass. You can see here—cellulose, hemicellulose, and lignin biopolymers that are highly complex oxygenates. But, in comparison, the fuels that we might wish to compare from this biomass are quite simple alkanes/alkenes, some minimal branching, maybe some aromatics, but, at the molecular level, these structures are quite different.
So, it's difficult to envision a direct pathway that can convert biomass to these alkanes and alkenes. So, rather, what we seek to do is identify intermediates along the way that we can access, and in this case, thermochemically. What I'll focus on is the gasification route. These intermediates, we want to have—we want to be able to access them in high yield, and we want to balance the stability of these intermediates with the reactivity to be able to further convert these and tailor the structures to the fuel molecules that we want to make.
So, if we're considering gasification technologies, the intermediate is syngas—carbon monoxide and hydrogen—which is fairly well accessed from feedstock, gasification, and clean up. From here, there are three routes that are flight mature, we would consider, in the conversion of syngas to fuels. First is Fischer-Tropsch, which directly converts the syngas intermediate to gasoline and distillates. The second route goes through a methanol intermediate, a secondary intermediate. Quite well understood—conversion of syngas to methanol via copper zinc oxide on [inaudible] catalyst. And from methanol, two routes: MTG—methanol-to-gasoline, and MODG—Mobil olefins-to-gasoline-and-distillates—are quite well known to access a variety of different fuels. Because these three technologies are pretty mature, we also understand the drawbacks associated with them that limits their industrial relevance.
So, in the case of Fischer-Tropsch, there's costly catalytic upgrading to produce quality fuels. You get a lot of straight chain and you get some waxes, so you need to further break those down to fuel molecules. And, in the case of the methanol routes, they tend to be capital intensive with either a high aromatics content in gasoline, or also, a high number of process steps to get your products. So, using process models and techno-economic analysis, we can look at these types of processes starting from biomass—including biomass gasification cleanup, and then further conversion. And, if we consider the net cost of production from biomass, we can see that even in the case of these mature technologies, the Fischer-Tropsch or MOGD were at $3.82 per gallon or $4.80 per gallon—not really cost competitive with today's market, and really highlighting the difficulty in generating cost-competitive biofuels.
And so, this really motivates us to seek new advanced catalyst and advanced processes to produce cost-competitive biofuels. And so, one route that we've been exploring as an alternative to FT or MTG or MOGD, but still utilizing the methanol intermediate is what we call the high-octane gasoline pathway—or the HOG pathway. And some methanol, we can develop a market responsive biorefinery concept. And what this means is that we can access all three of our common fuels from methanol. So, for example, dehydrative coupling of methanol-to-DME generates DME as a diesel fuel, which has been approved for use—especially in California.
You can also take that DME and convert it to hydrocarbons to generate a high-octane gasoline product, and that's what I'll focus most of our talk on today. But, you can also take the olefins from that high-octane gasoline product and couple those to a jet fuel. And so, we have the opportunity to balance the production of each of these types of fuels to meet market needs. And, alternatively, we're not confined only to biomass. Going through a syngas intermediate and methanol intermediate opens up routes to use other waste sources such as MSW or other renewable sources such as biogas.
So, we're looking at the overview of this DME to hydrocarbons process. [inaudible], we didn't develop this here at NREL. This was work that was funded out of BP, and some really excellent, ground-breaking work about 10 years ago out of Enrique Iglesia’s lab showing the conversion of dimethyl ether over a large-pore acidic zeolite such as beta-zeolite under mild conditions of just 200 Celsius, and usually closer to the one or two bar than high pressure. The product is a branch hydrocarbon mixture of C4 and C7s—typical product selectivities shown in the bar chart there—high selectivity to C4 and C7 with good selectivity to C5 and C6 as well, where the C5+ products would be considered the high-octane gasoline product that we desired to produce.
A few key points here. As I mentioned, DME-to-methanol—we have a variety of routes to get there, and quite versatile. The total product is great as a fuel mixture because of the variety of isomers that are produced. And, I'll highlight a little bit later, there are just some key differences between this type of process and MTG. And finally, this high-octane product is really attractive as a refinery alkylate blendstock, really bringing this high-octane fuel to the market. And again, the opportunity to couple the olefins to distillate.
So, we're often asked, "How does this really differ from MTG? What are some of the key differences?" And so, we highlight here—on the left—I must say, the methanol-to-gasoline pathway is a very excellent example of being able to convert methanol to what really resembles a finished gasoline product, having a mixture of alkanes, alkenes, and aromatics. But, it turns out, those aromatics can kind of work against you. In what we call the HOG pathway—the high-octane gasoline pathway—we favor branch hydrocarbon products with very minimal aromatics.
We get a different product because we use a different catalyst—in this case, utilizing a beta zeolite catalyst—and we also utilize lower severity conditions, that which also has the added benefit of a lower coking rate. Again, this leads to a product with the higher octane numbers—both in research octane and motor octane—and it really resembles a high-octane synthetic alkylate, which is one of the more valuable streams at a refinery. And, importantly, because of these lower severity conditions, lower coking rate, if we look at these two processes from a biomass front-end, we actually have the opportunity to access a higher yield of 65 gallons per ton in the HOG pathway versus 55 in MTG.
So, we can work with our process modeling group here at NREL and techno-economic analysis group, put together an Aspen model for the full conversion from biomass to this HOG product. And the great thing about this gasification technology is we can leverage commercial technologies all the way around from biomass to the production of DME in the bottom right-hand corner of the screen.
And then, we can assert this new technology of DME to high-octane gasoline, and we can look at some long-term targets, making assumptions about catalyst performance. And so, in particular, we can target 65 gallons per dry ton of biomass at a cost of production of $3.41 per gallon. Now, as you note, this $3.41 is lower than those traditional syngas conversion routes—even Fischer-Tropsch. But, if you look at putting the typical performance metrics of HBEA catalyst into this process model, we see that we only get a yield of about 40 gallons per dry ton biomass, and the cost of production is $5.20 per gallon. So, we don't really come anywhere close to these long-term targets with the parent beta catalyst.
So, if we look more closely into the TEA model, we understand some of the sensitivity around the catalyst development and the catalyst performance. So, here are four of the most—of the highest contributing cost to the hydrocarbon synthesis. And if you notice, the yield has the highest contribution. So, what we're looking at here is that the center zero line would be the base case scenario. So, shown on the right—65 gallons per ton. If we were able to increase that to 70—in this case, green is good, and we would reduce the production cost by 7.2 percent.
However, if that yield drops to just 60 gallons per ton, we would expect an 8 percent increase in the cost of production. We see the catalyst cost is important as well as the catalyst lifetime. Interesting, from this analysis, is that the single-pass DME conversion has only a minor effect on the cost of production. So, if we model it at 40 percent single pass, if that drops to just 25 percent, we see that only has about a 1.7 percent increase in the expected cost of production from this process. So, what this analysis really highlights is the importance of developing an inexpensive catalyst with a long lifetime that demonstrates a high selectivity to C5 products to increase the product yield. And selectivity is much more important than conversion.
So, considering the HBEA catalyst and the chemistry that's occurring over that catalyst, what really limits the performance to only 40 gallons per ton, for example? Well, the first off, when you look at the chemistry, it's hydrogen deficient. If you take a DME molecule and you pull water out of it—which you know the catalyst does—you're left with two CH2 units. And so, you're short on hydrogen to produce the alkanes that we see coming out of the reactor. And the way the BEA catalyst compensates for this is by generating a hexamethylbenzene as a key hydrocarbon aromatic product.
So, if you write a balanced equation out, you would start with 33 DME molecules; you can generate six of these very high-value trimethylbutane C7 compounds; 33 waters, which would make 2 hexamethylbenzene. So, this contributes significantly to yield loss that is not coming out of your C5+ products. Some really excellent work has been done—in particular, from Aditya Bhan’s group—looking at the mechanism of this type of reaction. These are details of the mechanism. They're not really important for today's talk, but mostly to highlight we do need to generate aromatics in the catalyst on the left hand side of the aromatic's carbon tool so that we can generate propylene that enters in the right-hand side of the mechanism, in order to grow the methylation to grow the hydrocarbon chain up to the C5, C6, and C7 products that we want to form.
And so, if we consider some of the limitations of HBEA and how we would address these from catalyst development, we need to shift products away from that aromatic cycle and push more of the carbon into this olefin cycle. We initially have [inaudible] we could do this and address the hydrogen deficiency by co-feeding a molecular hydrogen and that we could activate that with catalyst modifications to get it to participate in the reaction, but key to that would be maintaining that C5+ selectivity, as highlighted by the sensitivity analysis. And then finally, I showed earlier, there's a high selectivity to C4, and those are really key. We need to be able to reactivate those and get those back into the chain growth cycle to really realize this maximum of C5+ yield.
So, through the more traditional side of catalyst development—couple linked synthesis theory and catalytic testing—well, we generated a number of different catalyst targets. We tested those, and screened those, and what we found was that modification of the beta-zeolite with copper resulted in a two- to threefold increase in the hydrocarbon production rate—as shown in the top left chart there—and an extended lifetime. So, the red markers in the top left are the typical HBEA performance, and we have to stop the reaction after about 20 or 24 hours, because we've built up too much hexamethylbenzene in the reactor. However, after copper modification, we see an increase in the total hydrocarbon productivity, and we can run this reaction for longer than 100 hours. In this particular case, we stopped at 100, regenerated the catalyst, and the light blue markers—you can see that those match quite well to the beginning of the reaction.
Very important—we haven't changed the selectivity too much, and, if anything, we have a slightly higher selectivity to the C5, C6, and C7 products, which is advantageous. Looking more at the mechanism and pulling out some of the product ratios to understand where the product are coming from relative to the aromatics or the olefin cycle, we can see that for HBEA or HBEA with co-fed hydrogen, only about 20 to 25 percent of the products come out of the aromatic cycle, and almost 80 come out of the olefin cycle. However, when we add copper modification, we drive down the products from the aromatic cycle down less than 10 percent with over 90 percent of the carbon coming from the olefin cycle. We have a decrease in the HMB, as we expected, being able to run for a much longer time in the reaction and we've really favored these olefin cycle products. So, this copper catalyst seems to achieve the first two goals that we set out in catalyst improvement.
We've looked very deeply to try and figure out the role of copper is in this catalyst, and we worked closely with collaborators at Argonne National Lab to do X-ray absorption spectroscopy. And interestingly, we see contributions from both metallic and ionic copper in the catalyst. And so, we have this multifunctional catalyst now, where we have metallic copper that we hypothesize, activates the hydrogen, and we see an increased paraffin-to-olefin ratio resulting from that reaction. But, we also have cationic copper that we think facilitates the hydrogen transfer and dehydrogenation inside the catalyst cores. We explored this experimentally doing the reaction with deuterium, and as you can see in the left here, running just a para-beta catalyst with deuterium and looking at this major C7 triptane product. There's no deuterium in the product. Deuterium's not activated over the regular zeolite catalyst. But, after copper modification running the reaction with deuterium, we now see deuterium all over this triptane product and throughout the mass spectrum.
With respect to the cationic copper, we also looked at this dehydrogenation reaction feeding just isobutane over the copper BEA catalyst or the HBEA catalyst, and monitoring hydrogen production. And, as you can see, the HBEA catalyst doesn't produce any hydrogen under these conditions, but the copper BEA catalyst produces hydrogen quite extensively over—even up to 8 or 9 hours time on stream.
So, considering this dehydrogenation, we're quite intrigued by that, and it's a very interesting reaction. It's something that's very popular in literature, so, we went back to the process model to understand what kind of importance this would have within the TEA. And, it turns out, it's a critical component to the TEA. Because of the high selectivity to the C4 products, we have to be able to recycle those and get those back into the chain growth pathway. Looking on the right, we've put metrics now to what that importance would mean. This is just the high-octane gasoline synthesis cost—just that last portion of the process model.
With an HBEA catalyst, that would contribute $1.01 to the cost of the finished fuel. And that's with no C4 recycle. If we have a moderate amount of reactivation of the C4 products and reincorporation of the C5+ products, over the copper BEA catalyst, we can drop that down to $0.67. And, if we can really recycle this all the way to extinction, we could reach our target goal of just $0.38 per gallon for the hydrocarbon synthesis cost. So, this recycle is a critical component in the TEA to achieve these high yields and really realize those lower costs in production.
But why is it so difficult? Well, over the parent beta-zeolite catalyst, mechanisms have been studied extensively. Some really nice work out Enrique Iglesia’s group has shown that alkanes are considered terminal products and no re-incorporation. So, if you look at the highlighted area in green, that highlights you're doing a chain growth in propylene, and once you get the HT step to the butanes—isobutane or n-butane—that arrow only points one way, and those products do not come back into the cycle for further chain growth—the terminal products. Well, we've already shown that we can get some activation of isobutane over our copper BEA catalyst, so we wanted to study this more deeply—more on that more traditional left side of catalyst development looking at computational aspects and experimental aspects over the copper BEA catalyst.
So, what we did was we tried to isolate the different functionalities present in this multifunctional catalyst. We know we had metallic copper and we have cationic copper from the X-ray absorption. So, we made five materials to explore the difference between metallic copper and copper oxide on silica without any Brønsted acid sites, and we can control either copper oxide particles or copper metal just by the pre-treatment—whether oxidative or reductive. We can look at our regular HBEA catalyst, and then, we prepared two catalysts with just ionic copper. And we wanted to explore both copper(II) and copper(I). And again, we can control the copper(II) or copper(I) speciation just by the pre-treatment of the same ion exchanged copper BEA, where now, we have the Brønsted acid sites, copper ionic sites, but no metallic copper.
Looking just at the dehydrogenation reaction, feeding isobutane—which is major product from the DME-to-hydrocarbon C4 products—we can do this reaction at 300 degrees in a u-tube type reactor. And what we see is—I already showed the data that the HBEA catalyst was inactive, but it turns out that both copper and silica or copper oxide and silica are also inactive for this reaction, and no hydrogen was observed. However, both catalysts with ionic copper generate hydrogen in this reaction, and so, this tells us these ionic copper species are really the active sites for isobutane dehydrogenation. The copper oxide particles, copper metal, or Brønsted sites are not active.
To look more closely at copper(II) versus copper(I), we again went back to Argonne National Lab, and we had them do operando XAS experiments under essentially identical conditions as to what we did in our lab—feeding isobutane and monitoring the products. I show here on the top right the representative characteristic spectra for copper(I) and copper(II)—copper(II) in red; copper(I) in blue. So, you can see this clear shift in the XANES, so, these are indicative, and we'll be able to follow the copper fraction during the course of the reaction, which is shown on the bottom right. And these are the results over the copper(II) catalyst, and you can see that almost immediately, we have conversion from copper(II) to copper(I). In just 2 minutes, 80 percent of the copper is in the copper(I) state, and after two hours, 100 percent of it is.
If we load the copper(I) catalyst and do that pre-treatment, it remains copper(I) throughout the whole reaction. And at no point was metallic copper observed for these two materials. So, this really allows us to state that ionic copper(I) species are responsible for the observed dehydrogenation. And so, we can take this information now and feed it back into these theoretical calculations and we can look at reaction mechanisms and energetics to look at the reaction of isobutane to isobutene, only now that we know, it's a copper(I) site in the beta-zeolite.
So, we look at the mechanism of isobutane and explored both breaking at the primary CH bond first versus the tertiary CH bond first. Turns out, the primary route is slightly lower in energy, because the alkane is carbanion not a carbocation in this case. But we can go through this mechanism where we can see primary CH bond activation. We can put that—that goes to a Brønsted acid site on the zeolite. And then, that reaction with the tertiary CH to form isobutene ad hydrogen.
And what's important from these is we can look at these energetics, and we can compare to known dehydrogenation catalysts such as gallium or zinc in a zeolite, and what we find out is that these copper(I) species have higher barriers, and therefore, we would expect them to be less active in something like gallium or zinc. And this is important, because now we can move forward in the catalyst design and we can think about trying to improve this dehydrogenation performance by making bimetallic catalysts that maybe incorporate the species like gallium or zinc in these cation [inaudible].
So, this is really simulated to understand how dehydrogenation was occurring over this copper catalyst and at these ionic copper sites. So, we wanted to now look at experiments where we can simulate the recycle of isobutane in the presence of DME and hydrogen and try to get information that we can feed back into the process model in the techno-economic analysis.
So, what we did was we took our regular reaction conditions feeding DME and hydrogen, and now, we fed one percent of isobutane, representative of a recycle stream. And we did experiments with and without isobutane co-feed. We did them at a regular reaction conditions of 200 degrees and at a low pressure or a high pressure. And, importantly, we used isotopically labeled isobutane, because this would allow us to track 13-label carbon in the progress. And so, if we look at the conversion and the selectivity, there's not a very large effect in going with or without the isobutane co-feed, only a minor decrease in the conversion yield.
But what we see—particularly in the case of the isobutene product—if we do this reaction with just the HBEA catalyst, we see the mass spectrum that matches the missed spectrum for isobutene. However, when you use the copper BEA catalyst, now we see evidence of this isotopically labeled carbon in the isobutene product. So, this confirms the dehydrogenation activity and that this reaction is occurring, even in the presence of DME and hydrogen over this catalyst. But, it's not good enough just to dehydrogenate isobutene. We have to show this product gets into the C5+ products.
So, again, we can look at the mass spectra of C5 and C6 products from this reaction, and we see that in the cases of all three of these, we see the evidence of the isotopically labeled carbon in the C5 and C6 products when we use the copper BEA catalyst, and no evidence of isotopically labeled isobutane incorporation with just the HBEA catalyst. So, this indicates not only the reactivation of isobutane under these conditions, but reincorporation into these C5+ products. So, now, I'm gonna go back to our process design. We can take the performance metrics that we've measured over copper data, we can put them in, and to remind you, with the regular parent beta catalysts, we achieve about 40 gallons per ton yield at $5.20 per gallon. But now, with the performance of the copper BEA, we can achieve a yield of 56 gallons per drive time at $4.54 per gallon.
This is a 13 percent reduction versus the parent zeolite, a 40 percent increase in yield versus the parent zeolite, and it really requires both high productivity of the copper BEA catalyst and this unique ability to reactivate isobutane and reincorporating the C5+ products. And remember, I mentioned at the beginning that a process like MOGD for methanol, from the biomass front end, is around $4.80 per gallon. So, again, we're starting—we're even competing with these more mature technologies.
It's interesting to look at these C4 conversions and what we measured and what we would have expected to measure. So, under both of our conditions—at low pressure and high pressure—we measure isobutane conversion of about 14.5 percent and just over 23 percent at higher pressure. And these are remarkably high compared to the thermodynamic limitations of isobutane dehydrogenation that you would expect at 200 degrees. We would expect less than one percent conversion. So, this really suggests to us that the reactivity is kinetically controlled. We have a lot more to do to track this down and follow-up experiments to understand this, but presumably, it's through the consumption of both products isobutene—as evidenced by the heavy carbon in the C5+ products—and hydrogen, which we know the catalyst also activates at the metallic copper sites. This would be similar to other product removal concepts to drive reactions.
And, we could also—understanding now the structure of the catalyst, having cationic copper inside the zeolite core, we can start to hypothesize that maybe having that copper inside the core is important for dehydrogenation and generation of that alkene in close proximity to a Brønsted acid site that can really rapidly convert it and drive this reaction forward. We have to do a little bit of a stand and check and compare it to what's out there about isobutene methylation rates and high through transfer rates. And there's some great work from Dante Simonetti when he was with Enrique Iglesia’s group, where they looked at these two rates of isobutene over the HBEA catalyst. And they had values at 33 and 38 micromoles per mole aluminum per second. If we look at our isobutane conversion rates, they're somewhat lower at about 7 or 11.5 micromoles per mole aluminum per second, but this intuitively makes sense, because we have to do the dehydrogenation stuff as well. And it shows that we're roughly in the same order of magnitude for this type of conversion over this catalyst.
So, let's shift gears in adjust this last step of process design where we can take these olefins and couple them to a jet fuel. What we've done is taken a representative mixture of the DME to hydrocarbon's reaction where the mole percents here are shown near their representative structures—high NC4, NC7—and we've developed a process over a simple commercial Amberlyst catalyst under mild conditions to generate a couple of products that have hydrocarbon numbers in the distillate range. We've looked at carbon distribution analysis of this product, and the raw material coming out of the reactor still has some light contribution from unreactive C7 to C8, possibly dimerizations of C4. But a very simple back end distillation removes those light ends, and we get the distribution in red, which matches very nicely with their representative jet fuel example.
Might be a little bit off right in the very center of the distribution, but what's important is that there's no heavy molecules out past C22. That would really be problematic for something like jet fuel. We took this mixture and we did a variety of different ASTM tests for it, and it meets the specifications for density, viscosity, heat of combustion—very importantly, freeze point—and distillation curve. And so, while these branch compounds wouldn't be very attractive as a diesel fuel, it is, actually, quite attractive as a jet fuel blendstock.
So, again, we can go back to our process model, and we can update, just again, that keeping all the front end the same, starting just from our DME conversion, we can insert a couple new process steps where we take the C4s and rather than recycling them, we can put them through a dehydrogenation unit and we can put all of our olefins through an olefin coupling unit. And now, we can generate both a high-octane gasoline and a distillate product. Again, in the HOG only case, we’re at 56 gallons per ton—$4.54 per gallon. In the case of HOG plus distillate, we still get 29 gallons of the HOG product, 20 gallons of jet product, at a price of $4.71 per gallon. So, we do suffer a slight decrease in the total yield and an increase in the cost versus HOG only. It makes sense, we're adding additional capital. This is a whole new process and additional steps along the process. And inherently, we have a yield on the distillate—distillate yield is limited by the paraffin olefin ratio that degenerate in the HOG product. You can only convert the olefins that degenerate. But still, we're competitive with the MOGD as a benchmark at $4.80. So, in summary, I hope I've highlighted how we've utilized TEA and really coupled this with our R&D to both direct topics of research, but also, understanding the value of catalyst improvements that we've made.
We've developed this inexpensive copper BEA catalyst with a higher productivity and extended lifetime than the parent BEA catalyst, and importantly, we've shown that reactivates and re-incorporates isobutane, even in the presence of DME and hydrogen. This unique reactivity results in a 40 percent increase in yield and 13 percent reduced cost versus the HBEA in the process model. We've also shown that distillates can be produced from the olefins, but with an additional cost, since it requires additional capital. And ongoing R&D in this project—again, we could take these results from the computation that I showed that zinc and gallium should be more active for isobutane dehydrogenation than our copper(I) species that we identified. So we've been looking to develop bimetallic catalysts that can help control that paraffin-to-olefin ratio in the HOG product, and this benefits us twofold. We can control the HOG fuel properties by having olefins present that can help tune in the motor octane number, but also, we can direct more olefins to the distillate yield if we can generate more olefins in that process.
So, all the research that I presented today has been a great team effort—both the people here at NREL and collaborators at our National Labs. Great effort from our techno-economic analysis team and fuel property analysis collaborators here at NREL, as well as, again, our X-ray absorption spectroscopy collaborators at Argonne. This research was funded by the Department of Energy Bioenergy Technologies Office, and also, through ChemCatBio—a division of the Energy Materials Network. Thank you.
>>Moderator: We can take questions if anyone has them. Just let me take a second to come over to you with a mic.
>>Audience: Dan, can you comment on what you think might be limiting you to further increase the productivity of the high-octane gasoline? Like, what fundamental reaction steps could we further look at to improve that yield?
>>Dan Ruddy: Right. So, we can take a little bit from the sensitivity case there where the single-pass DME conversion might not be as important, because we can recycle the DME. But, the key areas are increasing that C4 recycle conversion. Right now, at around 23 percent, there's still quite a bit of a recycle stream. So, we can push that recycle conversion higher—say, to 40 percent, 50 percent. Now, we can start to generate a higher yield, less recycle cost, and that would drive down the cost and increase, essentially, again, the productivity, as you mentioned.
>>Audience: Yeah—follow on question to that. What is your theoretical maximum yield you could get if everything was perfect?
>>Dan Ruddy: I think it's pretty close to the 65 gallons per ton. Just based on carbon efficiency, known carbon efficiency losses through the gasification cleanup step. We're kind of in that 65 range.
>>Moderator: If there are no other questions, I will thank everyone for joining us today. Sorry about the technical difficulties. So, thank you, Dan.
>>Dan Ruddy: Okay. Thank you.